laser-induced breakdown spectroscopy

Elsevier AMS Job code: LRN Prelims-N51734 25-7-2007 8:20p.m. Page:xiii Trim:165×240MM TS: Integra

Font: Times Size:10/12pt Margins:Top:13MM Gutter:20MM T.Area:125mmX198mm 4 Color COP: Recto Floats: Inline

Preface

The acronym LIBS has a history almost parallel to the more popular acronym LASERwith the difference that the former is about twenty years younger and its first letterstands for laser. Although the production of sparks in air by a focused beam from apulsed ruby laser was observed in 1963, the use of spark emission for elemental analysisbecame a reality only around 1983, due to the pioneering spectroscopic investigationsof D. A. Cremer and L. J. Radziemski at Los Alamos National Laboratory in U.S.A.They also coined the name Laser-Induced Breakdown Spectroscopy (LIBS) for thistechnique in which spectra of laser-produced plasmas were used for qualitative as wellas quantitative spectrochemical analysis of condensed and gaseous samples without anyelaborate preparation. A thorough description of the physics of laser breakdown processesand laser-produced plasma is given in Laser Induced Plasma & Applications co-editedin 1989 by the two pioneers of LIBS. The authors not only summarized the work carriedout in this field during the previous 25 years but also pointed out its advantages anddisadvantages. In view of the rapid developments in laser and detection technologies,they predicted its widespread use in the future. During the past decade and a half,technology has produced more reliable lasers, charge coupled detectors, and miniaturespectrographs with its capabilities of recording spectra over a wide range of wavelengths.The combination of these technologies has produced unprecedented enhancements in thesignal-to-noise ratio. LIBS has rapidly developed into a major analytical technology withthe capability of detecting all chemical elements in a sample without any preparation,of real-time response, and of close-contact or stand-off analysis of targets. The presentbook includes the latest developments in the experimental techniques and applicationsof LIBS. It should be useful to analytical chemists and spectroscopists as an importantsource of information and also to graduate students and researchers engaged in the fieldsof combustion, environmental science and planetary and space exploration.

Understanding the major components in a LIBS experiment and the physics oflaser-target interactions are essential to appreciate the new vision of LIBS performancecapabilities. These basic ingredients are discussed in Part I (Basic Physics and Instru-mentation) of the book, comprising the first five chapters. The first chapter contains theeffects of laser beam characteristics on its focusing behavior and on the production oflaser sparks in gaseous samples and of plasma plumes from solid samples. The principleof charge-coupled detectors (CCD) and their incorporation in compact spectrographsfor broad-band detection are also briefly described. In Chapter 2, a brief account ofthe electronic structure of atoms and their quantum states is given. The allowed andforbidden transitions are discussed in the electric-dipole approximation, and the originsof continuum as well as line emission from atoms are explained. The broadening ofspectral lines is related to the physical conditions around the radiating atoms and theeffects of electric fields in a plasma environment are discussed in detail. Applications

Elsevier AMS Job code: LRN Prelims-N51734 25-7-2007 8:20p.m. Page:xiv Trim:165×240MM TS: Integra


xiv Preface

of atomic emission spectroscopy in determining electron density, electron temperatureand qualitative as well as quantitative spectrochemical analysis of the source are brieflydiscussed. Ablation forms the subject matter of Chapter 3. The term ‘ablation’ describesthe explosive vaporization of material irradiated by a laser beam. In general, the abla-tion rates depend on the material, the laser wavelength, the ambient atmosphere andthe geometry of the laser beam. Two different types of ablation mechanisms can bedistinguished: the photochemical and the photothermal. Regardless of the mechanism,the dominating effect of every kind of ablation is an extreme short-term temperatureincrease of the irradiated material surface which is the starting point of different physicaland chemical reactions. The characteristics of a radiating plasma produced by a laser isstrongly dependent on its pulse duration and irradiance. The influence of laser ablationon LIBS is discussed in this chapter. The physics of LIBS involves many processes ofwhich ablation and plasma formation are of great significance, but the process of opticalemission from the plasma is the crucial one for obtaining spectroscopic informationabout the constituent atomic species. A significant fraction of the incident laser pulseenergy is absorbed in the expanding plasma plume, causing the atoms and ions to reachdifferent states of excitation and subsequent optical emission. The physics of these opti-cal processes involving absorption of laser radiation and emission from the plasma plumeforms the subject matter of Chapter 4. The origins of continuum and line emission fromlaser-produced plasmas is discussed with a view to emphasize the importance of spatialas well as temporal resolution of the optical emission. Designs of experimental setupsfor obtaining maximum sensitivity of measurement are described in great detail. Thecontents of Chapter 5 deal with LIBS instrumentation. It has three major components:the laser, the ablation chamber, and the detection system for optical emission. LIBS is,however, a versatile technique for detection and identification of elements in a varietyof environments. Each one of these situations requires some kind of modification of thestandard LIBS instrumentation. Some of these unusual experimental arrangements arediscussed with emphasis on remote detection systems and portable detection systems.

Part II of the book comprises of Chapters 6 to 8, dealing with New LIBS Techniques.The technique of dual laser pulse LIBS (described in Chapter 6) has proved fruitful inimproving the signal-to-background ratio (S/B) and signal-to-noise ratio (S/N) relativeto conventional single-pulse technique. Experimental configurations with collinear andorthogonal impact of two laser pulses separated by a few microseconds have beenused resulting in a ten-fold or more enhancement in the LIBS signal. Although themechanism of increase in signal is not very well understood at present, this new LIBStechnique has been very widely used in the recent past to improve the reproducibilityand limit of detection. The use of femtosecond lasers in LIBS is discussed in Chapter 7.The interaction of such ultra-fast laser pulses with materials is very different from thatof nanosecond laser pulses commonly employed in LIBS. The fundamental physicalprocesses involved during laser ablation and applications of this technique are presentedin this chapter. Chapter 8 contains the results of the new technique of micro-LIBS,employing laser pulses with energy in the range of micro-joules to less than a milli-joule.Such low energy laser pulses permit two-dimensional microanalysis of material surfaceswith spatial resolutions approaching a micron. This technique has great potential in thedevelopment of portable LIBS systems.

The variety of LIBS applications are covered in Part III of the book in Chapters 9through 18. LIBS can be utilized in the detection of trace metals in the off-gases from

Elsevier AMS Job code: LRN Prelims-N51734 25-7-2007 8:20p.m. Page:xv Trim:165×240MM TS: Integra


Preface xv

industrial plants, traffic, volcanoes, wild fire and combustion processes. The resultspresented in Chapter 9 demonstrate that LIBS can be used as continuous emission monitorand also for the metallic species in the exhaust of rocket engines. Analysis of liquidsamples is of great importance in the context of environmental studies and of moltenmetals. The use of LIBS for determining the chemical composition of such samplesand many others, and also methods of enhancing the precision of such measurementsare described in Chapter 10. One of the major problems, in glass, aluminum, and steelindustries is the need for real-time measurement of constituents of the melt. LIBS canprovide rapid, in-situ melt composition measurements. It also allows chemical additionsto be made to the melt so that an acceptable product composition is achieved prior todraining a furnace. In Chapter 11 experimental arrangements are described, based onfiber optic LIBS sensor to measure in-situ elemental composition of solid and moltensamples. Chapter 12 deals with the elemental analysis of powder samples using LIBS.Powder materials, both granular as well as fine powders, represent the most commonform of raw material in industries, like chemical, pharmaceutical, glass and ceramics,food and others. It has been shown with examples, that LIBS can be used for on-linemonitoring of the elemental composition of the powder material before it is fed intoa process. The detection of chemical and biological agents that pose threat to humanlife, form the subject matter of chapter 13. Such agents are complex molecules andintricate living structures and it is not readily clear as to how an elemental analysistechnique should be useful in their analysis. The application of broadband spectrometersto LIBS in recent years has led to very accurate data on elemental ratios making itpossible to determine stoichiometry of a broad range of compounds. The use of LIBSin the analysis of chemical and biological agents in air, water and particulate matterhas been discussed in detail. Life science applications of LIBS discussed in chapter 14deals with the analysis of elemental composition of biological samples. The capabilityof LIBS for estimating trace elements in a single cell has potential medical applications.The relative concentrations of major as well as trace elements in normal and malignantcells have been determined. Physical parameters during laser ablation of teeth have alsobeen studied. Chapter 15 is concerned with the determination of total carbon contentof soil using LIBS. This has great significance in view of suggestions that soils andvegetation could be managed to increase their uptake and storage of CO2 and thusbecome ‘land carbon sink’ to reduce anthropogenic emissions of carbon dioxide. LIBSfor space exploration, one of the most exotic and exciting applications is described inChapter 16. This application is based on the stand-off capability of LIBS and resultsof measurements on atmospheric conditions simulating Mars are discussed. Chapter 17is devoted to the detection and analysis of chemical composition of aerosol particles –a complex mixture of nitrates, sulphates, chlorides, water etc- originating from bothnatural and anthropogenic sources. Quantitative aerosol analysis is presented in termsof the aerosol-sampling problem followed by direct and indirect aerosol measurements.It is always a difficult task to predict the future course of developments in the field ofscience and technology but one can point out the existing shortcomings and possibledirections for future researches in LIBS. This has been done in chapter 18 of this book.Many existing applications have not been put to practical use due to insufficient accuracyand precision of LIBS. The extension of echelle spectrometer to VUV range will permitthe detection of non-metals (S, P, Cl, Br) which are very important in process analysis.There is no suitable theoretical model at present to explain the laser ablation and plasma

Elsevier AMS Job code: LRN Prelims-N51734 25-7-2007 8:20p.m. Page:xvi Trim:165×240MM TS: Integra


xvi Preface

formation in a LIBS experiment. Developments along these and some other directions areexpected to make LIBS a very important field for science and technology in the future.

We gratefully acknowledge the imaginative contributions by the authors who sparedtime from their busy schedule of research and teaching for this book. We are also gratefulto the members of Laser and Spectroscopy Laboratory at the Banaras Hindu Universityand of the Institute for Clean Energy Technology at the Mississippi State University fortheir enthusiastic help during the preparation of the manuscript. Special thanks are dueto Mr. Sushil K. Singh, Dr. Vineeta Singh, Dr. Rajamohan R. Kalluru and Dr. S. B.Rai for their valuable editorial suggestions and help. We wish to record our gratitude toour wives Mrs. Shila Singh and Mrs. Vaidehi Thakur for their exemplary cooperation,patience and understanding.

Elsevier AMS Job code: LRN Prelims-N51734 25-7-2007 8:20p.m. Page:xvii Trim:165×240MM TS: Integra


Contributors

S. Michael AngelDepartment of Chemistry and BiochemistryUniversity of South CarolinaColumbia, SC 29208, USA

Steve BuckleyDepartment of Mechanical and Aerospace EngineeringUniversity of California, San DiegoLa Jolla, CA 92093, USA

David A. Cremers4300 San Mateo BoulevardApplied Research Associates, Inc.Albuquerque, NM 87110, USA

I. V. CravetchiDepartment of Electrical and Computer EngineeringUniversity of AlbertaEdminton, Alberta, T6G2V4, Canada

R. FedosejevsDepartment of Electrical and Computer EngineeringUniversity of AlbertaEdminton, Alberta T6G2V4, Canada

C. T. Garten Jr.Environmental Sciences DivisionOak Ridge National LaboratoryOak Ridge, TN 37831, USA

J. J. Gonzalez1 Cyclotron RoadLawrence Berkeley National LaboratoryBerkeley, CA 94720, USA

Elsevier AMS Job code: LRN Prelims-N51734 25-7-2007 8:20p.m. Page:xviii Trim:165×240MM TS: Integra


xviii Contributors

David HahnDepartment of Mechanical Engineering and Aerospace EngineeringUniversity of FloridaGainesville, FL 32611, USA

Akashya KumarDepartment of PhysicsTuskegee UniversityTuskegee, AL 36088, USA

Bansi LalCenter for Laser TechnologyIndian Institute of TechnologyKanpur 208016, India

X. L. Mao1 Cyclotron RoadLawrence Berkeley National LaboratoryBerkeley, CA 94720, USA

M. MartinEnvironmental Sciences DivisionOak Ridge National LaboratoryOak Ridge, TN 37831, USA

A. V. PalumboEnvironmental Sciences DivisionOak Ridge National LaboratoryOak Ridge, TN 37831, USA

Ulrich PanneDepartment of ChemistryHumboldt-Universitaet zu BerlinRichard-Willstaetter-Str. 1112489 Berlin, Germany

A. K. RaiDepartment of PhysicsAllahabad UniversityAllahabad 211002, India

V. N. RaiLaser Plasma DivisionCentre for Advanced TechnologyIndore 452 013, India

Elsevier AMS Job code: LRN Prelims-N51734 25-7-2007 8:20p.m. Page:xix Trim:165×240MM TS: Integra


Contributors xix

D. K. RaiDepartment of PhysicsBanaras Hindu UniversityVaranasi 221005, India

R. E. Russo1 Cyclotron RoadLawrence Berkeley National LaboratoryBerkeley, CA 94720, USA

Mohamad SabsabiNational Research Council CanadaBoucherville, Québec, J4B 6Y4 Canada

Louis St-OngeNational Research Council CanadaBoucherville, Québec, J4B 6Y4 Canada

J. ScaffidiDepartment of Chemistry and BiochemistryUniversity of South CarolinaColumbia, SC 29208, USA

Jagdish P. SinghInstitute for Clean Energy Technology (ICET)Mississippi State UniversityStarkville, MS 39759, USA

M. T. TaschukDepartment of Electrical and Computer EngineeringUniversity of AlbertaEdminton, Alberta, T6G2V4, Canada

Surya N. ThakurDepartment of PhysicsBanaras Hindu UniversityVaranasi 221005, India

Y. Y. TsuiDepartment of Electrical and Computer EngineeringUniversity of AlbertaEdminton, Alberta, T6G2V4, Canada

S. D. WullschlegerEnvironmental Sciences DivisionOak Ridge National LaboratoryOak Ridge, TN 37831, USA

Elsevier AMS Job code: LRN Prelims-N51734 25-7-2007 8:20p.m. Page:xx Trim:165×240MM TS: Integra


xx Contributors

J. Yoo1 Cyclotron RoadLawrence Berkeley National LaboratoryBerkeley, CA 94720, USA

F. Y. YuehInstitute for Clean Energy Technology (ICET)Mississippi State UniversityStarkville, MS 39759, USA

Elsevier AMS Job code: LRN Prelims-N51734 25-7-2007 8:20p.m. Page:xxi Trim:165×240MM TS: Integra


Acronyms

Å AngstromAAID Advanced Analytical Instrumentation DemonstrationAAS Atomic Absorption SpectrometryAES Atomic Emission SpectroscopyAPI Active Pharmaceutical IngredientAPXS Alpha Proton X-Ray SpectrometerArF Argon FluorideBBO Barium BorateBKG BackgroundCBE Conduction Band ElectronCCD Charge Coupled DeviceCE Coronal EquilibriumCEM Continuous Emission MonitoringCFFF Coal Fired Flow FacilityCIR Cumulative Intensity RatioCM Corona ModelCM Chlorpheniramine MaleateCPA Chirped Pulse AmplificationCPM Colliding Pulse Mode (Locked laser)CRM Collisional-radiative ModelCW Continuous WaveDCP-AES Direct Current Plasma Atomic Emission SpectrometryDIAL Diagnostic Instrumentation and Analysis LaboratoryDM Dichroic MirrorDMA Differential Mobility AnalyzerDOE Department of EnergyDU Depleted UraniumEP European PharmacopeiaEPA Environmental Protection AgencyEr: YAG Erbium Yttrium Aluminum GarnetESAWIN Echelle Spectra Analyzer software for WINdowsFDA Food and Drug AdministrationFO Fiber OpticFRAS Facility for Remote Analysis of Small BodiesFRC Field Research Centerfs femtosecond

Elsevier AMS Job code: LRN Prelims-N51734 25-7-2007 8:20p.m. Page:xxii Trim:165×240MM TS: Integra


xxii Acronyms

FTIR Fourier Transform InfraredFWHM Full Width at Half MaximumGRIN Gradient IndexHEPA High Efficiency Particulate Air (filter)HMX High Melting eXplosive (octogen and cyclotetramethylene

tetranitramine)HPLC High Performance Liquid ChromatographyIB Inverse BremsstrahlungICCD Intensified Charge Coupled DeviceICP-AES Inductively-Coupled Plasma Atomic Emission SpectrometryICPES Inductively Coupled Plasma Emission SpectroscopyICP-MS Inductively Coupled Plasma Mass SpectrometryID Inner DiameterIDAD Intensified Diode Array DetectorIPCF Instituto per I Processi CHimio FisiciIR InfraredJPL Jet Propulsion LaboratoryKrF Krypton FluorideKTP Potassium Titanium Oxide Phosphate (KTiOPO4)LA Laser AblationLASER Light Amplification by the Stimulated Emission of RadiationLASIK Laser Assisted in situ KeratomileusisLBR Line-to-Background RatioLDRD Laboratory Directed Research and DevelopmentLEAF Laser Enhanced Atomic FluorescenceLEAFS Laser Excited Atomic Fluorescence SpectrometryLIBS Laser Induced Breakdown SpectroscopyLIDAR Light Detection and RangingLIF Light Induced FluorescenceLIP Laser Induced PlasmaLIPS Laser Induced Plasma SpectroscopyLMA Large Mode AreaLNR Line-to-Noise RatioLOD Limit of DetectionLPF Laser PhotofragmentationLSC Laser Supported CombustionLSD Laser Supported Detonation (waves)LSR Laser Supported RadiationLTE Local Thermodynamic EquilibriumLTSD Lens-to-Surface DistanceMACT Maximum Achievable Control TechnologyMALDI Matrix Assisted Laser Desorption/IonizationMALIS Mars elemental Analysis by Laser Induced Breakdown

SpectroscopyMER Mars Exploration RoverMHD MagnetohydrodynamicsMIP-AES Microwave Induced Plasma Atomic Emission Spectrometry

Elsevier AMS Job code: LRN Prelims-N51734 25-7-2007 8:20p.m. Page:xxiii Trim:165×240MM TS: Integra


Acronyms xxiii

MIS Metal Insulated SemiconductorMIT Massachusetts Institute of TechnologymJ milli-JouleMPI Multiphoton IonizationMS Mass SpectrometryMSE Mountain States EnergyMSL Mars Science LaboratoryMSU Mississippi State UniversityMW MegawattNA Numerical ApertureNABIR Natural and Accelerated Bioremediation ResearchNASA National Aeronautics and Space AdministrationNd: YAG Neodymium Yttrium Aluminum GarnetNd:YLF Neodymium: Yttrium Lithium FluorideNETL National Energy Technology Laboratoryng nanogramNIR Near InfraredNIST National Institute of Standards and Technologynm nanometerNMR Nuclear Magnetic ResonanceNRC National Research Councilns nanosecondOD Outer DiameterORNL Oak Ridge National LaboratoryP/B Peak-to-BasePAT Process Analytical TechnologyPC Personnel ComputerPCA Principal Component AnalysisPDA Photodiode ArrayPETN Pentaerythritol TetranitratePLD Pulsed Laser DepositionPM Particulate MatterPMT Photomultiplier Tubeppb part per billionppm part per millionps picosecondPVA Polyvinyl AlcoholPVC Polyvinyl ChlorideRA Relative AccuracyRBC Red Blood CellRCRA Resource Conservation and Recovery ActRDX Royal Demolition eXplosive (1,3,5-trinitro-1,3,5-triazine)R-FIBS Remote Filament Induced Breakdown SpectroscopyRKIS Rotary Kiln Incinerator SimulatorRLIBS Resonance Laser Induced Breakdown SpectroscopyRM Reference MethodRMS Root Mean Square

Elsevier AMS Job code: LRN Prelims-N51734 25-7-2007 8:20p.m. Page:xxiv Trim:165×240MM TS: Integra


xxiv Acronyms

rpm revolutions per minuteRSD Relative Standard DeviationRSP Repetitive Spark PairS/B Signal-to-Background RatioS/N Signal-to-Noise RatioSAIC Science Applications International CorporationSEM Scanning Electron MicroscopySEM-EDX Scanning Electron Microscopy – Energy Dispersive X-raySESAM Semiconductor Saturable Absorber MirrorSHL Superheated LiquidSMA SubMiniature version A (fiber connector)SMR Surface Map-ping RateSOM Soil Organic MatterSSC Stennis Space CenterSSTB Slip Stream Test BedSTP Standard Temperature and PressureTE Thermodynamic EquilibriumTEM Transverse Electric ModeTFP Thin Film PolarizerTNT Tri-nitro TolueneTOF Time-of-FlightTTL Transistor-Transistor LogicTW Tera-WattU.S. United StatesUSN Ultrasonic NebulizerUSP United States PharmacopeiaUV UltravioletUV-VIS Ultraviolet VisibleVUV Vacuum UltravioletXeCl Xenon ChlorideXRF X-ray Fluorescence�J Micro-Joule�s microsecond

Elsevier AMS Job code: LRN Ch01-N51734 24-7-2007 7:53p.m. Page:3 Trim:165×240MM TS: Integra


Chapter 1

Fundamentals of Laser Induced Breakdown Spectroscopy

S. N. Thakura and J. P. Singhb

aLaser and Spectroscopy Laboratory, Department of PhysicsBanaras Hindu University, Varanasi-221005, INDIA

bInstitute for Clean Energy Technology, Mississippi State UniversityMississippi State, U.S.A

1. INTRODUCTION

The devastating power of the laser was demonstrated soon after its invention whena focused laser beam produced a bright flash in air similar to the spark produced bylightening discharge between two clouds [1]. Another spectacular effect involved theproduction of luminous clouds of vaporized material blasted from a metallic surfaceand often accompanied by a shower of sparks when the laser was focused on a metalsurface [2,3]. These laser effects have found many technological applications in the fieldsof metalworking, plasma production, and semiconductors. When a pulsed laser beam ofhigh intensity is focused, it generates plasma from the material. This phenomenon hasopened up applications in many fields of science from thin film deposition to elementalanalysis of samples. The possibility of using a high-power, short-duration laser pulseto produce a high temperature, high-density plasma was pointed out by Basov andKrokhin [4] as a means of filling a fusion device by vaporizing a small amount ofmaterial. Laser ablation of solids into background gases is now a proven method ofcluster-assembly [5,6]. In this method, a solid target is vaporized by a powerful laserpulse to form partially ionized plasma that contains atoms and small molecules. Notmuch is known about the formation and transport of particles in laser ablation plumes.In recent years there has been notable interest both in an increased understanding oflaser induced plasmas (LIP) and in the development of their applications. Emissionspectroscopy is used for elemental analysis of targets from which the luminous plasmais generated and it can also be applied to determine the temperature, electron densityand atom density in the LIP [7].

The history of laser spark spectroscopy runs parallel to the development of high-power lasers, starting with the early use of a ruby laser for producing sparks in gases [8].In subsequent years the spectral analysis of LIP became an area of study that hassignificantly matured at present. The current developments of this technique for chemicalanalysis can be traced to the work of Radziemski and Cremers [9] and their co-workers

Laser-Induced Breakdown SpectroscopyJagdish P Singh, Surya N Thakur (Editors)© 2007 Elsevier B.V. All rights reserved.



4 S. N. Thakur and J. P. Singh

at Los Alamos National Laboratory in the 1980s. It was this research group that firstcoined the acronym LIBS for laser induced breakdown spectroscopy. During the lasttwo decades, LIBS has undergone a dramatic transformation in terms of hardware,software and application areas. It has become a powerful sensor technology for bothlaboratory and field use. In order to obtain a reliable quantitative elemental analysis ofa sample using LIBS, one needs to control several parameters that can strongly affectthe measurements. Some of these parameters are the laser wavelength, its irradiance,the morphology of the sample surface, the amount of ablated and vaporized sample,and the ability of the resulting plasma to absorb the optical energy. If these and relatedparameters are properly optimized, the spectral line intensities will be proportional tothe elemental concentration. In the following sections we briefly describe the basiccomponents and the underlying physical processes that are essential to appreciate therange of applications and power of LIBS.

2. LASERS FOR LIBS

The main properties of laser light which distinguish it from conventional light sources arethe intensity, directionality, monochromaticity, and coherence. In addition the laser mayoperate to emit radiation continuously or it may generate radiation in short pulses. Somelasers can generate radiation with the above mentioned properties and that is tunableover a wide range of wavelengths. Generally pulsed lasers are used in the production ofplasmas and also in laser induced breakdown spectroscopy (LIBS). We consider onlythose properties of lasers relevant to plasma production in gaseous, liquid and solidsamples so that the role of various types of laser systems used in LIBS experiments isclearly understood in the later chapters of this book. It is possible to generate short-duration laser pulses with wavelengths ranging from the infrared to the ultraviolet, withpowers of the order of millions of watts. Several billions to trillions of watts and morehave been obtained in a pulse from more sophisticated lasers. Such high-power pulsesof laser radiation can vaporize metallic and refractory surfaces in a fraction of a second.It is to be noted that not only the peak power of the laser, but also the ability to deliverthe energy to a specific location is of great importance. For LIBS, the power per unitarea that can be delivered to the target is more important than the absolute value of thelaser power. The power per unit area in the laser beam is termed “irradiance” and isalso called “flux” or “flux density.” Conventional light sources with kilowatts powerscannot be focused as well as laser radiation and therefore are not capable of producingeffects that lasers can.

The next property of laser radiation that is of interest is the directionality of the beam.Laser radiation is confined to a narrow cone of angles which is of the order of a fewtenths of a milliradian for gas lasers to a few milliradians for solid state lasers. Becauseof the narrow divergence angle of laser radiation, it is easy to collect all the radiationwith a simple lens. The narrow beam angle also allows focusing of the laser light to asmall spot. Therefore the directionality of the beam is an important factor in the abilityof lasers to deliver high irradiance to a target. Coherence of the laser is also related tothe narrowness of the beam divergence angle and it is indirectly related to the ability ofthe laser to produce high irradiance. However, coherence is not of primary concern inLIBS. Provided that a certain number of watts per square centimeter are delivered to a



Fundamentals of LIBS 5

surface, the effect will be much the same whether the radiation is coherent or not. Themonochromaticity of the laser as such plays very little role as far as plasma productionis concerned because it is the power per unit area on the target that matters irrespectiveof the fact whether the radiation is monochromatic or covers a broad band. In specialcases, one may require highly monochromatic laser radiation to probe the plasma usingresonance excitation of atomic species. The frequency spread of gas lasers is of theorder of one part in 1010 or even better and for solid lasers, it is of the order of severalmegahertz. In specifying the frequency spread, we have taken the width of a singlecavity mode of the laser, although most lasers operate in more than one cavity mode sothat the total frequency spread may cover the entire line width of the laser transition. Thefrequency spread of each of the cavity modes is much narrower than the line width of thelaser transition and the former is used to characterize the frequency stability of the laser.

2.1. Mode Properties of Lasers

The optical cavity of a laser is determined by the configuration of the two end mirrors.The stationary patterns of the electromagnetic waves formed in the cavity are calledmodes. For a cavity formed by two confocal spherical mirrors separated by a distanceL, the frequency � of a mode (mnq) is given by

�mnq = �c/2L��q +1/2�m +n +1�� (1)

where c is the velocity of light, q is a large integer, and m and n are small integers.The axial modes correspond to m = n = 0 and involve a standing wave pattern with anintegral number q of half wavelengths with q �/2 = L, between the two mirrors with anode at each mirror. The separation between frequencies of two consecutive axial modes(c/2L) is of the order of gigahertz for typical solid state lasers. The transverse modesof the laser are designated by TEMmn. They affect the focusing properties of the laserbeam. The smallest focal spots and highest irradiance is obtained with beams containingthe lowest transverse modes with the smallest pair of values m, n. The higher transversemodes have radial intensity distributions which are less and less concentrated along theresonator axis with increasing values of m or n. These modes are also known as off-axismodes and their diffraction losses are much higher than that of the fundamental modesTEM00q. Some of the patterns for transverse modes are shown in Fig. 1.

The presence of higher transverse modes (large m, n) increases the divergence angleand affects the focusing of the laser beam. If there is no control over the mode properties,the modes present in the laser pulse can change from one shot to the next and differentpulses from a high power laser would be focused differently. High brightness is essentialfor delivering high irradiance. The brightness of a source is the power emitted per unit

TEM11TEM10TEM00 TEM20

Fig. 1. Transverse modes showing off-axis intensity distribution in selected higher modes.




area per unit solid angle. As laser power increases, the number of transverse modesincreases with little increase in brightness. The technology to produce high irradiance ina laser beam thus involves decreasing beam divergence as much as increasing power.

2.2. Spatial Intensity Distribution and Focusing of Laser Beam

In order to determine the irradiance produced by a laser, it is necessary to know thespot size to which the beam can be focused. It is impossible to focus the beam to ageometrical point and the minimum spot size is dependent on diffraction. Since opticalsystems are not perfect, the actual spot size is larger than the limit set by diffraction.Maximum irradiance is obtained with minimum focal area of the laser spot.

The spatial distribution of the output of a continuous gas laser follows the modepatterns shown in Fig. 1 and for the lowest transverse mode, the intensity distribution isgiven by

I00�r� = exp�−2r2/w2� (2)

where w is called the Gaussian radius of the TEM00 mode. The output of a high-powersolid state laser has a complicated spatial intensity distribution and does not exhibitthe recognizable mode patterns shown in Fig. 1. The output is a superposition of manymodes along with distortions caused by inhomogeneities of the crystal. This irregularspatial distribution leads to problems in focusing the laser beam to the minimum size.A schematic spatial profile of a solid laser is shown in Fig. 2.

The spatial profile of the laser beam can change during the course of the laserpulse [10]. Many methods have been employed for improvement of the mode propertiesof high power solid state lasers. One is shown in Fig. 3 where an aperture was introducedat the focus of a lens system contained within the laser cavity. This arrangement can

Fig. 2. Schematic contour of irradiance in unfocused high power ruby laser.

Mirror Ruby rod Aperture Output mirror

Lens Lens

Fig. 3. Schematic diagram of laser cavity with an aperture to remove higher order transversemodes.




reduce the off-axis mode content significantly, because the high order modes have largediffraction losses at the aperture. The output from a ruby laser can be made morespatially uniform than that shown in Fig. 2 and can have a divergence angle close tothe diffraction limit [11].

The number of axial modes in a laser output can be reduced with an optical cavityin which one mirror is made up of a number of uncoated interferometric flats. Laseroscillation occurs at those wavelengths which are simultaneously modes of the totalcavity and of the individual interferometers formed by each pair of flat parallel surfaces.Since the gain of the laser is nonlinear, the output power is funneled into a single ora few axial modes. In an optical cavity, introduction of a dye mode-selector has beenused to produce a single TEM00 axial mode output from a ruby laser [12].

The design and fabrication of a single mode laser oscillator followed by amplifiershas led to diffraction-limited lasers of high brightness [14]. An important concept in thecontext of lasers is the distinction between near and far field spatial patterns. In the nearfield, the intensity pattern is the same as at the output mirror of the laser and it followsthe mode patterns shown in Fig. 1. If ‘a’ is the aperture diameter of the output mirror andthe laser beam is approximated by a Gaussian beam, then the near field pattern persistsfor a distance of the order of a2/� where � is the wavelength of laser light. However forlarger distances from the output mirror, the well defined mode pattern in the near fieldwould be washed out by diffraction effects; the spreading angle is of the order of �/a.Gaussian beams have the same phase across the entire wave front, and they are capableof being focused to the minimum possible size [13].

Ruby lasers with an ordinary optical cavity can be focused with a simple lens toproduce spots with diameters of the order of 300 microns whereas those with an aperturedoptical cavity have focal spot sizes of the order of a few microns.

A focal area of 10−3cm2 is typical for a ruby laser focused with a simple lens and thefollowing peak power and irradiance can be obtained by different types of ruby lasers:

Laser Type Peak Power Irradiance

Normal Pulse Laser 105 W 108 W cm−2

Q-Switched Laser 108 W 1011 W cm−2

Picosecond Pulse Laser 1011 W 1014 W cm−2

2.3. Time Behavior of Laser Pulses

Solid state lasers, such as ruby lasers, Nd: glass and Nd: YAG lasers that produce highpowers are generally pulsed with widely different pulse durations and with differentmethods of pulsing. If the laser is pumped by a flashlamp, pulse widths in the range of100 to1000 microseconds are typical. In many cases, the laser emission is not uniform,but consists of many microsecond duration spikes called relaxation oscillations whoseamplitudes and spacing are not uniform. The presence of these spikes in the laser pulsecauses heating and cooling of the target surface and is not suitable for producing auniform plasma plume.

Laser pulse durations in the range of 10 to 1000 nanoseconds can be produced byQ-switching techniques, where laser operation is suppressed and population inversion




in the solid rod increases greatly over the normal threshold condition [14]. If theQ-switching component in the laser cavity is changed to a transparent condition, thelaser rod, now in a highly inverted state, gets coupled to the two mirrors of the cavityand the stored energy is emitted in a pulse of much higher power and much shorterduration than without Q-switching.

It is possible to produce laser pulses of picoseconds duration by the phenomenon ofmode locking. If there are N resonant axial modes of the cavity simultaneously present inthe linewidth of the lasing transition, then these can be coupled by using a Q-switchingdye with nonlinear transparency. This coupling of modes leads to a locking of the phasesof different axial modes. In the time domain, a single ultrashort pulse circulates in thecavity with a time period equal to the round trip transit time (2L/c). The laser outputis in the form of identical pulses whose spacing is equal to 2L/c. The width � of eachpulse is approximately the inverse of the frequency spread of the laser output. Thus, forN axial modes in the lasing transition linewidth, the pulse duration is given by

� = 1/�Nc/2L� = 2L/Nc (3)

Femtosecond laser pulses are produced by the technique of colliding pulse mode lock-ing (CPM) which utilizes the collision of two counter-propagating pulse trains in a thinsaturable dye jet. The interaction of the counter-propagating pulses creates a transientgrating of the population of dye molecules, which synchronizes, stabilizes and shortensthe pulse. The operation of femtosecond pulses is very sensitive to mirror coatings. Theshort duration and high electric field intensities encountered in amplifying femtosecondpulses introduce new problems in amplifier design. Improvement in amplification tech-niques has permitted generation of femtosecond laser pulses of gigawatt intensities [15].

2.4. Measurement of Laser Power and Energy

In order to study the physics of laser induced plasmas, reliable measurements of laserbeam power, beam energy, beam divergence angle, and spatial intensity distribution ofthe beam cross-section are needed.

The most common detectors used in the measurement of laser power are referred toas square law detectors because they respond to the square of the electric field. Photo-multiplier tubes (PMTs) and single stage vacuum photoemissive detectors are sensitivein the ultraviolet (UV), visible, and near infrared, whereas photoconductive detectorsare used for lasers emitting at wavelengths longer than one micron. For pulsed lasers,the phototube output is displayed on a fast oscilloscope to determine the pulse shape.The response speed of the photodetector must be fast and circuitry must be carefullydesigned to preserve the pulse shape. Intense laser output tends to saturate the output ofthe detectors, so absorbing filters are used to keep the detector in the linear portion oftheir operating range and to make them blind to background radiation.

Another widely used detector is the semiconductor photodiode which is a photovoltaicdevice. Laser radiation incident on the detector produces a voltage across the p-n junctioneven in the absence of an external bias. When light falls on a back-biased diode, thereverse current increases sharply. Room temperature devices are used in the visible regionand up to 3.6 microns whereas liquid nitrogen-cooled devices operate up to 5.7 microns.




The total energy in a laser pulse is measured by calorimetric methods using blackbodyabsorbers of low thermal mass in contact with thermocouples or other temperaturemeasuring devices. In one common form, the absorber is a small hollow cone of carbonsuch that radiation entering the base of the cone cannot be reflected out of the cone.Thermistor beads, forming an element of a balanced bridge circuit, are placed intimatelyin contact with the cone. As the cone is heated by a pulse of energy, the resistance of thethermistors changes, resulting in an imbalance of the bridge and a voltage pulse whichdecays as the cone cools to ambient temperature. The magnitude of the voltage pulsegives a measure of the energy in the laser pulse.

2.5. Varieties of Lasers

A laser is not one single device, but there are a wide variety of different lasers with manydifferent characteristics. Each type has its own properties of wavelength and operatingparameters. Even within one type, there are many varieties of construction. Now severalthousands laser lines are known which span a whole spectral range from extreme ultra-violet to the far-infrared region. Developments in LIBS have taken place by using thelaser wavelengths provided by existing technology. In 1962 a ruby laser at 694 nm wasused by Breach and Cross [16], but its pulse-to-pulse stability was very poor and LIBSwas not considered to be a very reliable technique for spectrochemical analysis. The nextphase of LIBS development was marked by the sophisticated pulsed-laser technologyof the 1980s which led to very reliable Nd: YAG lasers in the near-IR, visible, and UVregions and to excimer lasers in the UV region. At present many more laser wavelengthshave become available to study their effects on LIBS measurements [17–19]. Laserscommonly employed in LIBS are listed in Table 1 along with properties associatedwith these lasers. These are representative values, not necessarily the highest or the bestever achieved.

Table 1. Characteristic properties of some lasers for LIBS[20]

Laser Type Wavelength Pulse Duration Energy/Pulse

CO2 Repetitive 106 m 10–100 s 0.1–5 JCO2 Q-switched 106 m 200 ns 0.1 JEr:YAG Q-switched 294 m 170 ns 25 mJNd:YAG 106 m 5–10 ns 1–3 JRuby Normal Pulse 694.3 nm 0.2–10 ms 1–500 JRuby Q-switched 694.3 nm 5–30 ns 1–50 JRuby Picosecond Pulse 694.3 nm 10 ps 0.01–0.5 JNd:YAG Second Harmonic 532.0 nm 4–8 ns 0.5–2 JNd:YAG Third Harmonic 354.7 nm 4–8 ns 0.2–0.7 JN2 Laser 337.1 nm 3–6 ns 0.1–0.6 mJXeCl Excimer 308 nm 20–30 ns 0.5–1 JNd:YAG Fourth Harmonic 266 nm 3–5 ns 0.1–0.3 JKrF Excimer 248 nm 25–35 ns 0.5–1 JArF Excimer 193 nm 8–15 ns 8–15 mJ




3. LASER INDUCED PLASMAS

To produce a spark in air or a gas requires laser intensities of the order of 1011 W cm−2.Sparks are caused by the breakdown of the gas due to the electric field associatedwith the light wave. Breakdown thresholds are of the order of 106 to 107 V cm−1. Thespark is accompanied by production of charged particles, absorption of laser light, andre-radiation of light from the spark. If the temperature of the plasma at the position ofthe gas breakdown becomes high enough, X-ray emission is also observed, in additionof visible and UV radiation. This phenomenon was termed laser induced breakdown inanalogy with the electrical breakdown of gases [21]. The breakdown results from strongionization and absorption by gases that are usually transparent to light. The breakdownis marked by a threshold irradiance below which virtually no effects are observed.The onset of the breakdown is a sudden, dramatic phenomenon occurring at an easilydetermined threshold. Its spatial as well as temporal profiles make interesting study [22].

The breakdown in the focal volume of the lens in which the peak laser irradianceoccurs can be understood as occurring in two steps. First, the production of the initialionization and the subsequent cascade by which the ionization grows resulting in thebreakdown. Multiphoton ionization, where simultaneous absorption of many quanta byan atom produces an ion-electron pair, is considered to be a plausible mechanism for theinitial ionization. An alternative possibility is multiphoton excitation of an atom to anexcited state with many other excited states between it and the free electron continuum.Single photon absorption processes may rapidly ionize the atom from this excited level.A free electron in the focal volume absorbs photons and gains enough energy to ionizeadditional atoms by collisions. In each such ionization process, the colliding electronis replaced by two electrons with lower energy in the free electron continuum. Thesein turn absorb photons so that an avalanche or cascade of ionization will occur. Theabsorption of a photon by an electron may be visualized in two equivalent ways. (1) Itcan be considered as an inverse Bremsstrahlung process in which a single light quantumis absorbed by an electron in the field of an atom or ion. (2) Secondly it can be consideredas analogous to microwave-generated breakdown, in which the electron oscillates in theelectric field of the incident radiation.

3.1. Laser Induced Breakdown in Gases

One of the most striking observations is the extinction of laser light by a plasma plumeproduced by breakdown in gases. When the laser irradiance is less than the threshold,no significant attenuation is observed, but with laser irradiance exceeding the threshold,the absorption is so strong that it is often used as a critical test of whether breakdownhas actually occurred. Fig. 4 shows the shape of the original laser pulse and the pulsetransmitted through the plasma when breakdown occurs. It is evident that early in thetransmitted profile, there is little attenuation, but at later times, after breakdown occurs,the plasma becomes very opaque. The abrupt shutoff of the transmitted light occurssimultaneously with initiation of the spark. When light transmission is studied for aseries of laser pulses with increasing energy, breakdown occurs earlier in the pulse aslaser irradiance increases. The time to breakdown as a function of intensity depends




2010 30 40 50 ns

Original laser profile

Laser profiletransmitted

through plasma

Fig. 4. Schematic temporal profile of laser pulse in the absence and in presence of gas breakdownshowing attenuation of laser beam.

on the focal area [23]. For a small focal volume, higher laser intensity is required toproduce breakdown within the same time.

Although the laser induced blue-white spark appears spatially uniform to the nakedeye, it is indeed elongated along the direction of the incoming laser beam. For laserpowers of the order of 100 MW, the spark may be 1 cm long and a few millimeters indiameter. A schematic shape of the spark is shown in Fig. 5 where its expansion backtoward the laser essentially fills the converging cone of laser radiation. The growth ofthe spark in the direction opposite to the light flux has led to the model of a radiation-supported detonation wave. A detonation wave is a shock wave which is fed by release ofenergy behind the shock wavefront. In this case the energy is supplied by the absorptionof the incoming laser beam. This is analogous to the detonation of reacting gases, withthe reaction energy of the gases replaced by the absorbed laser energy. A shock wavepropagates from the focal region into the undisturbed gas and absorption of energy fromthe laser beam drives the shock wave, causing it to spread. The motion of the luminousfront has been measured as a function of time and two time regions have been identified.The plasma front has been found to move faster before the end of the laser pulse but itsexpansion is slowed down after the end of the pulse. This is schematically shown in Fig. 6.

Fig. 5. Schematic shape of laser-produced spark in air. The intense core is indicated by the whitecontour. Arrows indicate the propagation of the focused laser beam.




Laseroff

Time (ns)1

1

3

5

10

2 4 6 810 20 40

Distance

Fig. 6. Relative displacement of expanding luminous front as a function of time.

Spectroscopic investigations of laser induced spark in air show that its emissionconsists of spectral lines of N and O atoms as well as a strong continuum [24]. It hasbeen found that in the early part of the development of the spark, the continuum is thedominant component of emission in addition to broad lines of ions and neutral atoms.When the spark has expanded and cooled, less broadened lines from neutral atoms areobserved.

3.2. Plasma Production from Solid Targets

When a high-power laser beam strikes a solid surface, it produces a plasma plume due torapid melting and/or vaporization of the sample surface. The vaporization of a tungstensurface by a Nd:glass laser pulse was found to be accompanied by a shower of sparkscharacteristic of molten material expelled along with vaporization, whereas a plume ofglowing material was emitted by a pulse from a ruby laser beam on a carbon target [25].The plasma is produced by vaporization of the opaque target surface and subsequentabsorption of laser light in this vaporized material. The phenomena observed in thisinteraction are in many ways similar to the phenomena accompanying gas breakdown,but the initial density of the material is much lower in the latter case. Plasma productionstudies are carried out at laser irradiances of the order of 109 W/cm2 or greater whichproduce a denser, more absorbing blow-off material. There is a great difference in thebehavior of surfaces struck by laser pulses with millisecond duration as compared tothose with pulse durations in the nanosecond region. The short pulses of very high powerdo not produce much vaporization, but instead remove only a small amount of materialfrom the surface, whereas longer, low-power pulses produce deep, narrow holes in thetarget. For laser pulses of picosecond (ps) and femtosecond (fs) duration there is noreheating of the plasma due to absorption of laser radiation as in the case of nanosecond(ns) laser pulses (see Fig. 4). Thus the volume of plasma produced in the cases of psand fs laser pulses is much smaller than in the case of ns laser pulses. The plasma plumeproduced by ns laser pulses gets elongated towards the incident laser beam as a resultof reheating as shown in Fig. 5.




The interaction of the laser with a target surface is considerably modified by the pres-ence of material emitted from the surface by ns pulsed high power laser irradiation [26].It exerts a high pressure on the surface and changes the vaporization characteristicsof the surface. Since the laser flux density is very high, the ejected material can beheated further by absorption of incoming laser radiation. It becomes thermally ionizedand opaque to the incident radiation. The absorbing plasma prevents light from reachingthe target surface, which is effectively cut off from the incoming radiation for a largefraction of the laser pulse. At the end of the laser pulse, the blow off material becomesso hot that it begins to radiate thermally and some of this radiation may reach thesurface, causing further vaporization. The temporal evolution of the depth vaporized bythe high-power laser pulse is schematically shown in Fig. 7.

The processes involved in vaporization by a ns laser can be understood in terms of asimple model [27]. It takes into account the pressure produced by a small amount of theblow-off material early in the laser pulse. This recoil pressure raises the boiling pointof the target above its usual vaporization temperature. If the increase in vaporizationtemperature is sufficiently high, the surface will be prevented from vaporizing further andthe material will continue to heat to a high temperature (above the normal vaporizationtemperature) as more and more laser light from the pulse is absorbed by the targetsurface. Eventually, the target surface will reach the critical point and at that pointvaporization can occur. This model has been used to estimate the maximum depth atwhich the critical temperature is exceeded. At depths greater than this, removal of thematerial which is heated above the critical point will continue to exert a sufficientlyhigh pressure so that no vaporization will occur. The heat will eventually be conductedinto the interior of the target. This model does not take into account the shielding ofthe target surface from the incoming laser light as the blow-off material becomes hot,ionized and opaque.

The absorption of laser radiation at a surface can produce large pressure waves in thetarget material. One mechanism is evaporation of material from the surface, with recoilof the heated material against the surface leading to motion of the target as a whole.

Laser pulse

Arbitraryunits

Depth vaporized

10 20 30

Time (ns)

Fig. 7. Schematic representation of depth vaporized in a metal target as function of time showingeffect of shielding by the blow-off material.




There is another mechanism which does not necessarily involve removal of any materialfrom the surface. In this case, as laser radiation is absorbed in a thin layer near the surface,the internal energy of that layer increases and it will expand by thermal expansion. Thethermal energy is deposited very rapidly by a short pulse laser and the expanding layerof material exerts pressure on the adjacent layer, thus sending a compressive shock waveinto the target.

3.3. Radiation from Laser Induced Plasmas

The plasmas produced from solid targets also exhibit strong anisotropy in their expansion.The flow of the plasma has maximum velocity perpendicular to the surface, and it isindependent of the angle at which the laser beam is incident on the surface. Photographicmeasurements determine the motion of the plasma which emits light by recombinationor de-excitation of atoms [9].

X+ + e− → X +h� �recombination� (4)

X+∗ → X+ +h� (de-excitation of ions) (5)

X∗ → X +h� (de-excitation of atoms) (6)

A schematic diagram of expanding plasma is shown in Fig. 8 where the radius ofits outer luminous edge is plotted as a function of time. The results of one the earlieststudies of a plasma production by a Q-switched ruby laser from a carbon target indicatedthat a bright plume of emission began somewhat after the peak of the 45-ns laser pulse,reaching its maximum intensity about 120-ns after the start of the laser pulse [28].

Optical spectroscopic studies of laser-produced plasmas reveal both continuum andline radiation. The continuum radiation originates near the target surface and covers thespectral range from about 2 nm to 600 nm. The line spectrum shows the presence of highlyionized atoms as well as neutral atoms. The most highly ionized species are present nearthe plasma center, while lines of lower ionization and neutral species are observed near

Radius inarbitraryunit

Laser pulse

0 50 100 Time (ns)

Luminous edge

Fig. 8. Size of the luminous edge of expanding plasma produced by a short-pulse Q-switchedlaser as a function of time.




the outer regions of the plasma plume. The spectra of neutral atoms are found to originatein a larger spatial region, indicating that neutral atom emission dominates after the plasmahas expanded and cooled. The time variation of spectral line intensities indicates thatthe highest ionized states are present fairly early and lower ionized states appear later.

4. PROGRESS IN DETECTION OF LIBS

The early measurements of spectral emission from laser induced plasmas employedphotographic detection using prism or grating spectrographs. The system was far fromsatisfactory due to the fact that emission spectra consist of lines as well as continuum.Photometric detection with provision for time resolving the emission signals was notwidely available and spectrally resolved light could be detected using a photomultipliertube (PMT) only since the early 1980s [29]. The availability of a gated integrator madeit possible to integrate the PMT current only during a time period selected by a gatepulse. The gate pulse is synchronized in time to the arrival of the laser pulse at the target.To suppress the detection of continuum from laser induced plasmas (LIP) present inthe early part of the pulsed emission, the high voltage to the dynodes is gated so thatfull gain of the PMT is not realized until several microseconds after plasma formation.The disadvantage of gated PMT detection is that its gain does not remain constant. Theother detector for spectrally resolved emission is a photodiode array (PDA) consistingof a series of photosensitive silicon detector elements known as pixels, lined up ina row. Time resolution in this case is achieved by time-gating the voltage applied tothe microchannel-plate image-intensifier in front of the PDA. Time resolution of a fewnanoseconds could be obtained with a PDA and its gain could be controlled by a factorof 106. Due to the nature of the photodiode detectors, the PDA can only be cooled totemperatures reached by thermoelectric devices. At such temperatures, the dark currentis on the order of 500 counts/pixel/sec, which is relatively high so that PDAs are bestsuited for medium and high intensity signals.

The selection of the spectral range of plasma emission to be recorded in an experimentcould be made by using (i) a narrow bandpass filter, (ii) a monochromator, or (iii) aspectrograph between the detector and the plasma plume. If only a single emissionline is to be recorded at a time, the filter-detector combination or the monochromator-detector combination would be used depending on the presence of isolated or closelyspaced emission lines. The simultaneous recording of several lines with high resolutionwould require a spectrograph-detector combination. The PMT can be used in all of thesecases, and it is positioned behind the filter or the exit slit of the monochromator. Inthe latter case, the wavelength of the monochromator can be scanned to record severalspectral lines in emission from the LIP provided the nature of plasma plume does notchange from one pulse to another during the period of the scan. In the case of a samplecontaining many elements, either PMTs or PDAs can be used with the spectrographto record several spectral lines simultaneously. A slit-PMT combination located in thefocal plane of the spectrograph has to be used for each spectral line and the number oflines to be recorded is limited by the length of the focal plane. Another disadvantage ofPMT detection is that continuous wavelength coverage of the spectrum is not possible.PDA detection is more versatile because it has continuous wavelength coverage over the




Laser spark

PDA

Display

Pulsed laser

Spectrograph

Time control

Fig. 9. Schematic diagram for spectral analysis of plasma plume with time-gated PDA.

array length and it can record a spectrum from a single laser shot. A typical detectionsystem is shown in Fig. 9.

There has been a tremendous growth in the range and sophistication of photo-detectorssince 1990 due to progressive research and improvements in optical technology. Theadvent of high quality solid-state detectors has led to a quantum leap in applications ofLIBS [30–34].

4.1. CCD and ICCD Detectors

A charge-coupled device (CCD) is a micro-electronic device that is used in memory,signal processing and imaging applications. CCDs were initially conceived as an elec-tronic analogue of the magnetic bubble device. To function as memory, there must bea physical quantity that represents a bit of information, a means of recognizing thepresence or absence of the bit, and a means of creating and destroying the information.In the CCD, a bit of information is represented by a packet of electrons. These chargesare stored in the depletion region of a metal insulator semiconductor (MIS) capacitor andmoved about in the CCD circuit by placing the MIS capacitors so as to allow the chargeto spill from one capacitor to the next and hence the name charge-coupled device.

The CCD must perform four tasks in generating an image, viz. charge generation,charge collection, charge transfer, and charge detection. The first step occurs when freeelectrons are liberated due to incident photons. In the second step, the photoelectronsare collected in the nearest collecting site, referred to as pixels. Pixels are defined byelectrodes called gates formed on the surface of the CCD. The third operation is accom-plished by manipulating the voltage on the gates in a systematic way so that signalelectrons move down vertically from one pixel to the next. At the end of the columns is ahorizontal register of pixels. This register collects a line at a time and then transports thecharge packets in a serial fashion to an output amplifier. The final operating step is per-formed by the CCD when the charge packet from the horizontal register is converted to an




output voltage by the on-chip amplifier. This voltage is amplified, processed and digitallyencoded off chip and stored in a computer to reconstruct image on a television monitor.

CCDs provide the multichannel advantage of array detectors and since it is a two-dimensional array, it can record multiple spectra simultaneously. The large format,two-dimensional nature of CCDs is ideal for high-resolution or echelle spectroscopy.High-resolution spectra with overlapping orders are produced by a grating; each ordercontains information in successive spectral regions. The different order spectra areseparated in the orthogonal direction by a cross-dispersing element. The resulting two-dimensional spectrum is imaged onto the CCD. In this way, it is possible to obtainspectra covering the UV to the near IR range with 0.01-nm resolution.

The CCD is the most sensitive multichannel detector. It can be cooled with liquidnitrogen to 140 K where the dark current is less than 1 electron/pixel/hour. At thistemperature the detector can be exposed to a signal for hours without any significantcontribution from the dark current. CCDs have a large dynamic range which is definedas the ratio of the smallest distinguishable measurable charge to the largest beforesaturation. A 16-bit converter used with the CCD will allow the measurement of signalsthat are 1/65536th of the full scale signal. CCDs also offer variable gain which isimportant in the measurement of weak signals. By increasing the gain to measure signallevels which are very close to the noise, the signal-to-noise ratio (SNR) can be improvedwhile maintaining the same integration time. In other words one can achieve the samesignal-to-noise ratio in less time.

Experiments involving rapid kinetic measurements require an intensified CCD. TheICCD is a CCD with a multichannel plate intensifier attached. Light hits the photocathodeon the front of the multichannel plate and is converted to electrons which are multipliedand hit a phosphor to produce photons which are detected by the CCD.

Since the intensifier adds noise to the signal, causes blurring of the image and hasa non-uniform photocathode response, ICCDs are used for time-domain measurements.The intensifier is gated and the time between the pulsing of the laser and opening of themultichannel plate can be set to within better than 5 ns accuracy.

4.2. The Spectrograph-Detector Combination

LIBS makes use of the atomic emission from plasma plumes generated by a laserfrom solid, liquid or gaseous samples to identify the constituent elements present in thesample. An ideal experimental system should be capable of simultaneous multielementalmonitoring of both high- and low-Z elements. In many applications, rapid, near real-timestandoff detection capabilities are required. Typically a lens or a fiber optic collectsthe radiation from the plasma and couples it to a spectrograph. Emission from differentatomic species may occur at different times during the pulsed laser spark and time-resolved detection is necessary to obtain a spectral fingerprint of the atomic speciesthat are present in the sample. The wavelengths of the atomic emission lines mostcommonly analyzed with LIBS range from 190 to 850 nm. Detection below 190 nmis limited by atmospheric absorption but some elements with nonmetallic characterhave their strongest lines in the near vacuum ultraviolet (110–190 nm). Special effortsare required to minimize attenuation due to ambient air in the VUV region [35–38].




An ideal spectrograph-detector combination to detect all possible elements in a sampleshould have the following features:

1. Wide wavelength coverage (130–950 nm) to record simultaneously severalelements.

2. High resolution (0.003–0.01 nm) to resolve closely spaced spectral lines and toavoid interferences.

3. A large dynamic range (6–7 orders of magnitude) for the detector to provide theoptimum SNR for a large range of elemental concentrations.

4. High quantum efficiency of the detector particularly in the near IR and UV.5. Short readout and data-acquisition time (less than the time lapse between laser

pulses) for rapid analysis.

The ICCD array detector coupled to a grating spectrograph or integrated into a com-pact high-resolution Czerny-Turner spectrometer has been widely used as the detectorplatform for a great variety of LIBS applications. In some applications ensemble-averaged spectra are used to smooth pulse-to-pulse variations frequently seen inLIBS [39,40]. In applications that require rapid sorting or emission from single parti-cles, single-shot spectral measurements have to be made [41,42]. The use of non-ICCDarrays in LIBS is not common although correlation analysis in the identification ofstainless-steel standards has been carried out using this detector [43,44]. The much lowercosts of non-ICCD detectors are an important factor in their increasing use in researchlaboratories. The performance and sensitivity of a non-ICCD array and an ICCD arraydetector system have been compared in a recent publication [45].

Many applications of LIBS require remote and rapid multichannel analysis in hostileenvironments which implies large spectral coverage with high resolution [46]. Conven-tional Czerney-Turner spectrometers provide high resolution only in a limited spectralrange and it takes many laser shots to make sequential measurements for the analysisof many elements. In contrast an echelle spectrometer coupled with an ICCD detectorcan cover a large spectral range. Bauer and coworkers [47] were the first to couple anICCD camera to an echelle spectrometer, but they had to use a mobile mirror to obtainlarge spectral coverage. The efforts of several workers have shown that a large ICCDcamera is necessary for wide spectral coverage without any moving parts [48–51].

5. APPLICATIONS OF LIBS

The technological developments leading to the emergence of broadband high-resolutionspectrometers has led LIBS into the 21st century with unprecedented capabilities toextract spectral information from microplasmas. It is now possible to detect almost allchemical elements in the periodic table by analyzing the UV, visible and IR emissionprevalent in laser-generated sparks. Broadband high-resolution detection enables simul-taneous analysis of multiple component elements of targeted samples. For the first timein the history of LIBS [52], there is a hope to obtain qualitative as well as quantitativeinformation on complex biological molecules in a sample. LIBS-based technologies aredeveloping rapidly. It is not inconceivable that it would be possible to develop LIBS-sensors capable of the detection and identification of almost all forms of matter. In such




a case, it is difficult to make future predictions about the course of LIBS applications.In the following paragraphs, we will attempt to briefly summarize some novel featuresof this rapidly expanding field.

The use of femtosecond laser pulses in LIBS experiments has led to better preci-sion and better reproducibility in emission measurement as compared to nanosecondpulses. This improvement is attributed to high peak powers in the range of 1014 W/cm2.Femtosecond lasers consistently create well-defined craters and lead to better ablativereproducibility than nanosecond lasers [53]. Extremely short fs-laser pulses account forsome remarkable features as atomizers. In contrast to ns-lasers, the impact of the fs-laserenergy on the sample has ceased before the plasma is formed. There is no shieldingby the plasma and hence no dissipation of laser energy by it. The ablation threshold islower than for ns-lasers and the energy is more localized in the sample leading to betterspatial resolution [54]. Femtosecond-LIBS is being used for enhancement of signal andmeasurement of atom density distributions in the laser induced plasma [55,56].

The analysis of single microscopic particles, aerosols and cells has received greatinterest in recent years. A novel feature of LIBS for single particle analysis is its abilityto provide elemental mass composition and size data for individual particles [57]. Thepresence of aerosols in ambient air has been cause of great concern because of theirhazardous effects on human health, visibility, and climate change [58]. LIBS has foundincreasing application in studies on aerosols including effluent waste and real-timemonitoring [59,60]. Bioaerosols which include pollen, fungi, bacteria, and viruses arefound nearly everywhere; although their concentration is not high, they can cause diseaseor allergic reaction when inhaled even in very minute amounts [61–63].

The use of LIBS technology in field-portable instruments has given rise to a spurt ofresearch activity in order to deal with social problems arising from criminal and terroristactivities [64–67]. LIBS is the preferred detection and identification technique becauseof its many characteristic features, including flexibility of point detection or operationin a stand-off mode, and fast, real-time response.

The performance of LIBS can be enhanced with the use of an array of Geigerphotodiodes as the detector in echelle spectrometers. Single photon detection in roomtemperature conditions is possible without complex gating-timing circuitry [68]. A com-pact design and high sensitivity would make this instrument very handy for standoffdetection when low levels of plasma emission are to be collected. This developmentis very attractive in view of earlier work which shows that LIBS would provide anextremely useful tool for space and planetary exploration [69].

REFERENCES

[1] P. D. Maker, R. W. Terhune and C. M. Savage, Quantum Electronics, Eds. P. Grivet andN. Bloembergen, Columbia Univ. Press, New York (1964) p. 1559

[2] S. Namba and P. H. Kim, Jap. J. Appl. Phys. 3 (1964) 536[3] R. H. Fairbanks and C. M. Adams, Welding J. 43 (1964) 97s[4] N.G. Basov and O.N. Krokhin, Sov. Phys. JETP 19 (1964) 123[5] H. W. Kroto, J. R. Heath, S. C. O’Brien, R. F. Curl and R. E. Smalley, Nature 318 (1985) 162[6] F. Kokai, K. Takahashi, K. Shimizu, M. Yudakasa and S. Iijima, Appl. Phys. A69 (1999) S691[7] D. A. Rusak, B. C. Castle, B.W. Smith and J. D. Winefordner, Crit. Rev. Anal. Chem. 27

(1997) 257




[8] R. G. Meyerand and A. F. Haught, Phys. Rev. Lett. 9 (1963) 401[9] L. J. Radziemski and D. A. Cremers. Eds., Laser-Induced Plasma and Applications, Marcel

Dekker, New York (1989)[10] E. S. Dayhoff and B. Kessler, Appl. Opt. 1 (1962) 339[11] J. A. Baker and C. W. Peters, Appl. Opt. 1 (1962) 674[12] F. J. McClung and D. Weiner, IEEE J. Quantum Electron. QE-1 (1965) 94[13] A. L. Bloom, Gas Lasers, Chapter 4, Wiley, New York (1968)[14] F. J. McClung and R. W. Hellwarth, Proc. IEEE 51 (1963) 46[15] C. V. Shank, R. L. Fork and R. T. Yen, Picosecond Phenomena III, Eds. K. B. Eisenthal,

R. M. Hochstrasser, W. Kaiser and A. Laubereau, Springer Verlag, New York (1982) p. 1[16] F. Brech and L. Cross, Appl. Spectrosc. 16 (1962) 59[17] W. Sdorra, J. Brust and K. Niemax, Mikrochim. Acta 108 (1992) 1[18] C. W. Ng, W. F. Ho and N. H. Cheung, Appl. Spectrosc. 51 (1997) 976[19] M. Martin and M. D. Chang, Appl. Spectrosc. 54 (2000) 1279[20] Table of Laser Lines in Gases and Vapors, 3rd Revised and Enlarged Edition by R. Beck,

W. Englisch, and K Gürs (Springer-Verlag, Berlin, 1980)[21] E. Damon and R. Thomlinson, Appl. Opt. 2 (1963) 546[22] Y-L. Chen, J. W. L. Lewis and C. Parigger, J. Quant. Spectrosc. & Rad. Transfer. 67

(2000) 91[23] R. W. Waynant and J. H. Ramsey, J. Opt. Soc. Amer. 55 (1965) 602[24] S. L. Manel’shtam, Sov. Phys. JETP 20 (1965) 1344[25] J. F. Ready, Effects of High-Power Laser Radiation, Academic Press, New York (1971) p. 95[26] G. A. Askar’yan and E. M. Moroz, Sov. Phys. JETP 16 (1963) 1638[27] J. F. Ready, J. Appl. Phys. 36 (1965) 462[28] J. F. Ready, Appl. Phys. Lett. 3 (1963)[29] Y. Talmi, Ed, Multichannel Image Detectors, ACS Symp., Series No. 102, ACS, Washington,

D.C. (1983)[30] M. J. Pilon, M. B. Denton, R. G. Schleicher, P. M. Moran and S. B. Smith, Appl. Spectrosc.

44 (1990) 1613[31] S. Vasile, P. Gothoskar, R. Farrell and D. Sdrulla, IEEE Trans. Nucl. Sc. 45 (1997) 720[32] Q. S. Hanely, C. W. Earle, F. M. Pennebaker, S. P. Madden and M. B. Denton Anal. Chem.

68 (1996) 661A[33] J. M. Harnly and R. E. Fields, Appl. Spectrosc. 51 (1997) 334A[34] F. M. Pennebaker, D. A. Jones, C. A. Gresham, R. W. Williams, R. E. Simon, M. F. Schappert

and M. B. Denton, J. Anal. At. Spectrom. 13 (1998) 821[35] C. J. Lorenzen, C. Carlhoff, U. Hahn and M. Jogwich, J. Anal. At. Spectrom. 7 (1992) 1029[36] V. Strum, L. Peter and R. Noll, Appl. Spectrosc. 54 (2000) 1275[37] M. A. Khater, J. T. Costello and E.T. Kennedy, Appl. Spectrosc. 56 (2002) 970[38] S. Kaski, H. Hakkanen and J. Korppi-Tommola, Appl. Opt. 42 (2003) 6036[39] F. Colao, R. Fantoni, V. Lazic and V. Spizzichino, Spectrochimica Acta B57 (2002) 1219[40] A. K. Rai, F-Y. Yueh J. P. Singh and H. Zhang, Rev. Sci. Instrum. 73 (2002) 3599[41] C. Aragon, V. Madurga and A. J. Aguilera, Appl. Surface Sci. 197–198 (2002) 217[42] J. E. Carranza, B. T. Fisher, G. D. Yoder and D.W. Hahn, Spectrochimica Acta B56

(2001) 851[43] I. B. Gornushkin, B.W. Smith, H. J. Nasajpour and J.W. Winefordner, Anal. Chem. 71

(1999) 5157[44] S. I. Gornushkin, I. B. Gornushkin, J. M. Anzano, B. W. Smith and J. D. Winefordner, Appl.

Spectrosc. 56 (2002) 433[45] J.E. Carranza, E. Gibb, B. W. Smith, D. W. Hahn and J. D. Winefordner, Appl. Opt. 42

(2003) 6016[46] P. Fichet, P. Mauchien and C. Moulin, Appl. Spectrosc. 53 (1999) 1111




[47] H. E. Bauer, F. Leis and K. Niemax, Spectrochimica Acta B53 (1998) 1815[48] P. Lindblom, Anal. Chim. Acta 380 (1998) 353[49] S. R. Goode, S. L. Morgan, R. Hoskins and A. Oxsher, J. Anal. At. Spectrom. 15 (2000) 1133[50] S. Florek, C. Haisch, M. Okruss and H. Becker-Ross, Spectrochimica Acta B56 (2001) 1027[51] P. Fichet, D. Menut, R. Brennetot, E. Vors and A. Rivoallan, Appl. Opt. 42 (2003) 6029[52] L. Radziemski, Spectrochimica Acta B57 (2002) 1109[53] K. L. Eland, D. N. Stratis, D. M. Gold, S. R. Goode and S. M. Angel, Appl. Spectrosc. 55

(2001) 286[54] V. Margetic, T. Ban, F. Leis, K. Niemax and R. Hergenroder, Spectrochimic. Acta B58

(2003) 415[55] J. Scaffidi, J. Pender, W. Pearman, S. C. Goode, B. W. Colson, Jr., J. C. Carter and

S. M. Angel, Appl. Opt. 42 (2003) 6099.[56] O. Samek, F. Leis, V. Margetic, R. Malina, K. Niemax and R. Hergenroder, Appl. Opt. 42

(2003) 6001[57] D. W. Hahn and M. M. Lunden, Aerosol Sci. Technol. 33 (2000) 30[58] J. H. Seinfeld and S.N. Pandis, Atmospheric Chemistry and Physics: From Pollution to

Climate Change, Wiley, New York (1998)[59] J. P. Singh, F-Y. Yueh, H. Zhang and R. L. Cook, Process Control Qual. 10 (1997) 247[60] J. E. Carranza, B. T. Fisher, G. D. Yoder and D. W. Hahn, Spectrochimica Acta B56

(2001) 851[61] W. C. Hinds, Aerosol Technology: Properties, Behavior and Measurement of Airborne

Particles, 2nd Ed., Wiley, New York (1999)[62] A. R. Boyain-Goitia, D. C. S. Beddows, B. C. Griffiths and H. H. Telle, Appl. Opt. 42

(2003) 6119[63] A. C. Samuels, F. C. DeLucia, Jr., K. L. McNesby and A. W. Miziolek, Appl. Opt. 42

(2003) 6205[64] J. M. Anzano, I. B. Gornushkin, B. W. Smith and J. D. Winefordner, Polym. Eng. Sci. 40

(2000) 2423[65] R. T. Wainner, R. S. Harmon, A. W. Miziolek, K. L. McNesby and P. D. French, Spec-

trochimica Acta B56 (2001) 777[66] C. R. Dockery and S. R. Goode, Appl. Opt. 42 (2003) 6153[67] F. C. DeLucia, Jr., R. S. Harmon, K. L. McNesby, R. J. Winkel, Jr., and A. W. Miziolek,

Appl. Opt. 42 (2003) 6148[68] R. A. Myers, A. M. Karger and D. W. Hahn, Appl. Opt. 42 (2003) 6072[69] A. K. Knight, N. L. Scherbarth, D. L. Cremers and M. J. Ferris, Appl. Spectrosc. 54

(2000) 331



Chapter 2

Atomic Emission Spectroscopy

S. N. Thakur

Laser and Spectroscopy Laboratory, Department of PhysicsBanaras Hindu University, Varanasi-221005, INDIA

1. INTRODUCTION

The light emitted from a gaseous discharge when examined by a spectrometer to form aspectrum, is found to consist of discrete lines, bands and sometimes an overlying contin-uum. Discrete lines (and sometimes accompanying continuum) are characteristic featuresof emission from neutral atoms and ions in the discharge source. The spectral lines arecharacterized by three properties: wavelength, intensity and shape. These properties aredependent on the structure as well as the environment of the emitting atoms.

Atomic emission spectroscopy can be used to determine the identity, the structureand the environment of atoms by analyzing the radiation emitted by them. From themeasurement of wavelengths we may deduce the energy levels (or stationary states) ofthe atom and it provides experimental basis for the theories of atomic structure. If weknow the characteristic lines emitted by an atom then their appearance in the spectrumestablishes the presence of that element in the source. Measurement of intensities ofspectral lines of different atoms in a given source provides information about their num-ber densities. The physical parameters of the discharge source, such as temperature andpressure, affect the intensities and also the shape of spectral lines and these parameterscan be determined by analyzing the shapes of the spectral lines.

The Bohr theory of hydrogen in 1913 established the first link between the spectraand structure of atoms. The theoretical developments in quantum mechanics during1920s have their roots in the accurate experimental measurements on the fine structureand the hyperfine structure of spectral lines. The experimental measurement of Lambshift in the spectrum of hydrogen atom in 1947 added a new dimension to theoreticalphysics. In recent years, the availability of fast computers has established a high degreeof cooperation between theory and atomic spectroscopy.

The aim of this chapter is, however, not to discuss the details of atomic structurebut to provide the basis for the other two applications of atomic spectroscopy, namelythe identification of atoms together with their relative abundance and the determinationof the physical conditions of plasma discharge in which these atoms are located. In thefollowing sections we briefly discuss the measurement of spectral lines, the electronicstructure of atoms and nomenclature of atomic states, the radiative transitions in atoms




24 S. N. Thakur

and intensities of spectral lines. The environment of the atom affects its stationary statesby the presence of electric and magnetic fields due to moving electrons and ions inaddition to the electron-atom, ion-atom and atom-atom collisions which result in thebroadening of spectral lines emitted by the atom. In the case of plasma, Stark broadeningis the major cause of change of atomic lineshapes which depend on the electron densityand temperature of the plasma. The last section contains a brief account of applicationof atomic emission spectroscopy of the light emitted from a plasma source.

2. MEASUREMENT OF SPECTRAL LINES

The first step is the recording of the spectra using a prism or grating as the dispersingelement to spread the light spatially according to its wavelength. A typical spectrometeris shown in Fig. 1. A narrow slit S allows the light from the source to be collimated bylens L1 so as to form a beam of parallel rays incident on the dispersing element. Parallelrays of light corresponding to different wavelengths come out at different angles fromthe dispersing device and are focused at different points in the focal plane of lens L2.The capacity of the spectrometer to separate two closely spaced wavelengths is knownas the spectral resolving power and depends on the narrowness of the slit for a givendispersing device. A narrow exit slit can be placed in the focal plane of lens L2 witha photomultiplier tube behind it and the spectral lines can be recorded by rotating thedispersing device. Alternatively a photographic plate (or a CCD plate) can be placedin the focal plane of lens L2 to record many spectral lines simultaneously. Intensitymeasurements from photographic plates are cumbersome and in modern spectrometersphotomultiplier tubes or CCD plates are used for more reliable intensity measurements.In case a concave grating is used as the dispersing element, lenses L1 and L2 are notrequired because the concave grating acts as its own collimator and focusing lens.

The wavelengths of spectral lines have to be measured accurately. The primarystandard is the red line of the isotope of Krypton �Kr86� whose vacuum wavelengthis 6057�80210 × 10−10 meter. A number of spectral lines measured interferometricallyagainst the primary standard have been accepted as secondary standards. These aremostly lines of neon, argon, iron, thorium etc and are updated from time to time by acommission of the International Astronomical Union [1]. There are also many tertiarystandards and wavelengths of many atoms [2] quite accurate enough to calibrate the

Dispersing elementL1 L2 Focal plane

Sλ2

λ1

Fig. 1. Schematic diagram of a prism or grating spectrometer.



Atomic Emission Spectroscopy 25

spectra recorded on any spectrometer. The unit of wavelength is nanometer (nm) butmany spectroscopists prefer Angstrom (A)

3. ELECTRONIC STRUCTURE OF ATOMS

An atom consists of positively charged nucleus and a number of negatively charged elec-trons. In a neutral atom, the total negative charge of all the electrons is equal to the totalpositive charge of the nucleus. The forces holding the atom together are predominantlyelectrostatic, consisting of attractions between each electron and the nucleus and repulsionsamong all the atomic electrons. A simple theory of electrostatics, however, fails to accountfor the stability of atoms and for the characteristics of their spectra. The early attempts ofBohr and de Broglie to give a plausible theory of atomic structure led to the much moresophisticated, probabilistic quantum mechanics of Heisenberg, Schrödinger and others.

The object of the theory of atomic structure is the study of stationary states of isolatedatoms. At first sight, such a study seems to be unrealistic since an atom is never totallyisolated and in order to be observed it must interact with photons. However, in conditionsof low pressure and low photon density, the effects of atom-atom and photon-atominteraction are weak enough to be neglected. Even in the case of an isolated atom, itsenergy depends not only on the charge of the nucleus but on its volume and the chargedistribution inside this volume. Therefore the problem is very complicated and cannotbe solved without approximations. The first approximation is to consider the nucleus asa point charge with infinite mass. The relative velocities of electrons in atoms are smallenough to be neglected and the interactions can be described as electrostatic Coulomb. Toreproduce important experimental features of atomic spectra, it is necessary to introducea magnetic term: the spin-orbit interaction. As regards the remaining effects, they aresmall enough to be introduced by means of perturbation theory.

Although we do not want to get involved with the mathematical details of quantummechanics, it is necessary to use its language and its results to describe atomic structure.The atom is described by the fundamental differential equation of quantum mechanicscalled Schrödinger equation

H� = E� (1)

In Eq. (1), � is a function of the coordinates of the system called the atomic wavefunctionand characterizes the state (electronic configuration) of the atom. The H is a differentialoperator called the Hamiltonian operator of the atom and is composed of the operatorsfor kinetic energy of electrons and the nucleus, the attractions between electrons and thenucleus and the inter-electronic repulsions. Except in the case of hydrogen atom Eq. (1)does not have an exact solution and the average energy of atom in the state � is given by

E =∫ �

−� �∗H�d�∫ �

−� �∗�d�(2)

In Eq.(2) �∗ represents the complex conjugate of the wavefunction � and d� representsthe volume element in three dimensional space. The Schrödinger equation is insolublefor all but hydrogenic atoms and the energy E is based on the approximate solutions for



26 S. N. Thakur

atoms with two or more electrons. It is within the framework of these approximationsthat we will describe the structure of atoms with many electrons.

3.1. Hydrogenic Atoms

The atomic wavefunctions �n�m for hydrogenic atoms are expressed in terms of thecoordinates of the system and depend on three parameters called quantum numbers n, �and m which result from the exact solution of the Schrödinger equation. The principalquantum number n determines the average energy and radius of the hydrogenic atomin the corresponding stationary state, whereas � and m determine the average valueof angular momentum and its component along a fixed direction respectively. For agiven value of n the value of � for the wavefunction can be: 0� 1� 2� � �n − 1�.Similarly the values of quantum number m for a given � are: −�� −�+1�� −�+2�� 0� 1� 2� ��−1�� . Thus, for any value of angular momentum quantum number �,there are �2�+1� possible values of magnetic quantum number m and for a given valueof n there are n different values of �.

The wavefunction �n�m is said to represent an atomic orbital. Atomic orbitals belong-ing to a given value of n define a shell and those associated with a given value of � con-stitute a subshell. Hydrogenic orbitals belonging to the same shell have the same energyand are said to be degenerate. Atomic orbitals corresponding to � = 0� 1� 2� 3� 4� � 5are called s, p, d, f, g, h .orbitals respectively. Thus the atomic orbitals associatedwith n = 3 are, 3s, 3p, 3d and those for n = 5 are, 5s, 5p, 5d, 5f and 5g.

Dirac included the effects of special relativity in the solution of Schrödinger equationand was able to derive not only the quantum numbers n, �, and m, but also a fourthquantum number s, called the spin quantum number. The spin quantum number is relatedto the magnetic moment of the electron and it can only have two values +1/2 and −1/2.Thus an electron occupying a hydrogenic orbital �n�m can have either s = +1/2 or s = −1/2.The four quantum numbers n, �, m and s play a fundamental role in the electronicconfigurations of atoms with many electrons.

If we neglect the fine structure, the stationary states of hydrogen atom are as shown inFig. 2. The zero of energy is defined for the electron and proton at rest at infinite separa-tion. The energy corresponding to the principal quantum number n is −RH/n2 where RH

is the Rydberg constant for hydrogen. The ground state energy of hydrogen atom is −RH

corresponding to n = 1. The separation between consecutive energy states decreases as nincreases till the ionization limit corresponding to the complete removal of the electron. Thestates of positive energy correspond to proton + electron + kinetic energy, the energy is nolonger quantized and there exists a continuum of states. The transition of atom between apair of discrete states is possible under certain conditions resulting in a spectral line. Tran-sitions between the continuum and a discrete state and transitions within the continuumgive rise to continuum and are known as free- bound and free-free transitions respectively.

3.2. Many Electron Atoms

The model for the ground state of a neutral atom of nuclear charge Z is constructedby assigning Z electrons to the hydrogenic orbitals in such a way that electronic




Continuum

Ionization limit

3

2

Emission Absorption

n = 1

Fig. 2. Energy level diagram of hydrogen atom.

configuration of lowest potential energy is obtained. In a many- electron atom, there arerepulsions between each pair of electrons but the hydrogenic orbitals do not account forthis. The degeneracies of occupied orbitals of the same n but different � may be removedin many-electron atoms. However, the degeneracies of orbitals having the same valuesof n and � but different values of m are not removed in the absence of external magneticfields.

The orbital capacities and order of assigning electrons to atomic orbitals are gov-erned, respectively, by the Pauli’s exclusion principle and the Hund’s rule of maximummultiplicity. In its simplest form, the exclusion principle states that no two electrons inthe same atom can have four identical quantum numbers. Thus an orbital specified byn, �, m can accommodate a maximum of two electrons one with s = +1/2 and the otherwith s = −1/2. The Hund’s rule has its basis in Coulomb’s law and states that in thecase of degenerate orbitals, the configuration of minimum potential energy would beobtained by allowing the electrons occupying these orbitals to stay as far apart as possi-ble. Thus in filling degenerate orbitals, each orbital accepts one electron before doubleoccupancy (pairing) occurs, because the separate orbitals occupy different regions ofspace whereas, two electrons, paired (opposite spins) in one orbital are close together,resulting in greater repulsive potential energy.

The application of Pauli’s principle together with Hund’s rule of maximum multi-plicity has been quite successful in predicting the ground state electronic configurationsof lighter atoms. This is not so, however, in the assignment of ground state electronicconfigurations of heavier atoms. Thus iron (Fe) with Z = 26 would be assigned a configu-ration 1s22s22p63s23p63d8 with two unpaired electrons in d orbitals. This configuration isnot consistent with chemical and magnetic properties of iron which require four unpairedelectrons in 3d orbitals with the ground state configuration 1s22s22p63s23p63d64s2. Thisdiscrepancy is due to repulsions produced on the n = 3 electrons by the inner shell



28 S. N. Thakur

Table 1. The Periodic Table

*Atomic weight increases as we read the table from left to right and then upward

113 114………118 7p Rf Ha Sg Ns Hs Mz………...112

Fr Ra Ac 6d Th Pa U Np Pu Am Cm Bk Cf Es Fm Md No Lw 7s 5f

Tl Pb Bi Po At Rn 6p Hf Ta W Re Os Ir Pt Au Hg

Cs Ba La 5d Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu 6s 4f

InRb Sr 5p Y Zr Nb Mo Tc Ru Rh Pd Ag Cd

5s 4d Ga

K Ca 4p Sc Ti V Cr Mn Fe Co Ni Cu Zn 4s 3d

Al Si P S Cl Ar Na Mg 3p

3s B C N O F Ne

Li Be 2p

2s H He

1s

Sn Sb Te I Xe

Ge As Se Br Kr

electrons. The occupied d orbitals are repelled to energies greater than that of the unoc-cupied 4s orbital and as a result the 4s orbital becomes occupied by two electronsin preference to the 3d orbitals. A modern form of periodic table due to Longuet-Higgins [3] is given in Table 1 where each group contains elements with similar electronconfiguration.

3.3. Classification of Electronic States

Atomic electrons produce an orbital magnetic moment as a result of their orbital motion,which is located at the nucleus and is directed at right angles to the orbit plane,collinearly with the orbital angular momentum vector associated with the occupiedorbital. The magnitude of magnetic moment due to a single electron is directly pro-portional to its orbital angular momentum �. Similarly the magnetic moment due tothe spin angular momentum is located at the position of the electron and is directedeither up or down along the direction of the spin angular momentum vector of theelectron depending on whether s = +1/2 or s = −1/2. The individual spin and orbitalmagnetic moments may be added to give a resultant magnetic moment for the atom.This is synonymous with the addition of spin and orbital angular moments of electronsin the atom.

There are two ways in which individual electronic � and s values may be addedvectorially to give the resultant atomic angular momentum quantum number J. In thefirst scheme the � and s of each electron may be added vectorially to give a resultant




one-electron angular momentum quantum number j. The j values of all atomic electronsmay then be added vectorially to give J. This is known as j-j coupling scheme. In thesecond scheme, the individual � values of each electron may be added vectorially to givea total orbital angular momentum quantum number L. The s values of of each electronmay similarly be added vectorially to give a total spin angular momentum quantumnumber S. The values of L and S may then be added vectorially to give J for the atom.This is known as L-S (or Russell Saunders) coupling scheme.

The two addition methods for electronic angular momenta correspond to differentphysical situations. In the L-S coupling, L, S as well as J can be used to describeelectronic states but in the j-j coupling, L and S have no physical meaning and J is the onlygood angular momentum quantum number. The orbital angular momentum vector of anelectron is located at the nucleus whereas the spin angular momentum vector is located,roughly, on the electronic orbit, the coupling of � and s is very weak unless the electronspends a considerable portion of its time near the nucleus. The probability density ofelectron is large near the nucleus in atoms with high nuclear charge (heavy atoms) whichexhibit appreciable spin-orbit coupling and their atomic angular momentum is given byj-j coupling. On the other hand, interelectronic repulsion is strong if several electronshave high probability densities in the same region of space. If the atom has a nucleusof low nuclear charge, all the individual electronic orbital angular momenta, located atthe nucleus, will couple strongly and similarly will the individual spin momenta, locatedon the electronic charge distribution. The coupling of � and s will be secondary to thecoupling of �’s and s’s and hence L-S coupling will adequately describe the situation.For most atoms, even when appreciable spin-orbit interaction occurs, the L-S couplingis retained, with an appropriate perturbation treatment to account for the interaction. Inthe remainder of this discussion we will describe the stationary states of atoms in theframework of L-S coupling.

The general nomenclature of an atomic state is based on the representation 2S+1LJ,where the state is labeled as S, P, D, F, G corresponding to L = 0, 1, 2,3, 4 respectively. Thus, with L = 2, and S = 1, the values of J will be 3, 2 and 1and the resulting states will be represented as 3D3�

3D2 and 3D1 respectively. It is tobe noted that the total angular momentum quantum number J and the correspondingmagnetic quantum number M are always good quantum numbers, irrespective of thecoupling scheme and the atomic wavefunction is represented as �JM. The correspondingatomic state is �2J + 1� fold degenerate in the absence of external electric or magneticfield. For further information on atomic structure the reader is advised to see some ofthe excellent books [4–9].

4. RADIATION FROM ATOMS

The intensity of a spectral line depends on the atomic population of the initial level andalso on the intrinsic probability of transition to the final level. The transition probabilityis defined in terms of Einstein’s A and B coefficients shown in Fig. 3 where E1 and E2

are two discrete quantum levels of the atom with populations of N1 and N2 atoms/cm3

respectively. The frequency of the spectral line resulting from a transition between thetwo levels is given by h12 = E2 −E1



30 S. N. Thakur

________________________________________________________

________________________________________________________

ρ B12N1 ρB21N2 A21N2

N1

N2

E1

E2

Fig. 3. Emission and absorption processes between a pair of energy levels.

There are three kinds of radiative processes that transfer atoms between the energylevels E1 and E2:

(1) An atom with energy E2 may spontaneously make a transition to energy state E1

with emission of energy h12�The probability of this transition per second is A21 and thenumber of such transitions per second per cm3 is A21N2

(2) Under the influence of external radiation of density ��12� an atom may makea transition from state E1 to E2 with absorption of energy h12. The probability of thistransition per second is �B12 and the number of transitions is �B12N1 sec−1cm−3.

(3) An atom in state E2 may undergo a stimulated (or induced) transition to state E1

in the presence of external radiation of density ��12�. The probability of this transitionis �B21 per second and the number of transitions is �B21N2 sec−1cm−3.

The Einstein’s coefficients for spontaneous emission A21, stimulated emission B21

and absorption B12, are intrinsic properties of the atom and they can, in principle, becalculated if the wave functions of the two states are known. If the energy states E1 andE2 are degenerate with degeneracy parameters g1 and g2 respectively then the relationsbetween Einstein’s coefficients are as follows:

g1B12 = g2B21 (3a)

A21 = �8 h3/c3�B21 (3b)

where = 12 = 21 is the frequency of the spectral line resulting from the transitionbetween the two states.

4.1. Electric Dipole Selection Rules

If we assume the external electromagnetic radiation as a time dependent perturbationto the atom, it is easy to visualize that the perturbation should be as effective for thestimulated emission E2 → E1 as for the absorption E2 ← E1, so that B21 = B12 if the twostates are non-degenerate. The B coefficients can be calculated from quantum mechanics




by treating the oscillating electric and magnetic fields as a time dependent perturbationleading to the following expression:

B12 = �2 2/3�0h2� �Mx212 +My2

12 +Mz212� (4a)

where �0 is the permittivity of free space and Mx, My and Mz are the components ofthe transition moment vector M12 given by

M12 =∫

� ∗J2M2 er �J1M1d� (4b)

In Eq. (4b) �J2M2 and �J1M1 are the wavefunctions of the upper and lower states respec-tively e is the electronic charge and r is the operator corresponding to the displacementof electronic charge in the atom as a result of the transition (er is the dipole momentvector) and the transition between the two states is said to be allowed if the integralis different from zero. The quantum numbers that define the wavefunctions of the twostates satisfy certain relations for a non- zero transition moment and these relations arecalled selection rules.

The selection rules for the total angular momentum quantum number J for dipoleallowed transitions are given by

�J = J2 − J1 = 0�+1�−1 (5a)

If the spin-orbit coupling is so weak that the orbital motion is practically the same as inthe absence of electron spin then the orbital quantum number L retains its significanceand the corresponding selection rules are

�L = L2 −L1 = +1�−1 (5b)

The selection rules for M are

�M = M2 −M1 = 0�+1�−1 (5c)

The selection rule �M = 0 corresponds to emission of light from the atom with itselectric vector oscillating along the z-axis. This radiation can not be seen along thez-direction and it gives rise to linearly polarized light when viewed along a direction inthe x-y plane. The selection rules �M = +1 and �M = −1 correspond to right and leftcircular polarization when viewed in the z-direction and linearly polarized when viewedin a direction in the x-y plane. The selection rules for the quantum number M and thecorresponding polarizations of the emitted radiation are physically meaningful if a fixeddirection in space is defined. Thus the direction of the magnetic field in Zeeman effector that of the electric field in Stark effect defines the polar axis for the atom. In thecases of scattering or fluorescence from the atom in the presence of a linearly polarizedincident beam, the electric vector of the incident radiation can be used to define thepolar axis.



32 S. N. Thakur

4.2. Parity Selection Rules

The energy states of free atoms can be divided into two classes according to theirparities. If the orbital angular momentum quantum number �i of individual electrons istaken into consideration, the parity of the sum ��i is the parity of the resulting energystate. The energy state is said to have odd or even parity depending on the odd or evenvalue of ��i. Odd parity of a state is indicated by a superscript ‘o’ such as 2Po

3/2. Theparity property is well defined for an atom with any number of electrons with any kindof coupling of their orbital and spin angular momenta. The rigorous selection rule basedon parity is known as Laporte rule:

Only transitions between even and odd states are allowed for a dipole radiation. Itcan be shown that all the selection rules in Eq. (5) obey parity selection rule.

4.3. Forbidden Transitions

Transitions which are forbidden in the dipole approximation may appear very weaklyas quadrupole radiation or magnetic dipole radiation. The matrix elements for electricquadrupole radiation take the following form

Q12 =∫

� ∗J2M2 exy �J1M1d� (6a)

and it does not vanish when �J2M2 and �J1M1 have the same parity because the parity ofthe operator xy is clearly even. The selection rules for quadrupole allowed transitions are

�L = 0�+2�−2� �J = 0�+1�−1�+2�−2 (6b)

The Laporte rule in this case states: Only transitions from even to even or from odd toodd terms (states) give rise to quadrupole radiation.

Magnetic dipole transitions appear with the same strength (with intensity 10−6 timesthe intensity of electric dipole allowed transitions) as electric quadrupole transitions.They result from the fact that radiation field produced by a system of moving charges inthe atom cannot be adequately described in terms of electric dipole, quadrupole, octopoleradiation but also need terms containing magnetic dipole, quadrupole etc. The selectionrules for magnetic dipole allowed transitions are:

�L = 0� �J = +1�−1 (6c)

The forbidden transitions which are forbidden as dipole radiation only to a firstapproximation may become allowed under special conditions. Inter-combination linesresulting from transitions between states of different multiplicities fall in this category.The electric dipole selection rules hold strictly only for free atoms and external fieldsdue to ions in a crystal or those in a discharge or even the fields resulting due to neutralatoms can give rise to forced electric dipole transitions of observable intensity.




4.4. Line Strength

If we adopt the following notation for the x-component of the electric dipole transitionmoment:

∫� ∗

2 ex �1d� = e�x12�, Eq.(4a) takes the following form

B21=�2 2e2/3�0h2� ��x12�2 +�y12�2 +�z12�2� (7a)

The electric dipole line strength S of the transition is defined as

S = S21 = S12 = e2 ��x12�2 +�y12�2 +�z12�2� (7b)

From Eq.(3) and (7b) we get the following relations for non degenerate states

B21 = �2 2/3�0h2�S12 (7c)

A21 = �16 33/3�0hc3�S12 (7d)

If the two states involved in the transition are degenerate and the degeneracies of E1 andE2 are g1 and g2 respectively then the line strength is given by

S = S21 = S12 = e2∑g1

M1

∑g2

M2��x12�2 +�y12�2 +�z12�2� (8a)

The degeneracies will cause the atomic populations to be divided amongst g1and g2

number of sublevels for E1 and E2 respectively and the number of atoms in E1 willreduce by a factor 1/g1 and those in E2 by a factor 1/g2. The expressions for Einstein’scoefficients will take following form:

B12 = �2 2/3�0h2�S/g1 (8b)

B21 = �2 2/3�0h2�S/g2 (8c)

A21 = �16 33/3�0hc3�S/g2 (8d)

4.5. Oscillator Strength

The classical model of emission from an atom has the electron performing simpleharmonic motion at a characteristic frequency v but with amplitude decreasing with timebecause of the energy radiated away. The emission of radiation acts as the dampingagent for the electron’s motion and the damping constant is given by

� = �2 e22�/�3�0mc3� (9)

where m is the electron mass. The lifetime of a classical oscillator is the inverse ofdamping constant �� = 1/�� and it is of the order of 10−8 sec for emission in the visibleregion.



34 S. N. Thakur

When light passes through an absorbing medium, the energy absorbed by atoms isproportional, to the thickness of the medium and also to the incident flux of light.The transmitted light intensity at frequency across a thickness � of the absorbingmedium is given by

I�� = I�0� exp �−kv�� (10)

where k is in units of cm−1 and is called the absorption coefficient. Absorption isalso expressed in terms of atomic absorption coefficient (or atomic cross section) �

�= k/N1� where N1 is the population density of the lower energy state E1. � has thedimension of area. It is to be noted that k is dependent on the profile of the spectraltransition (line shape) and its relation with B12 is given by

∫kdv = �h0/c�N1B12 (11a)

where the integration is to be carried over the line profile and v0 is the frequencycorresponding to the peak of the line profile.

If the number density N2 of atoms in the upper energy state E2 is appreciable, therewould be stimulated emission, putting photons back into the incident beam and therelation between k and B12 would become

∫kdv = �h0/c�N1B12�1−g1N2/g2N1� (11b)

The quantum mechanical and classical models of light emission are related through aquantity called oscillator strength. This is achieved by equating the light absorbed by N1

atoms in the transition E2 ← E1 to that absorbed at the same frequency by N classicaloscillators so that N = f12N1. The oscillator strength ‘f’ is related to the Einstein’scoefficient B by

f12 = �4mh�0/e2�12B12 (12a)

If we replace �0 by 1/4 the relation between f and B (in c.g.s. units) becomes

f12 = �mh/ e2�12B12 (12b)

The normal meaning of ‘f’ is the oscillator strength f12 but the emission ‘f’ value f21 isa negative quantity given by

f21 = −�g1/g2�f12 (12c)

The oscillator strength ‘f’ is also interpreted as the effective number of electrons peratom for a particular transition. Thus f should be approximately 1 for one valence electronatoms (hydrogen and alkalis) and 2 for the alkaline earths. Since each electron canparticipate in several different transitions, the total oscillator strength is accordingly splitbetween several spectral lines. Thus f represents the fraction of the available electronsparticipating in a particular transition and has values in the range 1 to 0.01 for strongspectral lines.




Ec

Eu

El

Ei

Fig. 4. Transitions involved in f-sum rule.

The f-sum rule states that the sum of all transitions from a given state should be equalto the number of optical (or valence) electrons z. Thus

�ufiu +�lfil +∫

cficdc = z (13a)

where iu refers to transitions upwards from a particular state Ei, il refers to transitionsdownwards and ic refers to transitions to the continuum Ec as shown in Fig. 4.

The downward il transitions represent stimulated emission and if we use oscillatorstrengths for absorption only, Eq (13a) takes the following form:

�u>ifiu − �gl/gi��l<ifli +∫

cficdc = z (13b)

The f-sum rule has limited use in converting relative oscillator strengths to absolutevalues but partial f-sum rules involving f-values for multiplets from a given configurationare of much practical importance.

4.6. Intensities of Spectral Lines

The observed intensity of a spectral line depends on two factors. The first factor is theintrinsic property of the atom and it can be described in terms of transition probabilityor line strength or f-value. The second factor depends on the conditions of excitation.When the emitting medium is such that a photon produced by spontaneous emission hasno appreciable chance of being re-absorbed, the medium is termed optically thin. Thiscondition holds in a gas discharge at low pressure where kinetic energy of the electronsis distributed approximately according to Maxwell’s distribution law. The concept of anelectron temperature is valid in such conditions which has a very much higher value thanthe temperature defined by the kinetic energy of atoms or ions. Since the probability ofelectron impact excitation of atoms is very high in these conditions, the population ofexcited states tends to be in accordance with the electron temperature.

Equilibrium conditions exist in a discharge source whenever the exchange of energybetween atoms and between atoms and electrons is rapid compared to the rate of



36 S. N. Thakur

excitation. For gas discharges at or above atmospheric pressures, the gas temperaturemay approach the electron temperature. If two spectral lines originating from upperenergy levels E1 and E2 in the condition of thermal equilibrium are observed to haveintensities I1 and I2 respectively then these are related to the transition probabilitiesA1, A2 and line strengths S1, S2 as follows:

I1/I2 = A1N1/A2N2 = �A1g1/A2g2� exp �E2 −E1�/kT = �S13

1/S23

2� exp �E2 −E1�/kT(14)

where k is Boltzmann constant and T is the gas temperature.Spectral lines with lowest excited level of the atom as their upper state and ground

state of the atom as the lower state are called resonance lines. The light emitted in suchlines has a very large probability of re-absorption before leaving the plasma dischargeand this phenomenon is called self absorption. Self absorption tends to broaden theprofile of the spectral line and in extreme cases makes the peak appear greatly flattened.If the excitation temperature drops in the outer regions of the discharge, the passage oflight from the central region not only broadens the profile but also shows an intensitydip at the line center. This phenomenon is called self reversal and such lines give thewrong impression of being doublets.

The deviations from thermal equilibrium can occur at very low gas densities andemission of spectral lines may exhibit peculiar intensity features if metastable levels areinvolved. A level is called metastable if all transitions to lower levels occur with verysmall probabilities due to selection rules. The concentration of atoms in a metastablelevel can be abnormally high and in the extreme condition they lose their excitationenergy by collision and not by radiation. Metastability of a level is, however, a matterof degree depending on the values of the transition probability and gas density. In thecase of interstellar space or in stellar nebula, the gas density is extremely low and ametastable level with lifetime of 0.1 sec, which may emit only quadrupole radiation, willlose all its excitation energy by radiation. The intensity of such a line will only dependon the rate at which the metastable level is excited and the line may appear quite strongwhile it is not observed in laboratory sources.

4.7. Continuous Emission & Bremsstrahlung

It was pointed out in section 3.1 that there exist free (unbound) states of an electron-ionsystem for which the energy is positive. The transitions between a free state and thestationary state Ej of the atom, gives rise to a continuum emission whose frequencies are

h = �−Ej + 1/2mv2 (15)

where 1/2mv2 is the kinetic energy of the free electron and � is the ionization energy ofthe atom.

This emission process is also called radiative recombination, which occurs whenan ion captures an electron and makes a transition to the bound energy state Ej. Thiscontinuum is characterized by discontinuities whenever hv becomes large enough toreach the next bound level. The rate at which the binary reaction involving electronand ion occurs per unit volume, is proportional to the densities of both electrons and




ions. It can be calculated directly from quantum mechanics in analogy to spontaneousemission of line radiation. However, it is more conveniently done by use of Kirchhoff’slaw, which says that in thermal equilibrium the emission can be obtained from thenet absorption (difference of absorption and induced emission) by multiplying by thePlanck’s function.

Free-free emission transitions correspond to loss of kinetic energy by an electronin the field of an ion. This emission results from deceleration of the electron and isknown as bremsstrahlung, meaning ‘braking radiation’. The time between emissions bythe electron is much longer than the time between its collisions with other electrons. Thebremsstrahlung emission does not lead to significant distortion of the electron velocitydistribution which may always be assumed to be near-Maxwellian.

It is difficult to separate emission by radiative recombination from that due tobremsstrahlung except to say that the former is dominant at higher frequencies and thelatter at lower frequencies.

5. BROADENING OF SPECTRAL LINES

The intensities of spectral lines are greatly dependent on the environment of the atomthat emits the radiation. In the ideal case of a free atom the radiated intensity of a lineis spread over a frequency dependent Lorentzian profile having the form

I�� = I0��/4 �2/��−0�2 + ��/4 �2� (16a)

where I0 is the intensity at the center of the line profile 0 and � is the radiation dampingconstant of Eq.(9). This spread of intensity over a range of frequencies is called naturalbroadening of the spectral line and ��/2 � is called full width at half maximum (FWHM)Except at very low atomic densities the ideal condition is never realized in practice andnatural broadening is always accompanied by Doppler broadening which dominates theline shape near its center. Doppler broadening arises due to random thermal motions ofthe emitting atoms and the resulting profile has a Gaussian profile with FWHM given by

�D = �20/c��RT log 2/M�1/2 = 7�16×10−70�T/M�1/2 (16b)

where 0 is the frequency of line center, M is the atomic mass and T the equilibriumtemperature.

When the radiating atom is surrounded by dense plasma both of the above broadeningmechanisms are completely negligible in comparison to the broadening caused by thecharged particles. Since the atom interacts with the charged particles through the electricfields produced by them, this type of broadening is called Stark broadening. There aretwo major reasons for determining the line shapes of spectral lines originating in plasma.The first reason is the use of measured line shapes to determine physical properties ofthe emitting plasma such as the charged particle densities and temperature. The secondreason is to determine the absorption and induced emission coefficients which dependon the oscillator strength and densities of emitting atoms in addition to the line shape.



38 S. N. Thakur

5.1. Stark Broadening

The emitting species (atoms or ions) in plasma are under the influence of electric fieldsby fast moving electrons and relatively slow moving ions. The perturbing electric fieldacting on an atom at a distance ‘r’ from an ion or electron is F = �e/4 �0r2�. Theinteraction between the atom and this electric field is described by the Stark effectwhich splits and shifts the energy levels of the atom. The perturbation to the energylevels caused by the electric field is proportional to F only in the case of hydrogen atomleading to linear Stark effect. For all other atoms the perturbation to the energy levelsis proportional to F2 and the resulting shift and splitting is called quadratic Stark effect.It is obvious that the extent of Stark effect will be negligibly small for large values of‘r’. If we assume that the shift in the energy level of the emitting atom depends on itsexcitation energy as well as on its separation from the ion (or the electron) then thecenter of the spectral line 0 shifts to �r� such that (see Fig.3)

��r� = 0 −�r� = �1/h��E2�r�−�E1�r�� (17a)

It can be seen from the above relation that a +ve value of �E amounts to a downwardshift of the corresponding energy level. A positive value of ��r� corresponds to a shiftof the line center towards red and a −ve value indicates a shift to the blue.

An interaction causing a frequency shift � which depends on ‘r’ and tends to zerofor large r can be written as

��r� = Cn/rn (17b)

where the value of n and the interaction constant Cn depend on the type of the interaction.Thus in the case of hydrogen and hydrogen like atoms exhibiting linear Stark effect

[proportional to F = �e/4 �0r2�], n = 2 and interaction constant is C2. The linear Starkeffect splits the energy levels symmetrically resulting in a symmetrically broadened butunshifted line. In the case of all other atoms the quadratic Stark effect is proportionalto F2 so that n = 4 and corresponding interaction constant is C4 in Eq.(17b) It is tobe noted that because of the F2 dependence the Stark shift is the same for electronsand ions. The quadratic Stark effect splits the energy levels asymmetrically and alsoshifts their centre of gravity downwards (to lower energies). Since the shift of energylevels is larger for higher excitation energies the frequencies of the transitions arereduced. This implies that in Eqn (17b) � > 0 and hence C4 > 0. A spectral linebroadened by quadratic Stark effect becomes asymmetric and its center shifts to longerwavelengths.

5.2. Theory of Stark Effect

We present a qualitative explanation of Stark broadening in terms of interaction of theemitting atom with fast moving electrons and the slowly moving ions in plasma.




5.2.1. Impact approximation

The atom emitting radiation is assumed to collide with electrons and the duration of col-lision tc is taken to be extremely small in comparison to the time between two successivecollisions. The collision is assumed to cause a phase change in the optical wave trainemitted by the atom before and after the collision and not to stop it altogether by knockingout the atom from its excited state. The interaction between the electron and the emittingatom is thus regarded as an optical collision which results in a sudden phase change ofthe light wave emanating from the atom and there is no effect on the state of the atom.

Let us assume that the colliding electron is moving with velocity u along the x-axisand the impact parameter for collision with the atom is �. Since there is a frequencychange ��r� in the radiation emitted by the atom when it is located at a distance ‘r’from the electron, the phase change of the light wave during a small interval of time dtwould be 2 ��r�dt and the total phase change during the collision period tc is

�� =∫ tc

02 ��r� dt =

∫ �

−�2 ��r� dt (18a)

Where ��r� is given by Eq.(17b) and limits on the integral have no effect on thecollision because ��r� is zero outside the interaction duration tc.

From Fig. 5 we have � = r cos �, x = rsin� = � tan �� dt = dx/u = ��/u�sec2�d� andfrom Eq.(17b) we have ��r� = Cn/rn = Cn cosn �/�n

Hence from Eq.(18a) we get

�� = �2 Cn/u�∫ /2

− /2cosn−2 �d� = �2 Cn/u�n−1�an (18b)

Where an is a numerical factor of the order of one and depends on the value of n.In a well known theoretical treatment of collision induced phase change of light wave

by Weisskopf [10], the coherence of the wave train is completely destroyed if the phasechange �� = 1 and the corresponding impact parameter �0 known as Weisskopf radiusis obtained from Eq.(18b)

�0 = �2 Cn/u�1/n−1 (19a)

According to Weisskopf theory the line shape is Lorentzian with a FWHM given by

w = �02uNe = �2 Cn/u�2/n−1uNe (19b)

where Ne is the number density of electrons in the plasma.

x u-

ρ rθ

Fig. 5. Optical collision of electron with the radiating atom.



40 S. N. Thakur

Experimentally it is found that atomic lines in plasma sources are generally redshifted with a Lorentzian profile whereas the Weisskopf theory leads to a broadenedbut unshifted line. This comes from the fact that the theory ignores small phase shiftsoriginating from impact parameters � > �0. Lindholm [11] was the first to include thecontributions of small as well as large phase shifts and later Foley [12] and Anderson [13]included the effects of inelastic collision to obtain a Lorentzian profile with phase shift� and width ‘w’. The spectral line shape in the light of these modifications is given by

I�� = I0/��−0 −��2 + �w/2�2� (20a)

Where � = 2 Nu∫ �D

0� sin � ��d� (20b)

and w = 4 Nu∫ �D

0� �1− cos � �� d� (20c)

The upper limits of the above integrals �D is called Debye shielding radius such that theemitting atom is shielded from the effects of all charged particles located at distancesgreater than or equal to �D. This limit on � is necessary to avoid the electron-atominteraction time ��/u� from becoming very very large. Thus the upper limit for theduration of electron-atom interaction is �D/u and its reciprocal is the plasma frequency�p. The Debye radius is given by

�D = ��0kT/2e2Ne�1/2 ≈ 50�T/Ne�

1/2 (20d)

where T is in Kelvin (K), Ne is in m−3 and �D is in m.For very large values of �� is very small and sin � ≈ � has the same sign

as Cn for all electrons interacting with the atom, but for � in the neighborhood of (both below and above), sin � has both positive and negative values and its averagecontribution to � is very small (see Eq.(20b).

Thus, for 0 ≤ � ≤ � sin � has a non- zero average and it accounts for a shift of linecenter towards red if Cn>0 and towards blue if Cn< 0.

When the values of � are very large cos � ≈ 1 and its contribution to line width w isextremely small (see Eq.(20c). For � → � cos � → −1 and contributes a large valueto ‘w.’

The above description leads to the conclusion that electron collisions for which� ≈ �0 are responsible for most of the line broadening (w) while collisions with �>>�0

contribute greatly to the shift of the line center ��. We can group electron-atom collisionsinto two types: weak collisions (large �, small �) produce line shift and the strongcollisions (small �, large �) produce most of the broadening.

5.2.2. Quasi-static approximation

The interaction between slowly moving ions and radiating atoms can be approximatedby a perturbation which remains nearly constant over the whole time that the atomis radiating. Following Holtsmark [14] the motions of ions are neglected and theirperturbing action is included in an electric field F that produces static Stark effect.In the next step the statistical average of Stark effect over various values of the ionfield-strength F, is taken. In the final step of the calculation, each element of the line




is considered to be broadened and shifted by the electron impacts. The quadratic Starkeffect produces asymmetrically broadened lines whereas the linear Stark effect gives riseto symmetrically broadened line shapes. A large amount of theoretical and experimentalwork has been carried out in the case of spectral line broadening by charged particles.With the help of existing theoretical models it is possible to calculate line profiles thatfit the experimental ones [15–17].

6. APPLICATIONS

In the preceding sections we have seen that the intensities, shapes and widths of atomiclines depend on the atomic structure as well as on the temperature, pressure and electrondensity of the discharge plasma. Analysis of the spectral lines can give information aboutthe physical state of the the emitting gas without in any way interfering with the plasmaThe use of lasers based optical techniques in recent years have replaced spectroscopyas a diagnostic tool to some extent. Nevertheless spectroscopy still plays a major partin determining the physical processes going on in the plasma. It is the most reliablemethod of detecting and identifying trace elements in a source. In the following sectionswe present a brief account of these applications.

6.1. Determination of Electron Temperature

When the temperature of a molecular vapour is increased, molecules tend to dissociateinto atoms and atoms into ion plus electrons; some of the molecules, atoms and ionsare excited to higher energy states; and the kinetic energy of all these particles and ofthe free electrons increases. The spectroscopic determination of electron temperatureof a source of radiation is based on the assumption that local equilibrium conditionsmust exist in each small volume that contributes to emission. Complete thermodynamicequilibrium (TE) exists when all forms of energy distribution are described by the sametemperature.

In the following sections we first discuss the approximate conditions that may prevailin plasma and then describe methods of determining temperature of the source from themeasurements of spectral lines emitted by it.

6.1.1. Temperature and Equilibrium

Maxwell’s distribution for velocities of particles of mass ‘M’ gives the number ofparticles dN with velocity between v and v + dv in terms of their number density N andtemperature T as follows:

dN �v� = N �M/2 kT�3/2 exp�−mv2/2kT� 4 v2dv (21a)

The Boltzmann distribution of particles Nj having excitation energy Ej is

Nj = N �gj/U�T�� exp�−Ej/kT� (21b)

where U�T� = �gj exp�−Ej/kT� is called the state sum or partition function.



42 S. N. Thakur

The condition of equilibrium for ions, electrons and neutral atoms is given by Saha’sequation

NeNi

N0

= 2�2 mkT�3/2

h3

Ui�T�

U0�T�exp �−�/kT� (21c)

where Ne is the number density of electrons and Ni and N0 that of the ions and neutralatoms respectively, regardless of the energy levels they occupy, m is the mass of electronand � is the ionization energy of atom. The factor of 2 represents the state sum for thetwo possible spin states.

The equilibrium distributions for kinetic energy, excitation energy and ionizationenergy are represented by Maxwell, Boltzmann and Saha equations respectively andit may happen that there is equilibrium distribution of one of these forms of energybut not for the others. In that case the temperature parameters for Eq. (21a), (21b) and(21c) are all different. Complete thermodynamic equilibrium exists when all forms ofenergy distribution are described by the same temperature parameter. Thermodynamicarguments require that for equilibrium to hold, for every photon emitted by the system,a photon of the same energy be absorbed and for every excitation by electron collisionthere must be a de-excitation by electron collision. In practice, however, photons do leakout from the plasma, no matter how large or dense the plasma is, otherwise we wouldbe unable to observe the plasma. Thus the condition of near thermodynamic equilibriumrequires that such losses be small compared to the total energy. Many plasmas can bedescribed by a state known as local thermodynamic equilibrium (LTE), in which it ispossible to find a temperature parameter for every point in a region of space that fits theBoltzmann and Saha relations for the population density of excited and ionic states andthe Maxwell distribution of velocities among the electrons. The criterion for LTE is thatcollisional processes must be much more important than radiative, so that the deficit ofradiative energy is extremely small. In other words the probability of de-excitation byinelastic collision for an excited state must be very large compared to that of spontaneousemission. This is possible at very high electron densities in the plasma such that

Ne >> A21/�v�21� (22a)

If S is the line strength, the excitation cross-section corresponding to electron velocity vat threshold is given by �12�v�∼ �e/4 �0h�2S/v2. Putting this value of �12 in Eq. (22a)makes the right hand side proportional to vA21/S. Since A21 is proportional to S3

and v is proportional to T1/2, the value of Ne is proportional to 3T1/2 and numericalrelationship for LTE is given by

Ne >> 1�6×1012T1/2�E2 −E1�3 cm−3 (22b)

where T is electron temperature in K and E2 − E1 is the energy difference in electronvolts between the two neighboring states with an allowed transition.

It is possible to determine the relative population of atoms in various excited stateseven in the absence of LTE provided the collisional cross-sections and radiative transitionprobabilities are known. Another approximation is to assume that level E2 in Fig. 3is populated entirely by electron collision and depopulated entirely by spontaneous




radiation. This is called coronal equilibrium (CE) since it is applicable to sun’s coronawhere temperature is high �106 K�, electron density is low �108cm−3� and radiationdensity is also low. The populations of higher states are much lower in CE than in LTE.Coronal equilibrium can hold only if Ne is below a critical value and it holds good onlyfor the lower excited states. It is quite possible that the atomic populations of higherstates follow LTE while those of the lower states follow CE.

6.1.2. Temperature from Relative Intensities of Lines

The method for determination of temperature in LTE plasma is based on the factthat the number densities in various excited states follow Boltzmann distribution. Thetemperature in terms of relative intensities of lines from the same element and samestate of ionization is given by

kT = �E2 −E1�/ loge�I1�31g2f2/I2�

32g1f1� (22c)

where I1 is the total intensity integrated over the line profile, �1 is the wavelength andf1 the oscillator strength of the spectral line with excitation energy E1 and I2� �2 andf2 are the corresponding quantities for the line with excitation energy E2, The statisticalweights for energy states E1 and E2 are g1 and g2 respectively.

Relative intensities can be measured with an accuracy of better than 10 per cent buterrors in the oscillator strengths are more. Since kT is of the order of the largest energyseparation E2 − E1 between excitation energies of non-resonance lines, the uncertaintyof oscillator strengths is reflected in the errors associated with the temperature.

The method described above based on the relative intensities of lines from the sameatom and ionization stage generally leads to inaccurate temperature. The main reasonbeing the small separation between E1 and E2 which is typically smaller than or equal tokT and it renders the line-intensity ratio somewhat insensitive to temperature variations.This shortcoming is removed if spectral lines from successive ionization stages of thesame atom are compared. The effective energy difference is enhanced by the ionizationenergy leading to increased sensitivity to temperature changes. In LTE the relationbetween relative intensities and the source temperature is given by

�I′/I� = �f ′g′�3/fg�′3��4 3/2a30Ne�

−1�kT/EH�3/2 exp �E +�E� −E′ −E��/kT (22d)

where primed quantities correspond to the spectral line from the higher ionization stageand �E� is the reduction of the ionization energy E� of the lower ionization stage dueto plasma effect, a0 is the Bohr radius and EH is the ionization energy of hydrogen.

6.1.3. Temperature from Doppler Profile

The most reliable spectroscopic technique of measuring kinetic temperature of atoms andions is based on the measurement of the widths of Doppler broadened spectral lines. Inthe case of Maxwellian velocity distribution of emitting species such lines have Gaussianprofiles with FWHM given by Eq. (16b). One must make sure that the thermal Dopplereffect is the major cause of line broadening in the source before using this method oftemperature measurement. It has been found that even gross motions in the source could



44 S. N. Thakur

simulate a Gaussian profile. In order to avoid such misinterpretations it is advisable toobserve the plasma from different angles to watch for shifts of the intensity maxima.

6.2. Determination of the Electron Density

The most powerful spectroscopic technique of determining the electron density Ne ofdischarge plasma comes from the measurement of the Stark broadening of spectrallines. In this method absolute intensities of spectral lines are not required, merely lineshapes and FWHM are sufficient. Since broadening is quite appreciable for electrondensity N ≥ 1015 cm−3, standard spectrometers often suffice to record the spectra formeasurements of line shape. The electron density Ne is extracted by matching the linewidth (or the entire line shape) with the calculated one. Details of line shape calculationscan be found in a book by Griem [18]. In this section we will summarize only the salientpoints.

6.2.1. Hydrogen & hydrogen-like ions

Hydrogen and hydrogen-like ions exhibit linear Stark effect. The broadening of spectrallines is found to be dependent on the optical transition and a judicious choice is important.At low plasma densities a spectral line with large Stark broadening is desirable but athigh densities a line with relatively smaller broadening is useful so that its outer wingsdo not overlap with neighboring lines. The FWHM (in A) of the spectral line in thequasi-static approximation is given by:

�� = 8�16×10−19�1−0�7ND−1/3� �02 �n2

2 −n12� �Z1/3

p /Ze� N2/3 (23a)

where ND = �4 /3�N�D3 is the number of particles in the Debye sphere, N is in cm−3� �0

is the line center, n2 and n1 are the principal quantum numbers of the upper and lowerstates respectively, Zp is, the nuclear charge on the perturbing ion and Ze that on theemitting particle (atom or ion)

In the above discussion we have neglected contributions to FWHM from the plasmaelectrons. Although the line shapes do depend on the electron contribution, the FWHMare generally insensitive. Eq. (23a) represents a very good estimate of FWHM in thosehydrogenic lines that do not have a strong undisplaced Stark component as for example,the Lyman �, Lyman �, Balmer � and Balmer � transitions. On the other hand the FWHMof hydrogenic lines with strong central Stark components are dominated by interactionof the electron with the emitting atom such as Lyman � and Balmer � transitions. Suchlines have a Lorentzian line shape and FWHM for Lyman � transition in the impactapproximation is given

�� ≈ 1�62×10−17�N/�T 1/2��13�76− log �N1/2/T�� (23b)

where �� is in A, T is in K and N is in cm−3.It is seen from Eq. (23a,23b) that the ion broadening, in the quasi-static approxima-

tion, varies as N2/3 and is virtually independent of temperature whereas the collisionalbroadening, in the impact approximation, varies approximately as N and it is very much




temperature dependent. In case T is not reliably known, it is advisable to determineelectron density N from Balmer � line who’s FWHM is ion- dominated. It is to benoted that electron densities determined from Eq. (23) are only crude estimates of N andone must compute the entire line profile to extract the total line width for an accuratevalue of N. Some very precise hydrogen line shape measurements have been reportedby Wiese and coworkers [19,20].

6.2.2. Many-electron atoms & ions

Plasma generated by high power lasers focused on gas or solid targets, have electronenergies in the range of 50 eV to 10 keV and hydrogen as well as other low Z atomsare fully stripped and do not radiate. It is found that carbon is completely stripped whenT ≥ 100 eV and copper is stripped when T ≈ 1keV [21] which suggests that sufficientlyhigh-Z atoms should be involved in spectroscopic measurements on hot plasma. Spectrallines of neutral atoms and non-hydrogenic ions have Stark broadening mostly due toelectron impacts. For such high-Z ions the Stark widths are too small to be resolvedand they may be dominated by Doppler broadening. Non-hydrogenic high-Z ions aretherefore not very suitable for determining electron densities in very hot plasma.

6.3. Qualitative Emission Analysis

One of the major applications of atomic emission spectra is the identification of elementspresent in the source of light. The use of this technique for qualitative analysis ofsamples which could be fed into flames to emit characteristic light (yellow for Na, redfor Ca) dates back to Bunsen and Kirchoff and other spectroscopists of 19th century.Major developments in astrophysics have resulted from the studies of the spectra ofradiating stellar bodies which provided information about their chemical composition,temperature, distance, mass etc. Some elements are easily excited for their emissionspectra to be observed and recorded, than others. Thus non-metal atoms are more difficultto excite than metal atoms because of their high ionization potential. It is found thatpresence of easy to excite atoms in a sample, suppress the emission from atoms that arerelatively difficult to excite. Thus emission from helium is suppressed in the presenceof nitrogen, that of nitrogen in presence of mercury and the emission from mercury issuppressed if the sample also contains potassium.

When an element is excited in an arc or discharge source, a number of lines of varyingintensities are observed over a wide range of the spectrum. As one dilutes the amountof the element in the arc, the number of lines observable is reduced and ultimatelyonly a few lines of that element remain observable. These lines are known as persistentlines. It has been found that the persistent lines are also the lines of largest intensity.From this description of the persistent lines it is apparent that in the qualitative analysisfor a particular element we need look only for the persistent lines of that element. Ifthey are absent it may be safely assumed that the element is not present in the sample.Tables of persistent lines of many elements are to be found in a book by Brode [22]and in a publication by Meggers [23]. Persistent lines for some elements are given inTable 2. In most cases of qualitative analysis, the major constituents of the dischargeare readily determined by inspection of the strongest lines and other lines belonging to



46 S. N. Thakur

Table 2. Wavelengths (in Angstrom) of Persistent Lines of some Elements

Element Wave-length Element Wave-length Element Wave-length Element Wave-length

Ag I 3280�68 Cd I 2288�02 Be I 2348�61 Cu II 2192�263382�90 2312�84 2650�78 2246�995209�07 2573�09 3321�01 Fe I 3581�205465�49 2748�58 3321�09 3719�94

Ag II 2246�41 3261�06 3321�34 3737�132437�80 3403�65 Be II 3130�42 3745�56

Al I 3032�16 3466�20 3131�07 3745�903092�71 3610�51 Bi I 2061�70 3748�263944�03 6438�47 2276�58 Fe II 2382�043961�53 Cd II 2144�38 2780�52 2395�63

Al II 1671�0 2265�02 2809�63 2404�881856�0 Cl II 4794�54 2897�98 2410�521858�1 4810�06 2938�30 2413�311862�5 4819�46 2989�03 Hg I 1849�682631�55 Co I 3453�51 3067�72 2536�522669�17 3465�80 Bi II 1909�41 3650�152816�18 3529�81 4722�55 3654�836231�76 Co II 2286�16 Br II 4704�86 3663�286243�36 2307�86 4785�50 4046�56

Au I 2427�95 2363�79 4816�71 4358�352675�95 2378�62 C I 2296�89 5460�742802�19 2388�92 2478�57 K I 4044�14

B I 2496�78 2519�82 C II 2836�71 4047�202497�73 3405�12 2837�60 7664�91

B II 3451�41 Cr I 4254�35 4267�02 7698�98

Ba I 2304�23 4274�80 4267�27 Mg I 2852�132335�27 4289�02 Ca I 4226�73 3829�353071�59 5204�52 4425�44 3832�315424�62 5206�04 4434�96 3838�205519�12 5208�44 4454�78 5167�345535�55 Cr II 2835�63 Ca II 3158�87 5172�705777�67 2843�25 3179�33 5183�62

Ba II 3891�79 2849�84 3933�67 Mn I 4030�764130�66 2855�68 3968�47 4033�074554�04 2860�92 Cu I 5218�20 4034�494934�09 Cu II 2135�95 Mn II 2576�10

these constituents can then be identified by use of the tabulation of spectral lines of theelements [2,24, and 25]. After this procedure, the lines left unidentified are generallyweak and they may be hitherto unobserved lines of the major constituents or lines of theunknown minor constituents. In the latter case, they must be among the stronger lines ofthese elements. The unknown element can then be identified by comparing these lineswith the persistent lines of various elements.




The qualitative analysis of a sample may be divided into two parts:

(1) search for a definite element(2) identification of unknown elements

In search for a definite element, the identification of the persistent lines should besufficient to confirm the presence of the element. The identification of unknown elementsrequires a more complete identification of all spectral lines in the emission spectrum ofthe sample. At least three lines concerning whom there is no possible doubt as to originshould be identified so as positively identify the element.

6.4. Quantitative Emission Analysis

The determination of the elemental composition of a gaseous or condensed phase sampleby means of laser induced plasma requires the measurement of the intensities of thosespectral lines that are characteristic of the individual elements present in the sample. Theintensities must then be related to the number density of atoms or ions present in theplasma. The first step in the quantitative analysis is the preparation of standard samplesin which the concentration of the element of interest is varied in a precisely knownmanner. It is obvious that the intensities of spectral lines of this element in the emissionspectra recorded for different standard samples will be proportional to its concentration.The next step in the analysis is the measurement of intensities of one or more spectrallines of the said element from spectra of all the standard samples as well as the samplesto be analysed. The last step of the analysis is to plot a working curve of line intensityagainst known concentration of the element for the standard samples. The concentrationsof the element in the unknown samples are determined from the working curve from themeasured intensities of their spectral lines.

The procedure outlined in the previous paragraph for quantitative analysis is basedon the assumption that the excitation conditions for the unknown as well as the standardsamples are exactly identical. This is never realized in practice and fluctuations inemission from the plasma may cause the relative intensities of a line for two samplesof different elemental concentration not to reflect their relative concentrations. Thisshortcoming due to uncontrolled random fluctuations of emission intensity arising fromdifference in excitation conditions leads to a less accurate working curve and consequenterror in quantitative estimates of the element in unknown samples. To avoid this error itbecomes necessary to refer the intensities of lines in the unknown and standard samplesto some common line of another element which remains unchanged in both spectra.Such a spectral line is called internal standard because its intensity is also affected bythe random fluctuations as that of the element under investigation.

Two types of internal standard lines may be used:

(1) a weak line of the element which is the major constituent of the sample whoseintensity remains constant for all standard samples.

(2) a persistent line of an added small amount of an element which is known not tobe present in either the unknown or the standard samples.



48 S. N. Thakur

Since the lines from the element to be analyzed as well as the internal standard originatein the plasma from the same sample, it is obvious that any variation in the excitationconditions will not affect their relative intensities. The working curve is now plottedby using this relative intensity against the concentration of the element in the standardsamples. This procedure reduces the errors in quantitative analysis by a very largefactor. For further details of quantitative analysis the reader is referred to the books byBrode [22], Sawyer [26] and Radziemski and Cremers [27].

REFERENCES

[1] Transactions of International Astronomical Union, 12A (1964) 137[2] G. R. Harrison, M.I.T. Wavelength Tables, Wiley, New York (1959)[3] H. C. Longuet-Higgins, J. Chem. Edu. 34 (1957) 30[4] G. Herzberg, Atomic Spectra and Atomic Structure, Dover, New York (1944)[5] H. G. Kuhn, Atomic Spectra, Longman, London (1964)[6] V. Kondratyev, The Structure of Atoms and Molecules, Foreign Language Publishing House,

Moscow (1960)[7] B. Cagnac, J.C. Pebay-Peyroula, Modern Atomic Physics: Fundamental Principles, The

Macmillan Press, London (1975)[8] P. Bousquet, Instrumental Spectroscopy, Dunod, Paris (1969)[9] A. P. Thorne, Spectrophysics, Chapman & Hall and Science Paperbacks, London (1974)

[10] V. Weisskopf, Phys. Z. 34 (1933) 1[11] E. Lindholm, Arkiv Mat. Astron. Fysik. 28B (1941) no 3[12] H. M. Foley, Phys. Rev. 69 (1946) 616[13] P. W. Anderson, Phys. Rev. 86 (1952) 809[14] J. Holtsmark, Ann. Physik 58 (1919) 577[15] W. R. Hindmarsh, Prog. in Quantum Electronics 2 (1972) 143[16] D. D. Burgess, Space Science Reviews 13 (1972) 493[17] H. R. Griem, Plasma Spectroscopy, McGraw-Hill, New York (1964)[18] H. R. Griem, Spectral Line Broadening by Plasmas, Academic Press, New York (1974)[19] W. L. Wiese, D.E. Kelleher and D.R. Paquette, Phys. Rev. A6 (1972) 1132[20] W. L. Wiese, D.E. Kelleher and V. Helping, Phys. Rev. A11 (1975) 1854[21] D. Mosher, Phys. Rev. A10 (1974) 2330[22] W. R. Bride, Chemical Spectroscopy, Second Edition, John Wiley, New York (1952)[23] W,F. Meggers, J. Optical Soc. Am. 31 (1941) 44, 605[24] C. E. Moore, Selected Tables of Atomic Spectra, Washington, N.S.R.D.S.-N.B.S. 3, Section 1

(1965) and Section 2 (1967)[25] C. E. Moore, Bibliography on the Analysis of Optical Atomic Spectra, Section 1, H toV,

N.B.S. Special Publ. 306, Washington (1968)[26] R. A. Sawyer, Experimental Spectroscopy, Third Edition, Dover, New York (1963)[27] L. J. Radziemski and D. A. Cremers (Ed), Laser Induced Plasma and Applications, Marcel

Dekker, New York (1989)



Chapter 3

Laser Ablation

R. E. Russo, X. L. Mao, J. H. Yoo and J. J. Gonzalez

1 Cyclotron Road, Lawrence Berkeley National LaboratoryBerkeley, CA 94720, USA

1. INTRODUCTION

The definition of ablation from the Merriam-Webster Dictionary is “loss of a part bymelting or vaporization”. In order for ablation to occur, energy absorption is needed.The energy can be provided in the form of electrical discharges (e.g. an arc and spark)or in the form of light (e.g. as a laser). Laser ablation means using laser light energyto remove a portion of a sample by melting, fusion, sublimation, ionization, erosion,and/or explosion. Laser ablation results in the formation of a gaseous vapor, luminousplasma, and in the production of fine particles. By measuring the emission spectrum fromthe laser-induced plasma, qualitative and quantitative information about the sample’schemical composition can be obtained. This measurement technology is known as LaserInduced Breakdown Spectroscopy (LIBS). LIBS is an exciting field of study, boththeoretically and experimentally due to the wealth of diverse mechanisms underlyingthe physical processes and its significant potential for spectrochemical analysis. Thischapter will discuss the fundamental mechanisms of laser ablation processes and theirrelation to LIBS.

“The history of the interaction of high-power lasers with solid matter is as old asthe laser itself”[1]. In 1917 Albert Einstein [2] first proposed that stimulated emissionof light (process that makes lasing possible) should occur in addition to absorption andspontaneous emission. It then took over 40 years until the development of theoreticalprinciples of lasers were established by Arthur Schawlow and Charles Townes [3] in1958 and two more years to develop the first laser source by Theodore Maiman [4,5].The laser was built using a rod of synthetic ruby as the active medium. In 1962 the firstaccount of laser ablation was presented by Breech and Cross [6] at the InternationalConference on Spectroscopy held at the University of Maryland. A ruby laser was usedto vaporize and excite atoms from solid surfaces, and the plasma spectrum was usedto characterize the elemental composition of the sample. This paper began the field oflaser microprobe emission spectroscopy, which was one of the first real applicationsof laser ablation. Throughout 1963 and 1964, about a dozen publications detailed earlylaser ablation experiments [1]. Many phenomena, which were first observed in thoseyears, are still the subject of study today. During the 1970s and early 1980s, the




50 R. E. Russo et al.

development of lasers and the understanding of laser ablation were incremental andsteady. The fastest growing applications during the eighties were driven by the needsof materials science. Several laser ablation-based methods reached maturity in the late1980, including pulsed laser deposition (PLD) [7] for making high-Tc superconductorthin films, micro-machining [8], and laser-based medical applications such as laser-based ophthalmology (LASIK) [9], removal of birthmarks, tattoos and smoothing ofwrinkled skin in dermatology [10,11]; laser surgery for internal arthroscopic cutting andfor arterial angioplasty [12]; and for dental applications [13].

For analytical purposes, laser ablation became widely used for a range of microanaly-sis applications. Matrix-assisted laser desorption/ionization (MALDI) [14] revolutionizedthe identification and study of large molecular weight bio-molecules and polymers. Laserablation as a sampling technique coupled to established analytical techniques such asinductively coupled plasma atomic emission spectrometry (ICP-AES) [15] and induc-tively coupled plasma mass spectrometry (ICP-MS) [16], improved analytical capabilityfor direct solid analysis by minimizing the time of analysis, reducing hazardous chem-ical exposure and waste, and by providing reliable alternative to acid dissolution forchemical analysis. Laser ablation has been recognized for its powerful advantages; rapid,in situ, multi-element analysis of any kind of sample with no sample preparation. Diag-nostics and theoretical studies have advanced laser ablation research to a very activefield supported by a wide range of applications. The basis of LIBS is rooted in laserablation. Laser ablation is the first step in the LIBS process, and its influence will bereflected in the “figures of merit”, temporal and spatial resolution, sensitivity, precision,and accuracy. The influence of laser ablation on LIBS is addressed in this chapter.

2. FUNDAMENTAL ABLATION PROCESSES

Laser ablation is governed by a variety of distinct nonlinear mechanisms. Once the laserbeam illuminates the sample, mass leaves the surface of a sample in the form of electrons,ions, atoms, molecules, clusters, and particles, each of the processes separated in timeand space. An understanding of the fundamental mechanisms involved in each of theseprocesses is critical for efficiently coupling the laser beam to the sample and removingmass in the appropriate form for analysis. Understanding laser-material interaction willallow ablation of stoichiometric vapor and control of the laser-induced plasma propertiesfor optimum LIBS performance.

Laser ablation will be divided into three main processes for discussion in this chapter:bond breaking and plasma ignition, plasma expansion and cooling, and particle ejectionand condensation. These laser ablation processes occur over several orders of magnitudein time, starting with electronic absorption of laser optical energy �10−15 sec� to particlecondensation �10−3 sec� after the laser pulse is completed.

Fig. 1, shows a summary of these three processes and various mechanisms occurringduring each. During the plasma ignition process, the mechanisms and plasma propertiesstrongly depend on the laser irradiance and pulse duration. For a nanosecond laser pulsewith irradiances less than 108 W/cm2, the dominant mechanism is thermal vaporization:the temperature of the solid surface increases, and a well defined phase transitionoccurs, from solid to liquid, liquid to vapor, and vapor to plasma. For a picosecondlaser pulse with irradiance between 1010–1013 W/cm2, both thermal and non-thermal



Laser Ablation 51

ns laser (107 – 1011

W/cm2)

Thermal vaporization (10–9–10–8 s)

Non-thermal ablation (10–9–10–8 s)

Plasma shielding (10–9–10–8 s)

Plasma ignitionfs laser (1012

– 1017 W/cm2)

Electronic excitation and ionization(10–15–10–13

s)Coulomb explosion (10–13

s)Electron–lattice heating (10–12

s)

Particles ejectionand condensation

Nano particles formation (10–4–10–3 s)

Ejection of liquid droplet (10–8–10–6 s)

Solid exfoliation (10–6–10–5 s)

Plasma expansionand cooling

Shockwave propagationPlasma expansion (10–11–10–6

s)

Plasma radiation cooling (10–6–10–4 s)

Fig. 1. A summary of laser ablation processes and various mechanisms occurring during eachprocess.

mechanisms such as Coulomb explosion exist. For irradiances higher than 1013 W/cm2

with femtosecond laser pulse, Coulomb explosion is the main bond breaking mechanism.When the laser pulse duration is in the nanosecond time region, the later part of laser

pulse can be absorbed by the laser induced plasma, which is called plasma shielding.For picosecond pulsed laser ablation, the laser pulse is too short to be absorbed by theplasma. Plasma shielding will influence how much of the solid mass is converted intovapor and the properties of vapor. However, an air plasma can form during the picosecond laser pulse duration due to seed electrons from the target surface; this air plasmacan absorb part of the picosecond pulse. With femtosecond laser pulses, plasma shieldingcan be neglected because, to the best of our knowledge, no mass can be ejected fromthe surface during the short pulse duration.

Plasma expansion begins after the plasma ignition process. The plasma expansionprocess will be governed by the initial plasma properties (at the end of the laser pulse)and the expansion medium. The properties (electron number density, temperature, andexpansion speed) of the plasma initially are strongly dependent on the laser properties.Plasma expansion will be related to the initial mass and energy in the vapor plume, andthe gas environment. Plasma expansion will be adiabatic until approximately 1 micro-second after the laser pulse. After that time, line radiation will be a dominant energyloss influencing the temperature decrease.

Particle formation will be influenced by these primary processes. Nano-sized particleswill be formed from condensation of the vapor. Condensation starts when the vaporplume temperature reaches the boiling temperature of the material (∼3000 K) and stopsat the condensation temperature of the material (<2000 K). Liquid ejection of particles




can occur by high pressure gradient forces within the highly expanding vapor plumeacting as the molten surface. Solid sample exfoliation can occur from the large thermalstress gradients of the fast heating process; thermal stresses can break the sample intoirregular shaped particles, ejecting them from the surface. More detailed explanationsfor each of the processes diagramed in Fig. 1 are presented in detail in the followingsections.

2.1. Plasma Ignition Processes

Understanding plasma ignition processes will help to determine optimum conditions forLIBS measurements. The plasma ignition processes include bond breaking and plasmashielding during the laser pulse. The plasma conditions, after the laser pulse terminates,will determine the expansion and cooling.

Bond breaking mechanisms influence the amount and forms of energy (kinetic, ioniza-tion and excitation) that atoms and ions acquire. Plasma shielding can increase the energyby additional heating, before the laser pulse is finished. These mechanisms stronglydepend on the laser irradiance and pulse duration, as described previously for nano, picoand femto-second lasers. Plasma shielding can be dominant when the laser irradiancereaches certain thresholds. In the following sections, these different mechanisms will bediscussed in detail for the three (ns, ps and fs) laser pulse durations.

2.1.1. Femtosecond laser ablation

When a femtosecond laser pulse interacts with a solid sample, different electronicmechanisms are excited, depending on the sample material. For conducting samples,free-electrons inside the solid can directly absorb laser energy and form hot electron–hole plasma. For semi-conductor and wide bandgap dielectrics, the electron–hole plasmais created through nonlinear processes such as multi-photon absorption and ionization,tunneling, and avalanche ionization. At high energy, the electron-hole plasma createdon the surface of the solid will induce emission of x-rays, hot electrons, photoemission,and produce highly charged ions through a phenomenon called Coulomb explosion ornon-thermal melting.

For wide bandgap dielectrics, the simultaneous absorption of multiple photons resultsin a photoionization rate that is strongly dependent on the laser intensity [17]. The rateof multiphoton absorption can be expressed as �In, where I is the laser intensity and �is the n-photon absorption cross section for a valence band electron to be excited to theconduction band. The number of photons required is determined by the smallest n thatsatisfies the relation, nh�>Eg, where Eg is the bandgap energy of the dielectric material,and h� is the photon energy. A second photoionization process, tunneling ionization,may come into play under an extremely strong laser electromagnetic field interactionwith dielectrics. In the strong-field regime, the superposition of the nuclear Coulombfield and the laser electric field results in an oscillating finite potential barrier throughwhich bound electrons can tunnel, thus escaping the atom. In dielectrics, this mechanismallows valence electrons to tunnel to the conduction band in a time period shorter than thelaser pulse duration. Both multiphoton and tunneling ionization can be treated under the



Laser Ablation 53

same theoretical framework developed by Keldysh [18]. The transition from multiphotonto tunneling ionization is characterized by the Keldysh parameter � [18].

� = �(2meEg

)1/2

eEA

(1)

where me and e are the effective mass and charge of the electron and EA is the amplitudeof the laser electric field oscillating at frequency �. When � is much larger than one,which is the case for high intensity laser interactions with dielectrics, multiphotonionization dominates the excitation process.

For semiconductor samples, where the photon energy is larger than the bandgap,single photon absorption is the dominant mechanism for exciting valence electrons tothe conduction band [19,20]. In the case of semiconductors with an indirect bandgap,such as silicon, single photon absorption can still occur with photons of energy greaterthan the gap, but phonon assistance is necessary to conserve momentum.

Once an electron-hole plasma is formed inside the solid, the carriers can absorbadditional laser photons, sequentially moving to higher energy states. The absorptioncoefficient �0 depends on the imaginary part of the refractive index �, which is relatedto the dielectric function . According to the Drude model [21], can be expressed as:

= 1−�2p

[2

1+�22+ i

2

� �1+�22�

]

(2)

where � the scattering time is typically a fraction of a femtosecond and depends on theconduction electron energy. �p is the plasma frequency defined by

�p =√

e2N

0me

(3)

where N is the carrier density and 0 is the electric permittivity.Photon absorption increases the carrier energy of the electron–hole plasma; when the

energy of carriers is well above the bandgap (or Fermi level in a metal), collisionalionization generates additional excited carriers. A high energy electron can ionize anotherelectron from the valence band, resulting in two excited electrons with lower energy atthe conduction band [22,23]. These electrons can be heated by the laser through freecarrier absorption and impact additional valence band electrons. This process can repeatitself as long as the laser electromagnetic field is present and intense, leading to the so-called electronic avalanche. Avalanche ionization requires seed electrons to be presentin the conduction band, which can be excited by photoionization. The following rateequation can describe the injection of electrons into the conduction band of dielectricsusing femtosecond to picosecond laser pulses, under the combined action of multiphotonexcitation and avalanche ionization [24]:

dN

dt= aIN +�NIn (4)

where a is a constant. Within the fs timescale, a large number of excited electrons canleave the solid, the lattice modes remain vibrationally cold and the irradiated solid consistof charged ions and an electron-hole plasma. After about 10% of the valence electrons




are removed, the lattice is weakened and begins to melt. This is called non-thermalmelting (Coulomb explosion) caused by high energy electrons and ions [25–29].

Coulomb explosion processes were studied by time resolved reflectivity and x-raydiffraction, using pump-probe experiments [30]. Optical reflectivity spectra for fem-tosecond laser-excited silicon are plotted in Fig. 2. Negative times mean that the probebeam arrived at the surface before the pump ablation laser pulse. The dashed–dotted andthe dashed lines represent the well-known reflectivity spectra of crystalline and moltensilicon, respectively. When the probe pulse occurred 120 fs before the ablation pulse, thereflectivity spectrum was similar to that of crystalline Si. An abrupt increase in the opti-cal reflectivity occurred within three hundred femtoseconds after the ablation pulse, thereflectivity spectrum changes from crystalline to liquid character within this short time.

High energy femtosecond laser material interactions can produce x-ray radiation withduration comparable to that of the laser pulse. This x-ray pulse can be used to detect lat-tice changes by x-ray diffraction using a pump-probe method with femtosecond time res-olution. Fig. 3 shows time-resolved x-ray diffraction [30] measured from the (111) lat-tice planes of Ge as a function of the delay time between the x-ray probe and the laserexcitation pulse for two different laser fluences. Negative times indicate that the x-rayprobe pulse arrived before the laser excitation pulse. The diffraction intensity remainedunchanged when the x-ray pulse arrived before the excitation (ablation) pulse. A sharpdecrease in the diffraction was observed after the arrival of the ablation pulse. The ini-tial decrease in reflectivity takes approximately 300 femtoseconds and indicates latticemelting. Conventional electron phonon interactions occur on the picosecond time scale.Therefore, the fast decrease in diffraction (within a few hundred femtoseconds) and opticalreflectivity demonstrate that a portion of the Ge crystal underwent non-thermal melting.

2.1.2. Picosecond laser ablation

For picosecond laser ablation, the lattice could be melted through thermal and/or non-thermal processes, depending on the laser irradiance. Electrons are ejected from the

2000.0

0.2

0.4

0.6

0.8

1.0

400 600

cryst-Si

liquid-Si

300 fs

50 fs

–20 fs

–120 fs

Wavelength (nm)

Ref

lect

ivity

800 1000

Fig. 2. Spectra of the optical reflectivity of silicon. Dash–dotted curve: crystalline silicon. Dashedline: molten silicon. Data points: spectra measured at various time delays between the pump pulseand the optical probe pulse (optical pump/optical probe measurements). Ref [30]



Laser Ablation 55

–1

0.6

0.8

1.0

0 1

0.2 J/cm2

300 fs

0.4 J/cm2

2 ∞

Delay time [ps]

Inte

grat

ed r

efle

ctiv

ity

Fig. 3. X-ray diffraction from the (111) lattice planes of Ge versus time delay for two differentenergy fluences of the pump pulse. Ref [30]

target surface during the laser pulse. The free electrons can interact with the air andabsorb laser energy to initiate an air plasma during the ps laser pulse duration. Theplasma forms long before the plume forms [31–33]. Fig. 4 shows the measured airplasma electron number density Ne as a function of distance z from the target surface,at 150 ps delay time between the pump and probe beams. The electron density at z wasmeasured from the interference pattern (insert of Fig.4) using the expression

Ne�z� = 20me�2�q�z�

e2l�z�(5)

0

2.0 × 10

20

1.5 × 10

20

1.0 × 10

20

5.0 × 1019

0.050 100

z (μm)

z

TargetNe

(cm

–3)

150 200 250

Fig. 4. Electron number density profile along the incident laser axis. The solid curve is a leastsquare fit of the experimental data showing exponential decay. The inset is an interferogram ofthe picosecond laser ablation plasma. Ref [33]




where q(z) and l(z) are, respectively, the average phase shift and width of the plasmaat location z. � and � are circular frequency and wavelength of the probe beam. Theelectron number density of this air plasma was on the order of 1020 cm−3 which ishigher than the air density. The air plasma was observed immediately and expandedlongitudinally during the laser pulse. Its longitudinal extent remained approximatelyconstant after about 100 ps, after which the plasma expanded principally in the lateral(radial) direction. The evolution of the laser-ablated air plasma was simulated with atwo-fluid plasma model [31,34].

The air plasma above the sample would absorb a part of the incoming laser beamradiation. Unlike ns laser ablation, plasma shielding is not caused by absorption fromthe vapor plume; on the picosecond time scale, plasma shielding is caused by the airplasma. To confirm this plasma shielding mechanism in ps laser ablation, the lateralexpansion of early stage ablation plasma induced by a 1064 nm, 35 ps laser pulse on acopper target was measured. A relation of t1/2 was found for the lateral expansion of theair plasma. Measurements of energy absorption by the air plasma (∼10% of incominglaser energy) confirmed plasma shielding for picosecond laser ablation.

2.1.3. Nanosecond ablation

When the pulse duration is on the order of a few nanoseconds, and laser irradiance is onthe order of 107–1011 W/cm2, some of the mechanisms involved in ablation are: melting,fusion, sublimation, vaporization, ionization, etc. If the laser irradiance is high enough,non-thermal ablation is also important and can co-exist with these thermal mechanisms.

When the laser irradiance is less than 108 W/cm2, thermal processes are dominant.The temperature at the target surface will rise during the laser pulse, and eventually thetarget will melt and vaporize. The temperature distribution in the target can be calculatedwith the heat conduction equation [35].

�T �x� t�

�t= �

�x

[(�

Cp�s

)�T �x� t�

�x

]

+ �

Cp�s

I �x� t� (6)

where T represents the temperature inside the target, x is the position from the surface, �,Cp, �s and � denote the thermal conductivity, heat capacity, mass density and absorptioncoefficient of the solid target material, respectively.

The thermal evaporation rate Jv is the function of surface temperature. Assumingthermal equilibrium,

Jv = 1�06×106 exp(

−Lv

kB

(1Ts

− 1TB

))√M

2�kBTs

(7)

where Lv is the heat of vaporization and M is mass of vapor. kB is the Boltzmannconstant. Tb and Ts are the boiling-point temperature and surface temperature of thesample, respectively.

Vaporized mass can be ionized by absorbing the incoming laser beam, forming aplasma. Laser radiation is absorbed primarily by inverse Bremsstrahlung, which involves



Laser Ablation 57

the absorption of a photon by free electrons during the collision with heavy particles(ions and atoms). The inverse Bremsstrahlung absorption coefficient is given by:

�IB =[

QNeN0 + 4e6�3NeZ2Ni

3hc4me

×(

2�3mekBTe

)1/2]

×[

1− exp(

− hc

�kBTe

)]

(8)

where Q is the cross section for photon absorption by an electron during the collisionwith atoms, c is the speed of light, h is Planck’s constant and Z is the charge on ions.Ne� N0 and Ni are number density of electron, atoms and ions, respectively. Te is theelectron temperature. The first term on the right side of Eqn. (8) is the electron atominteraction and the second term is related to electron ions interaction. Multi-photonionization in the vapor also can contribute to this process, if the laser intensity is highand laser wavelength is short.

When the plasma plume is near the critical density, the later part of the laser beampulse energy would be partially absorbed before it reaches the target. Plasma shieldingwas observed by the transmitted laser-pulse temporal profile through a glass sam-ple (Fig. 5). The temporal profiles of the transmitted laser pulse were similar to theoriginal laser pulse at low laser irradiance. When the laser irradiance was greater than0�3 GW/cm2, the later part of laser pulse became truncated. [36].

0.6 GW/cm2

0.3 GW/cm2

0.15 GW/cm2

0.1 GW/cm2

0.05 GW/cm2

5 GW/cm2

57 GW/cm2

Time (ns)

0 100

Tra

nsm

itted

lase

r in

tens

ity (

Arb

. uni

ts)

Fig. 5. Transmitted laser beam temporal profiles through a glass sample at varies power densities.Ref [36]




2.2. Plasma Expansion Processes

After the laser pulse ends, the induced plasma plume will continue to expand intothe ambient. The electron number density and temperature of the plasma changes asthe plasma expands. Plasma expansion depends on the amount and properties of theablated mass, how much energy was coupled into the mass, the spot size of laser beam,and the environment (gases, liquid, and pressure). Most LIBS spectra are recordedfrom several hundreds of nanoseconds to several microseconds after the laser pulse.Understanding plasma expansion during this time period is critical for optimization ofLIBS measurements.

2.2.1. Expansion of the evaporated material plume and shockwaves

After the laser pulse ends, hot electrons, atomic, and ionic mass leave the sample surface.The expansion of the evaporated material into vacuum can be described by the Eulerequations of hydrodynamics, expressing the conservation of mass density, momentumand energy [35]:

��

�t= −� ��

�x(9)

� ��

�t= − �

�x

[p+��2

](10)

�

�t

[

�

(

Ed + �2

2

)]

= − �

�x

[

��

(

Ed + p

�+ �2

2

)]

+�IBI (11)

where � is the mass density, � is the velocity, Ed is the internal energy density, and p isthe local pressure. This theory governing plasma expansion can be used for both ns andfs laser ablation.

In a vacuum, the laser induced plasma-plume expands adiabatically. The expansionspeed can be expressed by [37]

�p =√

4�+103

E

M�

(12)

where �p is the velocity, � is the specific heat ratio, E is the energy supporting theexpansion, and M� is the total vaporized sample mass within the vapor plume. Most ofthe plume energy is kinetic energy.

When ablation occurs into a gas or liquid environment, the ejected mass compressesthe surrounding media and produces shockwaves. The plume is the ablated mass fromthe sample target. The plasma is a mixture of atoms and ions, and mass from boththe ablated target material and the ambient gas. The interaction between the plume andsurrounding media slows the expansion of the plasma. At the same time, the ambientmedia performs work on the vapor. The vapor temperature will be higher than that forfree expansion; temperature and number density of ablated mass depend on the propertiesof the surrounding media.



Laser Ablation 59

Once the external shockwave is formed, its expansion distance can be described bySedov’s theory. The expansion distance H, representing the location of the shockwavefront, can be calculated as a function of time [38]:

H = �0

(E0

�1

)1/�2+d�

t2/�2+d� (13)

where the parameter d is the dimensionality of the propagation (for spherical propagationd = 3, for cylindrical propagation d = 2, and for planar propagation d = 1). �0 is adimensionless constant. E0 has the unit of “energy per area” in the case of one dimen-sional expansion (planar propagation), “energy per length” for two-dimensional expan-sion (cylindrical propagation), and “energy” for three-dimensional expansion (sphericalpropagation). �1 is air density. By fitting the experimental data using Eqn. (13), thedimensionality of expansion can be determined.

Early stage plasma expansion from femtosecond laser ablation of stainless steel tar-gets was investigated by time resolved shadowgraph imaging (Fig. 6). At early times,the femtosecond laser-induced plasma expanded primarily in the direction perpendicularto sample surface; the expansion distance was approximately 10�m after 130 ps. Therewas no lateral expansion until nanoseconds after the ablation laser pulse, after which thelateral expansion slowly increased. If the laser has a nano-second pulse duration, the per-pendicular expansion distance of the plasma is proportional to t2/5, and can be predictedby Sedov’s blast wave theory for spherical propagation (three dimension expansion).The perpendicular expansion of the plasma generated by the femtosecond laser pulsedablation was proportional to t2/3, corresponding to one dimensional expansion (measuredin the shadowgraphs of Fig. 6 which showed no lateral expansion at early times).

100 μmfs laser

500 μmns laser

43 ns1130 ps

470 ps 17 ns

3.2 ns130 ps

Time (ns)

Hfs ~ t

0.66 Hns ~ t

0.4

fs plasma

Per

pend

icul

ar e

xpan

sion

dist

ance

(m

icro

n)

ns plasma

0.110

100

500

(a)

(b)

1 10 100

Fig. 6. (a) Sequence of shock wave images obtained by laser shadowgraph for femtosecondand nanosecond laser ablation. Please noted the scale for fs laser and ns laser is different.(b) Perpendicular expansion distance of shockwave as a function of time for femtosecond andnanosecond laser ablation. Ref [39]




Once the plume pressure equalizes to the pressure of the surrounding media theexpansion stops. The stopping time and distance of the vapor plume can be expressedas [40]:

ts = �st

(E

pg

)1/3 1cg

(14)

Rs = �st

(E

pg

)1/3

(15)

where �st and st are constants, pg is pressure, and cg is the sound velocity of the gas. Thestopping time is in the range of microseconds. LIBS is usually measured after microsec-onds, a time at which vapor plume expansion has stopped. The final distance determinesthe volume of the vapor plume. LIBS performance depends on the electron numberdensity and temperature of the plasma, which strongly depends on the plume volume.

2.3. Plasma Emission Spectra

2.3.1. Femtosecond pulsed laser plasma emission

When a femtosecond laser pulse is focused in air, optical emission of nitrogen molecularlines will exist in the air plasma several picosecond after the laser pulse; molecularstructure is preserved for this time period [41]. A spectrum, measured using gatedintegration with a delay slightly after the laser pulse (1 ns), is presented in Fig. 7. The laserenergy was 200 mJ. The two electronic systems observed are the 2+ N2 �C3!u → B3!g�

280

400

600

800 1-0 0-0

0-11-1

2-2

1-23-2

2-02-1

3-1

0-21-3

2-43-5

2-3

0-0

1-40-3

3-7

0-11-2

2-3200

0300 320 340 360

Wavelength (nm)

2+ system N2 1– system N2+

Am

plitu

de

380 400 420 440

Fig. 7. Molecular line emission spectrum obtained at laser energy of 200 mJ with the maximumof the gate pulse ∼1 ns before the laser pulse. The upper–lower vibrational levels of the transitionsare indicated. Ref [41]



Laser Ablation 61

and the 1− N2+ �B2"+

u→ X2"+

g� bands. No oxygen lines were observed in this early

time period. The bands are identified in Fig. 7 with notation indicating the upper (v′)and lower (v′′) vibrational levels. The estimated vibrational temperature �Tv� of the 2+

system was between 7000 and 8000 K with accuracy ∼1000 K. A streak camera wasused to obtain a fast time-resolution measurement of the total emission, which is shownin Fig. 8. A narrow 60-picosecond duration peak was observed just after the laser pulsefollowed by a slower decrease in amplitude lasting 0.5 nanosecond. The peak is dueto molecular line emission, as there were no other spectral features measured at thisshort delay.

Conventional LIBS measurements are made using nanosecond to microsecond delaysafter the laser pulse. Emission spectra at these times depend on the laser-induced plasmaproperties; when the plasma is hot and dense, the spectrum is mostly composed of contin-uum emission. During plasma expansion, the temperature and number density decrease;ionic lines then atomic emission lines appear. Continuum emission was observed withinone nanosecond after the fs laser ablation pulse; Fig. 9 shows continuum emissionspectrum measured at increasing time delays [41]. Within the time measurement reso-lution (4.5 ns), the amplitude of continuum emission increased ∼10 times after the laserpulse, and decreased by about two orders of magnitude in 80 ns. Ionic lines began toappear about 10 ns after the laser pulse. The femtosecond pulsed laser induced plasmahas a shorter overall lifetime compared to those plasmas initiated using longer laserpulses.

2.3.2. Nanosecond pulsed laser plasma emission

For the ns pulsed laser induced plasma, continuum emission appears during the laserpulse and lasts for several hundred nanoseconds. Ion emission also dominates on the nstimescale. Atomic and molecular line emission occurs after ∼1 microsecond. Molecularline emission measured at later times is from the recombination of species in the plasma.

3000

2000

1000

0–0.1 0 0.1 0.2 0.3 0.4 0.5

4000

5000

Time (ns)

60 psec

Am

plitu

de

Fig. 8. Streak camera traces line-outs of the light emitted on a fast time scale. The laser pulse isat t = 0. Streak speeds are 15 mm/ns. Ref [41]




10

1

100

300 400 500 600

2 ns

3 ns4.5 ns5 ns6 ns

700

1000A

mpl

itude

Wavelength (nm)

10

1

100

300 400 500 600

90 ns

60 ns

35 ns

23 ns16 ns10 ns

700

Am

plitu

de

Wavelength (nm)

(b)(a)

Fig. 9. Continuum emission spectra at successive delays after the laser pulse. The spectra in both(a) and (b) are all plotted to the same scale. The time values indicate the delay of the samplingpulse with respect to the laser pulse. (a) 2–6 ns; and (b) 10–90 ns. Ref [41]

C-N and C-C swan bands are often observed on the microsecond time scale. To thebest of our knowledge, there are no reports for the original molecular structure beingpreserved for ns-pulsed plasma emission.

2.4. Electron Density and Plasma Temperature

Plasma temperature and electron number density can be estimated from the continuumemission and peak width of atomic and ionic emission lines. A Lorentz function can beused to fit the line spectra (Fig. 10). Stark line broadening from collisions of chargedspecies is the primary mechanism influencing the emission spectra in conventional LIBSexperiments. [42]

1.5 × 104

1.0 × 104

5.0 × 103

0.0286 290 292

Wavelength (nm)

FWHM

y0

Inte

nsity

(a.

u.)

288 x0

Fig. 10. Lorentzian fitting of the Stark broadened profile. The full width half maximum (FWHM)was used for the calculation of the electron number density.



Laser Ablation 63

The FWHM of Stark broadened lines is related [42,43] to the electron number densityNe by Eq. (10):

#�1/2 = 2W

(Ne

1016

[

1+1�75A(

Ne

1016

)1/4(

1− 34N

−1/3D

)])

(16)

where ND is the number of particles in the Debye sphere and is estimated from

ND = 1�72×109 T3/2e

N1/2e

(17)

W is the electron impact parameter in nm and A is the ion impact parameter; W and A arefunctions of temperature and can be obtained approximated by second-order polynomialsfrom reference [43].

Under the assumption of local thermal equilibrium (LTE), the plasma temperatureT can be determined by the line-to-continuum intensity ratio c/l, where c is thecontinuum emission coefficient and l is the integrated emission coefficient over theline spectral profile. The line emission coefficient l can be expressed in terms of theelectron temperature and density [44]:

l =(h�l

4�

)

A21

g2

2Zion�T�

h3

�2�mekB�3/2 NeNiT

−3/2e exp

(Eion −E2

kTe

)

(18)

where A21 is the Einstein transition probability of spontaneous emission, and Eion is theionization potential. E2 and g2 are upper level energy and degeneracy, respectively. �l

is the frequency of the emission line. Zion�T� is the partition function for ions, which isgiven by [45]:

Zion�T� =∑

i

gi exp(

− Ei

kTe

)

(19)

with gi the degeneracy or statistical weight of the i-th energy level Ei.At early times (10–100 ns), the plasma temperature is relatively high and the sec-

ond ionization state can be important at this time. By including the second ionizationcontribution, the expression for continuum radiation can be rewritten as [44,45]:

c =(

16�e6

3c3�6�m3ekB�

1/2

)

Ne�N+i +4N++

i �T−1/2e

[

�

(

1−exp−h�c

kBTe

)

+G

(

exp−h�c

kBTe

)]

�

(20)

where N+i and N++

i are the number density of single and double charged ions. G is thefree-free Gaunt factor, which is assumed to be unity by Kramer’s rule [29]. �c is thefrequency of the continuum emission. � is the free-bound continuum correction factor.




From Eqns. (18) and (20), the line to continuum ratio is:

l

c

= 1

1+4(

N++i

N+i

)CrA21

g2

Zion

�2c

�lTe

exp(

Eion−E2kBTe

)

[�(

1− exp(

−h�ckBTe

))+G

(exp −h�c

kBTe

)] (21)

�c� �l are the continuum and center wavelength of the spectral line, respectively. Cr isa constant. The ion ratio N++

i

N+i

can be calculated using the Saha equation [46]:

N++i

N+i

= �2�mek Te�3/2

h3

2Z++�T�Z+�T�

1Ne

exp(

− E++ion

kBTe

)

(22)

N+i

Na

= �2�mekBT�3/2

h3

2Z+�T�Z0�T�

1Ne

exp(

− E+ion

kBTe

)

(23)

where E+ion and E++

ion are the first and second ionization potentials, respectively.Z++�T�� Z+�T� and Z0�T� are the partition function for second ionized, first ionized,ions and atoms, respectively.

The evolution of the plasma temperature and electron number density were evaluatedfrom the Si(I) line emission versus time (Fig. 11). Since continuum emission dominatesinitially, the electron number density and temperature could not be obtained for delaytime less than 30 ns. After 30 ns, but before 300 ns, the line and continuum intensitieswere comparable; good measurement precision for the temperature calculation couldbe obtained. For later times �t > 300 ns� the continuum was very weak, and the line-to-continuum ratio would be sensitive to the errors of continuum determination. Thedata in Fig.11 show that in the early stage of plasma evolution (30–300 ns), temperatureand electron number density decreased rapidly with time. Within 300 ns, the plasmatemperature and electron number density decreased following a power-law dependence,

8 × 104

5 × 104

3 × 1019

1019

104

30 100

T (K)

ne (cm–3)ne = A2 t

–0.99

T = A2 t –0.77

300

6 × 104

4 × 104

2 × 104

0

3 × 1019

1 × 1019

4 × 1019

2 × 1019

0

0 500 1000

Delay time (ns)

Pla

sma

tem

pera

ture

(K

)E

lectron number density (cm

–3)

Fig. 11. Temporal evolution of plasma temperature (T) and electron number density �ne�. Theinset shows in log/log scale that in the early phase(30 ns–300 ns) of the plasma.



Laser Ablation 65

proportional to tn (insert of Fig. 11). The exponent n, obtained by experimental dataregression, was −0�70 and −0�99 for the temperature and electron number density,respectively. The analytical performance of LIBS is determined by the temperature andelectron number density of laser induced plasma.

When the continuum emission becomes weak, the plasma temperature calculated fromratio of line to continuum is not accurate. The plasma temperature can be estimated byusing a Boltzmann plot based on measuring the relative intensities of lines with knowntransition probabilities and degeneracies. Temperature can be determined from the slopeof a Boltzmann plot using ln �I�/gA� vs. the upper energy level of the transition, whereI is the line intensity, g is the statistical weight, and A is the transition probability.

3. PARTICLE FORMATION PROCESSES

A significant quantity of the ablated mass is not excited vapor, but in the form ofparticles. Particle formation occurs from condensed vapor, liquid sample ejection, andsolid-sample exfoliation. The mass ablated as particles does not contribute to a LIBSmeasurement unless these particles can be re-evaporated and excited by the plume itself.Particles are important for laser ablation ICP analysis. For LIBS, laser parameters mustbe established to eliminate particle formation.

3.1. Particle Ejection

In most cases, laser ablated mass consists of primarily particles. Kelly et al. reported thathomogeneous boiling within the molten layer was a significant mechanism responsiblefor particle removal during high-power nanosecond pulsed laser ablation [47,48]. Time-resolved shadowgraph images show that the violent ejection of particles occurs on themicrosecond time scale (Fig. 12). After the laser pulse, there is a time period in whichno particle ejection is observed; approximately 400 ns after the laser pulse, mass leavingthe silicon surface begins to appear. The ejection of these particles lasts for about 30�s.The largest particles were estimated to be on the order of 10�m in size [49].

Much of the theoretical foundation on explosive boiling was established by Martynyuk[50,51]. A rapid heating rate is required to induce explosive boiling. For a 3-ns laserpulse duration, the heating rate can exceed 1012 deg/s. Since thermal diffusion takesplace on the order of 10–11 s, a melt layer readily forms and propagates into the bulksample (silicon in this case) during the laser pulse. The liquid silicon is heated above itsboiling temperature and becomes metastable. Near the critical point, density fluctuationscan generate vapor bubbles in the superheated liquid silicon. Vapor bubbles greater thana critical radius, rc, will grow; bubbles of size less than rc will collapse [17]. Once vaporbubbles of size rc are generated in the superheated liquid, they undergo a rapid transitioninto a mixture of vapor and liquid droplets. Rapid expansion of the high-pressure bubblesin the liquid leads to a violent ejection of molten droplets from the target surface [52].This phase explosion process is detrimental to LIBS in that a significant portion ofablated mass is not utilized for analysis.




1100 ns

12100 ns

Fig. 12. Liquid ejection from Si single crystal sample.

3.2. Nanoparticle Formation

When the laser induced vapor plume cools to below the boiling point of the samplematerial, atoms begin to condense and form nano-particles. The size of particles isdetermined by the cooling time and the density of the vapor plume. Currently, numerousgroups study particle formation mechanisms because of their influence for ICP-MS andnanotechnology applications [53–60]. For LIBS, the particles represent a loss of signal.However, by understanding particle formation mechanisms, laser parameters can beestablished to minimize this loss and ultimately increase LIBS sensitivity.

4. LASER ABLATION PARAMETERS

As discussed above laser ablation involves complex non-linear phenomena that dependon the laser and sample material properties. The experimental isolation and effect of eachlaser and material parameter is very difficult without considering the interaction withother parameters. A general description of each of these parameters and their influenceon ablation behavior is presented in this section.



Laser Ablation 67

The primary mechanisms operative during ablation depend on the laser irradiance. Ingeneral, for low intensity nanosecond laser pulses, the dominant mechanism is thermalvaporization. For picosecond laser pulses and high intensity nanosecond laser pulses,both thermal and non-thermal mechanisms exist. For high intensity femtosecond laserpulses, Coulomb explosion is the primary ablation mechanism. Therefore, the laserpulse duration and irradiance are the most important factor for defining experimentalconditions. In addition to the pulse duration, some of the other parameters consideredduring the selection of the laser are wavelength, energy, beam profile, repetition rate,fluence, etc. The influence of environmental ambient (gas and pressure) and the sample’sproperties on laser ablation processes also will be discussed.

4.1. Nanosecond Pulsed Lasers

Nanosecond pulsed lasers are the most commonly used for analytical applications,especially LIBS. Analytical performance using nanosecond lasers for LIBS has beenstudied in many papers, including the influence of the wavelength, energy, repetitionrate, dual and multi-pulse regime, and the ambient gas [61–63].

4.1.1. Laser wavelength

The laser wavelength effect on LIBS could be addressed from two points of view; a) thelaser-material interaction (energy absorption) and b) plasma development and properties(plasma-material interaction).a) Laser-sample interaction: Shorter wavelengths offer higher photon energies for bondbreaking and ionization processes. For example, the UV wavelength 193 nm has pho-ton energy of 6.4 eV compared to 1064 nm that provides 1.16 eV. For most materials,bonding energy is a few eV. When the photon energy is higher than the bond energy,photon-ionization occurs and non-thermal mechanisms will play an important role in theablation process. In addition, shorter optical penetration depth exits with UV-wavelengths,providing more laser energy per unit volume for ablation. In general, the shorterthe laser wavelength, the higher the ablation rate and lower the elemental fractionation.

Ablation rate is a parameter used to describe the amount of ablated mass per laser pulseper unit area. Ablation rate also is an indirect indicator of the coupling efficiency betweenthe laser energy and the target material, and a measurement of the spatial resolution(lateral and depth resolution). An example of different ablation rates was presented byGunther et al. [64]. They studied the ablation rate in metals using a 266nm-Nd:YAG laserand 193nm-excimer laser (Table 1). The ablation rate depended on the laser wavelengthfor samples with low optical absorbance; Fig. 13 shows the rate behavior for NISTstandard reference materials 610–614 (series of glasses). Different ablation rates using266 nm wavelength (Fig. 13a) implies different laser-material coupling efficiency foreach of these samples. By using 193 nm wavelength, the three glass samples exhibitedthe same ablation rate (Fig. 13b); optical penetration depth was very shallow for thesethree samples at this wavelength.

During the interaction of nanosecond laser pulses with materials, there is enough timefor a thermal wave to propagate into the sample and create a relatively large molten layer.Evaporation occurs from the molten liquid, which can cause preferential evaporation




Table 1. Ablation rates for metals and CaF2 (IR grade) and NIST silicate glasses at 23 J/cm2

Metal orcompound

Ablation rate (mm/pulse)Nd:YAG

Ablation rate (mm/pulse)Excimer

Reflectance at260 nm

Al 1�04 0�7 92Si 0�39 71Ti 0�20Cr 0�12 0�13 45�7Ni 0�13 0�30 47�5Cu 0�13 0�29 35�5Zn 0�55 0�65 35Mo 0�14 61Pt 0�18 0�25 44�5Au 0�54 0�75 35�6ZnSe 0�2CaF2 2�0SRM NIST 610 0�34 0�28SRM NIST 612 0�60 0�29SRM NIST 614 0�67 0�29

Number of pulses applied

NIST 614 (a)NIST 612

NIST 610

0 1000

20

40

60

80

100

120

140

160

200 300 400 500

266 nm23 J/cm2(3.8 GW/cm2)

Helium

Dep

th (μm

)

Number of pulses applied

(b)NIST 614

NIST 612

NIST 610

00

5

10

15

20

25

30

50 100

193 nm23 J/cm2(1.9 GW/cm2)

Helium

150

Dep

th (μm

)

Fig. 13. (a) Depth vs. applied number of pulse for the widely used standard reference materialsNIST 610, 612 and 614 (266 nm Nd:YAG). Despite their similarity in their major element compo-sition largely different ablation rates have been found ranging from 0.34 to 0�67�m/pulse usingenergy of 23 J/cm2 and from 0.49 to 0�96�m/pulse at 35 J/cm2 (inlet −5�4GW/cm2). (b) Depthvs. number of laser pulses for the standard reference materials NIST 610, 612 and 614 using the193 nm Excimer LA system with an energy density of 23 J/cm2 �3�8 GW/cm2�. Ref [64]

or elemental fractionation. Elemental fractionation is due to many factors, includingwavelength, laser energy, pulse duration, sample properties, etc; wavelength is not themost critical parameter influencing fractionation [65]. Fractionation can be minimizedor enhanced in any sample, depending on the laser beam irradiance.b) Plasma development and properties: The initiation of the plasma and its propertiesalso depend on the laser wavelength. Plasma formation requires vaporization of thesample surface as a first step. The initiation of the nanosecond laser-induced plasma overthe target surface could be promoted by two different photon absorption processes. One



Laser Ablation 69

is inverse Bremsstrahlung absorption by which free electrons gain kinetic energy fromthe laser beam and during collisions among sample atoms and ions, electron and gasspecies. The second mechanism is photoionization of excited species and, at sufficientlyhigh laser intensity, multiphoton ionization of excited or ground state atoms.

Laserna et al. [66] studied the influence of wavelength (1064, 532 and 266 nm) onthe plasma formation threshold for metal samples. Although laser energy coupling ismore effective at shorter wavelengths, the fluence threshold for plasma formation wasgreater for 266 nm compared to 532 and 1064 nm. These studies agree with assumptionof plasma ignition by inverse Bremsstrahlung, which is approximately proportional to�3, (see Eqn. 8) considerably more favorable for IR than UV wavelengths.

Russo et al. [65] studied the influence of laser wavelength on fractionation in laserablation by comparing three different UV wavelengths (266 nm, 213 nm and 157 nm).It was found that the shorter the wavelength, the more controlled and reproducible wasthe ablation rate. Also, the shorter the wavelength, the lower was the fluence required toinitiate ablation. These data support the proposed mechanism of photoionization and/ormultiphoton ionization due to the higher photonic energy provide by UV wavelengths(7.9 eV, 5.8 eV, and 4.7 eV for 157 nm, 213 nm and 266 nm, respectively). For thesewavelengths, the Inverse Bremsstrahlung process was less important.

When the inverse Bremsstrahlung process occurs, part of the laser beam heats theplasma. Reheating the plasma can increase lifetime and the intensity of the emissionlines, which would be beneficial to LIBS. However, an increase in the broadbandbackground emission also can occur. The overall effect of the inverse Bremsstrahlungplasma reheating needs to be better investigated and understood.

At high fluence, the efficiency of inverse Bremsstrahlung can be such that the plasmaacts to shield the laser pulse energy (plasma shielding), from reaching the sample surface.Longer wavelengths favor the inverse Bremsstrahlung plasma shielding processes, butlower the ablation rate and increase the chances of elemental fractionation[37,67–69].In general, most LIBS studies are based on the 1064 nm (IR) Nd:YAG wavelength,whereas most laser ablation ICP-MS studies are based on the fourth or fifth harmonic(266 or 213 nm) of the Nd:YAG laser.

4.1.2. Laser energy

The primary energy related parameters influencing the laser material interaction arefluence (energy per unit area, J/cm2) and irradiance (energy per unit area and time,W/cm2). Laser ablation processes (i.e. melting, fusion, sublimation, erosion, explosion,etc) are dependent on the laser energy and the pulse duration, these different processeshave different fluence (or irradiance) threshold [70–74]. These processes define thecharacteristics of the laser-induced plasma (temperature and number electron density)and the characteristics of the ablated mass. The effect of the laser energy alone is difficultto quantify. In general, the ablated mass quantity and the ablation rate increase withincrease of the laser energy (when compared to same pulse duration and spot size).

Russo et al. [37,75–79], investigated the plasma shielding effect on the mass ablationrate. By using different lasers, increased laser irradiance lead to increased ablated mass,as shown in Fig. 14 (∼0�3 GW/cm2 for a copper target). However, plasma shieldingcaused saturation in mass removal and constant ablation rate with further increase inirradiance. Studies of mass ablation rate dependence on experimental parameters (laser




1012

1011

1010

109

108

107

106

105

0.01 0.1 1 10 100

Power density (GW/cm2)

Cu

ICP

inte

nsity

/are

a

Fig. 14. ICP emission intensity / area versus laser power density showing the two distinct massablation rate and roll-off at 0.3 GW/cm2. Ref [36]

irradiance, spot size, pulse width, surface condition, etc) at atmospheric pressure showedthat, in general, the mass ablation rate increased with increasing irradiance and decreasingpulse duration. Shorter pulses (on the order of picoseconds or femtosecond) producehigher mass ablation rates, likely because they are not affected by plasma shielding.Also, the fraction of the pulse energy lost by thermal diffusion (heat effected zone) inthe sample is lower for shorter pulses [39,79,80].

Theoretical studies have been conducted to model the processes underlying nanosec-ond laser pulsed ablation. Bogaerts and Chen [81] presented a model to describe nanosec-ond pulsed laser ablation with “typical” experimental conditions; wavelength of 266 nm,a Gaussian-shaped laser pulse, pulse duration of 10 ns. Copper (Cu) was used as thesample. The model supported that laser-material interactions during nanosecond laserpulsed ablation include heating, melting, and vaporization of the material. For a laserirradiance of 107 W/cm2, target heating was moderate, and neither melting nor vapor-ization took place. When the laser irradiance increased to 108 W/cm2 or 2×108 W/cm2,target heating and melting became more pronounced, but evaporation was still limited;the vapor plume was short and cool and no plasma was formed. At higher laser irra-diance �5 × 108 W/cm2�, target evaporation became much more significant, the plumebecame longer and hotter and a plasma was established. Plasma shielding was predictedat this irradiance. For laser irradiances between 5×108 and 1010 W/cm2, target heating,melting and vaporization increased, and the plume lifetime became longer. The vapordensity, temperature, and the degree of ionization increased with irradiance. Althoughthis model did not include all mechanisms underlying nanosecond pulsed laser ablation,it qualitatively supported measured behavior.

Dumitru et al. [82] developed a numerical model based on enthalpy to describenanosecond laser ablation. The authors reported that vaporization stopped before theend of the laser pulse. The effect was related to shielding by the plume, which iscontinuously growing and whose adsorption is increasing. Differences in ablated volumeswere reported when the calculation was performed with different pulse durations (1 to100 ns); the lower the pulse duration, the higher the ablated volume.



Laser Ablation 71

4.1.3. Ambient

The simplicity of LIBS derives from the fact that sample preparation is not requiredand in situ analysis at atmospheric pressure can be performed. However the interestof seeking new applications for LIBS, for example, designing an analytical instrumentthat can operate in Venus or Mars atmospheres [83,84], has led to studies of LIBSperformance under different ambient conditions.

The laser induced-plasma size, propagation speed, stability, energy and emission prop-erties depend strongly on the gas ambient into which the plasma expands. Plasma prop-erties have been studied in different gases [81,85–89], as a function of pressure [85,90],even in liquids [91–93]. The ambient gas either helps or prevents the coupling of laserenergy into the plasma. For example, the ambient can shield the sample from the laserbeam (plasma shielding) if gas breakdown occurs before sample vaporization occurs [85].The ambient also can quench the luminous plume by collisional cooling; a shorter plasmalifetime with lower temperatures was found in air atmosphere compared to argon [94].This observation was explained by lower conductivity and specific heat of argon withrespect to air. Wisbrun et al. [95] found that the collisional translation energy also wasdependent on the atomic mass of the ambient gas, being less effective when the atomicmass increased, thereby causing a plasma with longer lifetime.

These opposite effects may be less significant for gaseous or aerosols samples, but willbe more important for solid samples in the form of reduced ablation rates (plasma shield-ing), higher continuum background, and shorter-lifetime (fast dissipation). The ambientgas also can be helpful as to confine the expanding plume and minimize backgroundemission by atmospheric elements. In a particular case where the UV spectrum rangedfrom 100–200 nm, the use of inert gas avoided the absorption by oxygen molecules [96].

Ambient pressure will influence plasma expansion and LIBS emission intensities. Forlow ambient pressure (<1 mbar), the ablated vapor expands almost freely and the outerpart of the plasma becomes colder due to energy loss. Confining the plasma to pressureshigher than 1 mbar causes a reduction in energy loss and more uniform distributionof the energy within the plasma. Different gases show different behavior at differentpressures. For example, comparison of iron emission line intensities (Fe I, 374.979 nm)at 760 torr, showed to be greater in helium than in argon or air (Fig. 15). However, atreduced pressure of 100 torr, an intensity increase of 10 fold in argon and 2–3 fold in airwas observed. These observations were attributed to the fact that denser plasmas wereformed in Ar compared to He [85].

An increase in plasma size with decreased pressure is commonly measured [97].However, further reduction of pressure down to 10−3 torr resulted in a decrease in LIBSsignal intensity, as a result of a reduction of collisional excitation of the emission lines.Free expansion (short lifetime) of the plasma in vacuum is not recommended for LIBS,although many other laser ablation applications find the use of vacuum suitable [7,98,99].

Several studies of LIBS at high pressures have been pursued. Deguchi et al. [100]used LIBS to detect carbon in fly ash, char and pulverized coal under high-pressure andhigh-temperature conditions typical of gasification thermal power plants. High-pressureenvironments are also interesting for performing in situ geochemical analysis in hostileenvironments such as the deep sea, volcanoes, or even on others planets [84,101].

One of the drawbacks of high-pressure environments is broadening of spectral line emis-sion as well as self-absorption, therefore diminishing LIBS performance. This behavior wasreported by Nyga et al. [91] during the ablation of samples submerged in water (Fig. 16).




105

106

104

103

1 10 100 1000

Inte

nsity

, cou

nts

Pressure, torr

Fig. 15. Dependence of the emission intensities of Fe I 374.949 nm on the pressure of the ambientgases; Ar (open circle), air (filled circle) and He (triangle). The sample was a standard Al alloycontaining 0.97% Fe. Ref [85]

(c)Cal Cal Cal Cal Cal

3900

40

80

120

160

200

240

410 430 450 470 490

Wavelength (nm)

Inte

nsity

(a.

u.)

Wavelength (nm)

(a) (b)Cal Cal Cal Cal Cal CalCal Cal Cal Cal

390 410 430 450 470 490 390 410 430 450 470 4900

40

80

120

160

200

240

0

40

80

120

160

200

240

Wavelength (nm)

Inte

nsity

(a.

u.)

Inte

nsity

(a.

u.)

Fig. 16. Optical emission spectra from calcite (a) in air, (b) in water using the single-pulsetechnique, and (c) in water using the double-pulse technique. The arrows mark the wavelengthsof the identified calcium lines. The pulse energy was 10 mJ for all spectra shown. Ref [91]



Laser Ablation 73

However, minimization of line broadening and quenching was reported by using a dou-ble pulse approach. The first pulse creates a gas environment “bubble” above the samplesurface. The second pulse fired into the “bubble” allows LIBS analysis in a gas envi-ronment provided by the first pulse. Arp et al. [84] studied the performance of LIBSunder pressures as high as 90 atm in nitrogen and 56 atm in CO2 (simulating Venusplanet conditions). Although some elements showed strong emission line broadeningand self-absorption, other lines appeared unaffected by the high pressure.

There is no single optimal ambient gas environment for all laser ablation applications.Characterization of pathways for energy transfer between the plasma, the ambient gas,and the analyte needs to be discerned and then exploited to optimize sensitivity anddetection limits for a particular application.

5. PICOSECOND PULSED LASERS

For picosecond pulsed laser ablation, the pulse duration is similar to the lattice heatingtime ��l ∼ �p�. In this case, laser ablation is accompanied by electron heat conduction andformation of a melted zone in the target. Evaporation occurs as a direct solid-vapor (orsolid plasma) transition. However, the presence of a liquid phase (as in nanosecond case)can still cause preferential evaporation or elemental fractionation. Free electron emissionis strong for picosecond laser ablation [102]. A comparison by Angel et al. [103] ofLIBS measurements using nanosecond vs. picosecond laser pulses, showed that themass removal from the sample was more reproducible when using picosecond comparedto nanosecond laser pulses. Less heat, stress damage, and redeposition of mass wereobserved in SEM images of the craters generated by picosecond laser pulses comparedto those by nanosecond laser pulses (Fig. 17). The emission line intensity-to-backgroundratio was many times higher when using picosecond laser pulses (Fig. 18). The volumeof the plasma was about an order of magnitude smaller in the picosecond case, becausethe plasma was not reheated by the laser pulse. In the case of nanosecond excitation,plasma reheating tends to elongate the plasma in the direction of the laser beam.

In spite the differences between nanosecond and picosecond laser pulses; a mix ofmechanism is expected depending on the fluence. Rieger et al. [104] found similar emis-sion behavior from the laser-induced plasma generated by picosecond and nanosecondlaser pulses using only micro joules of laser energy. The plasma emission properties with

(a) (b)

200 μm200 μm

Fig. 17. SEM images of holes produced from 50 consecutive laser shots of the glass sample byusing picosecond (a) and nanosecond (b) excitation. Ref [103]




Wavelength (nm)515 525505

2000

0

4000

6000

8000(a)

Inte

nsity

515 525505100

300

500

700 × 103 (b)

Wavelength (nm)

Inte

nsity

Fig. 18. LIBS spectra of a Cu metal sample with the use of one laser pulse and nongateddetector for picosecond (a) and nanosecond (b) excitation. The exposure time was 0.1 s for eachmeasurement. Ref [103]

the pulse durations of 50 ps and 10 ns at energies above approximately 3�J were found tobe similar. This behavior was attributed to the fact that even though the initial plasma con-ditions during the time of irradiation of the laser pulses are quite different, and the plasmaformation threshold for picosecond laser pulse is lower, after 10 ns the energy absorbedby the plasma for both laser pulses will create a similar blast wave expanding into thebackground air. Most of the observed emission occurs in this expanding blast wave whichprimarily depends on the absorbed energy. However, as one approaches to the energythreshold for plasma formation for 10 ns pulses (∼1 mJ) the emission rapidly diminishes.For the case of 50 ps pulses significant plasma emission is still observed for 1 mJ pulsessince the much higher intensities are still significantly above the plasma breakdownthreshold and significant plasma heating occurs for these ultrashort pulses (Fig. 19).

Energy (μJ)

50 ps10 ps

PM

T s

igna

l (A

rbitr

ary

units

) 108

107

106

105

104

0.1 1 10

Fig. 19. Time integrated emission of Si I 288 nm as measured with a photomultiplier at an averageangle of 77� from the target normal as a function of the laser pulse energy for single shots onsilicon wafer. For signals below approximately 3 × 107 on the vertical axis the arbitrary unitscorrespond to total photons emitted per steradian. Above approximately 3 × 107 arbitrary unitsthe signals start to become slightly saturated and the response is no longer linearly related to theemission. Ref [104]



Laser Ablation 75

6. FEMTOSECOND PULSED LASERS

For femtosecond laser pulse durations, the pulse length is similar to the electron coolingtime ��e ∼ �p) in a solid sample. Ablation with femtosecond laser pulses is considered asa direct solid-vapor (or solid plasma) transition. The lattice is heated on a sub-picosecondtime scale by non-thermal melting, followed by the creation of vapor and plasma withrapid expansion rates.

During the pulse, thermal conduction into the sample can be neglected. Some of thebenefits of this fast interaction are more reproducible ablation process, and very preciselaser-machining of solids. Chichkov et al. [105] showed a qualitative comparison ofholes drilled in 100�m thick steel foils (in vacuum) with 104 laser pulses in the threeregimes: femtosecond, picosecond and nanosecond (Fig. 20).

Russo et al. [39] compared the crater depth produced by 266 nm- nanosecond andfemtosecond ablation in silicon (Fig. 21). For nanosecond and femtosecond laser ablation,

(b)

St - 10000 10192 30 μm

(a)

St - 10000 10199 30 μm

(c)

St - 10000 10191 30 μm

Fig. 20. SEM photographs of pulse laser ablation holes drilled (at 780 nm) in a 100�m thick steelfoil with (a) femtosecond-pulse with 200 fs, 120�$� F = 0�5 J/cm2, (b) picosecond-pulse with80 ps, 900�J� F = 3�7 J/cm2 and (c) picosecond-pulse with 3.3 ns, 1 mJ, F = 4�2 J/cm2. Ref [105]

Pulse number

00

10

20

30

40

50

10 20 30 40 50

Cra

ter

dept

h (m

icro

n)

dept

h(m

icro

n)

dept

h (m

icro

n)

(mm)

(mm)

ns crater

ns laser

fs crater

fs laser

0.00 0.03–4

–2

0

2

4

0.06

0.00–7.5

–5.0

–2.5

0.0

2.5

0.03 0.06

0.09

Fig. 21. Ablation depth vs. pulse number for femtosecond and nanosecond laser ablation. Theinserts show the fs and ns crater profile after five laser pulses. Ref [39]




the crater depth increased linearly with the number of pulses. However, for the samenumber of pulses, the fs-craters were about two times deeper than the ns-crater usingthe same fluence.

Within the framework of LIBS, significant differences in the temporal evolutionof emission lines also were reported for various laser pulse durations. Le Drogoffet al. [106,107], reported that the plasma temperature increased with the laser pulseduration, while the electron density remained relatively constant for the typical delaysused in LIBS experiments. For pulses longer that 5 ps, the laser-plasma interaction resultsin plasma heating. Therefore, as the laser pulse duration increases (from fs to ns), theplasma takes longer to decay. This effect also was shown by Russo et al. [39]. For earlytimes (<30 ns), (Fig. 22) the emission intensity of the fs-plasma decreased, while the ns-induced plasma became hotter due to absorption of the trailing part of the laser pulse atthe beginning of the plasma process. Similar results were reported by Sirven et al. [108]where the characteristics of the temporal dynamics of the emission was similar in bothcases (ns and fs), except at early moments after the excitation. Although the resultsobtained in this study with non-gated detection for both regimes were satisfactory (interms of signal-to-noise ratio) the continuum background in the nanosecond regime wasabout three times higher than in the femtosecond pulsed case (Fig. 23).

Eland et al. [80] also reported an increase in mass ablated with the use of a hybridfemtosecond laser, 140 fs pulse width, and energies between 0.26–0.94 mJ. They showedthat at high laser irradiances, higher plasma temperatures and emission intensities werereached. Although the intensity increase appeared to be mostly related to the amount ofablated material, the increased temperature allowed a higher degree of dissociation andexcitation thus improving the analytical sensitivity (Fig. 24).

As was mentioned above, the plasma decays faster when it is induced by femtosecondlaser pulses compared to nanosecond pulses. The main reason for this observation wasthat the plasma-induced by nanosecond laser pulses absorbs part of the laser energy andis reheated, causing elongation of its life time and intensifying line emissions. However

Time (ns)

0

0

10000

20000

30000

35000

100 200 300

ns laserfs laser

400 500

Pea

k in

tens

ity o

f spe

ctra

l lin

e (a

.u.)

Fig. 22. Peak intensity of Si(I) 288.16 nm emission line as a function of time for femtosecondand nanosecond laser ablation at a distance of 0.6 mm above the sample surface. Ref [39]



Laser Ablation 77

λ (nm)

fsns

× 1

03

300

0

5

10

15

20

25

400 500

Fig. 23. Time-integrated spectra of a sample of metallic aluminum performed at 67 J/cm2 in twoLIBS experiments after a 100-fs excitation (plain line) and 7-ns excitation (dotted line). Ref [108]

(b)

Relative volume1 2 3 4

Rel

ativ

e em

issi

on in

tens

ity

0.2

0.4

0.6

0.8

1.0

Rel

ativ

e vo

lum

e

Energy (mJ/pulse)

(a)

1

0.2 0.4 0.6 0.8

2

3

4

Relative em

ission intensity

0.2

0.6

1.0

Fig. 24. (a) The estimated relative crater volume (filled triangles) and relative emission intensityof the 404.6 nm Fe line (open squares), vs. laser energy with 140 fs excitation at 10 Hz. Therelative volume was estimated by assuming a radius-cubed dependence. (b) A plot of relativeemission intensity for the 404.6 nm Fe line vs. the relative volume of the ablation crater. Ref [80]

an increase of the continuum background is also observed which decreases the techniquesensitivity for nanosecond laser pulses.

Further improvement in sensitivity during femtosecond laser ablation-based analysismay be possible by choosing suitable ambient gases and pressures. Yalcin et al. [109]with the use of a femtosecond laser in air reported that the characteristic emissionlines from Al, Mg, Si and Cu elements exhibited significant enhancement in signalintensity at a few torr background air pressure as compared to atmospheric air pressure.These observations were attributed to a longer lifetime of the plasma expanding to alarger size at lower background pressures. It was also observed that signal enhancementat low pressure was dependent on the measurement delay time and on the transitionbeing observed. With a delay time of 200 ns, the integrated intensity of the neutral Al I




380

25000

20000

15000

Rel

ativ

e in

tens

ity (

a.u.

)

10000

5000

0400

Td = 0(a)

410390

Rel

ativ

e in

tens

ity (

a.u.

)

Wavelength (nm)

150000

100000

50000

0

(d) Td = 0

380 400 410390 410

Wavelength (nm)

150000

100000

50000

0

(e) Td = 85 ns

380 400390

1500

1000

500

0

(b) Td = 85 ns

380 400 410390

1500

1000

500

0

(c) Td = 200 ns

380 400 410390

Wavelength (nm)

100000

75000

50000

25000

0

(f)Td = 200 ns

380 400 410390

Fig. 25. Al I 394.6 and 396.15 nm lines at three different delay times (0.85 and 200 ns) and twodifferent pressures: (a–c) for atmospheric pressure (d–f) for 4 Torr pressure. Laser pulse energy10�J. Ref [109]

line at 396 nm exhibited 67 times enhancement in signal intensity at 4 torr of pressurewith respect to atmospheric pressure. A signal enhancement was only 4 times when nomeasurement delay time was used (Fig. 25). The best signal to noise ratio of 850 wasobserved at 4 torr pressure for an 85 ns delay time.

Another way at improving the performance of femtosecond laser-induced plasma forLIBS could be by the use of combination of pulses, for example, femtosecond pulses togenerate a more efficient ablation and nanosecond or picosecond pulse to heat or reheatthe plasma. Optimal conditions for plasma reheating with ultrafast pulses was proposedby Semerok et al. [110]. The optimal plasma re-heating regime with the highest plasmaintensity and reproducibility was determined to correspond to the double pulse delay of100–200 ps. The pulses in the 50 fs – 2 ps range were observed to give approximately thesame ablation efficiency with a metal target. However in these cases the double pulseapproach is not being used to improve the ablation process (laser-material interaction)but to improve the plasma emission characteristics. A more detailed perspective of theseapproaches will be given in subsequent chapters.

7. PERSPECTIVES, FUTURE AND TRENDS

Laser ablation is the underlying process in LIBS; the pulsed laser beam must convertthe sample into a luminous plasma with sufficient lifetime for statistical data analysis.Therefore, the laser parameters will have a dramatic influence on LIBS analytical perfor-mance. A better understanding of fundamental laser ablation processes can enhance LIBSperformance by a-priori selecting the appropriate parameters for efficiently coupling



Laser Ablation 79

the laser energy into the plume temperature and electron number density. Dynamics ofthe plume/plasma also are significantly influenced by the ratio of laser energy coupledinto the sample versus the plume itself. Competing mechanisms for particle generation,heat loss and reflectivity will always exist and must be balanced to achieve efficientand accurate ablation as well as robust plasma conditions. With the varied availablelaser parameter space (energy, wavelength, pulse duration, fluence, etc.), and samplechemical systems, an empirical approach to LIBS optimization for each applicationwould be onerous. A balanced approach consisting of iterative theory, modeling, andexperimentation will be necessary to drive LIBS into new regimes.

ACKNOWLEDGMENT

The U.S. Department of Energy, Office of Basic Energy Sciences, Chemical Sciences,and the Office of Nonproliferation and National Security (NA22) supported this researchat the Lawrence Berkeley National Laboratory under Contract DE-AC02-05CH11231.

REFERENCES

[1] J.C.Miller. Laser ablation: Principles and Applications. 1 [28] (1994).[2] A.Einstein, in Henry A. Boorse and Lloyd Motz (eds.) The World of the Atom, Volume II,

Basic Books, 1966, pp. 884–901. Physika Zeitschrift 18 (1917) 121.[3] A.L.Schawlow and C.H.Townes, Phys. Rev. 112 (1958) 1940[4] T.H.Maiman, R.H.Hoskins, J.D.D’Harnens, C.K.Asawa, and V.Evtuhov, Phys. Rev. 123

(1961) 1151.[5] T.H.Maiman, Phys. Rev. 123 (1961) 1145.[6] F.Breech and L.Cross, Applied Spectroscopy 16 (1962) 59.[7] M.N.R.Ashfold, F.Claeyssens, G.M.Fuge, and S.J.Henley, Chemical Society Reviews 33

(2004) 23.[8] T.Otani, L.Herbst, M.Heglin, S.V.Govorkov, and A.O.Weissner, Applied Physics A79

(2004) 1335.[9] S.Taneri, J.D.Zieske, and D.T.Azar, Survey of Ophthalmology 49 (2004) 576.

[10] A.M.Troilius, Laser in surgery and medicine 22 (1998) 103.[11] B.Drnovsek-Olup, M.Beltram and J.Pizem, Laser in surgery and medicine 35 (2004) 146.[12] O.Topaz, E.A.Rozenbaum, A.Schumacher and M.G.Luxenberg, Laser in surgery and

medicine 19 (1996) 260.[13] A.Moritz, U.Schoop, K.Goharkhay, P.Schauer, O.Doertbudak, J.Wernisch and W.Sperr,

Laser in surgery and medicine 22 (1998) 302.[14] R.Kaufmann, Journal of Biothecnology 41 (1995) 155.[15] L.Moenke-Blankenburg, T.Schumann and J.Nolte, J. Anal. Atomic Spectrometry 9

(1994) 1059.[16] R.E.Russo, X.L.Mao, H.C.Liu, J.Gonzalez, and S.S.Mao, Talanta 57 (2002) 425.[17] S.S.Mao, F.Quere, S.Guizard, X.Mao, R.E.Russo, G.Petite and P.Martin, Applied Physics

A79 (2004) 1695.[18] L.V.Keldysh, Soviet Physics JETP (1965) 1307.[19] S.K.Sundaram and E.Mazur. Nature materials 1 (2002) 217.[20] P. Y. Yu and M. Cardona, Fundamentals of Semiconductor : Physics and materials prop-

erties (Springer, 3rd Edition, New York, 2001), p. 306.




[21] N.W.Ashcroft and N.D.Mermin. Solid state physics. (1976).[22] N.Bloembergen, IEEE J Quantum Electronics QE10 (1974) 375.[23] E.Yablonovitch and N.Bloembergen, Phys. Rev. Letts 29 (1972) 907[24] B.C.Stuart, M.D.Feit, S.Herman, A.M.Rubenchik, B.W.Shore, and M.D.Perry. Phys. Rev.

B53 (1996) 1749.[25] P.Saeta, J.K.Wang, Y.Siegal, N.Bloembergen, and E.Mazur, Phys. Rev. Letts 67

(1991) 1023.[26] J.S.Preston, H.M.Vandriel, and J.E.Sipe, Phys. Rev. Letts. 58 (1987) 69.[27] D.H.Lowndes and G.E.Jr.Jellison. Semiconductors and Semimetals. 23 (1984) 313.[28] Y.Siegal, E.N.Glezer, and E.Mazur, Phys. Rev. B49 (1994) 16403.[29] L. Spitzer, Physics of fully ionized gases (John Wiley & Sons, 2nd Edition, New York,

(1962), p. 120.[30] K.Sokolowski-Tinten and D.von der Linde, J. Phys.-Condensed Matter 16 (2004) R1517.[31] R.E.Russo, X.L.Mao, H.C.Liu, J.H.Yoo, and S.S.Mao, Appl. Phys.A69 (1999) S887.[32] S.S.Mao, X.L.Mao, R.Greif, and R.E.Russo, Appl. Phys. Letts. 76 (2000) 31.[33] S.S.Mao, X.Mao, R.Greif, and R.E.Russo, Appl. Phys. Letts. 77 (2000) 2464.[34] S.S.Mao, X.L.Mao, R.Greif, and R.E.Russo, Appl. Phys. Letts. 76 (2000) 3370.[35] A.Bogaerts, Z.Y.Chen, R.Gijbels, and A.Vertes, Spectrochim. Acta B58 (2003) 1867.[36] X.L.Mao and R.E.Russo, Applied Physics A64 (1997) 1.[37] Ya. B. Zel’dovich and Yu. P. Raizer, Physics of shock waves and high-temperature hydro-

dynamics phenomena (Dover Publications, New York, 2002), p. 97.[38] L. I. Sedov, Similarily and dimesional methods in mechanics (CRC Press, 10th Edition,

Boca Raton, London, 1993), p. 264.[39] Zeng X., Mao X.L, Greif R., and Russo R.E., Applied Physics A 80 (2004) 237.[40] N.Arnold, J.Gruber, and J.Heitz, Applied Physics A69 (1999) S87.[41] F.Martin, R.Mawassi, F.Vidal, I.Gallimberti, D.Comtois, H.Pepin, J.C.Kieffer and

H.P.Mercure, Applied Spectroscopy 56 (2002) 1444.[42] H.C.Liu, X.L.Mao, J.H.Yoo, and R.E.Russo, Spectrochim. Acta B54 (1999) 1607.[43] X.Zeng, X.L.Mao, S.Mao, J.H.Yoo, R.Greif, and R.E.Russo, J. Appl. Phys. 95 (2004) 816.[44] G.J.Bastiaans and R.A.Mangold, Spectrochim. Acta B40 (1985) 885[45] H. R. Griem, Principles of plasma spectroscopy Cambridge University Press, Cambridge

(1997).[46] G. Befeki, Principles of laser plasmas Wiley Interscience, New York (1976).[47] R.Kelly and A.Miotello,. Appl. Surface Sci. 96 (1996) 205.[48] A.Miotello and R.Kelly, Appl. Phys. Letts. 67 (1995) 3535.[49] J.H.Yoo, S.H.Jeong, R.Greif, and R.E.Russo, J. Appl. Phys. 88 (2000) 1638.[50] M.M.Martynyuk. Sov.Phys.Tech.Phys. 21 (1976) 430.[51] M.M.Martynyuk. Sov.Phys.Tech.Phys. 19 (1974) 793.[52] V. P. Carey, Liquid-vapor phase-change phenomena Hemisphere, Washington (1992),

p. 127.[53] B.S.Luk’yanchuk, W.Marine, and S.I.Anisimov, Laser Physics 8 (1998) 291.[54] J.Koch, H.Lindner, A.von Bohlen, R.Hergenroder, and K.Niemax, J. Anal. Atomic Spec-

trometry 20 (2005) 901.[55] J.Kosler, M.Wiedenbeck, R.Wirth, J.Hovorka, P.Sylvester, and J.Mikova, J. Anal. Atomic

Spectrometry 20 (2005) 402.[56] C.Y.Liu, X.L.Mao, J.Gonzalez, and R.E.Russo, J. Anal. Atomic Spectrometry 20 (2005) 200.[57] H.R.Kuhn and D.Gunther, J. Anal. Atomic Spectrometry 19 (2004) 1158.[58] D.B.Geohegan, A.A.Puretzky, G.Duscher, and S.J.Pennycook, Appl. Phys. Letts. 72

(1998) 2987.[59] M.Kuwata, B.Luk’Yanchuk, and T.Yabe, Japanese J. Appl. Phys. (Part 1) 40 (2001) 4262.[60] A.Pereira, A.Cros, P.Delaporte, S.Georgiou, A.Manousaki, W.Marine, and M.Sentis. Appl.

Phys. A79 (2004) 1433.



Laser Ablation 81

[61] L.St-Onge, M.Sabsabi, and P.Cielo, Spectrochim. Acta B53 (1998) 407.[62] F.Colao, V.Lazic, R.Fantoni, and S.Pershin, Spectrochim. Acta B57 (2002) 1167.[63] G.Adbellatif and H.Imam, Spectrochim. Acta B57 (2002) 1155.[64] I.Horn, M.Guillong, and D.Gunther, Appl. Surface Sci. 182 (2001) 91.[65] R.E.Russo, X.L.Mao, O.V.Borisov, and H.C.Liu, J. Anal. Atomic Spectrometry 15

(2000) 1115.[66] L.M.Cabalin and J.Laserna, Spectrochim. Acta B53 (1998) 723.[67] X.L.Mao, W.T.Chan, M.A.Shannon, and R.E.Russo, J. Appl. Phys. 74 (1993) 4915.[68] X.L.Mao, W.T.Chan, M.Caetano, M.Shannon, and R.E.Russo, Appl. Surface Sci. 96–98

(1996) 126.[69] J.A.Aguilera, C.Aragon, and F.Penalba, Appl. Surface Sci. 127–129 (1998) 309.[70] X.L.Mao, A.C.Ciocan, O.V.Borisov, and R.E.Russo, Appl. Surf. Sci. 129 (1998) 262.[71] H.C.Liu, X.L.Mao, J.H.Yoo, and R.E.Russo, Appl. Phys. Letts. 75 (1999) 1216.[72] D.von der Linde and K.Sokolowski-Tinten, Appl. Surf. Sci. 154 (2000) 1.[73] W.Marine, N.M.Bulgakova, L.Patrone, and I.Ozerov, Appl. Phys. A79 (2004) 771.[74] J.Reif, F.Costache, S.Eckert, and M.Henryk,. Appl. Phys. A79 (2004) 1229.[75] W.T.Chan and R.E.Russo, Spectrochim. Acta B46 (1991) 1471.[76] A.Fernandez, X.L.Mao, W.T.Chan, M.A.Shannon, and R.E.Russo, Analytical Chemistry 67

(1995) 2444.[77] R.E.Russo, X.L.Mao, W.T.Chan, M.F.Bryant, and W.F.Kinard,. J. Anal. Atomic Spectrom-

etry 10 (1995) 295.[78] X.L.Mao, W.T.Chan, and R.E.Russo, Appl.Spectrosc. 51 (1997) 1047.[79] O.V.Borisov, X.L.Mao, A.C.Ciocan, and R.E.Russo, Appl. Surf. Sc. 129(1998) 315.[80] K.L.Eland, D.N.Stratis, D.M.Gold, S.R.Goode, and S.M.Angel, Appl. Spectrosc. 55

(2001) 286.[81] A.Bogaerts and Z.Chen, J. Anal. Atomic Spectrometry 19 (2004) 1169.[82] G.Dumitru, V.Romano, and H.P.Weber. Appl. Phys. A79 (2004) 1225.[83] F.Colao, R.Fantoni, V.Lazic, and A.Paolini, Appl. Phys. A79 (2004) 143.[84] Z.A. Arp, D.A.Cremers, R.D.Harris, D.M.Oschwald, G.R.Parker, Jr, and D.M.Wayne, Spec-

trochim. Acta B59 (2004) 987.[85] Y.Iida, Spectrochim. Acta B45 (1990) 1353.[86] S.S.Harilal, C.V.Bindhu, V.P.N.Nampoori, and C.P.G.Vallabhan, Appl. Phys. Letts. 72

(1998) 167.[87] V.Detalle, M.Sabsabi, L.St-Onge, A.Hamel, and R.Heon. Appl. Optics 42 (2003) 5971.[88] S.Amoruso, B.Toftmann, and J.Schou. Applied Physics A79 (2004) 1311.[89] R.E.Russo, X.L Mao, M.Caetano, and M.A.Shannon, Appl. Surf. Sci. 96–98(1996)144.[90] G.Cristoforetti, S.Legnaioli, V.Palleschi, A.Salvetti, and E.Tognoni,. Spectrochim. Acta

B59(2004) 1907.[91] R.Nyga and W.Neu. Optics Lett. 18 (1993) 747.[92] A.E.Pichahchy, D.A.Cremers, and M.J.Ferris. Spectrochim. Acta B52 (1997) 25.[93] D.C.S.Beddows, O.Samek, M.Lizka, and H.H.Telle, Spectrochim.Acta B57(2002) 1461.[94] D.E.Kim, K.J.Yoo, H.K.Park, K.J.Oh, and D.W.Kim. Appl. Spectrosc. 51 (1997) 22.[95] R.Wisbrun, I.Schechter, R.Niessner, H.Schroder, and K.L.Kompa, Anal. Chem. 66

(1994) 2964.[96] L.J.Radziemski, D.A.Cremers, K.Benelli, C.Khoo, and R.D.Harris, Spectrochim. Acta B60

(2005) 237.[97] Y.-I.Lee, T.L.Thiem, G.-H.Kim, Y.-Y.Teng, and J.Sneddon, Appl. Spectrosc. 46

(1992) 1597.[98] R.J.Lade, F.Claeyssens, K.N.Rosser, and M.N.R.Ashfold, Appl. Phys. A69 (1999) 935[99] M.Tanaka, Y.Fujisawa, T.Nakajima, Y.Tasaka, K.Ota and S.Usami, J. Appl. Physi.83

(1998)3379.




[100] M.Noda, Y.Deguchi, S.Iwasaki, and N.Yoshikawa, Spectrochim. Acta B57 (2002) 701.[101] B.Salle, J.L.Lacour, E.Vors, P.Fichet, S.Maurice, D.A.Cremers and R.C.Wiens, Spec-

trochim. Acta. B59 (2004) 1413.[102] S.S.Mao, X.L.Mao, J.H.Yoo, R.Greif, and R.E.Russo, J. Appl. Phys. 83 (1998) 4462.[103] K.L.Eland, D.N.Stratis, T.S.Lai, M.A.Berg, S.R.Goode, and S.M.Angel, Appl. Spectrosc.

55 (2001) 279.[104] G.W.Rieger, A.Taschuk, Y.Y.Tsui and R.Fedosejevs. Spectrochim. Acta B58 (2003) 497.[105] B.N.Chichkov, C.Momma, S.Nolte, F.von Alvensleben, and A.Tunnermann, Appl. Phys.

A63 (1996) 109.[106] B.Le Drogoff, J.Margot, M.Chaker, M.Sabsabi, O.Barthelemy, T.W.Johnston, S.Laville,

F.Vidal, and Y.von Kaenel, Spectrochim. Acta B56 (2001) 98.[107] B.Le Drogoff, J.Margot, F.Vidal, S.Laville, M.Chaker, M.Sabsabi, T.W.Johnston and

O.Barthelemy. Plasma Sources Sci.Technol. 13 (2004) 223.[108] J.B.Sirven, B.Bousquet, L.Canioni, and L.Sarger, Spectrochim. Acta B59 (2004) 1033.[109] S.Yalcin, Y.Y.Tsui, and R.Fedosejevs, J. Anal. Atomic Spectrometry 19 (2004) 1295.[110] A.Semerok and C.Dutouquet. Thin solids films 453–454 (2004) 501.



Chapter 4

Physics of Plasma in Laser-Induced BreakdownSpectroscopy

V. N. Raia and S. N. Thakurb

aLaser Plasma Division, Raja Ramanna Centre for Advanced TechnologyP. O. CAT, Indore-452 013, INDIA

bLaser and Spectroscopy Laboratory, Department of PhysicsBanaras Hindu University, Varanasi-221 005, INDIA

1. INTRODUCTION

The interaction of high-power laser light with a target material has been an activetopic of research not only in plasma physics but also in the field of material science,chemical physics and particularly in analytical chemistry [1]. The high intensity laserbeam impinging on a target (solid, liquid or gas) may dissociate, excite, and/or ionizethe costituent atomic species of the solid and produces plasma, which expands eitherin the vacuum or in the ambient gas depending on the experimental conditions. Asa result of laser-matter interaction, various processes may occur such as ablation ofmaterial, target acceleration, high energy particle emission, generation of various para-metric instabilities as well as emission of radiation ranging from the visible to hardX-rays depending on the intensity of laser. These processes have many applications butwe are mainly interested here in the study of optical emission from the plasma. Forother applications of laser-produced plasma the readers can find detailed informationin a series of review articles and books [1–6]. Spectroscopic study of optical emissionfrom laser-produced plasma is known as laser-induced breakdown spectroscopy (LIBS),which was started in 1968 soon after the invention of the laser. In brief, LIBS is anatomic emission technique suitable for quick and on-line elemental analysis of any phaseof material (solid, liquid, gas and aerosols). It has several advantages over conventionallaboratory based analytical techniques such as a high spatial resolution (due to smallfocal spot), absence of sample preparation, studies of hostile environments like melt-ing or burning samples and remote detection. Several review articles are available inliterature on the LIBS [1,7–16]. The physics of plasma relevant to laser-induced break-down spectroscopy has been discussed in the following sections, which are such that abeginner can understand the basic physics of LIBS and can get up-to-date references atone place.




84 V. N. Rai and S. N. Thakur

2. BASICS OF LASER-MATTER INTERACTION

When a high-power laser pulse is focused onto a material target (solid, liquid, gasand aerosols), the intensity in the focal spot produces rapid local heating and intenseevaporation followed by plasma formation. The interaction between a laser beam anda solid is a complicated process dependent on many characteristics of both the laserand the solid material. Various factors affect ablation of material, which includes thelaser pulse width, its spatial and temporal fluctuations as well as its power fluctuations.The mechanical, physical and chemical properties of the target material also play animportant role in laser-induced ablation. The phenomena of laser-target interactionshave been reviewed by several authors [17–18], while the description of melting andevaporation at metal surfaces has been reported by Ready [18]. It has been found thatthe thermal conductivity is an important parameter for free vaporization of the solid intoa vacuum. The plasma expands normal to the target surface at a supersonic speed invacuum or in the ambient gas. The hot expanding plasma interacts with the surroundinggas mainly by two mechanisms: (i) the expansion of high pressure plasma compressesthe surrounding gas and drives a shock wave, (ii) during this expansion, energy istransferred to ambient gas by the combination of thermal conduction, radiative transferand heating by shock wave. The evolution of plasma depends on the intensity of laser,its wavelength, size of focal spot, target vapor composition, ambient gas compositionand pressure. It has been found that the plasma parameters such as radiative transfer,surface pressure, plasma velocity, and plasma temperature are strongly influenced bythe nature of the plasma. Since vaporization and ionization take place during the initialfraction of laser pulse duration, rest of the laser pulse energy is absorbed in the vapor andexpanding plasma plume. This laser absorption in the expanding vapor/plasma generatesthree different types of waves as a result of different mechanisms of propagation ofabsorbing front into the cool transparent gas atmosphere. These waves are (i) laser-supported combustion (LSC) waves, (ii) laser-supported detonation (LSD) waves, and(iii) laser-supported radiation (LSR) waves [18]. Each wave is distinguished on the basisof its velocity, pressure, and on the effect of its radial expansion during the subsequentplasma evolution, which is strongly dependent on the intensity of irradiation.

At low irradiation, laser-supported combustion waves are produced, which compriseof a precursor shock, that is separated from the absorption zone and the plasma. The shockwave results in an increase in the ambient gas density, pressure and temperature, whereasthe shock edges remain transparent to the laser radiation. The front edge of the expandingplasma and the laser absorption zone propagate into the shocked gas and give rise tolaser supported combustion wave (Fig. 1). In the early days thermal conduction wasassumed to be the primary propagation mechanism. However many investigators [19–21]studied the one-dimensional propagation using a variety of transport models and foundthat the major mechanism causing LSC wave propagation is radiative transfer from thehot plasma to the cool high pressure gas in the shock wave. The plasma radiation liesprimarily in the extreme ultraviolet and is generated by photo-recombination of electronsand ions into the ground-state atom.

At intermediate irradiance, the precursor shock is sufficiently strong and the shockedgas is hot enough to begin absorbing the laser radiation without requiring additionalheating by energy transport from the plasma. The laser absorption zone follows directlybehind the shock wave and moves at the same velocity. This is similar to the chemical



Physics of Plasma in LIBS 85

TargetHot plasma

Cold plasma

Bremsstrahlungcontinuum emission

Line emission

Laser

Shock wave inambient gas

Fig. 1. Schematic diagram of expanding laser produced plasma in ambient gas. Plasma plume isdivided into many zones having high-density hot and low-density cold plasma. The farthest zone fromthe target has minimum plasma density and temperature. Laser is absorbed in low-density corona.

detonation wave and has been modeled by Ramsden and Savic [22] and Raiser [23].The propagation of the laser-supported detonation wave is entirely controlled by theabsorption of the laser energy. Several workers have theoretically and experimentallystudied the ignition and propagation of LSD wave away from the metal surfaces [24–27].

At high irradiance, the plasma is so hot that, prior to the arrival of the shock wave,the ambient gas is heated to temperatures at which laser absorption begins. In the idealcondition, laser absorption is initiated without any density change, and the pressureprofile results mainly from the strong local heating of the gas rather than a propagatingshock wave. This configuration is nothing but an overdriven absorption wave [23].These supersonic waves have been modeled numerically by Bergel’son et al. [24] andit has been found that once the transient plasma initiation and formation processes arecompleted the quasi-steady approximation is suitable. The laser supported radiative wavevelocity increases much more rapidly with irradiance than those of the LSC and LSDwaves. The temperature and pressure increase, are quite low, which indicates that theLSR wave is effective in channeling the absorbed energy into heating a large amount ofgas rather than increasing the local enthalpy.

3. PROCESSES IN LASER PRODUCED PLASMA

As discussed in the previous section, the interaction of a high intensity laser light withsolid target initially increases the surface temperature of the sample such that materialtransfer across the surface becomes significant (Vaporization). As a result of materialvaporization and plasma formation, target erosion appears in the form of craters onthe sample surface. The theoretical considerations on plasma production and heating bymeans of laser beams have been proposed by several workers [25–28]. The initiationof plasma formation over a target surface begins in the hot target vapor. First of allabsorption of laser radiation takes place via electron-neutral inverse Bremsstrahlung,but when sufficient electrons are generated, the dominant laser absorption mechanismmakes a transition to electron-ion inverse Bremsstrahlung. Photo-ionization of excited




states can also contribute in the case of interactions with short wavelength radiations.The same absorption processes are responsible for the absorption by the ambient gasalso. The laser-produced plasma expands into the vacuum or into the surrounding gasatmosphere, where the free electrons present in the plasma [1–6] modify propagationof laser light. The plasma formed by a high intensity or small time duration laser hasa very steep density and temperature gradient in comparison to the plasma formedby the low intensity or long time duration laser. The density gradient in the plasmaplays a very important role in the mechanism of light absorption and in the partition ofabsorbed energy between thermal and non-thermal particle distribution. There are threebasic mechanisms through, which intense laser light may interact with plasma [5]. Thefirst mechanism is an inverse Bremsstrahlung, where electric field of the incident lightrattles electrons, which then lose this energy in collision with ions. This mechanismis important with shallow density gradients in the plasma. The parametric processesalso take part most efficiently, when the density gradient is shallow. There are three-wave parametric interaction processes in which intense laser field drives one or morelongitudinal plasma waves out of the noise and also parametric decay processes wherelaser light decays into a high frequency electron acoustic wave and a low frequency ionacoustic wave conserving energy and momentum. Another important short pulse laserabsorption mechanism is the resonance absorption. With a p-polarized light obliquelyincident on plasma surface, the radial component of electric field resonates with plasmafrequency and causes large transfer of energy to electrons near critical density �Nc�surface. Critical density for a given laser wavelength is

Nc = 1021/�2cm−3 (1)

where � is in micron. Energy absorbed at or below the critical density in plasma isthen conducted towards the target surface by various transport processes. The study ofenergy coupling to the target has many sub areas such as laser light absorption, nonlinearinteraction, electron energy transport and ablation of material from the target surface.One of the important processes, in laser-plasma interaction, is emission of radiationfrom the plasma ranging from visible to hard X-rays [1] and it is very relevant for theunderstanding of laser-induced breakdown spectroscopy. It has been found that X-raysare emitted from all parts of the absorption, interaction and transport regime. At densitiesnear and slightly above the critical, nearly 70% of the incident laser energy may bere-emitted as X-rays with energy ranging from 50 eV to 1 keV or above depending onthe temperature of the plasma. However, as the plasma expands away from the targetsurface, its density as well as the temperature decreases. As the plasma temperaturedecreases the wavelength of emission from the plasma increases, that is, emission shiftsfrom X-rays to visible region.

4. SPECTRAL EMISSION FROM PLASMA

The spectral composition of emission from plasma has line as well as continuum com-ponents. Study of characteristic line emission from the plasma can give informationabout the composition of target material, that is, the elements present in the target.Laser-Induced Breakdown Spectroscopy (LIBS) has proved to be a versatile techniqueof elemental analysis [1,8–16].




4.1. Continuum Emission

The continuum radiation is emitted by the plasmas as a result of free-free and free-boundtransitions. Free-free transitions (Bremsstrahlung emission) are due to the interactionof electrons with positively charged ions of charge Z and density NZ. Neglecting thecorrection factor (the Gaunt factor), the free-free spectral power density, for a Maxwellianelectron distribution at a temperature Te is given by [29–30]

Iv ≈ Zeff

NeNZ

T1/2e

exp(

−hv

Te

)

(2)

where Zeff is the effective charge for Bremsstrahlung emission, which is generallydifferent from Z due to the partial screening of the nuclear charge ZN by the �ZN −Z�electrons [31]. Equation (2) shows that the logarithmic slope of the spectrum can provideinformation about the electron temperature Te of plasma. The inverse process is calledinverse Bremsstrahlung and is responsible for the absorption of laser radiation at or belowthe critical density region. When the electron distribution function is not Maxwellian,it is in principle possible to deconvolute it from the high photon energy non-thermalspectrum (since free-free radiation is very sensitive to high-energy tails). In the case oflaser-produced plasmas the high-energy electrons (usually called suprathermal electrons)have long mean free paths and can therefore penetrate into the cold solid target. In thiscase Bremsstrahlung spectrum is modified, because the electrons (of all energies) areslowed down by their interaction with the cold solid matter [32].

In free-bound (recombination radiation) transition, a free electron (kinetic energy �e)is captured by an ion of charge Z in a bound level n of the ion of charge �Z −1� (ioni-sation energy �n) and as a result a photon of energy hv = �e +�n is emitted. The photonenergy is a function of �e, but only photons with hv > �n are emitted. The spectra there-fore show discontinuities, generally called recombination edges. Above the edges, thespectral dependence is the same as for free-free transition, and Te can therefore again bededuced from the logarithmic slope. Non-thermal electrons do not contribute to the free-bound spectrum, because the recombination cross-section decreases rapidly for �e >> �n.De Michelis and Mattioli [31] and references there-in have discussed this subject in detail.

4.2. Line Emission

The emissivity of a spectral line (i.e. number of photons emitted per unit time and per unitvolume) is equal to the product of its radiative transition probability times the emittingexcited level density. On the other hand, the absorption coefficient for line radiation isobtained from the radiative transition probability by using the Einstein relations. For theinterpretation of the line emission, the excited level population should be known.

4.3. Temporal and Spatial Resolution of Emission

The emission from laser produced plasma in general and optical emission in particular hasbeen studied extensively using temporally and spatially resolved spectroscopy [1,9–16].It has been observed that initially after plasma formation an intense continuum is emitted,




which remains close to the target surface. This emission from the dense plasma is likea black body continuum radiation. However, as the plasma expands away from thesurface of the target, it cools and the emission is dominated by the spectral lines. Duringthis expansion the spectral lines from highly ionized atoms are observed close to thetarget surface whereas those from neutral atoms are observed in the plume away fromthe target. Thus line emissions are superimposed on the continuum emission. The lineemission from the multiply ionized species occurs at the time of plasma formation,whereas emission from singly ionized and neutral species are observed nearly 500 nsafter the plasma-formation. The rates of decay of the continuum and line emissions aredifferent. The continuum emission due to Bremsstrahlung from hot plasma decays fasterin comparison to the line emission. Thus it is necessary to record the emission after acertain time delay to get clear information about the line emission from the cold plasmafor the purpose of elemental analysis.

5. THEORETICAL MODELS FOR PLASMA

Interpretation of the radiation emitted by the plasmas requires knowledge of both thecharge state distribution and the excited level populations of different ions. This ispossible by obtaining the solution of a complex system of rate equations, describingthe population and depopulation of all the levels by the processes such as ionization,recombination, collisional excitation and de-excitation, radiative decay and absorptionas well as the stimulated emission. Any given charge state is connected with its twoneighbouring states by the processes of ionization and recombination. Considering thedifficulties associated in solving these equations, the approximations used in order ofincreasing electron density are: the corona model (CM), the collisional-radiative model(CRM), the local thermodynamic equilibrium (LTE) and models suitable for ultra highdensity �Ne ≥ 1024 cm−3� plasmas [29,31,33].

5.1. Corona Model

In this approximation there is a balance between collisional ionisation (and excitation)and recombination (and spontaneous decay). This model, therefore, depends critically onthe knowledge of atomic cross sections. Assuming that free electrons have a Maxwellianvelocity distribution, it is required that only a negligible number of ions be in excitedlevels (as compared to the ground level). Two neighbouring ionisation states, of chargeZ and �Z +1�, are then connected by

NZNeSZ �Te� = NZ+1Ne�Z+1�Te�Ne�

or

NZ

NZ+1

= �Z+1 �Te�Ne�

SZ �Te�(3)

where Sz and Z+l are the ionisation and recombination rate coefficients, respectively.Eq. (3) is, to a first approximation, independent of Ne. In this approximation a balance




between the rate of collisional excitation from the ground level and the rate of sponta-neous radiative decay determines the population densities of excited levels. This modelrequires that the electron density be sufficiently low for collisions not to interfere withradiative emission.

5.2. Local Thermodynamic Equilibrium Model

In the LTE model it is assumed that mainly particle collision processes determine thedistribution of population densities of the electrons, which are so fast that the lattertake place with sufficient rapidity and the distribution responds instantaneously to anychange in the plasma condition. In such circumstances each process is accompanied byits inverse and these pairs of processes occur at equal rate by the principle of detailedbalance. Thus the distribution of population densities of the energy levels of the elec-trons is the same as it would be in a system in complete equilibrium. The populationdistribution is determined by the statistical mechanical law of equipartition among theenergy levels and does not require the knowledge of atomic cross sections. Actually theplasma temperature and density vary in space and time, but the distribution of popu-lation densities at any instant and point in space depends entirely on the local valuesof temperature, density and chemical composition of the plasma. The uncertainty inprediction of spectral line intensities from LTE model plasma depend mainly on theuncertainty in the values of these plasma parameters and atomic transition probabil-ities. For analytical plasmas the condition of LTE is considered very much vital forgetting any reliable quantitative information. In the case of thermal equilibrium all theprocesses in the plasma are collision dominated as discussed above and the plasmacan be considered as having a single temperature Ti = Te. However in the expandingplasma this is possible only locally and for specific time segment during the evolution.The following criterion must be satisfied by the plasma to be in local thermodynamicequilibrium [33–34]:

Ne ≥ 16×1012�E3T 1/2e (4)

where �E �eV� is the largest observed transition energy for which the condition holds,and Te is the excitation temperature (K). It should be noted that the choice of thetime delay is crucial for obtaining the best operating conditions in the LIBS plasmato ensure that LTE prevails during the measurements for obtaining the quantitativeresults. However, it has also been found that LTE is not an indispensable condition forqualitative analysis, once the measurement conditions are kept constant and are exactlyreproducible.

5.3. Collisional Radiative Model

In the case of high density plasma both of the above models can not be used safely(although LTE is sometimes marginally satisfied). Salzmann [35] has addressed the




applicability of the CM and LTE approximations to laser-produced plasmas. In theintermediate density range the rate equations have to be solved up to the thermal limit.The total ion densities of two successive ionisation states of charge Z and �Z + 1� areconnected by

NZSeffZ = NZ+1��

effZ+1 +Ne�

effZ+1� (5)

where SeffZ ��

effZ+1 and �

effZ+1 are the effective (or net) ionisation, recombination and

three-body recombination rate coefficients, respectively. The last term dominates onlyat high densities, when NZ

NZ+1tends to the Saha equation value. At low electron densi-

ties the three-body recombination is negligible, and Eq. (5) tends to Eq. (3) (excitedstate populations become negligible with respect to the ground-state population). TheCRM system can be solved locally for given Ne and Te values. It can however also becoupled to a hydrodynamic code describing the plasma evolution. Capitelli et al. [36]have studied non-equilibrium and equilibrium problems in laser induced breakdownplasma. Particularly they focused on the problem associated with the fluid dynamicsof the expanding plume with time dependent collisional-radiative models for describ-ing the population densities of excited states and with the time dependent Boltzmanequation for characterizing the electron energy distribution function in the LIBS plas-mas. It was found that the violation of equilibrium conditions in laser-produced plasmanear the surface could be caused by the decrease in the plasma temperature due toexpansion. This can occur if the characteristic time of such temperature decrease is lessthan or comparable with that corresponding to the ionization balance, which is esti-mated as ion ≈ �Nakion�

−1, where Na�cm−3� and kion�cm3s−1� are the number density

of heavy particles and ionization rate coefficients respectively. For typical laser plasmaconditions �Na ≈ 3 × 1019 cm−3� Te ≈ 2 eV� ion is in the range ≈ 10−6 − 10−5 s. Thisquantity has to be compared with the characteristics time of the laser plasma expansion,i.e. exp ≈ d/v, where d is the laser spot diameter and v ≈ 105 −106 cm/s is the plasmaexpansion velocity. It can be concluded that the equilibrium conditions are violated,whenever

ion ≥ exp (6)

which takes place at laser spot sizes d < 1 cm. In this condition, the deviation fromthe equilibrium state has a recombination character. LIBS plasmas, characterized bylarge electron densities and electron temperatures, apparently seem to satisfy the LTEconditions. However, the characteristics equilibrium times for the different phenomenacan occur in the same temporal scale during which LIBS measurements are performed.Dedicated experiments and the development of a unitary theory, which takes into accountthe fluid dynamics and the kinetic aspects, is necessary to completely master the exper-imental conditions for the development of a calibration–free LIBS [37].

6. MEASUREMENT OF PLASMA PARAMETERS

During the LIBS experiments some important parameters such as emission line shapefunctions, electron density and plasma temperature are required in order to produce




reliable spectral features. Some relevant information about the measurement of theseparameters has been given in the following sections.

6.1. Line Broadening

In the spectroscopic study of line emission from the plasma, intensity of the spectralline as well as its profile plays an important role. The emission line profile is importantbecause it contains information related with the emitter and surrounding plasma envi-ronment. An introduction to the theory of line broadening can be found in the booksby Griem [29,38] and Sobelman et al. [39]. Various types of line broadening have beenobserved in the plasma emission. Natural broadening occurs due to the finite lifetimeof excited states and results in a Lorentzian profile. It must be considered in particularcases of highly ionised ions like Fe25+. The Doppler broadening occurs due to thermalor directed motion of the emitting ions and has a Gaussian line shape with a width pro-portional to the square root of the emitter temperature. This is the dominant broadeningmechanism at low electron densities. Stark broadening which is also known as pressurebroadening is observed because the emitting ions experience an electric field due to thepresence of plasma electrons and ions around them. This field varies statistically forindividual emitters and fluctuates in time. The net result is an ensemble averaged lineshape with an overall width related to the average strength of the perturbation in thebound states.

During the early stage of plasma formation in the LIBS experiment the electrondensity remains very high �ne ∼1015−1018 cm−3�. As a consequence, the line profilesare dominated by Stark/pressure broadening for a considerable period of time [14].Doppler as well as natural broadening is generally negligible during this period. Whenthe expanded plasma cools and the electron density decreases, the dominance of the Starkbroadening is reduced and finally Doppler broadening starts playing a leading role. It iswell known that measured line profiles normally contain contribution from instrumentresolution width also, if spectrometers are used for wavelength analysis. This indicatesthat measured profiles must be deconvoluted prior to their analysis for the extractionof plasma parameters. The instrument width needs to be measured experimentally fora given spectrometer, which is dependent on the parameters such as slit width, thegrating dispersion, and the dynamic behavior of the photon detector. The central peakof the spectrometer slit function can be approximated to a good degree of accuracy by aLorentzian profile. Since both (the Stark-broadened) spectral line and the spectrometerexhibit Lorentz shape functions, convolution and deconvolution become easy. In thiscase line widths can simply be added or subtracted in the convolution/deconvolutionprocess as given below

��Total = ��line +��spectrometer (7)

This shows that actual line width can easily be extracted from the measured line widthby simply subtracting the instrument width.

Samek et al. [14] fitted the line shape for CaI and CaII for the highest concentrationin liquid sample and nearly perfect Lorenzian profiles were obtained even for delay timeof <1�s, which shows that problem of self absorption is less probable in liquid plasmathan in solid plasma due to comparatively less concentration.




6.2. Electron Density

The width of Stark-broadened spectral lines in plasmas depends mainly on the electrondensity Ne as discussed in chapter 2. Both the linear and the quadratic Stark effect areencountered in spectroscopy. However, only the hydrogen atom and H-like ions exhibitthe linear Stark effect, whereas all other atoms exhibit the quadratic Stark effect. Thisis the reason that ideally information about electron density is extracted from the linesof H or H like ions, where half width of the line profile can be calculated easily with agreater accuracy. In the case of linear Stark effect, the relation between electron densityand the line width is given by a simple relation [29]:

Ne = C �Ne�Te��3/2FWHM (8)

here �� is the “true” FWHM and the parameter C depends (only weakly) on Ne and Te,which can normally be treated as being constant. The constant C for the H Balmer linesis available in the literature [29]. The first choice for electron density determination inLIBS plasmas containing hydrogen is the H� line (with an error of 5%) [29] becauseof its large intensity and sufficiently large line broadening, which can be measuredprecisely using a spectrometer of moderate resolution. The possibility of self-absorptionin this case is relatively small. The second best choice among the Balmer series is theH� line. The H� line is suitable in the cases where the electron density is not too high�Ne ∼1017 cm−3�, because at higher electron densities this strong line is quite susceptibleto self-absorption, which severely distorts the line profile. In the case of non-H-likeatoms, where the quadratic Stark effect is dominant, the relation between the electrondensity and the line width [29] is

��FWHM ≈ 2[2+175×10−4N 1/4

e �(1−0068N 1/6

e T−1/2)]×10−16wNe (9)

The first term in the brackets gives the contribution from electron broadening, and thesecond term stems from ion broadening. Here w is the electron impact parameter atNe = 1016 cm−3, and � is the ion broadening parameter. The parameters w and � can befound easily from the literature [29]. Since the second term in Eq. (9) is normally small,so the expression reduces to

��FWHM ≈ 2×10−16wNe (10)

which is normally used for calculations in the case of plasmas generated from solidtargets.

It has been reported by Samek et al [14] that the electron density derived from the H�

line and H� line data were in good agreement with each other, while the electron densityestimated from the H� line was up to a factor of ten larger. This study suggested thatthe H� line is not purely Stark broadened. The additional broadening may be due to theonset of self-absorption. The electron densities extracted from the H� line data fluctuatesignificantly because the H� line overlaps with strong emission from a nitrogen line. Itis also found that in addition to the hydrogen lines, resonance lines of aluminum andcalcium at 396.15 and 422.67 nm respectively can be used to estimate electron-densitiesconsidering that these lines exhibit quadratic Stark broadening.




6.3. Plasma Temperature

Plasma temperatures are often determined from the measurement of ratios of the inten-sities of (a) ion to neutral lines and (b) neutral to neutral lines, usually for the sameelement. In case (a) the line intensities are combined with the Saha equation and theelectron density measurements to determine the ionization temperature of the plasma:

Iion

Iatom

= 2× �2�mekT�3/2

Neh3

(gA

�

)

ion

(�

gA

)

atom

× exp[− �V + +Eion +Eatom�

kTion

]

(11)

here I is the integrated emission intensity of the ion or atom, Ne is the electron density�cm−3�, gA is the product of the statistical weight and the Einstein coefficient forspontaneous emission of the upper level �s−1�� is the wavelength (nm), Tion is theionization temperature (K), V + is the ion potential of the atom (J), Eion is the excitationenergy of the ionic line (J), Eatom is the excitation energy of the atomic line (J), k is theBoltzman constant (J/K), and h is the Planck’s constant (J s).

In case (b) the line intensities are combined with the Boltzman equation to determinethe excitation temperature of the plasma and the relation is given by,

I1

I2

= g1A1

g2A2

�2

�1

exp(

−�E1 −E2�kTe

)

(12)

where 1 and 2 refer to the individual lines in the pair. The accuracy in the temperaturedetermination increases with an increase in energy difference E1 − E2 in the aboveequation. The accuracy may be improved by measuring a number of different line pairsand taking the average. However, it should be noted that the accuracy of the measurementlargely depends on values of A coefficients and their error ranging from 5% to 50%. Theelectron densities obtained from the hydrogen Balmer lines were used, along with lineintensities of Ca I at � = 4226 nm and Ca ll at � = 3933 nm, to calculate the ionizationtemperature from Eq. (11) [14].

Normally in the plasmas generated from liquids, very few (resonance) lines areobserved. Consequently, line pairs available for calculating the excitation temperature forsuch cases using the two-line Boltzmann method, as given in Eq. (12), are rather sparse.Samek et al. [14] attempted to determine the excitation temperature using line pairs fromCa II, Cu I, and the H� and H� lines. They relied on the temperature extracted fromH�/H� line-pair data because of the large errors in estimating the excitation temperaturefrom the Ca II or Cu I line pairs (a consequence of the small energy gap E1 − E2

and a small FWHM). From the evaluation of the hydrogen line shapes and intensitiesthey found that the excitation temperature agrees reasonably well with the ionizationtemperature in the LIBS plasma generated from the laminar liquid jet stream, in thetime window 1–5 �s. The errors for excitation and ionization temperature (10% and 8%,respectively) were estimated according to the procedure described by Simeonsson andMiziolek [40]. They noted during this study that the temperature observed for the plasmagenerated from an aqueous solution are generally lower than those obtained from solidsamples.

Abdellatif and Imam [41] used Eq. (12) for evaluating the spatial profile of plasmatemperature from aluminum plasma along the axial direction normal to the target surfaceand found a peak at 500 �m from the surface.




6.4. Optical Thickness and Self-absorption

In the case of very high plasma density conditions, the plasma itself absorbs its ownemission. This is mainly true for the resonance lines connected to the ground state, butother lines may also be affected. The absorption causes a distortion in the spectral lineprofile, resulting in a broadened line. This effect is known as self-absorption. Duringthe LIBS experiments the plasma temperature tends to drop towards the outer parts ofplasma- plume, and when light passes through these colder parts an effect known asself-reversal can occur. Such lines show a (severe) dip at the center of the ordinarilywell-behaved line-shape function, giving a doubt of there being two lines.

Above discussion shows that for evaluating the plasma parameters and extractingquantitative data from the line intensities, it is important to verify that the plasma is notoptically thick for the lines being used for this purpose. The ratio of emission intensitiesof resonant and non-resonant lines should be verified according to a procedure forthe “optically thin” limit described by Cremers and Radziemski [42], Simeonsson andMiziolek [40], and Sabsabi and Cielo [43]. It is necessary that the observed intensityratios are consistent with those predicted by the statistical weights of the upper levelsindicating that the plasma is optically thin. Normally during the measurement of majorelements in composite solid samples, or in pure samples, severe self- reversal and self-absorption are observed for many lines during few �s after the plasma formation. Theresonance line of calcium at 422.67 nm has shown huge self-reversal and self-absorptionover a long period of the plasma evolution [14]. This phenomenon has less effect in thecase of liquid and gaseous samples in comparison to solid but it is observable at higheranalyte concentration or at higher laser intensity.

7. CHARACTERISTICS OF LIBS PLASMA

The schematic diagram of LIBS setup is shown in Fig. 2. It consists of a laser, a focusingsystem, the target sample to be analyzed, and a telescope for imaging the plasma-plume on the entrance slit of spectrometer, where gated ICCD camera is used as adetector. Small differences in the LIBS instrumentation has-been noted in terms of laserwavelengths used for plasma formation. Ciucci et al. [44], Cremers [45] and Wisbrunet al. [46] used a Nd:YAG laser operating in its fundamental wavelength �� = 1064 nm�.Singh et al. [47] and Rai et al. [48] have used second harmonic of Nd: YAG, whereasBarbini et al. [49] used an Nd: YAG laser operating in the third harmonic �� = 355 nm�.In most of the cases, the laser duration is between 5–10 ns and the fluence ranges from1 to 50 J/cm2. When the laser is focused on solid samples of different nature, irradiancesof the order of a few GW/cm2 can be achieved. Much higher laser irradiances arenecessary, when the laser is focused on liquid, gas or aerosol mixtures [46–48,50–51].Yalcin et al. [50] have used a modified Nd:YAG laser �� = 532 nm� with pulse energiesof 40–150 mJ (pulse duration between 10 and 13 ns) focused to a 21�m spot size,thus reaching irradiance values ranging from 500 to 1500 GW/cm2. Normally theline emission intensities of selected elements are recorded during a temporal window(gate width) tb after a given time delay �td� from plasma formation [45–51]. Thedetection of the strong background continuum at early times can introduce noise in themeasurements, because in the beginning the plasma is hot and it emits Bremsstrahlung




IDAD – Intensified Charge couple Device

2XNd:YAG laser

Data acquisition/Analysis system

ComputerController

Spectrograph

Pulse generator

ICCD

L – Lens2x – KDP Doubler

FO – Fiber Optics

BD – Beam Dump

HS – Harmonic Separator

DM – Dichroic Mirror

BDHS

LDMLLFO

Solid sample

Fig. 2. Schematic diagram of experimental system for recording the LIBS spectra from solidsample. In fact any type of sample can be placed at laser focus to get LIBS spectra of that sample.

0.00E + 00

5.00E + 04

1.00E + 05

1.50E + 05

2.00E + 05

2.50E + 05

0 5 10 15 20

Delay time (µs)

Inte

nsity

Cr 425.4 nm

Background

Fig. 3. The variation in the atomic line and background emission intensity from chromium (Cr)in aqueous solution with gate delay �td�. Gate width �tb� = 10 �m.

radiation (Fig. 3) [52]. Thus a judicious selection of td and tb is necessary to optimize thesignal-to-noise ratio as well as signal to background ratio. Experimental investigationsshow that LIBS measurements are generally made for td and tb values ranging in themicrosecond regime. These characteristic times can be compared with the characteristictimes indicating the different kinetics in the plasma under consideration.

As discussed earlier the plasma is formed by nanosecond laser pulses as a resultof photon absorption processes such as inverse Bremsstrahlung and photoionization.




Theory and experiment [53] have shown that the electron densities as large as 1020 cm−3

and electron temperatures of the order of 2 eV are typical of laser ablation process in thenanosecond regime. However after nearly 5 ns, electron �Te� and ion temperature �Ti�become nearly equal as a result of thermalization process occuring due to collisions.This is one of the requirements for plasma to be in LTE and is the reason why LIBSmeasurements are made a few microseconds after the plasma formation, when theLIBS plasma is under typical recombination conditions. Lower electron density Ne andtemperatures Te exist during LIBS sampling in the microsecond regime. Typical densityand temperature lie in the range 1015 < Ne < 1018 cm−3 and 05 < Te < 2 eV respectively,which are strongly dependent on the time delay td. Higher electron densities are foundfor the plasmas produced at atmospheric conditions, because of confinement of freeexpansion of plasma by the air. However, these values also depend on the type of laserand the sample used for the experiment. Normally, lower electron densities and lowertemperatures are observed in the plasma plume away from the target sample surface. Thereported Ne and Te value ranges are sufficient to satisfy the LTE plasma condition, whereSaha and Boltzmann equations hold for the number densities of plasma constituentsincluding excited state number densities.

These equations also imply the validity of Maxwell distribution function for describingthe electron energy distribution function. Moreover, these distributions {Saha, Boltzmannand Maxwell) are thought to adapt themselves to the local values of Ne and Te duringthe process of plasma expansion. The choice of the experimental arrangement shouldbe aimed at optimizing the reproducibility of the measurements, while at the same timeachieving plasma conditions, where the hypotheses of LTE and thin plasma are fulfilled.

It has been noted that LIBS plasma as well as emission from it depend on theexperimental configuration as well as on the plasma parameters. This dependence willbe discussed in the following sections. A variety of different experimental configurationshave been reported in the literature, where different arrangements are adopted to solvedifferent analytical problems.

8. FACTORS AFFECTING THE LIBS PLASMA

Various types of lasers are used in laser-induced breakdown spectroscopy, typicallyranging from UV excimer lasers to infrared solid-state lasers. Each laser has a differentabsorption characteristic during plasma production, which affects the behavior of theresulting plasma. The shape and size of the laser-induced plasma plume is dependent onvarious other experimental conditions including the thermal and mechanical propertiesof the target material. Effects of these factors on the LIBS plasma has been discussedin the following sections.

8.1. Laser Characteristics

There are two main mechanisms for electron generation and their growth before theplasma formation. The first mechanism involves absorption of laser radiation by freeelectrons present in the target vapor, when they collide with neutrals. If the electronsgain sufficient energy, they can impact and ionize atoms or molecules present in the




sample vapor. The electron concentration will increase exponentially with time due tothe cascade breakdown. The second mechanism, called multiphoton ionization (MPI),involves the simultaneous absorption of a sufficient number of photons by an atom ormolecule to cause its ionization. Multiphoton ionization plays an important role only forthe short wavelength lasers �<1 �m�. Both cascade and multiphoton ionization requirehigh laser irradiances, usually ≥108 W/cm2. Finally plasma formation takes place, whenthe laser intensity exceeds this threshold value, which is easily attained by properfocusing of the laser beam. In the diffraction limit, the diameter w2 of the laser beam atthe focus of the lens can be calculated as:

w2 ≈ 244�f

w1

(13)

where w1 is the diameter of the laser beam before focusing, f is the focal length of thelens and � the laser wavelength.

Many authors have studied laser ablation in various materials as well as its relationwith the threshold intensity, which is also dependent on the properties of the target [10].Hwang et al. [54] have reported that the mass removed from the target depends notonly on the intensity of the laser but also on various other factors. A relation has beenfound for the ablated mass m (t) from the target surface on the basis of heat conductionmechanism as

m�t� = A�aIt�+B�AI�2t3/2 (14)

where A, B are proportional to the target thermal properties, a is energy coupling factor,I is laser irradiance and t is the laser pulse duration. Cabalin and Laserna [55] madea systematic study for determining the threshold intensity for plasma formation in agroup of nine metals, largely spread in physical and thermal properties, like meltingand boiling temperatures, thermal conductivity, etc. They found that the thresholdintensity is correlated fairly well with thermal properties such as melting and boilingtemperature, suggesting that thermal effects play a significant role during laser ablationwith nanosecond pulses. They also investigated the behavior of line emission intensityvs. incident intensity and found an initial linear correlation, probably due to an increasein the amount of ablated material. For higher laser intensity, emission signal reaches asaturation regime, attributed to the self-absorption of the emission by the plasma formedin front of the sample, or due to poor coupling of the laser because of plasma shielding.The starting point of the saturation regime is element-dependent, and correlated with thevalue of the threshold intensity. Similar behavior in respect of laser coupling efficiencywas found for nanosecond and picosecond lasers by Russo [56].

It has been found that spectral emission intensity is also dependent on the plasma tem-perature, which is significantly increased by higher laser power density. Yueh et al. [52]reported an increase in the sensitivity of calibration curve of Re, when laser energy wasincreased from 200 to 250 mJ (Fig. 4). Higher power density has been found beneficialfor improving the analytical sensitivity [57], but in the limit of self-absorption, severaltechniques have been used for manipulating the laser intensities. Chaleard et al. [58]spatially filtered the laser beam before focusing it on the sample surface, in order toobtain a more homogeneous energy distribution and consequently a more regular crater.Ideally a ‘flat-top’ distribution of the laser beam energy is supposed to be better for




0

5000

10000

15000

20000

25000

30000

35000

40000

45000

0 50 100 150

Re concentration (ppm)

Inte

nsity

E = 250 mJE = 200 mJ

Fig. 4. The LIBS calibration curve for Re obtained from liquid jet measurement with delay timeof 8 �s and gate width of 15 �s.

target ablation, because each point of the target surface would receive the same amountof energy, which is generally absent in the lasers used for this purpose.

The effect of pulse-to-pulse laser stability on the emission intensity has been inves-tigated by Castle et al. [59] by simultaneously monitoring the analyte (Cu) signal andthe laser pulse energy by means of a photodiode. The absence of significant correlationindicated that the laser pulse variance played a small role in the overall variance ofLIBS measurement. It seems that along with the fluctuation in laser intensity, instabil-ities in the plasma also add up in the standard deviation of the LIBS signal. Wisbrunet al. [46] measured the behavior of the signal-to-noise ratio in sand and soil samplesdepending on the laser pulse energy, which was varied between 0 and 320 mJ. In theirexperimental conditions, the signal-to-noise ratio increased with laser pulse energy untilapproximately 100 mJ, and then remained almost constant.

8.2. Wavelength and Pulse Duration of Laser

As stated earlier various types of lasers have been used for the study of LIBS and theresulting plasma having different properties [60–61]. In a large number of studies Nd:YAG lasers �� = 1064 nm� have been used, and in many cases the fundamental radiationhas been converted into second �� = 532 nm�, third �� = 355 nm� or fourth �� = 266 nm�harmonic depending on the application. Fabbro et al. [62] used an Nd: YAG laser thatwas frequency doubled and quadrupled to study the effect of wavelengths and founda relation for the mass ablation rate m�kg/s cm2� in terms of wavelength �, and theabsorbed laser flux IL �W/cm2�

m = 110�I

1/3L �

�1014��−4/3 (15)




Another scaling for mass ablation rate has been reported by Dahmain [63] for thelaser energy lower than 1013 W/cm2 and Z ≤ 13 as

m�kg/scm2� = 65[

IL�W/cm2�

1013

]5/9

×�−4/9��m�× z1/4 (16)

These relations show that mass ablation rate will increase strongly with shorter wave-lengths. Eq. (1) also shows that the critical density of the plasma will increase for alaser of lower wavelength, which is an indication that laser energy will be absorbedin the plasma more efficiently (up to higher plasma density). In a nanosecond laserablation process, target evaporation begins just after the impact of the leading edgeof the laser pulse on the surface. The interaction of the following part of the laserpulse with the vapor in the vicinity of the target surface leads to a strong heating andionization of the vapor resulting in plasma formation. Although some species can bedirectly vaporized as ionized particles, plasma formation can be mainly ascribed tothe processes involved in the laser-vapor interaction. Amoruso et al. [64] evaluatedthe efficiency of the mechanisms of energy absorption in the plasma for a visible anda UV laser during ablation of an aluminum sample. They observed that the primarymechanism of laser absorption and ionization of the relatively cold neutral vapor formedby the leading edge of the ablating laser pulse is very strong for the radiation wave-length � = 532 nm. This ionization can be mainly ascribed to electron-neutral inverseBremsstrahlung (IB) processes. However for � = 355 nm it was found that direct photo-ionization of excited states in the vapor was the most effective process (the IB processis less efficient in the UV than in the visible part of the spectrum, because of the �3

dependence of IB).Berman and Wolf [65] compared the analytical results obtained in the detection of

Ni in water by using alternatively the fundamental Nd: YAG wavelength and the thirdharmonic UV radiation. They observed lower continuum intensity in the UV generatedspectrum, leading to a better signal-to-noise ratio. The calibration curves for the samespectral lines obtained with UV irradiation revealed a higher slope and allowed a betterlimit of detection (LOD) value. These results are in close agreement with the resultsobtained on liquids by Ng et al [66], who found that the main difference in the plasmaplume generated by visible or ultraviolet laser of the same fluence is in the Te value,which is significantly lower in the second case, while the Ne value is approximatelythe same.

Shorter pulses (picosecond) are reported to produce higher mass ablation rates, prob-ably because the fraction of the pulse energy loss by thermal diffusion in the sample ismuch lower than in the case of nanosecond pulses [67]. A comparison of emission fromplasma produced by nanosecond, pico-second and femto-second laser has been made bymany authors [68–69]. They found that both the line emission, continuum backgroundemission intensity as well as plasma temperature decay very rapidly after excitationusing short time duration laser pulses (ps and fs) compared to the nano second pulseexcitation. It was also noted that gated detection is not necessary with short laser pulses,because background emission is much lower in the case of LIBS using shorter laserpulse than the LIBS using nano second laser.




8.3. Properties of Target Material

The physical and the mechanical properties of the target have an important influenceon the shape and size of the crater on its surface. The reflectivity of the target materialdetermines the fraction of laser energy coupled to the target. It has been found that laserenergy can be absorbed effectively by a reflecting surface due to the phase change of thematerial at high temperature, which is possible only at very high intensity of laser [70].Allemand [71] has indicated that the reflectivity of the sample surface, density, specificheat and boiling point of the pure metal target have an important influence on the shapeand size of the craters and he obtained the following relation:

D = A�1−R�

�cTb

(17)

where D is diameter of the crater; A is proportionality constant; R the reflectivity of thesurface; � density of material; C specific heat and Tb is the boiling temperature.

A comparison of the crater size of the homogeneous material revealed that the thermalconductivity is an important parameter. The volume heated by laser pulse is found todepend on the thermal conductivity of the material [72]. The heating of material aroundthe crater increased with increase in incident laser intensity, because evaporation dependsmainly on the boiling point of the material at fixed pressure.

Dimitrov et al. [73–74] have investigated the dependence of evaporation processesand the dynamics of the plasma plume on the orientation of the target surface withrespect to the laser beam. When a metal target is irradiated by laser, the ablated productsexpand nearly perpendicular to the target surface. When the target surface is inclinedwith respect to the direction of laser beam, the path length of the radiation in the plasmais shortened, and results in decreased absorption by the laser-produced plasma.

Mao et al. [67,75] have reported that the absorbed energy in copper target is muchhigher, when irradiated by UV laser for producing plasma-plume. They experimentallydemonstrated that the transition between thermal and explosive regimes in laser ablationoccurs, for the different wavelengths, at power density values scaling inversely as the�1 − R� factor, where R is the reflectivity of copper. Russo et al. [76] analyzed theinfluence of laser wavelength on the specific problem of fractionation in laser ablationand found that shorter the wavelength, the more controlled and reproducible is theablation rate. The intensity requirement to initiate the ablation process is also found tobe lower for shorter wavelength lasers.

Cabalin and Laserna [55] found that the thresholds for plasma formation and theonset of saturation regime were shifted towards lower fluence for IR wavelength, whilethe energy threshold values were higher for IR radiation. They found that reflectivitydoes not seem to be a relevant parameter at high fluences (i.e. above the threshold),because the plasma formation changes the properties of the target surface.

8.4. Time Window of Observation

Initially the laser-induced plasma has very high electron temperature. However, plasmatemperature decreases as the plasma expands away from the target surface. Early stages




of plasma evolution are dominated by the Bremsstrahlung continuum emission. Theevolution of line emission has been observed after the expansion of plasma followedby decrease in plasma temperature. These line emissions are observed superimposed onthe continuum emission. Only the line emission from the plume is important for thecompositional analysis of the sample target. Normally the continuum emission from theplasma decays very fast in comparison to the line emission. So recording of line emissionfor the purpose of elemental analysis would depend on choosing a proper delay time,when the ratio of line emission to background (continuum) emission is very high. Atlarge delay times the absolute intensity of the line emission can be too low for efficientdetection. The best compromise between high line intensity and low background isdetermined on a case by case basis. Depending on the density of the plasma, ambientgas and other factors, the plasma lifetime ranges from approximately 300 ns to morethan 40 �s [77].

Ciucci et al. [44] compared the time evolution of the plasma emission obtained byirradiating in air the same sample with the fundamental wavelength of an Nd: YAGlaser and the UV radiation of an excimer laser. They observed that in the case ofUV excitation the plasma emission was initially dominated by the background up toapproximately 400 ns, when the atomic line transitions started to appear. The continuumemission generated by Nd: YAG laser showed comparatively longer times ranging inseveral microseconds. The time scales of the plasma induced by the two different sourcesdiffer remarkably due to a more rapid decay of continuum emission in the experimentperformed with UV excitation as compared to the IR excitation. It has been observed thatthe delay time (gate delay) and integration time (gate window) both play an importantrole in optimization of signal to background and signal to noise ratio. The optimum valueof gate delay and gate width changes due to the composition of the target as well as theother plasma parameters that affect the plasma. Wisbrun et al. [46] reported a systematicanalysis of two elements (Zn and Cd) in a sand sample. They found an optimal valuefor the delay time of approximately 05 �s with an integration time of approximately15 �s for these elements.

During the plasma evolution, ion density remains higher near the target surface,whereas neutral density dominates as a result of recombination as the plasma expandsaway from the target surface. This is the reason, why the ratio between the populationof neutral and ionized species changes with time. Yueh et al. [52] reported that theintensities of line emission from magnesium ion and background emission were foundto be high in comparison to that of line emission from neutral Mg, when the LIBSspectra were recorded at 2 �s gate delay. However ion emission and background emissionintensity decayed fast as the gate delay was increased to 4–5 �s (Fig. 5). Leis et al. [78]reported the study of evolution of the emission intensity of two iron lines (Fe I 285.2 nmand Fe II 288.4 nm), in a pure iron sample, recorded during the first 15 �s of plasmalifetime. During the first 3 �s, the intensity of the ionic line exceeded the neutral by 50times, whereas at a delay of approximately 10 �s, the neutral line intensity was eighttimes that of the ionic. Aragon et al. [79] studied the evolution of line intensity andthe line-to-continuum ratio and found dependence on the specific line and on the pulseenergy. They concluded that line emission intensities and line-to-continuum emissionratios cannot be simultaneously maximized for all the elements investigated by using asingle detection time window. They suggested to use a large integration time �2–15 �s),




0

200000

400000

600000

800000

1000000

1200000

1400000

312 317 322 327 332 337

Wavelength (nm)

Inte

nsity

2 µs

Mg+

279

.53

nm

Mg+

280

.27

nm

Mg

285.

29 n

m

4 µs 5 µs

Fig. 5. Typical LIBS spectra of Mg recorded at different delay times from liquid jet.

which allows to collect most of the LIBS signal for all the elements, while eliminatingthe undesirable initial part of the emission.

Sabsabi and Cielo [80] have analyzed the LIBS spectrum of aluminum alloys recordedafter a delay time of 10 �s, and a gate width of 10 �s. They have also reported thetemperature evolution in a nanosecond time duration Nd: YAG laser-induced plasmafrom aluminum and copper targets in air [81]. A quick decrease in plasma temperaturewas found during first few microseconds (from approximately 1 eV to approximately0.5 eV) in both the cases, whereas a small change was noted at later times in themicrosecond scale (two decay rates). Leis et al. [78] determined the time-resolvedtemperature for a series of binary Fe-Cr alloys with iron content ranging from 10 to100%. The time evolution in plasma temperature was found similar for all the samples,whereas the absolute value differed by approximately 50% among the samples, with thepure iron showing the higher temperature.

8.5. Geometric Set-up

The geometrical shape of the plasma and spatial emission intensity profile are stronglydependent on the laser power density and on other parameters such as optical alignmentfor focusing the laser and collecting the emission from plasma plume for recording thespectrum. Therefore, it is necessary to understand the dependence of LIBS signal on theoptical alignment and the collection of emitted light for recording the spectrum.

Eppler et al. [82] have compared the precision of the results obtained using a sphericaland a cylindrical lens for focusing the laser pulse on the surface of a soil samplewhere more precise results were found in the case of cylindrical lens. In the case ofLIBS of solid sample normally laser is focused perpendicular to the target surface.The high intensity of laser may produce a breakdown in air before the focus at thesurface, particularly when some dust particles are occasionally irradiated. This problemis avoided by setting the distance between the focusing lens and the target a little shorter




than the focal length to produce a stable breakdown, while maximizing the interactionarea [46,79]. This perpendicular irradiation configuration may cause some problems inthe case of liquid samples due to the splashing of liquid on the lenses, which stronglyreduces the sensitivity of LIBS as plasma plume (ejected droplets) expands in thedirection perpendicular to the sample surface (liquid surface) [77,83]. Fichet et al. [84]suggested a configuration with the laser beam directed on the surface with a tilt angleof 15� (75 � with respect to the perpendicular), which keeps the optics protected fromsplashing.

The collection of plasma emission and detection is normally performed axially, thatis, along the direction perpendicular to the target surface because of its simplicity andreproducibility. This configuration was found less sensitive to changes in the surface-to-lens distance, which occur when several shots are fired at the same place on thesample surface and a crater is formed. With on-axis collection, the change in lens tosurface distance causes minimum perturbations on the LIBS signal, because of the depthof focus of the detection optics, which is typically longer than crater depth [58]. Thisis why lenses having focal length of a few centimeters are typically used for LIBSmeasurements. There are certain other types of changes, which may occur in the plasmaemission, i.e. those produced by time evolution of the plasma plume as the plumeexpansion is governed by the strong explosion law. If we assume a one-dimensionalmodel, where the plasma expands along the x-direction and the target surface is at x = 0,the distance x (t), reached at time t can be expressed as

x�t� = k

(EL

�0

)m/2

tm (18)

where EL is the energy deposited by the laser during the ablation process and �0 is theambient gas density. The coefficient m depends on the plasma expansion geometry andis equal to 2/3 for planar propagation and 2/5 for spherical symmetry [85–86]. Eq. (18)shows that the plasma is moving fast in the initial stages, but slows down at latertimes. This indicates that the time-gates and the region of the plasma to be inspectedmust therefore be matched in order to intercept the emission signal at a certain spatiallocation [57].

Another arrangement used for collecting the plasma emission is in the directionperpendicular to the laser axis. Several studies have investigated the dependence ofemission intensity on the distance of the collection axis from the target surface. Kimet al [87] studied the spatial distribution of the emission intensity by scanning thedirection perpendicular to the target surface and found highest intensity at a distance of3 mm from the target surface for a time delay of 30 �s in air. Liu et al. [88] and Ciucciet al. [89] studied the early phase of the plasma, and reported a double-peaked intensitydistribution along the normal to the irradiated surface. They collected the emission signalfrom the region, where the maximum intensity was found, avoiding at the same time theregion close to the target surface and at the plasma-air interface, where self-absorptionwas dominant even for non-resonance lines.

Lee et al. [90] have studied the spatial features of laser-induced plasmas from differentmetallic targets by measuring the emission intensity along the axial direction perpen-dicular to the target using a time-integrated system. It was noted that for copper plasmathe spatial extension was limited to approximately 2 mm in the axial direction, with a




peak intensity lying at a distance lower than 1 mm from the target surface, whereas leadplasma extended for approximately 5 mm from the target, with peak intensity at approx-imately 2 mm from it. This indicates that the spatial location sampled for analyticalmeasurements should be identified carefully, which is dependent on the target material.In this experiment lead plasma expanded more inspite of being heavier than copper.It was found that thermal conductivity of the element played important role. Copperbeing more conducting showed less plasma temperature and exhibited less expansion incomparison to lead plasma.

Normally LIBS measurements are performed by integrating the emitted intensity overthe line of sight in the plasma, which takes care of the inhomogeneity of the plume arisingdue to different temperatures and electronic densities present in the path. During thismeasurement the whole of plasma plume is imaged on the entrance slit of spectrometerfor obtaining a better reproducibility. Aragon et al. [79] used a demagnification factor offive, in the optical collection optics in order to form an image of the whole plume on theentrance slit of the spectrometer. Similarly, Chaleard et al. [58] used a demagnificationfactor of 30, for the collecting optics in order to image the whole volume of the plasmaon the 200 �m entrance slit of the spectrometer to improve the reproducibility of themeasurements. It is to be noted that shot-to-shot variations in the atomic densities andtemperature distribution inside the plasma affect the reproducibility of the measurement,particularly when the central part of the plasma is probed.

One of the important techniques of collecting emission from the plume is by placinga fiber optic near the plume, at a distance of a few millimeters, in order to avoid damageby excessive heating. The collection angle of the fiber end allows gathering of light froma broad volume of plasma plume. The use of optical fiber bundles has also been proposedfor the simultaneous acquisition of spectral emission from different regions inside theplasma plume [91]. Various configurations have been proposed, including the use of twooptical fibers [92–94] one for delivering the laser beam and one for collecting the plumeemission or one single fiber for both the purposes [95–96]. A comparison has been carriedout between the performances of the fiber-coupled LIBS and lens-coupled LIBS [96].After optimization of different experimental parameters, nearly similar detection limitwere achieved for spectral emission of some of the elements. The use of single opticalfiber for delivering the laser pulse and collecting the emission from plume makes theLIBS system suitable for the study of a sample present at a remote distance, whichpresents a unique advantage for its industrial applications.

8.6. Ambient Gas

The laser produced hot plasma expands away from the target. The presence of ambientgas around the plasma affects the dynamics of plasma plume as well as the emissioncharacteristics of the plasma. Piepmeier and Olsten [97] have reported the effects ofsurrounding atmosphere on the emission spectra, on the crater size and on the amount ofthe sample ablated, by changing the ambient air pressure. Grant and Paul [98] measuredthe spatially resolved emission intensity from an excimer (XeCl) laser produced plasmaon a steel target along the axial direction in the presence of different gas atmospheres(air, helium and argon) at pressures of 0.5, 50 and 760 torr. The behavior of the emissionintensity with change in the three variables (gas, pressure and axial distance) was too




complex to reach a unique interpretation. However, the best signal-to-noise ratio wasobserved for argon atmosphere at 50 torr pressure, when the LIBS signal was recordedapproximately 6 mm from the target surface. Leis et al. [78] found an argon ambientpressure of 140 torr for yielding a higher intensity values for the Si I 288.2 nm linein a steel sample. Sdorra and Niemax [99] compared the effects of different ambientgases (argon, neon, helium, nitrogen, air) in the generation of plasma by focusing ananosecond Nd: YAG laser on a copper target. In a similar experimental condition(pressure lower than atmospheric), argon was found to produce a comparatively higherplasma temperature, higher electron density, and higher emission intensity, but a lowermass ablation rate for some of the elements. The decay rate of the temperature duringthe first 40 �s after the plasma formation was found slower for argon than for theother gases due to its low thermal conductivity. The same behavior was also found byIida [100]. However, helium atmosphere produced a comparatively lower temperature,electron density and emission intensity. This indicates that argon is most efficientlyheated by inverse Bremsstrahlung and produces buffer plasma, which optically shieldsthe target surface and as a result reduces the amount of ablated mass. The shieldingeffect takes place in the presence of other gases also but for pulse energy greater than20 mJ, whereas for argon it was observed for 10 mJ energy per pulse. It was noted thatargon yields the highest analyte emission intensity, except at high pressures, whereasneon offers the best performance.

Kim et al. [87] also observed an increase in emission intensity along with longerplasma lifetime in argon atmosphere and have attributed this phenomenon to smallerconductivity (0.0387 cal/cm s deg in STP) and specific heat (0.0763 cal/g deg in STP)of argon gas in comparison to those of air. Such differences in the thermal propertiesof ambient gas result in higher temperature plasma leading to stronger and longeremission and slower cooling. Another important effect produced by the ambient argonis protection of the excited atoms from forming stable compounds such as oxides,which might reduce the LIBS emission from the analyte. Wisbrun et al. [46] foundthat the argon atmosphere was most favorable both in terms of higher analyte emissionintensity (1.8 times the intensity obtained in air for the Zn line at 481.1 nm) and betterreproducibility (R.S.D. = 12% over 18 measurements, compared to 18% in air). It isimportant to note that as the atomic mass of the ambient gas increases the collisionaltranslational energy transfer is less effective and the plasma life becomes longer.

Normally at low ambient pressure (<1 torr), the ablated vapor (plasma) expandsalmost freely, and the outer part of the plasma becomes colder in comparison to itscore, because of higher energy loss. An increase in the pressure to approximately 1 torrconfines the plasma and causes a reduction in energy loss and produces a more uniformdistribution. Hermann et al. [101] studied the time and space evolution of the electrondensity and temperature by changing the laser power density and the ambient pressure. Intheir experimental conditions, the electron density was found to be more sensitive to laserpower density changes than the electron temperature. Measurements of line emissionintensity, plasma density and temperature decay rate indicated an increase in plasmalifetime with an increase in laser power density. These observations can be explainedby the assumption that the plasma density increases with laser power density due toincreased ablation, and self-absorption of the spectral lines becomes more importantleading to reduced radiation loss and enhanced plasma lifetime. In the case of higherpressure of ambient gas elastic and inelastic collisions occur between target vapor and




ambient gas species become more frequent in comparison to those in free expansion invacuum. Confinement of plasma by ambient gas makes the decrease of plasma densityslower and at the same time the kinetic energy of the plasma particles is partiallytransformed into excitation energy as a result of inelastic collisions.

Thiem et al. [102] have noted that analysis of experimental data at atmosphericpressure is difficult due to the presence of a broadband background emission spectrumgenerated by atmospheric elements. They preferred the use of a vacuum chamber toobtain a better signal-to-noise ratio. Special ambient gases are necessary in particularcases, such as for observing UV spectrum ranging down to 180 nm, where the experi-mental housing must be filled with nitrogen or some other inert gas, in order to avoidthe absorption of the LIBS signal by oxygen molecules.

9. METHODS OF ENHANCING LIBS SENSITIVITY

A comparison of LIBS with the other elemental analysis techniques has shown that poordetection limits are the most important limitations of the LIBS technique. Various tech-niques have been used to improve the sensitivity of LIBS, such as oblique incidence oflaser on the sample surface [84], introduction of purge gas around the plasma [103–104],application of pulsed and dc magnetic field [105–108] as well as double laser pulseexcitation of plasma [109–112]. Mason and Goldberg [105–106] used tens of kilogausspulsed magnetic field for enhancing the emission from laser produced plasma by 2–5times. An enhancement of 1.5–2 times in the emission from the laser produced plasmawas obtained by Rai et al. [107–108] using a steady magnetic field of ∼5–6 kG. Themagnetic field system used in the later experiment was found very simple to handlein comparison to the generation and synchronization of pulsed magnetic field with theLIBS experiment. A detailed study was performed to better understand the mechanismof enhancement in emission from plasma in the presence of magnetic field. The enhance-ment factor was found dependent mainly on the nature of target material (solid or liquid)as well as on the transition probability of the elements. Saturation in the emission fromplasma was also noted towards higher laser energy in the absence as well as in the pres-ence of a magnetic field. Saturation became pronounced in the presence of a magneticfield in both types of samples (solid and liquid) for higher laser energy. Simple analysisof plasma emission in the presence of a magnetic field explained the experimental find-ings of enhancement in emission and showed that this enhancement is dependent mainlyon the plasma � = 8�NekTe/B2 (ratio of plasma kinetic energy to magnetic energy)parameter, where Ne� Te� k and B are plasma density, temperature, Boltzman constantand strength of magnetic field respectively. No enhancement in plasma emission waspossible, when the plasma � was high either due to high plasma temperature or due todensity. This condition was found for the delay time <2 �s. A correlation between anenhancement in the plasma emission and deceleration in the plasma expansion due toconfinement of plasma in magnetic field was noted through plasma �. As the plasmadecelerated under the effect of a magnetic confinement, the emission from it startedincreasing due to an increase in plasma density, which affected the rate of radiativerecombination. A simple relation was found, which explained the phenomenon well andshowed that enhancement in emission was mainly dependent on plasma �, which is a




function of plasma density, its temperature and the external magnetic field. The relationcan be expressed as

I2

I1

=[

V1

V2

t1

t2

]3

=(

1− 1�

)−3/2(t1

t2

)3

(19)

where I, V, t represent the emission intensity, plasma velocity and duration of emissionrespectively. Indices 1 and 2 denote the condition of plasma, that is, in the absence andin the presence of magnetic field respectively. Here I2/I1 denotes the enhancement in theemission from plasma, when the plasma is expanding across the external magnetic field.

Double laser pulse excitation technique was also successfully used to enhance thesensitivity of LIBS by many workers [109–112]. In a double laser pulse experiment onaqueous solution of Mg and Cr, Rai et al. [112] varied the inter-pulse separation betweentwo lasers from 0 to 20 �s and reported an enhancement in the emission by a factor ofmore than six for the inter pulse separation of 2–3 �s. Further increase in the inter-pulseseparation between the lasers made neutral magnesium line emission dominant overthe ion line emission. Observation of maximum enhancement in emission at 2–3 �sinter-pulse intervals indicates that an optimum expansion of pre-plasma was necessaryfor better absorption of the second laser pulse in the plasma plume, which ultimatelyprovided an increased emitting plasma volume, which contributed to the enhancement inemissions. An increase in the background emissions, just after the interaction of secondlaser pulse with pre-formed plasma from first laser, indicated an increase in plasmatemperature due to better absorption of second laser pulse in the plasma. An increasein plasma temperature may increase the ablation of the target material and as a resultthe density of the emitting volume of the plasma. Finally it was concluded that anincrease in the emitting plasma volume, plasma temperature and ablation of the samplematerial after second laser pulse contributed towards enhancing the emission from theplasma. Enhancement in the emission from the plasma was found directly related withimprovement in its sensitivity. The limit of detection of Cr under double laser pulseexcitation was improved by an order of magnitude.

Another technique to improve detection limit has also been developed using selectiveelemental excitation of laser-induced plasma by a different tunable laser and recordingthe fluorescence from it. Many combinations of LIBS and laser-induced fluorescence(LIF) have been described in the literature, which make the LIBS a highly performingtechnique [113–115].

10. CONCLUSION

Laser-induced breakdown spectroscopy is now a very active field in analytical science.Considerable progress in the area of basic and applied research of LIBS has been madeduring the last two decades. However, LIBS is presently being used in a limited numberof applications for which the analytical requirements are low and the advantages of LIBSas a rapid, non-contact, in-situ method have been realised. To fully utilize the potentialof LIBS, much attention is to be paid to the development of theoretical models thatprovide improved understanding of LIBS events and incorporation of the latest advances




in lasers, spectrometer and detection technologies to facilitate acquisition and analysisof data in the field.

From theoretical point of view, there is a need to simulate and understand the completeLIBS phenomenon including the initial laser-material interaction, the ignition and growthof micro-plasma, the fluid dynamics, plasma physics, as well as the detailed chemicalkinetics involving excited atoms, ions and electrons. LIBS plasmas characterised bylarge electron densities and electron temperatures, normally satisfies LTE condition butit is quite likely that non-equilibrium plasma is also present during the LIBS experiment.Such experimental conditions cannot be handled theoretically in the absence of suitablemodels. This indicates the requirement of a multi disciplinary team of scientists engagedin experimental and theoretical research in the related fields. This level of understandingof LIBS process would prove valuable in advancing the LIBS sensing that will movethe technology forward. Further study will be beneficial in advanced chemometrics toimprove the LIBS for identification of materials and quantification of plasma parameterssuch as electron density and temperature to improve pulse-to-pulse reproducibility ofthe LIBS signal.

For the use of LIBS in a broader field of applications, it is necessary that thesensitivity of this system be raised considerably. Accuracy of the measurements needsto be improved by reducing pulse-to-pulse fluctuations and suitable statistical methodshave to be developed to estimate the standard deviation in the data points. Remotesensing LIBS experiments involving large distances require a detailed understanding ofatmospheric propagation as well as innovative approaches for using LIBS hardware.Finally, it seems that LIBS has lot of scope for its improvement and optimization.

REFERENCES

[1] L. J. Radziemski, D. A. Cremers, “Laser induced plasma and applications”, Marcel Dekker,New York, (1989).

[2] G. Bekefi, “Principle of laser plasmas”, John Wiley & Sons, New York, (1976).[3] R. E. Cairn, J. J. Sanderson, “Laser plasma interaction”, Institute of Physics Publishing,

Edinberg (1980).[4] C. E. Max, “Laser plasma interaction” (Ed) R. Balian, J. C. Adams, North Holland,

Amsterdam (1982).[5] W. L. Kruer, “The physics of laser plasma interaction”, Addison- Wesley, New York (1988).[6] S. F. Jacob, M. O. Scully, Sergent III and C. D. Cantrel III, “Laser induced fusion and

X-ray laser studies” Addison-Wesley, New York (1976).[7] L. J. Radziemski, R. W. Solarz and J. A. Paisner, “Laser spectroscopy and its applications”

Marcel Dekker Inc. New York & Basel (1987).[8] R. S. Adrain and J. Watson, J. Phys. D: Appl. Phys. 17 (1984) 1915.[9] V. Majidi and M. R. Joseph, Crit. Rev. Anal. Chem. 23 (1992) 143.

[10] T. L. Thiem, Y. Lee and J. Sneddon, Microchem. J. 45 (1992) 1.[11] F. Y. Yueh, J. P. Singh and H. Zhang, “ Laser-induced breakdown spectroscopy: Elemental

analysis,” in Encyclopedia of Analytical Chemistry, Wiley, New York, Vol. 3 (2000)pp. 2065–2087

[12] K. Song, Y. Lee and J. Sneddon, Appl. Spectrosc. Rev 32 (1997) 183.[13] D. A. Rusak, B. C. Castle, B. W. Smith and J. D. Winefordner, Crit. Rev. Anal. Chem. 27

(1997) 257.




[14] O. Samek, D. C. S. Beddows, J. Kaiser, S. V. Kukhlevsky, M. Liska, H. H. Telle andJ. Young, Opt. Eng. 39 (2000) 2248.

[15] A. K. Rai, V. N. Rai, F. Y. Yueh and J. P. Singh, “Trends in Applied Spectroscopy” Vol.4 Research Trends, Trivendrum, India, (2002) pp. 165–214.

[16] E. Togoni, V. Palleschi, M. Corsi and G. Cristoforetti, Spectrochim. Acta B 57 (2002) 1115.[17] C. Gray-Morgan, Rep. Prog. Phys. 38 (1975) 621.[18] J. F. Ready, “Effect of High Power Laser Radiation,” Academic Press, New York (1971).[19] R. G. Root, A. N. Pirri, AIAA paper 79–1489, AIAA 12th Fluid and Plasma dynamics

Conference, Williamsburg, Virginia, July 23–25 (1979).[20] N. H. Kemp and R. G. Root, J. Energy 3 (1979) 40.[21] D. R. Keefer, H. L. Crowder and R. E. Elkins, AIAA paper 82–0404 (1981).[22] S. A. Ramsden and P. Sevic, Nature 203 (1964) 1217.[23] Y. P. Raiser, JEPT 21 (1965) 1009.[24] V. I. Bergel’son, T. V. Loseva, I. V. Nemechinov, T. I. Orlova, Sov. J. Plasma Phys.

1 (1975) 498.[25] G. Weyl, A. Pirri and R. Root, AIAA J. 19 (1981) 460.[26] W. E. Maher, R. B. Hall and R. R. Johnson, J. Appl. Phys. 45 (1974) 2138.[27] J. B. Steverding, J. Appl. Phys. 45 (1974) 3507.[28] A. N. Pirri, R. G. Root and P. K. S. Wu, AIAA J. 16 (1978) 1296.[29] H. R. Griem, “Plasma Spectroscopy”, McGraw Hill, New York (1964).[30] Ya. B. Zeldovich and Yu. P. Raizer, “Physics of Shock Waves and High Temperature

Hydrodynamic Phenomenon”, Vol. I Academic Press, New York (1966).[31] C. De. Michelis and M. Mattioli, Rep. Prog. Phys. 47 (1984) 1233.[32] M. H. Key and R. J. Hutcheon, “Advances in Atomic and Molecular Physics” Vol. 16 (Ed.)

D. R. Bates and B. Bederson, Academic Press, New York (1980).[33] R. H. Huddlestone, S. L. Leonard, “Plasma Diagnostics Techniques” Academic Press, New

York (1965).[34] A. P. Thorne, “Spectrophysics”, Chapman and Hall, London (1988).[35] D. Salzmann, J. Quant. Spectrosc. Radiation Transfer 27 (1982) 359.[36] M. Capittelli, F. Capittelli, A. Eletskii, Spectrochim. Acta B 55 (2000) 559.[37] A. Ciucci, M. Corsi, V. Palleschi, S. Rastelli, A. Salvetti and E. Togoni, Appl. Spectrsc.

53 (1999) 960.[38] H. R. Griem, “Spectral Line Broadening by Plasmas”, Academic Press, New York (1974).[39] I. I. Sobelman, L. A. Vainshtein and E. A. Yukov, “Excitation of Atoms and Broadening

of Spectral Lines”, Springer Verlag, Berlin (1981).[40] J. B. Simeonsson and A. W. Miziolek, Appl. Opt. 32 (1993) 939.[41] G. Abdellatif and H. Imam, Spectrochim. Acta. B 57 (2002) 1155.[42] D. A Cremers and L. J. Radziemski, Anal. Chem. 55 (1983) 1252.[43] M. Sabsabi and P. Cielo, Appl.Spectrosc. 49 (1995) 499.[44] A. Ciucci, V. Palleschi, S. Ratelli, R. Barbini, F. Colao, R. Fantoni, A. Palucci, S. Ribezzo

and H. J. L. Van der Steen, Appl. Phys. B 63 (1996) 185.[45] D. A. Cremers, Appl. Spectrosc. 41 (1998) 572.[46] R. Wisbrun, I. Schechter, R. Niesner, H. Schroder, K. L. Kompa, Anal. Chem. 66

(1994) 2964.[47] J. P. Singh, F. Y. Yueh, H. Zhang and R. L. Cook, Process Control and Quality 10

(1997) 247.[48] V. N. Rai, H. Zhang, F. Y. Yueh, J. P. Singh and A. Kumar, Appl. Opt. 42 (2003) 3662.[49] R. Barbini, F. Colao, R. Fantoni, A. Palucci, F. Capitelli, Appl. Phys. A 69 (1999) 175.[50] S. Yalsin, D. R. Crosley, G. P. Smith and G. W. Faris, Appl. Phys. B 68 (1999) 121.[51] V. N. Rai, J. P. Singh, C. Winstead, F. Y. Yueh and R. L. Cook, AIAA J. 41 (2003) 2192.[52] F. Y. Yueh, V. N. Rai, J. P. Singh and H. Zhang, AIAA-2001–2933, 32nd AIAA Plasma-

dynamics and Laser Conference, 11–14 June (2001) Anaheim, CA USA.




[53] S. Amoruso, Appl. Phys A 69 (1999) 323.[54] Z. W. Huang, Y. Y. Teng, K. P. Li, J. Sneddon, Appl. Spectrosc. 45 (1991) 435.[55] L. M. Cabalin and J. J. Laserna, Spectrochim. Acta. B. 53 (1998) 723.[56] R. E. Russo, Appl. Spectrosc. 49 (1995) 14A.[57] R. E. Russo, X. L. Mao, “Chemical analysis by laser ablation”, M. C. Millar (Ed.) Laser

Ablation and Desorption, Academic Press, San Diego (1998) pp. 375–412.[58] C. Chaleard, P. Mauchien, N. Andre, J. Uebbing, J. L. Lacour, C. Geertsen, J. Anal. At.

Spectrom. 12 (1997) 183.[59] B. C. Castle, K. Talabardon, B. W. Smith, J. D. Winefordner, Appl. Spectrosc. 52

(1998) 649.[60] H. T. Buscher, R. G. Tomlinson, E. K. Damon, Phys. Rev. Lett. 15 (1965) 847.[61] R. A. Bingham, P. L. Salter, Anal. Chem. 48 (1976) 1735.[62] F. Fabbro, E. Fabre, F. Amiranoff, C. Garbeau-Labaune, J. Virmont, M. Weinfield,

C. E. Max, Phys. Rev. A 26 (1980) 2289.[63] F. Dahmain, Phys. Fluid B 4 (1992) 1585.[64] S. Amoruso, M. Armenante, V. Berardi, R. Bruzzesse, N. Spinelli, Appl. Phys. A 65

(1997) 265.[65] L. M. Berman, P. J. Wolf, Appl. Spectrosc. 52 (1998) 438.[66] C. W. Ng, W. F. Ho, N. H. Cheung, Appl. Spectrosc. 51 (1997) 976.[67] X. L. Mao, A. C. Ciocan, O. V. Borisov, R. E. Russo, Appl. Surf. Sci. 127–129 (1998) 262.[68] K. L. Eland, D. N. Stratis, T. Lai, M. A. Berg, S. R. Goode and S. M. Angel, Appl.

Spectrosc. 55 (2001) 279.[69] K. L. Eland, D. N. Stratis, D. M. Gold, S. R. Goode and S. M. Angel, Appl. Spectrosc.

55 (2001) 286.[70] E. H. Piepmeier, “Laser ablation for atomic spectroscopy”, Analytical Applications of Lasers

(Ed. E. H. Piepmeier) Wiley, New york (1986).[71] C. D. Allemand, Spectrochim Acta B 27 (1972) 185.[72] A. M. Prochorov, V. A. Batanov, F. V. Bunkin, V. B. Fedorov, IEEE, J. Quant Electron.

QE-9 (1973) 503.[73] G. Dimitrov, Ts. Maximova, Spectrosc. Lett. 14 (1981) 734.[74] G. Dimitrov, Ts. Zheleva, Spectrochim. Acta B 39 (1998) 1209.[75] X. L. Mao, A. C. Ciocan, R. E. Russo, Appl. Spectrosc. 52 (1998) 913.[76] R. E. Russo, X. L. Mao, O. V. Borisov, H. C. Liu, J. Anal. At. Spectrom. 15 (2000) 1115.[77] Y. W. Kim., “Laser-Induced Plasmas and Applications” Ed. L. J. Radziemski and

D. A. Cremers, Marcel Dekker Inc. New York (1989)[78] F. Leis, W. Sdorra, J. B. Ko, K. Niemax, Mikrochim. Acta II (1989) 185.[79] C. Aragon, J. A. Aguilera, F. Penalba, Appl. Spectrosc. 53 (1999) 1259.[80] M. Sabsabi and P. Cielo, Appl. Spectrosc. 49 (1995) 499.[81] M. Sabsabi, P. Cielo, J. Anal. At. Spectrom. 10 (1995) 643.[82] A. S. Eppler, D. A. Cremers, D. D. Hickmott, M. J. .Ferris, A. C. Koskelo, Appl. Spectrosc.

50 (1996) 1175.[83] R. A. Multari, L. E. Foster, D. A. Cremers, M. J. Ferris, Appl. Spectrosc. 50 (1996) 1483.[84] P. Fichet, P. Mauchien, J. F. Wagner, C. Maulin, Anal. Chim. Acta 429 (2001) 269.[85] L. I. Sedov, “Similarity and Dimensional Methods in Mechanics”, MIR Publishers,

Moscow (1981).[86] M. Gatti, V. Palleschi, A. Salvetti, D. P. Singh and M. Vaselli, Opt. Commun. 69 (1988) 141.[87] D. E. Kim, K. J. Yoo, H. K. Park, K. J. Oh, D. W. Kim, Appl. Spectrosc. 51 (1997) 22.[88] H. C. Liu, X. L. Mao, J. H. Yoo, R. E. Russo, Spectrochim. Acta. B 54 (1999) 1607.[89] A. Ciucci, V. Palleschi, S. Rastelli, A. Salvetti, D. P. Singh, E. Tognoni, Nuovo Cimento-D

20D (1998) 1469.[90] Y. I. Lee, S. P. Sawan, T. L. Thiem, Y. Y. Teng, J. Sneddon, Appl. Spectrosc. 46 (1992) 436.




[91] R. Krasniker, V. Bulatov, I. Schechter, Spectrochim. Acta B 56 (2001) 609.[92] R. E. Neuhauser, U. Panne, R. Niessner, Anal. Chim. Acta 392 (1999) 47.[93] B. J. Marquardt, P. N. Stratis, D. A. Cremers, S. M. Angel, Appl. Spectrosc. 52 (1998) 1148.[94] R. E. Neuhauser, U. Panne and R. Niessner, Appl. Spectrosc. 54 (2000) 923.[95] A. I. Whitehouse, J. Young, I. M. Botheroyd, S. Lawson, C. P. Evans, J. Wright, Spec-

trochim. Acta. B 56 (2001) 821.[96] A. K. Rai, H. Zhang, F. Y. Yueh, J. P. Singh, A. Weisburg, Spectrochim. Acta B 56

(2001) 2371.[97] E. H. Piepmeier and D. E. Olsten, Appl. Spectrosc. 25 (1971) 462.[98] K. J. Grant and G. L. Paul, Appl. Spectrosc. 44 (1990) 1349.[99] W. Sdorra, K. Niemax, Mikrochim. Acta. 107 (1992) 319.

[100] Y. Iida, Spectrochim. Acta. 45 (1990) 1353.[101] J. Hermann, C. Baulmer Leborgne, D. Hong, J. Appl. Physics 83 (1998) 691.[102] T. L. Thiem, R. H. Salter, J.A. Gardner, Y. I. Lee, J. Sneddon, Appl. Spectrosc. 48 (1994) 58.[103] Y. Ito, O. Ukei and S. Nakamura, Anal. Chim. Acta 299 (1995) 401.[104] S. Nakamura, Y. Ito, K. Sone, H. Hiraga and K. L. Kaneko, Anal. Chim. 68 (1996) 2966.[105] K. J. Mason and J. M. Goldberg, Appl. Spectrosc. 45 (1991) 370.[106] K. J. Mason and J. M. Goldberg, Appl. Spectrosc. 45 (1991) 1444.[107] V. N. Rai, A. K. Rai, F. Y. Yueh and J. P. Singh, Appl. Opt. 42 (2003) 2085.[108] V. N. Rai, J. P. Singh, F. Y. Yueh and R. L. Cook, Laser and Particle Beam 21 (2003) 65.[109] J. Uebbing, J. Burst, W. Sdorra, F. Leis and K. Niemax, Appl. Spectrosc. 45 (1991) 1419.[110] R. Sattmann, V. Sturn and R. Noll, J. Phys. D, Appl. Phys. 28 (1995) 2181.[111] L. St. Onge, V. Detalles and M. Sabsabi, Spectrochim. Acta B 57 (2002) 121.[112] V. N. Rai, F. Y. Yueh and J. P. Singh, Appl. Opt. 42 (2003) 2094.[113] I. B. Gornushkin, S. A. Baker, B. W. Smith, J. D. Winefordner, Spectrochim. Acta B 52

(1997) 1653.[114] F. Hilbk Kortenbruck, R. Noll, P. Wintjens, H. Falk, C. Becker, Spectrochim. Acta B

56 (2001) 933.[115] H. H. Telle, D. C. S. Beddows, G. W. Morris, O. Samek, Spectrochim. Acta B 56 (2001)

947.



Chapter 5

Instrumentation for Laser-InducedBreakdown Spectroscopy

V. N. Raia and S. N. Thakurb

aLaser Plasma Division, Raja Ramanna Centre for Advanced TechnologyP. O. CAT, Indore 452 013, INDIA

bLaser and Spectroscopy Laboratory, Department of PhysicsBanaras Hindu University, Varanasi- 221 005, INDIA

1. INTRODUCTION

Laser-induced breakdown spectroscopy (LIBS) is a laser diagnostics, where a laser beamfocused onto a material generates transient high density plasma as the laser intensityexceeds the breakdown threshold of the material (∼1–10 MW/cm2). The UV and visibleemission from the plasma can be spectrally resolved and recorded for qualitative andquantitative analysis of the sample. LIBS was first used for the determination of elementalcomposition of materials in the form of gases, liquids and solids during 1960’s [1,2].Research on LIBS continued to grow and reached a peak around 1980 and field-portableinstruments capable of in-situ and real time analysis of samples have been developed inrecent years with the availability of reliable, smaller and less costly laser systems alongwith sensitive optical detectors, such as the intensified charge-coupled device (ICCD).Several review articles have been published on this topic [3–14].

A short duration laser pulse of sufficient energy focused onto the surface of a materialsample instantly increases its temperature above the vaporization temperature, regardlessof the type of material. Compared with the rate of energy delivery from the laser pulse,the energy dissipation through vaporization is relatively slow and the underlying layer ofmaterial reaches critical temperatures and pressures before the surface layer vaporizes,which forces the surface to explode. Generally material ablation and plasma formationtake place during the initial period of the laser pulse, whereas rest of the laser energyis absorbed by the ablated material to form luminous plasma. The temperature of theplasma emitting UV and visible radiation is in the range of 104 − 105 0K, whereasthe electron number density ranges from 1015 to 1019 cm−3 and the plasma-plume maylast from a few microseconds to several milliseconds. The laser-induced plasma maybe coupled with various detection systems, such as mass spectrometry (MS), atomicemission spectrometry (AES), atomic absorption spectrometry (AAS), and laser excitedatomic fluorescence spectrometry (LEAFS). In some experiments laser-induced plasma





is used as an atom or ion reservoir directly, or the ablated material is transported to asecond source for further excitation or ionization. The direct detection of atomic emissionfrom the LIP is referred to as laser-induced breakdown spectrometry (LIBS), which hasproved its potential in elemental analysis and feasibility for miniaturization.

LIBS has many advantages as an analytical technique. There is no need of samplepreparation, which avoids further contamination of the material to be analyzed [10–11].The analysis process is fast and can be used for both non-conducting and conductingsamples, regardless of their physical states, i.e. aerosols, gases, liquids or solids. LIBSis applicable to the analysis of extremely hard materials that are difficult to digest ordissolve, such as ceramics and semi/super-conductors as well as biological samples.Its capability for simultaneous multi-element determination, localized microanalysis,and surface analysis are also of great importance and it has been used successfullyin hazardous and difficult environmental conditions to study remotely located samplesfor online and real time information about their spectra. LIBS has been found usefulin elemental process monitoring and in field-portable analyzers for in situ trace metalanalysis of real samples, where accuracy and precision are not the main requirement [11].

2. TYPICAL LIBS SET-UP

Various types of LIBS experimental set up have been used which differ mainly inthe form of collection optics for the radiation emitted by the plasma plume. In one ofthe arrangements the emission from plasma is collected in the direction perpendicularto the direction of the incident laser. In addition to the difficulties of alignment andreduced sensitivity the collected emission exhibits spatial dependence leading to loss ofspectral information about emission from the whole plasma plume. These shortcomingsare removed in another arrangement where focusing lens itself acts as the collecting lensfor the plasma emission. In this case collected emission corresponds to the integratedvalue of radiation from all the spatial locations of the plasma. The typical schematicdiagrams of the experimental set-up [14–16] for recording the laser-induced breakdownemission from the solid is shown in Fig. 2 of chapter 4 and those for liquid and gaseoussamples are shown in Figs 1 and 2 respectively.

The LIBS experimental set up for studying solid samples consists of a Q-switched,frequency-doubled Nd: YAG laser (Continuum Surelite III) that delivers energy of∼300 mJ at 532 nm in 5-ns pulse. This laser was operated at 10 Hz and was focusedon the target with the help of a dichroic mirror and quartz-focusing lens of 20 cm focallength. The combination of dichroic mirror and the same focusing lens (Fig. 1) wasused to collect the optical emission from the laser-induced plasma. Two UV gradequartz lenses of focal lengths 100 mm and 50 mm were used to couple the plasmaemission to an optical fiber bundle. The fiber bundle consists of 80 single fibers of0.01 mm core diameter. The rectangular exit end of the optical fiber was coupled tothe spectrograph (Model HR 460, Instrument SA, Inc., Edison, NJ) and used as anentrance slit. The spectrograph was equipped with 1200 and 2400 lines/mm diffractiongratings of dimension 75 mm×75 mm. A 1024×256 element intensified charge-coupleddetector (ICCD) (Princeton Instrument Corporation, Princeton, NJ), with a pixel widthof 0.022 mm, was attached to the exit focal plane of the spectrograph and used to detectthe dispersed light from the laser-induced plasma. The detector was operated in gated



Instrumentation for LIBS 115

2XNd: YAG Laser

Data acquisition/Analysis system

Computer Controller

Spectrograph

Pulse generator

IDAD

L – Lens2x – KDP DoublerIDAD – Intensified Diode Array Detector

FO – Fiber Optics

BD – Beam Dump

HS – Harmonic Separator

DM – Dichroic Mirror

BDHS

LDMLLFO

Peristalticpump

Jet

Beaker

Solution

Beaker

Solution

Bulk liquid experiment

Liquid jet experiment

Prism

Lens

Fig. 1. Schematic diagram of experimental system for recording LIBS of liquid samples. In theseexperiments plasma is produced on the surface of bulk liquid or liquid jet.

(a)

Ultrasonicnebulizer

Lens

LIP

Beam dump

Dry metalaerosol

Laserbeam

Fig. 2. LIBS calibration system for gaseous samples (a) Open system (b) Closed system. Rest ofthe instrumentation is similar as shown in Fig. 1.




(b)

Ultrasonicnebulizer

Exhaust

Dry metalaerosol

Beam dump

Lens

Laserbeam

Window

LIP

Fig. 2. (Continued)

mode with the control of a high voltage pulse generator (PG-10, Princeton InstrumentsCorporation, Princeton, NJ) and was synchronized to the output of the laser pulse. Dataacquisition and analysis were performed using a personal computer. The gate delay timeand gate width were adjusted to maximize the signal-to-background (S/B) and signal-to-noise (S/N) ratios. Emission spectra were recorded mainly using 2400 lines/mm gratingfor a better spectral resolution. Around 100 pulses were accumulated to obtain onespectrum and 30 such spectra were recorded for each experimental condition in orderto increase the sensitivity of the system and to reduce the standard deviation in therecorded data.

3. LIBS INSTRUMENTATION

The principles of LIBS are similar to those of conventional plasma atomic emissionspectrometry, such as ICP-AES, microwave induced plasma (MIP)-AES, direct currentplasma (DCP)-AES, arc-AES and spark-AES. The main difference between LIBS andconventional AES is that there is no need to transport the sample to the plasma in LIBS.As discussed above, plasma is formed in or on the sample in situ by the use of a focusedlaser beam. LIBS instrumentation consists of three major parts: (i) a laser to generatethe LIP; (ii) a sample container (ablation chamber) to house the samples in an inertgaseous environment, under a vacuum, or simply in air; and (iii) a detection systemto collect, resolve and measure the atomic emission lines from the LIP. The detectionsystem usually consists of a dispersing element (a monochromator or a grating), anoptical detector, detection electronics, and a computer.

The basic purpose of a laser for LIBS is to produce sufficient and stable pulsedenergy to generate the plasma. Various lasers with wavelengths ranging from IR to UVregions of the spectrum have been used in LIBS and have been summarized by Leeet al. [13]. These include solid-state lasers such as the Nd: YAG laser (1064 nm, 532 nm,




and a pulse duration of 5–10 ns); and the ruby laser (693 nm, and a pulse duration of20 ns); gas lasers such as the CO2 laser (10�6 �m, and a pulse duration of 100 ns), andthe N2 laser (337 nm, and a pulse duration of 30 ps-10 ns); and excimer lasers (193 nm[ArF], 248 nm [KrF], 308 nm [XeCI], and a pulse duration of 10–20 ns). Among allthese, Nd: YAG lasers are the most widely used. The typical output energies for theselasers are tens of mJ to hundreds of mJ per pulse, and peak power is in the rangeof MW. These laser beams are focused to spots few tens of micrometers in diameterproducing 1010 −1012 W/cm2 irradiance. The characteristics of a laser, such as energy,energy stability, wavelength, pulse duration, beam quality, and mode quality, togetherwith the properties of the target material, affect the production and characteristics of theplasma. Typically, 100 mJ/pulse energy is sufficient to generate plasma for the analysisof most of the materials. With the use of laser wavelengths associated with gaseousatomic transitions of analyte elements, much lower laser energy at sub-mJ is adequate forthe generation of the plasma, where enhanced signals at the resonant wavelengths can beobserved, when the laser is scanned across them [17]. This technique is known as resonantlaser-induced breakdown spectrometry (RLIBS). RLIBS requires less laser energy andmay result in less spectral interference in real sample analysis, but it requires a tunablelaser, which makes the whole instrumental system more complex. Laser wavelength isa major influential factor for LIBS along with the laser energy. UV laser radiation forLIBS has advantage because of its low UV reflectivity from most of the metal surfaces,which usually leads to more efficient energy coupling and high optical resolution [3].

LIBS experiments can be performed in air, low-pressure inert gas, and vacuum.Some considerations concerning the construction of vacuum and gas chambers mainlyinclude two aspects: (i) to extend the analytical spectral range to the deep-UV region forelements such as carbon, phosphorous, sulfur, chlorine, bromine, iodine, oxygen, andnitrogen; and (ii) to improve the detectability by inert gas purging. Research on laboratoryapplications of LIBS is frequently carried out on a sample kept in vacuum chamber butfield applicable LIBS is usually performed in the atmospheric environment. Detaileddiscussions about sample chamber can be found in the literature [5,12]. The detectionsystem consists of focusing optics (used with an optical fiber cable for remote sensing),a dispersing element, a detector, signal processing electronics, and a computer for dataprocessing and storage. Many types of detectors have been used for recording the LIBS.In the early days, a photographic plate was used as a detector which had the advantage ofwide wavelength range with relatively low cost but had the disadvantage of being timeconsuming with low reproducibility. Photographic detection has been gradually replacedby detectors for spectrally resolved emission such as a photomultiplier tube (PMT),a photodiode array (PDA), or a charge-coupled device (CCD) which provide fast andaccurate measurements. The PMT is placed behind the exit slit of the dispersive unit,and produces a photocurrent proportional to the intensity of the incident radiation. Theproblem of this system is that many PMTs have to be used for simultaneous multi-elementanalysis. Optical multichannel analyzers can be used for multi-element determinations,but at the cost of more complicated instrumentation with a limited coverage of number ofelements. Photodiode array detectors are good for simultaneous multi-element analysis.In early studies of LIBS, Radziemski et al. [18] and Cremers et al. [19] have usedvarious types of PMTs sensitive over the range of 200–900 nm. They have also used atime-gated linear diode array coupled to a multichannel analyzer. The array consistedof 1024 diodes in 2.54 cm length, where each channel recorded the signal seen by one




diode during each shot. The array was time-gated by switching on the high voltageof the intensifier during the time of interest, typically 200 ns to several microseconds.Time gating was essential because of the strong continuum emission from the plasmaduring early stages (500 ns) after the plasma formation. The system was sensitive for thewavelength range of 350–800 nm. In fact, PDAs [20–23], gated and/or intensified andmultichannel analyzers [24–25] are now frequently used as detectors in LIBS. However,sensitive and imaging scientific CCD cameras as detectors have been gaining morepopularity in recent years and can be time-gated to isolate temporal intervals, duringthe evolution of the plasma, for optimum signal measurements in different applications.It has been reported that CCD is nearly three orders of magnitude more sensitive thanPDA [26], but the dynamic range is slightly less than that obtained with PDA. However,the spectral range is more limited because the light is detected through a transparentelectrode. Castle et al. [27] and Zhang et al [16] used an ICCD as detector for LIBSwhere a programmable pulse delay generator was used to gate the ICCD to obtain anoptimum signal-to-background noise ratio and the firing of the laser and data collectionwere under the control of a computer. In another case, Castle et al. [28] employed a linearCCD as the detector which was non-gated, but had the provision of external triggeringand a timing circuit was designed to control laser firing and data collection. This linearCCD had 2046 pixels, and the system covered the spectral range of 339–462 nm.

3.1. Echelle Spectrometer

Czerny- Turner spectrographs are usually employed to disperse the emission collectedfrom LIBS plasma, where suitable detectors coupled to it offer the possibility of timeresolved measurements. These detection systems are intrinsically limited in resolutionas well as in spectral coverage. The multi-elemental detection capability offered bythe LIBS technique demands a spectrograph with a wider spectral coverage. In themulti-elemental analysis, sequential measurements of parts of the spectrum of interestare performed, inspecting each time a different sample of the material ablated from thetarget surface. In principle this procedure limits the LIBS application to homogeneoussamples but most of the samples are inhomogeneous, which is why the spectra varyfrom shot to shot, as a result of changes in the sample composition as well as due tostochastic fluctuations in the plasma [29]. Therefore simultaneous measurement of thecomplete optical spectrum is necessary for getting optimum information for analyticalpurposes. Instruments, which allow simultaneous measurements, are Paschen-Rungespectrometers or the more compact Echelle spectrometers as discussed in a reviewarticle by Detalle et al. [30]. Echelle spectrometers offer excellent spectral resolvingpower (�/�� ≥ 10�000 and more) in combination with a spectral coverage of severalhundred nanometers. In combination with intensified charge coupled devices, Echellespectrometers represent a very powerful tool for elemental analysis as demonstrated byHaisch et al. [31] who found substantial improvement in the detection limits obtainedfor several elements with an Echelle system as compared with those obtained with aconventional Czerny- Turner system.

The principle of Echelle spectrometer has been described by Detalle et al [30]. It hasfocal length of 25 cm with a numerical aperture 1:10 and a quartz prism positioned in frontof the grating separates the different orders of spectra and produces a two dimensional




pattern. The flat image plane is 24�85 × 24�5 mm2. This system provides maximumresolution in the wavelength range between 200 to 780 nm. The linear dispersion perpixel ranges from 0.005 nm (at 200 nm) to 0.019 nm (at 780 nm), which is based on thespectral resolution �/�� = 40� 000. The detector in this system is an ICCD camera,having a CCD array of 1024 × 1024 pixels �24 × 24 �m2� and a microchannel plate.A fast pulse generator delivers a 5 ns pulse to the intensifier to ensure synchronizationof the measurements with the laser pulse. The spectral response in a particular orderof diffraction of Echelle spectrometer is non-linear, when measured using the black-body radiation from a deuterium lamp and maximum sensitivity is found in the centerof the given order. Each diffraction order has similar shape but a different sensitivitywhich requires a correction factor when the measurement is made in different spectralrange with different sensitivity. Normally a black-body radiation calibration spectrumis recorded to obtain the intrinsic response of the Echelle/ICCD system, which is thenused to normalize the acquired spectrum.

3.2. Specialty of Echelle Spectrometer

In recent years the Echelle spectrometer has proved to be very successful in the acqui-sition of spectral data from which relevant physical or chemical information can beextracted. Calibration curves of various elements have been obtained, limits of theirdetection determined, and the excitation and ionic temperatures of laser induced plasmaas well as the electron density have been measured. The reproducibility of experimentalresults show that the dynamic range of the detector makes it possible to simultaneouslymeasure the intensities of spectral lines of the major elements and the trace elementsin the sample. The detectability of elements at low concentration is facilitated by thevery high resolution of this system. One has to be aware of the advantages as well asshortcomings of this system for its judicious application.

3.2.1. Advantages

The main advantage of the very compact Echelle system is its excellent resolution, whichis comparable to that of a Czerny- Turner spectrometer of 1-m focal length having agrating of 3600 lines/mm. This resolution is useful in avoiding spectral interference. Thesecond important feature of this system is its coverage of a very broad spectral range(200–780 nm), which makes it possible to record several lines of the same element.Thus, a large dynamic range of concentrations can be measured, because a saturated linecan be replaced by another one of the same element in the spectrum. The system hasbeen found appropriate for the analysis of complex matrices, such as alloys containinga great number of elements. The broad spectral range, combined with the good spectralresolution of the system is also found suitable for multi-elemental analysis. An Echellespectrometer is an excellent tool for instantaneous recognition of the elements present inan unknown sample as long as the element emits in the plume of laser-induced plasma.The multi-elemental analysis in various types of samples, solids (conductor or not),liquids or gases can be easily carried out with an Echelle spectrometer and it has beenused in classifying different types of alloys of varying characteristics [32].




The main difficulties faced by the LIBS technique are the variation in the intensitiesof spectral lines, which depend on the complex process of laser-matter interaction.The solution often used to solve this problem is internal standardization. Generally,the ratio of the line intensity of the impurity or trace element to that of the majorelement of the matrix is used in determining the concentration of the trace element.Some research groups showed the possibility of normalizing the analyte signals usingthe variation in the excitation temperature of the plasma [33] that is difficult to realizewith a conventional spectrometer, but becomes easy with an Echelle system. One canalso follow the evolution of emission spectra of all the elements present in the matrixand carry out a compositional assessment considering that the sum of concentrations isalways 100%.

3.2.2. Limitations

The mode of dispersion in an Echelle spectrometer involves a change in resolution withwavelength and the response of the intensifier is not equal throughout the spectral range(maximum at ∼400 nm). Thus spectral sensitivity becomes low for wavelengths above600 nm resulting in a loss of luminosity and resolution of the system for the longestwavelengths. The Echelle dispersion also involves loss of zones of the spectrum locatedbetween the different orders, called dead zones. Although this has little consequence forthe ultraviolet region, the dimensions of these dead zones increase with wavelength toreach several nm beyond 700 nm. Thus, some spectral lines falling in the dead zone maynot be accessible. As discussed above the Echelle dispersion is non-linear even in thesame order of dispersion which makes it necessary to correct the observed spectra withthe help of a reference spectrum.

Although a CCD detector can receive only a limited number of electrons on eachof its pixels. With increase in the gain or the gate-width the emission lines with verylarge intensity may saturate some pixels of the CCD. The electrons accumulated onthe saturated pixels jump, mainly onto the closest pixels located below leading to thephenomenon of blooming discussed by Detalle et al. in the case of Al emission lineat 308.22nm [30]. The pixels located below the saturated one correspond to a differentorder and different wavelengths get affected by the signal of the saturating emissionline. Detalle et al. observed the appearance on the rebuilt linear spectrum of a linecorresponding to the theoretical position of Nb (316.37 nm), whereas the sample wascompletely free of this element. The spectral lines observed as a consequence of bloomingare called ‘ghost lines’ and to avoid them the experiments must be performed under theconditions of non-saturation. The dynamic range of the system thus presents a limit, notdue to a lack of gain of the intensifier, but due to the impossibility of eliminating thesignal from the major lines.

Another limitation of the Echelle system is related with its low rate of data acquisition.The CCD detector has a dimension of 1024 × 1024 pixels, whereas the rate of transferof the PC board is 500 kHz for 16-bit resolution. The transfer time for all data fromthe CCD is thus higher than 2 seconds and one acquisition is actually possible every 3seconds. Therefore, the present configuration of the system does not make it possible tocarry out rapid sampling. Of course, the speed of analysis is a relative concept, but it isone of the main factors in the choice of one technique or method over another.




4. FIBER OPTIC LIBS

LIBS is most suitable for field based industrial applications, which include real time, online analysis of material for process control and monitoring. Most of the experimentaltechniques discussed so far are laboratory based, where plasma is generated by focusingthe high intensity laser beam on the sample surface with an assembly of lenses andthe light emitted from the plasma is collected by either the same assembly of lensesor a separate assembly of lenses to be focused on the entrance slit of the spectrometerfor further analysis [10]. Such an experimental set up is not well suited for fieldmeasurements, which require a flexible optical access to the test facility and minimalon-site alignment.

Recent advances in fiber optic materials have opened up many new areas of appli-cations for the LIBS technique. Using optical fiber, we can send the laser beam to thedesired location and perform remote measurements. The low breakdown threshold ofthe optical fiber material did not permit delivery of laser radiation on the target andapplications of optical fiber in LIBS were initially limited to delivering the plasmaemission to the detection system [16,23,34]. The next development in LIBS studies wasthe use of two optical fibers, one of the optical fibers was delivering the laser beam forcreating spark by focusing the laser radiation on the surface of sample, and other onewas collecting the radiation from the spark emission. Adjustment of two optical fibersis a very delicate and difficult task, especially when LIBS is being used in harsh andhazardous environmental condition, such as those found in aluminum, glass and steelindustries. Therefore, it is more desirable to use only one optical fiber to transmit thelaser beam as well as to collect the emission from laser-induced plasma [35–36].

A simple and robust fiber-optic probe that uses one optical fiber both for deliveringthe laser power to produce a spark and for collecting the resulting radiation from thespark for quantitative elemental analysis with greater accuracy and a lower detectionlimit has been developed [33]. In order to obtain maximum emission intensity anda better signal-to-background ratio, several parameters have to be optimized such asdetector gain, damage threshold of fiber optics, focal lengths of different lenses, gatedelay and gate width, sample surface, etc.

4.1. Fiber Optic LIBS Probe

A schematic diagram of the fiber optic LIBS probe is shown in Fig. 3 [11–36]. The secondharmonic (532 nm) of a pulsed Nd: YAG laser (Big Sky, Model CFR 400) operating at10 Hz, with pulse duration 8 ns, beam diameter 7 mm, and the full angle divergence 1.0m rad was directed into the optical fiber by a 532/1064-nm beam splitter and a 532 nmdichroic mirror. A specially coated 45� dichroic mirror (DM), which reflects at 532 nmand transmits in wavelength ranges 180–510 nm and 550–1000 nm, was used to reflectthe laser beam and to transmit the LIBS signal to the detection system. This simpledesign avoids any damage to the detector from the reflected laser light. To transmitsufficient laser energy below the damage threshold of fiber optic cable, the laser beamwas focused 5 mm away from the fiber tip via a 10 cm focal length lens. A cap with a0.8-mm pinhole was placed at the fiber input end to avoid the possibility of damage tothe core and cladding of the fiber. The laser beam transmitted through the optical fiber




Nd: YAGLaser

Fiber optics

Cap with a 0.8 mm pinhole

Dichroicmirror

LensLens LensDichroicmirror

Pulsegenerator Computer

Fiberbundle

ICCDSample

Spectrograph

Detectorcontroler

Fig. 3. Schematic diagram of fiber optic probe for remote analysis of LIBS signal from a sample.Similar system is used for recording LIBS from samples submerged under water or molten liquidsamples in industries. In the molten metal case optical fiber is passed through a ceramic tubeand dry air, N2 or argon gas is passed through it for creating a bubble at the place of plasmaformation. In some cases fiber is kept close enough to the sample for creating plasma without afocusing lens.

was collimated with a 10 cm focal length lens and then focused on the sample with a5 cm focal-length lens. The same lenses and optical fiber assembly were used to collectthe emission from the laser-induced plasma and the collimated emission was passedthrough the dichroic mirror and focused onto an optical fiber bundle with a 20-cm focallength lens. The fiber bundle, a round-to-slit type, consists of 78 fibers, each havinga 100 �m diameter and a 0.16 numerical aperture (NA). The slit-type end of the fiberbundle delivers the emission to the entrance slit of a 0.5-m focal length spectrometer(Model HR 460, Jobin Yvon-SPEX) equipped with a 2400 lines/mm grating blazed at300 nm. An intensified charge coupled device (Model ITF/CCD, Princeton Instruments)was used as the detector with its controller (Model ST 133, Princeton Instruments).A programmable pulse delay generator (Model PG-200, Princeton Instruments) wasused to gate the ICCD and the data acquisition was under the control of a computer(Dell Dimension M 200a) running the WinSpec/32 software (Princeton Instruments).Multiple (100) laser shots spectra were stored in one file, where fifty such spectra wererecorded for analysis to get an average area/intensity value for the spectral lines underinvestigation.

LIBS spectra of different Al alloys recorded with an optical fiber probe were correctedfor baseline spectral intensity by integrating the peak area under each line [11,36]. Thequantitative spectral analysis involves relating the spectral line intensity of an element inthe plasma to the concentration of that element in the target. The most important minorelements in Al samples were analyzed, which included copper, magnesium, manganese,nickel, chromium, and iron. To optimize the signal for the quantitative analysis of theseelements in the aluminum alloys, the LIBS spectra were recorded by changing the




various experimental parameters (laser energy, sample surface, detector gain, gate delay,width, etc.).

4.2. Transmission Property of Optical Fiber

The LIBS system used by Rai et al [36] had a silica core/silica cladding multimodefiber (FG-I.0-UAT from ThorLabs Inc.). The stability of silica cladding allows thefiber to transmit high laser energy and the low OH silica-core design provides superiorUV transmission, required to transfer the LIBS signal. The length of the fiber was ∼3 m,and it was fitted with SMA 905 stainless steel fiber connectors (ThorLabs Inc.) at boththe ends. The fiber was polished with a 0.3-mm grain size aluminum oxide powder in thefinal step. The core and cladding diameters were kept as 1 and 1.25 mm respectively andits maximum power capability was ∼5 GW/cm2. The low numerical aperture of 0.16enables the fiber to produce low beam divergence and uniform spot size that facilitatesfocusing the beam after transmission through the fiber. A spherical plano-convex fusedsilica lens of 10 cm focal length was used to couple the laser beam into the fiber. Withthis lens, a 30-mJ-laser beam causes breakdown in air, which is the maximum laserenergy that can be transferred through the fiber. A metal cover with a 0.8-mm pinhole atthe center was placed just in front of the fiber to avoid any damage to the boundary ofthe core cladding during alignment. The fiber was placed about 5 mm behind the focalpoint in the diverging beam, where only about 0.6–0.7 mm of the core diameter wasilluminated. A simple calculation indicates that a 30-mJ-pulse energy with a spot size of0.5-mm diameter would produce an energy density of ∼2 GW/cm2. This is lower thanthe damage threshold of the fiber. However, at this laser energy level, damage to theinput surface of the fiber is still possible by randomly occurring hot spots in the laserprofile. This system had an energy transmission efficiency of about 88%, which is fairlyhigh. In order to improve the S/B ratio, various experimental parameters were testedand the optical fiber was damaged several times during these tests. Most of the damageoccurred inside the fiber when the laser energy input was more than 20 mJ. Later it wasfound that as long as the laser energy remains below ∼20 mJ at the fiber input end, thefiber does not get damaged and most of the experiments were performed by using laserenergy below this threshold The core cladding was also damaged due to breakdownseveral times and it occurred between 2 to 5 cm away from the fiber input end, verylikely at the location of the first reflection of the laser beam inside the fiber. Precautionsare therefore necessary to avoid any damage in the optical fiber during the experiments.It was noted that while the laser power at the input end remained approximately constantwith time (30–40 minutes), the laser power decreases slightly at the output end of thefiber. This suggests that the recording time for one set of data should be kept as shortas possible if the laser energy is near threshold.

5. PORTABLE LIBS DEVICES

In recent years, miniaturizing the LIBS instrumentation has become a necessity becauseof efforts to move LIBS systems for field applications and for at-site analysis, particularlyof environmental samples [37]. This has been possible by use of optical fiber to deliver




a laser beam from a miniaturized laser system providing the radiation necessary for theplasma production. LIBS instruments using fiber optic cables have gained popularity forlocal and remote sensing in hostile environments, such as urban and industrial dumps orheavily polluted zones [24,38]. The fiber optics system is used both to deliver the laserradiation to the sample and to collect the emission signal from the laser-induced plasmaas discussed above, which allow in-situ analysis of relatively inaccessible samples, suchas painted surfaces and nuclear reactors. Barbini et al. [34] and Ciucci et al. [25] haverecently developed a mobile instrument to apply the LIBS technique in conjunction withlight detection and ranging (LIDAR) apparatus. The single-laser-shot mobile device,equipped with the relevant data analysis software, was able to provide a real-timeresponse regarding the presence of hazardous species (such as antimony, barium, copper,chromium, lead, and mercury) in a variety of polluted environmental solids (such asrock, soil, sand, and ashes). A signal-to-noise ratio of 2 was obtained for mercury witha concentration as low as 80 ppb. This type of LIBS instrument is suitable for remotesensing and on-site analysis in some cases.

The fiber optic probes mentioned in previous sections are not truly field-portable LIBSinstruments, due to the non- portability of the lasers employed and the length of opticalfibers which is not unlimited {several meters in the above cases) due to signal and laserradiation attenuation in the optical fibers. A real field-portable LIBS instrument can onlybe realized by using a battery power supply, optical fibers and a miniature laser. Sucha device was first developed in the research group of Cremers at Los Alamos NationalLaboratory [23]. It had a weight of 14.6 kg and a compact size of 46 × 33 × 24 cm3 tofit into a small suitcase. The hand-held probe employed a compact, low cost, passivelyQ-switched Nd: YAG laser for making it portable. The laser had a low pulse energy(15–20 mJ/pulse at 1064 nm, 4–8 ns duration) and repetition rate �<1 Hz�, but it had theability to operate from 12 Volt D.C. batteries. A spark was produced on the sample byfocusing the laser with a 50-mm focal length lens of 12 mm diameter. A fused silicafiber optic bundle of 2 m length was used to collect the emission from the LIP andtransmit it to a 1/8-m spectrograph. The end of the fiber optic bundle was positioned5 cm from the LIP. It was not necessary to focus the LIP light onto the fiber with alens due to the already sufficient emission collected in this configuration. The spectrallyresolved emission was recorded with a compact CCD system. A compact computer wasused for data processing and storage. The performance of the portable LIBS device wascompared with that of a laboratory-based system using lead-containing paint samplesand soil samples containing barium, beryllium, lead, and strontium. The results wereidentical in all aspects, indicating that downsizing the instrument did not affect itsanalytical performance. The same samples were also analyzed with ICP-AES and withportable X-ray fluorescence (XRF) and it was found that results were in agreement witheach other.

A more compact portable LIBS instrument has been built in Winefordner’s researchgroup at the University of Florida [28] using rechargeable batteries. This device isuseful for field application, where regular power supplies are not available. It consistsof a Kigre Nd: YAG laser (1064 nm, 21 mJ, 3.6 ns duration, 1/3 Hz), spectrometercoupled with detector, computer, electronics, optics, and a rechargeable battery, allarranged in a suitcase of 48�3×33×17�8 cm3, with a total weight of 13.8 kg. An averageincident laser power density of 0.92 GW/cm2 was achieved for the production of theLIP. A linear CCD (2046 pixels) was used as the detector having a spectral range




of 339–462 nm. This system was optimized using the battery power supply, and afterparametric studies of spatial development of plasma, lens-to-sample distance, and spatialfiltering its performance was evaluated on paint, steel, and biological samples for thedetermination of lead, manganese, and calcium.

A portable LIBS instrumentation has shown advantages [9,23] over portable X-rayfluorescence (XRF) system, which is currently the choice for many types of fieldscreening measurements. In the case of LIBS one may choose different analytical linesto avoid spectral interferences. A portable LIBS device provides analytical results threeto 30 times faster than a portable XRF unit for homogeneous samples. The microprobecapability of LIBS makes it possible to analyze very small samples (e.g. lead in solderjoint) and uneven surfaces (e.g. irregular rock surfaces and uneven ground surfaces).Finally a LIBS probe can provide fiber optic delivery of the laser pulses to sub-surfacesoil for remote, in situ monitoring, which is not possible with XRF.

6. SENSITIVE LIBS TECHNIQUES

It has been realized that poor detection limit is the most serious limitation of the LIBStechnique in comparison to other analytical techniques. Several research groups havemade modifications in the experimental setup to improve the LIBS detection limits.The combined use of a pair of laser pulses to ablate the material and further excite theresulting plasma to enhance the sensitivity of LIBS has been found most promising. Thistechnique is known as Dual-Pulse LIBS, or Repetitive Spark Pair (RSP), or Double-PulseExcitation. In some cases the dual pulse is delivered by a unique laser [39–40] whilein other experiments use of two different lasers has been reported [41–45]. The secondtechnique is more flexible with spatial arrangement of the two laser beams, their pulseenergies and the time delay between two pulses. The use of a single laser, however,makes the system more compact and avoids the problems of alignment between the twolaser pulses, ensuring better reproducibility.

In 1997 Pichahchy et al. [39] applied the RSP technique to analyse the metals underwater. The use of two laser pulses separated in time by tens of microseconds producedhotter plasmas (90000 K compared to 30000 K for a single spark) and resulted in betterdetection limits for the elements. The reason of the strong excitation produced by thesecond pulse was attributed to the formation of a gas bubble in the water at the solidsurface. The presence of a solid-gas interface would allow the interaction of the secondpulse with the plasma in a way similar to the gaseous environment.

In 1998 St-Onge et al. [44] reported the study of some parameters affecting theperformance of dual-pulse LIBS on metal samples in air. They found that the volumeof the emitting plasma increases under the effect of the second laser pulse resulting insignal enhancement. This is due to more uniform absorption of the second laser pulse,whose energy is then distributed over a larger volume. St-Onge et al. [45] also used aUV laser pulse to increase the ablation of the sample and an IR laser pulse to maximizethe heating efficiency. In this configuration a significant signal enhancement was notedthe extent of which varied depending on the ionization state and energy levels giving riseto the spectral line of interest. A correlation has been established between the observedincrease in intensity and the theoretical increase expected as a result of the higher plasmatemperature generated by a combination of the UV-IR pulses. The enhancement in the




signal was found greater than predicted by the increased temperature. This shows that anincrease in plasma volume is also contributing the enhancement in intensity of emission.In contrast to the above-mentioned studies, where the plasma obtained by the firstablating pulse is reheated by a second pulse, Stratis et al. [43] used a pre-pulse parallelto the sample surface and focused it to form air-plasma, followed by a second ablatingpulse perpendicular to the sample surface and delayed in time by a few microseconds. Inthis case increase in intensity of the spectral line was correlated with an increased massablation. They simultaneously measured the time-resolved, spatially integrated emissionintensity from two directions- perpendicular to the target surface; and parallel to the targetsurface- resulting in a slight difference, which indicates the importance of the collectiongeometry in the LIBS measurements. Rai et al. [46] used two lasers (Nd: YAG) forLIBS experiments and the spatially integrated emission in the direction opposite to thedirection of the laser beams was collected (Fig. 4). LIBS signal was enhanced by morethan 6 times, when the time separation between two laser pulses was ∼2–3 �s. Smithet al. [47] used a different technique to improve detection limits by applying selectiveelemental excitation with a tunable diode laser to the LIP. Tunable diode laser inducedatomic fluorescence was used for selective isotope detection of uranium containingsamples. In order to detect the fluorescence signal, two techniques were employed:(i) the fast wavelength scanning of the diode laser during the lifetime of the plasmaproduced by each shot of the ablating laser and (ii) the time-integrated measurement withthe diode laser wavelength fixed at the isotope line center. The optimal experimentalconditions were found by means of a systematic scanning of the pressure of argon in theexperimental chamber. The limit of detection in the optimal conditions was of the orderof 0.6 ppm. Many other combinations of LIBS and laser-induced fluorescence (LIF)

Nd: YAG Laser

FO

L L DM

Liquid jetTFP

M

M

MM

HS

BD

P-polarized

S-polarized

Beam dump

BDHS

TFP – Thin-Film Polarizer

DM – Dichroic Mirror FO – Fiber Optics

HS – Harmonic Separator M – Mirror

L – Lens 2x – KDP Doubler

Triggerpulsegenerator

BD – Beam Dump

2x

Nd: YAG Laser 2x

L

Fig. 4. Schematic diagram of optical system for recording the LIBS spectrum of liquid sample(Jet)in double laser pulse excitation mode. Rest of the system such as dispersing device, detectors anddata acquisition components are same as shown in Fig. 1.




have been reported in recent years [48,50] to enhance the performance of LIBS as anextremely sensitive technique.

7. VARIETY OF LIBS INSTRUMENTATION

LIBS is a versatile technique for detection and identification of elements in a varietyof samples that cannot be easily analyzed by other spectroscopic methods. Each one ofthese situations requires a modification of the standard LIBS instrumentation to givethe best results. In the following sections we describe some of the unusual experimentalarrangements.

7.1. Environmental Monitoring

7.1.1. Off gas emission

The detection of hazardous and toxic trace elements in the off gas from waste processingsystem is very important for public health. LIBS has been used for in-situ off gasmonitoring by focusing the laser beam in the gas stream through a window and collectingthe optical emission through an optical fiber. Neuhauser et al. [51] have tested an online lead (Pb) aerosol detection system with aerosol diameters ranging between 10 and800 nm and a detection limit of 155 �g m−3 has been achieved. LIBS has also beendemonstrated as a process monitor and control tool for waste remediation [15]. The toxicmetals from three plasma torch test facilities were monitored and it was found that LIBScan be integrated with a torch-control system to minimize toxic metal emission duringplasma torch waste remediation. The possibility of using metal hydride to calibratemetals in off gas emission was also investigated [52] by using a static sample cell toperform LIBS measurements and the signal was found to be affected by gas composition,gas pressure and laser intensity.

The use of LIBS as continuous emission monitor (CEM) requires the quantitativetrace level determination of the toxic metals. A system has been developed to monitorthe concentration of selected toxic metals in near real time [16]. The concentrations ofBe and Cr were measured at all the tested metal levels while that of Cd was measuredduring medium as well as high metal feed tests and the concentration of Pb was measuredonly at high concentrations. It was concluded that the LIBS system can be used as aCEM to monitor only the concentrations of Be, Cr and Cd but further improvements inthe sensitivity of this system were needed for the monitoring of Pb, Hg, As and Sb.

7.1.2. Study of soil, concrete and paint

Detection of contaminated soil and concrete is an important area of environmentalapplications of LIBS. Yamamoto et al. [23] used a portable LIBS system to detect toxicmetals in soil and detection limits of Ba, Be, Pb and Sr were found to be 265, 9.3,298 and 42 ppm, respectively. Cremers et al. [38] detected Ba and Cr in soil using anoptical fiber probe for remote operation with limit of detection of 26 ppm and 50 ppmfor Ba and Cr, respectively. The effect of matrix was also studied in a soil sample.




The limits of detection for Pb and Ba in a sand matrix were found as 17 and 76 ppm (byweight), respectively with a precision of 7% RSD whereas those in the soil were 112and 63 ppm respectively with 10 % RSD. The LIBS signal was found to be affected bychemical speciation as well as matrix composition and its accuracy could be degradedif calibrations were not matrix specific.

Pakhomov et al. [22] have applied LIBS for the detection of Pb in contaminatedconcrete. A time resolved LIBS spectrum was recorded for the quantitative measurementof the Pb content in concrete. Pb calibrations were obtained by using the ratio of theintegrated emission of lead line (405.78 nm) and that of an oxygen line (407.59 nm). Itwas found that the absolute Pb signal was independent of the laser pulse energy for laserenergy between 250 and 400 mJ. The presence of Pb in the paint is a potential healththreat, especially to children and Yamamoto et al. [23] have successfully demonstratedthe feasibility of using LIBS to determine Pb in the paint surface.

7.1.3. Study of radioactive elements

LIBS has also been used to monitor the level of radioactive elements in a process stream.Watcher and Cremers [53] found a detection limit of l00 ppm for uranium in solution.LIBS is preferable to other radiological measurements because nuclear detector may notbe able to differentiate the radionuclides U, Pu and Np. The LIBS spectra of U, Pu an Npwere recorded in a globe box and the emission lines suitable for the detection of theseradioactive elements were identified by Singh et al. [54]. The preliminary studies showthat LIBS is suitable for the measurement of radioactive elements in waste stream. LIBShas also been used as a tool for detection of radiation embrittlement [9] in a nuclearpower plant by determining the copper concentration in A533b steel. As copper is a keyimpurity contributing to radiation embrittlement, the Cu concentration in the steel maybe an indicator of radiation embrittlement and expected material lifetime.

7.2. LIBS in Space Research

7.2.1. Rocket engine health monitor

Detection and characterization of metallic species in the exhaust plume of hydrocarbon-fueled rocket engines can indicate the onset of wear and/or corrosion of metal in therocket engine. This information on engine wear obtained during engine operation isvery useful, allowing the possibility of engine shutdown before any catastrophic failure.It has been observed that a catastrophic engine failure is generally preceded by a brightoptical emission, which results from the erosion of metal from the engine parts. This isbecause of high temperature in the rocket plume �∼2000 K�, which partially vaporizesand atomizes the metal species, leading to atomic emission in the near ultraviolet andvisible region (300–760 nm). The performance of LIBS was evaluated by Rai et al. [55]in detecting the trace of elements in the fuel plume of a hybrid rocket engine simulatorat Stennis Space Center, USA. Copper wire was inserted in the ignition chamber ofengine and its vaporized trace was recorded in the rocket plume outside the exit nozzle.The trace of copper was recorded near the nozzle exit during an initial fraction of asecond, when the burnt-fuel plume started building up. However it decreased when plume




attained its full length, high temperature and high speed. It was interesting to note thatcopper emission was observed throughout the plume away from the exit nozzle as wellas in the luminous zone. This observation was attributed to better mixing of the metalvapor (away from the exit nozzle) along with decrease in background emission (due toluminous zone). It was found that the measurements made away from the luminous partof the plume could provide more meaningful information about the health of the rocketengine.

7.2.2. Probe for Mars expedition

Cremers et al [56] are involved in evaluating the use of LIBS for future use on Landerand Rover to Mars. The main interest is the use of LIBS for stand off measurementsof geological sample nearly up to 20 meters from the instrument. The objective is todevelop a very compact instrument operating at a remote distance from the target todetect at least 10 species in the rocks with detection limits <100ppm including Ba, Li,Rb, and Sr with detection limit <20ppm. The ability to measure these separately in dustand pristine rocks is required. Minor and trace element composition are also importantin determining the provenance of rocks and dusts. In an experiment under the simulatedMartian atmospheric condition (5–12 mbar CO2), it was noted that bulk matrix affectedthe calibration for Sr. Accuracy and precision were obtained in the detection of variousother elements. Another project called MALIS (Mars elemental analysis by laser inducedbreakdown spectroscopy) is also in progress to demonstrate LIBS capability in Martianatmospheric condition [57].

7.3. LIBS in Industry

Now a-days all the metal producing industries are facing a major challenge of increasingproductivity at reduced cost and maximizing the benefits from existing equipment.During refining, it is critical that operating parameters be adjusted and controlled so thatthe chemistry of the molten metal remains within predetermined limits. LIBS has beensuccessfully used to get the composition of alloys along with their quantitative analysisin solid as well as in molten state [36,58–59]. The analyte lines of Cu, Cr, Mn, Fe andZn were used to obtain the calibration curves for their quantitative analysis. This type ofanalysis was performed by recording the spectra of alloy samples in the laboratory butLIBS has been found most suitable for field-based industrial applications, which includereal time and online analysis of molten material for process control and monitoring.Many groups [60–64] have used LIBS probe that uses a single optical fiber for deliveringthe laser pulses to the target at a remote place for producing a micro-plasma as well asfor collecting the resulting radiation from the LIP for quantitative elemental analysis.

8. LIBS EXPERIMENTS & ANALYTICAL PERFORMANCE

The observed analytical figures of merit (precision, accuracy, and LOD) in LIBS exper-iments are inferior to those other atomic spectrometric techniques such as ICP-AES andICP-MS. The main reason for this deficiency is the multitude of experimental parameters




that influence the analytical signal. These factors include laser wavelength, laser power,incidence angle, pulse-to-pulse variation, beam profile, beam shape, pulse duration, sam-ple matrix, freshness of surfaces for solid samples, purging gas, ambient pressure etc.Even the experimental arrangement and sampling geometry affect the LIBS measure-ments significantly [65]. These factors have to be identified and optimized for betteranalytical performance. Spatial and temporal dependence of emission signals from thelaser- induced plasma (LIP) have been studied by several research groups [66–67]. It hasbeen realized that the investigation of methods for measuring relative mass removal inthe laser-induced plasma would probably continue for a long time and will be valuable,not only for LIBS experiments, but also for LIP-ICP-AES and LIP-ICP-MS experiments.It is expected that the fundamental studies of ablation and the excitation processes in thelaser-induced plasma using different wavelengths and ultra-short pulses would enhancethe analytical capabilities of LIBS. A good review on these aspects has been publishedby Rusak et al. [7].

The major limitations of LIBS for practical applications result from self-absorption,line broadening, and the high intensity of the background continuum along with strongmatrix effects [68]. Some of these limitations can be minimized or avoided by workingin a controlled atmosphere and using time-resolved spectroscopic measurements or time-integrated and spatially resolved measurement techniques. In time-resolved spectroscopy,the temporal evolution of the plasma is obtained by recording the plasma emissionspectra at various delay times. The LIBS spectra of magnesium in liquid matrix werereported by Rai et al. [69] who found that 500 ns after the laser irradiation, the observedspectrum consisted of continuum and ion emission lines, as the plasma temperature washigh, but after 10 �s the intensity of the continuum and ion lines decreased, and that oflines due to neutral atoms increased as a result of electron ion recombination.

The purpose of time-integrated and spatially resolved spectroscopy is to measurethe emission from the LIP at different positions of the plasma. Lee et al. [12,21,70]have carried out experiments on copper and lead using ArF, XeCI and Nd: YAGlasers. They found that the plasma consisted of two distinct regions when the ambientpressure was reduced below 50 torr in air or argon atmosphere. The region near thetarget surface referred to as the inner sphere plasma, emitted a strong signal of copperions and continuum background. The other region referred to as the outer sphere plasma,surrounded the inner sphere plasma and emitted blue-green copper atomic lines witha relatively low background continuum and without ion lines. They reported the LODvalues in the range of several parts per million to hundreds of parts per billion in solidsamples, while the relative standard deviation (RSD) values varied from a few percent to80%. Wachter and Cremers [53] have examined the effects of the laser pulse repetitionrate, the detector gating, and the number of averaged laser shots on the precision. It wasfound that the precision increased with repetition rate and total number of laser pulsesaveraged, but was independent of the gating parameters. It was also found that the RSDdecreased from 13.3% for 50 laser shots to 1.8% for 1600 laser shots. The freshnessof sample surface also affects precision and the lowest RSD has been obtained wheneach shot samples a totally new portion of the material [71]. Eppler et al. [72] foundan increase in the precision using a cylindrical lens instead of a spherical lens, but theRSD was independent of the choice of lens. The reduction of the RSD was attributed tothe greater amount of material sampled by the cylindrical lens. Castle et al. [27] havesystematically studied variables that influence the precision of LIBS measurements with




special emphasis on the effect of temporal development of the emission, the sampletranslational velocity, and the number of spectra accumulated, laser pulse stability,detector gate delay, surface roughness, and the use of background correction.

9. CONCLUSION

Developments of LIBS techniques have been very rapid in the recent years and commer-cial instruments are coming up for application in process monitoring in many industriesas well as in various field applications. The prototypes of miniaturized version of LIBShave already been demonstrated and it is hoped that these will be commercially availablein near future. Inspite of many advantages LIBS techniques lag behind the conventionalanalytical techniques in terms of sensitivity but work is in progress to circumvent thisshortcoming by a careful assessment of the experimental parameters that influence theLIBS signal. Finally it seems that applicability of LIBS will increase many fold afterthe development of a miniaturized LIBS system with enhanced sensitivity.

REFERENCES

[1] F. Brech and L. Cross, Appl. Spectrosc. 16 (1962) 59.[2] E. R. Runge, R.W. Minck and F.R. Bryan, Spectrochim. Acta B20 (1964) 733.[3] L. J. Radziemski and D. A. Cremers, Laser-induced plasma and applications, Marcel Dekker,

New York, (1989) pp 295–325.[4] L. Moenke-Blankenburg, Laser microanalysis, (Eds.) J. D. Winefordner and I. M. Kolthoff,

John Wiley & Sons, New York, (1989).[5] L. Moenke-Blankenburg, Laser in analytical atomic spectroscopy, (Eds.) J. Sneddon,

T. L Thiem, Y. I. Lee, Wiley-VCH, New York, (1997) pp. 125–195.[6] S. A. Drake and J. F. Tyson, J. Anal. At. Spectrom. 8 (1993) 145.[7] D. A. Rusak, B. C. Castle, B. W. Smith and J. D. Winefordner, Crit. Rev. Anal. Chem.27

(1997) 257[8] K. Song, Y. I. Lee and J. Sneddon, Appl. Spectrosc. 32 (1997) 183.[9] F. Y. Yueh, J. P. Singh and H. Zhang, Encyclopedia of analytical chemistry Ed. R. A. Meyers,

Vol. 3, Wiley, New York, (2000) p. 2065[10] O. Samek, D. C. S. Beddows, J. Kaiser, S. V. Kukhlevsky, M. Liska, H. H. Telle and

J. Young, Opt. Eng. 38 (2000) 2248.[11] A. K. Rai, V. N. Rai, F. Y. Yueh and J. P. Singh, Trends in Applied Spectroscopy Vol. 4

Research Trends, Trivandrum, India, (2002) p. 165.[12] Y. I. Lee, K. Song and J. Sneddon, Laser in analytical atomic spectroscopy, (Eds.) J. Sneddon,

T. L. Thiem and Y. I. Lee, Wiley-VCH, New York, (1977) p. 197.[13] Y. I. Lee, Y. J. Yoo and J. Sneddon, Spectrosc. 13 (1998) 14.[14] F. Y. Yueh, V. N. Rai, J. P. Singh and H. Zhang, AIAA-2001-2933, 32nd AIAA Plasmady-

namics and Laser Conference, 11–14 June (2001), Anaheim, CA, USA.[15] J. P. Singh, F. Y. Yueh, H. Zhang and R. L. Cook, Process Control and Quality 10 (1997) 247.[16] H. Zhang, F. Y. Yueh and J. P. Singh, Appl. Opt. 38 (1999) 1459.[17] X. Hou, P. Stchur, T. Sun, K. X. Yang and R. G. Michel, 37th Eastern analytical symposium

and exposition, Paper # 37, Nov. 15–20, Somerset, NJ (1998).[18] L. J. Radziemski, T. R. Loree, D. A. Cremers and N. M. Hoffman, Anal. Chem. 55

(1983) 1246.




[19] D. A. Cremers and L. J. Radziemski, Anal. Chem. 55 (1983) 1252.[20] C. M. Davies, H. H. Telle, J. D Montgomery and R. E. Corbett, Spectrochim. Acta B50

(1995) 1059.[21] Y. I. Lee, S. P. Sawan, T. L. Thiem, Y. Y. Teng and J. Sneddon, Appl. Spectrosc. 46

(1992) 436.[22] A. V. Pakhomov, W. Nichols and J. Borysow, Appl. Spectrosc. 50 (1996) 880.[23] K. Y. Yamamoto, D. A. Cremers, M. J. Ferris and L. E. Foster, Appl. Spectrosc. 50

(1996) 222.[24] B. J. Marquardt, S. R. Goode and S. M. Angel, Anal. Chem. 68 (1996) 977.[25] A. Ciucci, V. Palleschi, S. Rastelli, R. Barbini, F. Colao, R. Fantoni, A. Palucci, S. Ribezzo

and H. J. L. Van der Steen, Appl. Phys. B63 (1996) 185.[26] R. E. Russo, W. T. Chan, M. F. Bryant and W. F. Kinard, J. Anal. At. Spectrom. 10

(1995) 295.[27] B. C. Castle, K. Talabardon, B. W. Smith and J. D. Winefordner, Appl. Spectrosc.

52(1998) 649.[28] B. C. Castle, A. K. Knight, K. Visser, B. W. Smith and J. D. Winefordner, J. Anal. At.

Spectrom. 13 (1998) 589.[29] H. E. Bauer, F. Leis and K. Niemax, Spectrochim. Acta B53 (1998) 1815.[30] V. Detalle, R. Heon, M. Sabsabi and L.St.Onge, Spectrochim. Acta B56 (2001) 1011.[31] C. Haisch, U. Panne and R. Niessner, Spectrochim. Acta B53 (1998) 1657.[32] S. R. Goode, S. L. Morgan, R. Hoskins and A. Oxsher, J. Anal. At. Spectrom. 15 (2000) 1133.[33] U. Panne, C. Haisch, M. Clara and R. Niessner, Spectrochim. Acta B53 (1998) 1957[34] R. Barbini, F. Colao, R. Fantoni, A. Palussi, S. Ribezzo, H. J. L. Van der Steen and

M. Angelone, Appl. Phys. B65 (1997) 101.[35] A. I. Whitehouse, J. Young, I. M. Botheroyd, S. Lawson, C. P. Evans and J. Wright,

Spectrochim. Acta B56 (2001) 821.[36] A. K. Rai, H. Zhang, F. Y. Yueh, J. P. Singh and A. Weisburg, Spectrochim. Acta B56

(2001) 2371.[37] X. Hou and B. T. Jones, Microchemical J. 66 (2000) 115.[38] D. A. Cremers, J. E. Barefield II and A. C. Koskelo, Appl. Spectrosc. 49 (1995) 857.[39] A. E. Pichahchy, D. A. Cremers and M. J. Ferris, Spectrochim. Acta B52 (1997) 25.[40] R. Sattmann, V. Sturn and R. Noll, J. Phys. D28 (1995) 2181.[41] D. N. Stratis, K. L. Eland and S. M. Angel, Appl. Spectrosc. 55 (2001) 1292.[42] D. N. Stratis, K. L. Eland and S. M. Angel, Appl. Spectrosc. 54 (2000) 1719.[43] D. N. Stratis, E. L. Eland and S. M. Angel Appl. Spectrosc. 54 (2000) 1270.[44] L. St.-Onge, M. Sabsabi and P. Cielo, Spectrochim. Acta B53 (1998) 407.[45] L. St.-Onge, V. Detalle and M. Sabsabi, Spectrochim Acta B57 (2002) 121.[46] V. N. Rai, F. Y. Yueh and J. P. Singh, Appl. Opt. 42 (2003) 2094.[47] B. W. Smith, A. Quentmeir, M. Bolshov, K. Niemax, Spectrochim. Acta B54 (1999) 943.[48] I. B. Gornushkin, S. A. Baker, B. W. Smith and J. D. Winefordner, Spectrochim. Acta B52

(1997) 1653[49] F. Hilbk Kortenbruck, R. Noll, P. Wintjens, H. Falk and C. Becker, Spectrochim. Acta B56

(2001) 933.[50] H. H. Telle, D. C. S. Beddows, G. W. Morris and O. Samek, Spectrochim. Acta B56

(2001) 947.[51] R. E. Neuhauser, U. Panne, R. Neissner, G. A. Petrucci, P. Cavalli and N. Omenetto, Anal.

Chim. Acta, 346 (1997) 37.[52] J. P. Singh, H. Zhang, F. Y. Yueh, and K. P. Karney, Appl. Spectrosc. 12 (1996) 764.[53] J. R. Watcher and D. A. Cremers, Appl. Spectrosc. 41 (1987) 1042.[54] J. P. Singh, F. Y. Yueh, H. Zhang and K. P. Karney, Rec. Res. Dev. Appl. Spectrosc.

2 (1999) 59.




[55] V. N. Rai, J. P. Singh, C. Winstead, F. Y. Yueh and R. L. Cook, AIAA Journal 41 (2003) 2192.[56] D. A. Cremers, R. C. Wiens, M. J. Ferris, R. Brennetot and S. Maurice, Trends in Optics

and Photonics Vol. 81 Laser-Induced Plasma Spectroscopy and Applications, OSA TechnicalDigest (2002) p. 5.

[57] R. Brennetot, J. L. Lacour, E. Vors, P. Fichet, D. Vailhen, S. Maurice and A. Rivoallan,Trends in Optics and Photonics Vol.81 Laser-Induced Plasma Spectroscopy and Applications,OSA Technical Digest (2002) p. 9.

[58] A. K. Rai, F. Y. Yueh and J. P. Singh, Rev. Sci. Instrum. 73 (2002) 3589.[59] A. K. Rai, F. Y. Yueh and J. P. Singh, Appl. Opt. 42 (2003) 2078.[60] C. Aragon, J. A. Aguilera, J. Campos, Appl. Spectrosc. 47 (1993) 606.[61] L. Pasky, B. Nemet, A. Lengyel and L. Kozma, Spectrochim. Acta B51 (1996) 27.[62] J. Gruber, J. Heitz, H. Strasser and N. Ramasedar, Spectrochim. Acta B56 (1981) 685.[63] J. Gruber, J. Heitz, N. Arnold, N. Ramsedar, W. Meyer, F. Koch, Appl. Spectrosc. 58

(2004) 457.[64] L. Peter, V. Sturn and R. Noll, Appl. Opt. 42 (2003) 6199.[65] R. A. Multari, L. E. Foster, D. A. Cremers and M. J. Ferris, Appl. Spectrosc. 50 (1996) 1483.[66] B. C. Castle, K. Visser, B. W. Smith and J. D. Winefordner, Appl. Spectrosc. 51 (1997) 1017.[67] K. Song, H. Cha, J. Lee and Y. I. Lee, Microchem. J. 63 (1999) 53.[68] M. Austin, A. Briand and P. Mauchien, Spectrochim. Acta B48 (1993) 851.[69] V. N. Rai, H. Zhang, F. Y. Yueh, J. P. Singh and A. Kumar, Appl. Opt. 42 (2003) 3662.[70] Y. I. Lee, T. L. Thiem, G. H. Kim, Y. Y. Teng and J. Sneddon, Appl. Spectrosc. 46

(1992) 1597.[71] R. Wisbrun, I. Schechter, R. Niessner, H. Schroder and K.L Kompa, Annal. Chem. 66

(1994) 2964.[72] A. S. Eppler, D. A. Cremers, D. D. Hickmott, M. J. Ferris and A. C Koskelo, Appl. Spectrosc.

50 (1996) 1175.

Elsevier AMS Job code: LRN Ch06-N51734 25-7-2007 6:41a.m. Page:137 Trim:165×240MM TS: Integra


Chapter 6

Dual-Pulse Laser-Induced Breakdown Spectroscopy

J. Scaffidia, D.A. Cremersb and S.M. Angela

a University of South Carolina, Department of Chemistry and Biochemistry,Columbia, SC, 29208, U.S.A.

bSDN Research, Santa Fe, NM, 87501, U.S.A.

1. INTRODUCTION

In laser-induced breakdown spectroscopy (LIBS), first introduced by Brech and Crossin 1962,[1] a high-powered laser pulse is focused to a sub-mm spot, yielding a peak fluxbetween 108 and 1010 W/cm2. Free and loosely-bound electrons interact with this intenseelectromagnetic field, absorbing energy from the temporally long nanosecond (ns) laserpulse through inverse Bremsstrahlung processes and freeing additional electrons throughcollisions within tens or hundreds of picoseconds. The newly-freed electrons also absorbenergy from the nanoseconds-long laser pulse, colliding with and freeing yet moreelectrons until a thermally hot, charge-neutral laser-induced plasma (LIP) with electrondensities as high as 1018 or even 1020/cm3 is produced. These initial stages of LIPevolution are referred to as ablation and plasma formation and (for the LIP produced witha high-energy nanosecond laser pulse) are followed by a microseconds-long period duringwhich atomic, ionic, and molecular emission characteristic of the plasma compositioncan be measured as electrons and ions recombine and cool to ambient temperatures.As emission by a LIP is in large part a function of the composition of the solid, liquid,or gaseous sample, LIBS has the potential to yield analytically-useful information aboutvirtually any sample under any conditions where a LIP can be formed, ranging fromroutine industrial [2–7] and environmental [8–10] settings to applications as challengingas deep oceanographic [11] and extraterrestrial [12–18] analyses.

The appeal of such an analytical technique is readily apparent—as LIBS uses onlylight to generate the LIP (see Fig. 1) and only photons need be collected to yieldanalytically-useful information, standoff or remote analysis of hazardous or difficult-to-reach samples becomes far simpler and safer than is possible with traditional laboratoryanalyses. Additionally, because LIBS can provide information on the elemental com-position of solids, liquids, and gases, the lack of time-consuming sample preparation(collection, transport, digestion, dilution, etc.) and the high data acquisition speeds pos-sible with current detection systems would make the technique seem ideal for rapidanalysis of a wide range of real-world samples. Recent work by a number of researchers




138 J. Scaffidi et al.

ICCD

(a) (b)

ICCD

Timing

TimingFO

FO

S

S

Laser

Laser 1

Laser 2

Spectrometer

Spectrometer

Fig. 1. A typical LIBS setup. Emission from the plasma formed by short-pulse laser-inducedbreakdown of a solid, liquid, or gaseous sample (S) is focused onto a fiber optic (FO), spectrallyseparated with a spectrograph, and recorded with a time-gated intensified CCD (ICCD) andcomputer. Use of a programmable timing generator allows system synchronization and easyincorporation of a second laser for dual-pulse LIBS (Fig. 1b).

has further illustrated this aspect of LIBS’ potential, especially in the areas of industrialand environmental analysis.

Despite the technique’s potential and many researchers’ hopes, however, real-worldlimitations currently restrict its applicability. Limits of detection for all but the mosteasily-determined elements are in the parts-per-million or high parts-per-billion for LIBSusing a single nanosecond pulse (ns single-pulse LIBS, Fig. 1a), and relative stan-dard deviations (in large part a function of laser instability, detector noise, ablativeirreproducibility, plasma formation irreproducibility, and sample inhomogeneity on themicroscopic scale) often fall between five and ten percent for even the simplest analyses.Further, matrix effects are often significant enough that attempts to apply LIBS with-out matrix-matched standards can be an exercise in futility. These difficulties, thoughbothersome, are not insurmountable. Matrix-matched standards can often be purchasedor produced for industrial analyses, and improved lasers, cameras, and spectrometershave become increasingly available and reliable in recent years. Additionally, modelingof both the ablative and emissive stages of plasma evolution (an attempt to addressthe irreproducibility of LIP formation and decay), and the discovery of up to hundred-fold atomic emission and signal-to-noise enhancements in both collinear and orthogonaldual-pulse LIBS (Fig. 2) may eventually allow LIBS to achieve its potential as a meansof rapid remote, on-site and in situ quantitative multielemental analysis.

2. DUAL-PULSE LIBS

The origins of dual-pulse LIBS lie in research performed by Cremers over twenty yearsago, [19] in which two collinear, spatially-overlapping ns-pulse plasmas produced withinforty microseconds of one another were used to improve limits of detection for a range ofanalytes (primarily alkali and alkaline elements such as lithium, potassium, calcium, mag-nesium, but also aluminum and boron) in bulk aqueous solution by orders of magnitudeover those seen with a single ns-pulse plasma. Despite the obvious advantages of enhanced



Dual-Pulse LIBS 139

(a)

400 × 103

300

200

100

0

840830820810Wavelength (nm)

Inte

nsity

(au

)(b)

170 × 103

125

80

35

Inte

nsity

(au

)

402 405 408396 399393390Wavelength (nm)

(c)

1.4 × 103

1.2

1.0

0.8

0.6

0.4

288284280Wavelength (nm)

Inte

nsity

(au

)

40 × 103

30

20

10

0

Intensity (au)

Fig. 2. Dual-Pulse LIBS Atomic Emission and Signal-to-Noise Enhancements. Above are shownsample spectra from Angel, et. al. (a, b) and Sabsabi, et. al. (c) showing the neutral atomic (a, b)and ionic (c) emission and signal-to-noise enhancements possible with orthogonal pre-ablativespark (a, b) and collinear (c) dual-pulse LIBS (top spectra) of dissolved sodium in aqueous solution(a), vitrified glass simulants in air (b), and an aluminum standard in air (c). The lower traces arethe corresponding ns single-pulse LIBS (bottom spectra).

atomic emission, larger signal-to-noise ratios, and orders-of-magnitude improvementsin limits of detection (LOD) simply via addition of a second laser pulse (Fig. 3), dual-pulseanalyses of solutions remained an interesting curiosity until Sattmann [20] and Uebbing [21]applied multi-pulse LIBS in collinear (Fig. 3a) and orthogonal reheating (Fig. 3b) dual-pulse configurations, respectively, for solids in air. Following this work, however, severalyears again passed before additional researchers took active interest in dual-pulse LIBS.It was not until very recently, in fact, that dual-pulse LIBS truly became a focusof research in 2000 and 2001, several authors examined dual-pulse LIBS using thecollinear pulse alignment [19,20,22–25] favored by both Cremers [19] and Sattmann, [20]and Stratis, Eland, and Angel [26–30] published a number of articles describing andcharacterizing the newly-developed orthogonal pre-ablative spark dual-pulse config-uration (Fig. 3c). Since that time, research regarding both sources and applicationsof dual-pulse LIBS enhancements has continued, with many advances in each area.




2

1

(a) (b) (c)

2

1

1

2

Fig. 3. Common Pulse Configurations in Dual-Pulse LIBS. Above are shown the collinear (a),orthogonal reheating (b), and orthogonal pre-ablative spark (c) dual-pulse LIBS configurations.Unlike the collinear case (a), in which the first and second laser pulses are both focused onto orinto the solid, liquid, or gaseous sample, the orthogonal dual-pulse configurations couple a singleablative pulse with either a post-ablative reheating pulse (b) or a pre-ablative air spark (c) up toseveral mm above the sample surface.

The purpose of this chapter is to survey the dual-pulse LIBS literature publishedduring the past several years, and failure to include any particular article should not beviewed by the reader as a suggestion that the research not included is less importantthan the work described here. Rather, the authors acknowledge that only a finite numberof articles can be incorporated into any discussion of cutting-edge research, and ask thereader’s tolerance regarding any omissions.

3. DUAL-PULSE LIBS APPLICATIONS

The greatly improved atomic emission, signal-to-background (S/B) and signal-to-noise(S/N) ratios seen in dual-pulse LIBS relative to its single-pulse counterpart can beobtained with what currently amounts to a minor one-time expense when comparedto the ongoing personnel, chemical, chemical transport, and hazardous waste disposalcosts incurred by more traditional analytical techniques. Additionally, rapid analy-sis of difficult- or impossible-to-reach environments and reduced human exposure topotentially-hazardous samples (radioactives and RCRA metals, [31–33] for example)both remain significant but unquantifiable advantages of both single- and dual-pulseLIBS. Given the above, then, it is not surprising that dual-pulse LIBS applications havebecome increasingly common in recent years.

As LIBS’ greatest potential in the near future lies in online, on-site, and in situanalyses, any work discussing recent trends yet failing to include the rapidly-growingmass of research in these areas would be inherently incomplete. Indeed, with continuingreductions in size and power requirements and improvements in reliability for lasers,spectrometers, and detectors (a trend which shows no indications of slowing), one canonly expect the pace of online, on-site, and in situ LIBS research and application to



Dual-Pulse LIBS 141

further accelerate. The recent commercial availability of lasers able to produce mul-tiple collinear ns pulses has further advanced the development of online, on-site andin situ dual-pulse LIBS applications. In the past several years alone, dozens of researchershave demonstrated LIBS’ usefulness in industrial and environmental analyses, as wellas a number of studies which have the potential to result in the use of LIBS as a meansof multielemental analysis in less conventional settings.

A number of authors have applied the collinear dual-pulse configuration (Fig. 3a) toelemental analysis throughout the steel-making process. In 2001, Barrette et. al. [34] usedsingle- and dual-pulse LIBS to examine silicon, graphitic and total carbon, magnesium,calcium, and aluminum in iron ore slurries at two pelletizing plants. In addition tofinding good correlation between LIBS and more traditional analytical techniques, theauthors noted that while the traditional analysis required a full hour, LIBS analysis tookless than two minutes. From a time savings and sample throughput perspective alone,dual-pulse LIBS’ usefulness is clear.

Peter, Sattmann, Sturm, and Noll [20,35–37] have also published several articlessummarizing their application of single- and dual-pulse LIBS in the steel industry,including blast furnace top gas monitoring, in-process analysis of molten steel, andexamination of finished products. Of special interest is the authors’ development ofa water-jacketed probe able to withstand immersion in molten steel for several hourson end at temperatures above 1600 �C, with limits of detection below 25 ppm for arange of light and heavy elements [35]. Sturm et. al. [37] have also used a modifiedNd:YAG laser operating at 1064 nm for collinear dual- and triple-pulse LIBS (Fig. 3a)of solid steel. The authors noted substantial improvements in atomic emission intensityand signal-to-noise ratios for a range of elements following addition of a second laserpulse, and were the first to successfully use LIBS to measure carbon, phosphorus, sulfur,and silicon in steel at concentrations below 10 ppm. Addition of a third collinear laserpulse improved analyte emission relative to the single- and dual-pulse cases, though theincrease was not as significant as that seen when going from single- to dual-pulse LIBS.

In other work applying LIBS to industrial analyses, Stepputat, et al. [38] used single-and collinear dual-pulse LIBS (Fig. 3a) for online analysis of heavy metals (cadmium,chromium, mercury, lead, and antimony) and brominated flame retardants in polymers.The dual-pulse technique improved limits of detection for cadmium, antimony, andbromine (whose atomic emission lines in this study all have excitation energies above5 eV), but worsened limits of detection for chromium, mercury, and lead (whose emis-sion lines in this work have excitation energies below 5 eV). Additionally, the authorsdemonstrated the usefulness of autofocusing for rapid online analysis with single- anddual-pulse LIBS.

Along with the above studies examining industrial applications for dual-pulse LIBSof solids, several authors have applied the technique to trace analysis of ions or colloidsin aqueous solution. Rai, et al. [39] focused paired collinear pulses (Fig. 3a) onto aliquid jet in air at an inter-pulse delay of 2.5–3.0 microseconds for determination ofmagnesium in water. The limit of detection was 230 ppb using a single nanosecond laserpulse, but addition of a second pulse lowered the limit of detection for magnesium to69 ppb. The authors attributed this improvement to the increased plasma volume seen inthe collinear dual-pulse configuration.

In other work using dual-pulse LIBS and a liquid jet, Pu, et al. [40] combined a1064 nm pulse focused on a liquid jet in air with a 193 nm ablative pulse in the orthogonal




dual-pulse configuration (analogous to Fig. 3b) for determination of lead colloidconcentrations in aqueous solution. Limits of detection improved 14-fold with respect tosingle-pulse LIBS when the 193 nm pulse was properly timed to ablate colloids ejectedfrom the liquid jet by the 1064 nm pulse and kinetically concentrated in the surroundingair, allowing the authors to measure concentrations as low as 136 ppb. Early optimizationof the technique reduced the limit of detection to 14.2 ppb, [41] and subsequent workcombining LIBS and fluorescence has improved this to less than 1 ppb. [42]

With a hybrid configuration that essentially amounts to using two orthogonally-oriented collinear pulse pairs (Fig. 4), Kuwako, et al. [43] applied dual-pulse LIBSto measurement of dissolved sodium concentrations in a specially-designed flow cell.Through optimization of laser power and inter-pulse timing for two 1064 nm, 3.5 nspulses, the authors were able to achieve an estimated limit of detection of 0.1 ppb forsodium when using their unique dual-pulse configuration and a falling aqueous film.

Pearman, et al. [44] used paired orthogonal pulses in bulk solution (Fig. 5) when revis-iting Cremers’ initial collinear dual-pulse work [19]. Examination of atomic emission byoxygen and various analytes (Zn, Cr, and Ca) supported Cremers’ hypothesis [45] thatdual-pulse LIBS enhancements in bulk solution appear to be related to the second pulseprobing cavitation bubbles formed by the first LIP, and the authors noted large enhance-ments (similar to those seen for sodium in Fig. 2a, for example) using the orthogonalconfiguration over a wide range of inter-pulse delays and plasma observation times.Calculated and experimentally verified limits of detection for the orthogonal dual-pulseconfiguration were 17 ppm for zinc, 1.04 ppm for chromium, and 41.7 ppb for calcium,

2

A

BBC

1

Fig. 4. Cell and Optics for Ultratrace LIBS of Aqueous Sodium Solutions. In Kuwako, et. al.’sresearch measuring sub-ppb dissolved sodium concentrations, two temporally-separated pulses(1, 2) are divided by a beamsplitter (A) and directed onto a falling aqueous film in their specially-designed cell using dichroic mirrors (B). Due to the pulse alignment and pulse timing, thearrangement yields two orthogonally-oriented collinear dual-pulse laser-induced plasmas whoseemission can be collected using built-in fiber-optic collection (C).



Dual-Pulse LIBS 143

11

2211

(b)(a)

Fig. 5. Single and orthogonal dual-pulse LIBS plasmas in aqueous solution. Both single- (a) anddual-pulse (b) LIBS plasmas can be formed quite easily in aqueous solution at atmospheric pres-sure, with the dual-pulse plasma (b) being somewhat larger and longer-lived than that seen in thesingle-pulse case (a). The light speckles surrounding the single-pulse LIBS plasma in (a) arethe result of cavitation bubbles produced by LIP formation in solution, and are thought to reducethe quenching effects of the dense aqueous solution. The numbered arrows indicate pulse firingorder, for reference.

all of which compared favorably to those seen in other dual-pulse LIBS analyses in bulksolution. Ongoing work by the same group aims to apply single- and dual-pulse LIBSto analysis of high-pressure, high-temperature sub-oceanic hydrothermal vent fluids.

Most recently, St-Onge, et al. [46] used saline solutions as model samples whileexamining the feasibility of single- and dual-pulse LIBS for in-process and end-productanalysis of pharmaceuticals. The authors noted their best results when using a singlens pulse focused on the surface of stationary or flowing samples, taking special care tominimize excessive splashing and using a gas purge to remove the aerosols producedduring surface ablation of solutions, but in-bottle analysis of saline solution requiredcollinear dual-pulse LIBS (Fig. 3a) with the pulses focused as closely as possible to thenear wall of the container.

4. DUAL-PULSE LIBS MECHANISTIC STUDIES

The large atomic emission and signal-to-noise enhancements generated by dual-pulseLIBS can be easily seen through direct comparison of single- and dual-pulse spectra(Fig. 2).What is less apparent, however, are the cause or causes of the analytical improve-ments seen in both the collinear and orthogonal dual-pulse configurations. Cremers’ earlywork in bulk solution [19] suggested the possibility of bubble formation by the first LIPfollowed with interrogation by the second pulse, yielding reduced plasma quenching by(and greatly enhanced plasma emission in) the dense liquid medium [45]. Sattman [20]and Uebbing [21] presented the possibility of energetic causes for the enhancements seenin orthogonal reheating and collinear dual-pulse LIBS of solids in air. Research sincethose early studies has also indicated the potential for sample heating by the first/pre-ablative LIP, as well as atmospheric pressure or number density reductions followingLIP formation in air [26–29,47,48]. Given that evidence exists supporting each of these




hypotheses as a source of dual-pulse LIBS enhancements, it seems unlikely that any oneis the sole source of these analytical improvements. Rather, it appears probable that dual-pulse LIBS atomic emission, signal-to-background, and signal-to-noise enhancementsresult from some combination of these (and potentially other) effects.

Soon after discovery of dual-pulse enhancements in the orthogonal pre-ablative sparkconfiguration (Fig. 3c), Stratis, et al. [26–29] published a series of articles examining andcharacterizing the phenomenon. In 2000, [27] the authors detailed 7- to 33-fold neutralatomic emission enhancements for iron, titanium, cadmium, zinc, lead, aluminum, andcopper for a 1064 nm, 5 ns ablative pulse combined with the pre-ablative air sparkformed with another 1064 nm, 5 ns pulse. The authors also found substantially improvedablation, but noted poor correlation between atomic emission enhancement and thethermodynamic properties (melting point and thermal conductivity) of the sample. Fromthis they concluded that some non-thermal mechanism is at least in part responsible fordual-pulse enhancements in the orthogonal pre-ablative spark configuration.

In that same year, Stratis, et al. [28] reported using two 1064 nm, 5 ns pulsesin the orthogonal pre-ablative spark configuration (Fig. 3c) for analysis of wastevitrification simulants, observing 20-fold neutral atomic emission enhancement for ironand aluminum, and 11-fold enhancement for titanium (Fig. 2b). Application of thesame setup to analysis of pure solids showed only 4- to 6-fold enhancements for iron,aluminum, and titanium, indicating that orthogonal pre-ablative spark dual-pulse LIBSatomic emission enhancements (like those in the collinear dual-pulse configuration)are both analyte- and matrix-dependent. As in the case of the pure solids used in theirprevious work, the authors noted substantial increases in sample ablation, reinforcingthe possibility that atomic emission enhancements in the orthogonal pulse configurationmay be related to increased sample introduction into the dual-pulse plasma. It is alsoin this article that Stratis, et al. presented two hypotheses regarding dual-pulse LIBSenhancements in air: That the first/pre-ablative spark heats the sample surface, and thatthe first/pre-ablative spark forms a short-lived, localized, low-pressure region abovethe sample surface (similar to the bubbles described by Cremers [19] for collineardual-pulse LIBS of aqueous solutions).

In 2001, [29] the same research group spatially and temporally resolved orthogonalpre-ablative spark dual-pulse plasma (Fig. 3c) emission, noting that enhancement ofspatially-integrated LIP emission is comparable to that seen for spatially-focused fiber-optic collection in the same pulse configuration. In contrast to the results shown byStepputat, et al. [38] while using the collinear dual-pulse configuration (Fig. 3a), no defi-nite trend was apparent when comparing dual-pulse LIBS atomic emission enhancementsin the orthogonal pulse configuration to the excitation energy for iron lines: though the426.1 nm line (excitation energy of 5.31 eV) showed greater atomic emission enhance-ment than lines with excitation energies below 5 eV it also showed greater enhancementthan the 411.9 nm line (excitation energy of 6.58 eV). The authors further observed thatthe effects of the pre-ablative air spark appeared to be transient in nature rather thanthe result of permanent sample modification (yielding atomic emission enhancementsfor ablation as late as 1 millisecond after air spark formation), and again raised thepossibility of sample heating, atmospheric ionization, and lingering shock wave effectsas potential sources of dual-pulse LIBS enhancements.

The following year yielded a pair of articles from Colao et. al. [22,24] examiningcollinear dual-pulse LIBS enhancements (Fig. 3a). In the first [22], the authors discussed



Dual-Pulse LIBS 145

the pulse power dependence of atomic emission intensity and stability, noting that higherpulse energies resulted in more stable and more reproducible plasma formation thanlower-energy pulses as well as narrowing of both atomic and ionic lines, with theapparent threshold for these effects being approximately 1�2 GW/cm2 at focus. In thesecond, Colao et al. [24] examined the sources of collinear dual-pulse LIBS (Fig. 3a)atomic emission enhancements by comparing crater volume, plasma temperature, andelectron densities for ns single-pulse and ns-ns collinear dual-pulse LIBS. Crater vol-ume and, therefore, per-shot analyte introduction into the analytical plasma increasedsomewhat relative to the single-pulse case, though not to the extent seen in Stratis’earlier work [27,29] using the orthogonal dual-pulse configuration (Fig. 3c). Plasmatemperatures and electron densities, though initially lower in the collinear dual-pulseconfiguration than in single-pulse LIBS, decayed far more slowly than in the single-pulsecase. In an observation with implications for later work by other researchers, the authorsnoted less intense ionic nitrogen emission in the collinear dual-pulse configuration thanwas seen for ns single-pulse LIBS.

St-Onge et al. [23] revisited their pre-2000 dual-pulse LIBS work in 2002, investi-gating the wavelength dependence of the collinear dual-pulse configuration (Fig. 3a) bycombining 1064, 532, 355, and 266 nm pulses. Optimal atomic emission enhancementwas observed when the authors paired a 1064 nm first pulse with a 266 nm secondpulse, producing 30-fold atomic emission enhancement for the 288.16 nm neutral siliconline at an inter-pulse delay of 0.1 microseconds, and more than hundred-fold emissionenhancement for the 281.62 nm aluminum(II) line at a 3 microsecond inter-pulse delay(Fig. 2c). From these results the authors developed a three-area model for plasma-plasmainteractions in the collinear dual-pulse configuration.

In 2003, preliminary work by Scaffidi, et al. [49] used the greatly differing propertiesof ns and fs LIPs (size, plasma lifetime, means of plasma formation, etc.) in the orthog-onal dual-pulse configuration (Fig. 3c) while attempting to separate the atmosphericpressure/number density, sample heating, and plasma-plasma coupling effects thought tocause collinear and orthogonal ns-ns dual-pulse LIBS enhancements. The large atomicemission enhancements seen with a 800 nm, 100 fs ablative pulse and a 1064 nm, 5 nspre-ablative air spark or reheating LIP could be attributed to the increased size andhigher temperature of the long-lived ns plasma (relative to the smaller, shorter-livedfs plasma), but the two- to three-fold atomic emission enhancements seen for neutralcopper emission at 510.6, 515.3, and 521.8 nm with a fs pre-ablative air spark and ans ablative pulse at inter-pulse delays as late as 30 microseconds (long after the fsspark’s six-microsecond emissive lifetime) were considered an indication of atmosphericpressure or number density effects following fs LIP formation in air.

Further work by Scaffidi, et al. [47] in 2004 spatially and temporally mapped theinteraction of a 1064 nm, 5 ns ablative pulse orthogonal to a 800 nm, 100 fs pre-ablativeair spark (Fig. 3c), noting the apparent existence of a region of reduced atmosphericpressure or number density following fs LIP formation in air (Fig. 6). Atomic emissionenhancement for copper in brass and aluminum in bulk aluminum showed good spatialand temporal overlap with this region of reduced pressure/number density, indicatingthat atomic emission enhancement in the orthogonal dual-pulse configuration may atleast in part result from an atmospheric “bubble” similar to that first hypothesized byStratis, et al. in 2001 [29].




2.5

2.0

1.5

1.0

0.5

0.0

–500 –400 –300 –200Inter-pulse delay, µs

Enh

ance

men

t vs.

ns

SP

B

A

–100 0

Fig. 6. Atmospheric and ablated analyte emission correlations in dual-pulse LIBS. Neutral atomicemission intensity for atmospheric oxygen (B, 777 nm line) and ablated copper (A, 521.8 nm line)show similar but inverse dependence on inter-pulse delay for orthogonal dual-pulse LIBS using afs pre-ablative air spark and a ns ablative pulse. This result (and similar results seen for the ns-nscase) is consistent with the hypothesis that the ablative LIP can be formed in a region of reducedpressure or atmospheric number density produced by the pre-ablative air spark.

A third study by the same authors [48] also published in 2004, compared neutralatomic emission enhancement and mass removal enhancement in the fs-ns orthogonalpre-ablative spark configuration (Fig. 3c) to examine the role of increased mass removalin generating dual-pulse LIBS enhancements. Whereas both aluminum and coppershowed three- to four-fold neutral atomic emission and signal-to-noise enhancementwhen combining a fs pre-ablative spark and a ns ablative pulse, per-shot mass removalshowed eight- to ten-fold enhancement. In addition, mass removal improvement andatomic emission enhancement showed only very general temporal correlation. Fromthese results, the authors concluded that although improved sample introduction intothe fs-ns dual-pulse LIP may play a role in generating improved atomic emission andsignal-to-noise ratios at some inter-pulse delays, there were likely additional effectswhich more generally contribute to fs-ns and ns-ns dual-pulse LIBS enhancements.

Corsi, et al. [50] pursued a similar line of research around the same time, focusing onspatially and temporally profiling single-pulse and collinear dual-pulse (Fig. 3a) LIBSplasmas and shock waves. In addition to significant though unquantified atomic emissionenhancements for minor constituents in brass (lead, tin, iron, aluminum, nickel, andmanganese), the authors observed reductions in ambient atmospheric density followingformation of the first LIP in air. Further, the authors noted that they were unable tospatially or temporally separate the shock waves resulting from formation of the firstand second LIPs, and theorized that this phenomenon was a result of the first plasmacreating a rarefied bubble into which the second plasma could expand unimpeded.

Most recently, Gautier, et al. [51] revisited Uebbing’s orthogonal reheating configura-tion [21] (Fig. 3b) using a 532 nm, 9 ns ablative pulse and a 1064 nm, 9 ns reheating pulse.The authors observed atomic emission enhancement for a range of analytes (manganese,



Dual-Pulse LIBS 147

copper, iron, magnesium, aluminum, and titanium) in aluminum alloys, and noted thatatomic and ionic emission from transitions with highly-energetic upper energy states wasenhanced more than emission from transitions with less-energetic upper energy states.While cautioning that direct quantitative comparison would be inappropriate due to thediffering pulse configurations, the authors went on to comment that their orthogonalreheating results paralleled those seen in collinear dual-pulse LIBS.

In examining the overall results of the various collinear, orthogonal reheating, andorthogonal pre-ablative spark dual-pulse LIBS research in the past several years, oneconclusion is particularly evident: Different pulse configurations suggest the existence ofdifferent sources for dual-pulse LIBS enhancements. Collinear (Fig. 3a) and orthogonalreheating work (Fig. 3b), for example, indicates the potential for energetic couplingbetween the first LIP and the second laser pulse. Orthogonal pre-ablative spark stud-ies (Fig. 3c), alternatively, indicate the possibility of sample heating and atmosphericpressure or number density effects in production of fs-ns and ns-ns dual-pulse LIBSatomic emission, signal-to-background, and signal-to-noise enhancements. Given that theorthogonal pre-ablative spark configuration is intentionally designed to prevent ablationby the first LIP (and thereby prevent direct excitation of already-ablated material by thesecond laser pulse), it is altogether possible that energetic coupling, sample heating andpressure/number density effects all play a role in generating the up-to-hundred-fold dual-pulse LIBS enhancements observed to date. Future mechanistic research will no doubtfocus on determining the relative importance of (and the potential to further improveenhancements due to) these and any additional effects in dual-pulse LIBS.

That said, research by Scaffidi, et al. [52] has recently begun the attempt to addressthese questions. While traditional single and dual-pulse LIBS studies tend to primarilyfocus on the signal-to-noise ratios and limits of detection seen under some “optimal”experimental conditions, the authors have instead chosen to extend their work examiningspatial and temporal correlations during dual-pulse LIBS optimization to the collinearpulse configuration. Limiting their initial study to a 100 fs, 800 nm first pulse and a5 ns, 1064 nm second pulse while varying pulse focus and inter-pulse delay, the authorsfound that the dominant source of dual-pulse LIBS enhancements at any given inter-pulse delay under these conditions may primarily be a function of pulse focusing. Whenthe fs and ns pulses were focused 2.5 mm beneath the sample surface, the temporaldependence of dual-pulse atomic emission enhancement for iron closely matched theemissive lifetime of the fs-pulse LIP, supporting the hypothesis that energetic couplingbetween the fs and ns plasmas may be one cause of dual-pulse LIBS enhancements.If focused 2.5 mm above the sample surface, atomic emission with a similar maximumintensity can be generated but with a temporal profile closely matching that seen foratmospheric nitrogen and oxygen emission reductions following LIP formation in air,thereby supporting the atmospheric pressure/number density hypothesis discussed above.Based on these results, it is evident that additional research will be necessary beforecollinear and orthogonal dual-pulse LIBS enhancements are fully understood.

5. FUTURE DIRECTIONS

Until the sources of the up to hundred-fold atomic emission, signal-to-noise, and signal-to-background enhancements seen in collinear and orthogonal dual-pulse LIBS are fully




explained, mechanistic studies will remain a significant area of LIBS research due tothe potential for further improvements in LIBS limits of detection and reproducibil-ity. At the same time, given the improved analytical merits seen in dual-pulse LIBSdespite the lack of understanding regarding the sources of these improvements, onewould expect to see continued interest in direct application of dual-pulse LIBS for rapidremote, online, on-site and in situ analyses, especially in exotic applications like oceano-graphic analysis (recently discussed at the LIBS 2004 international conference) and theextraterrestrial applications discussed above. Additionally, though LIBS is quite usefulfor multi-elemental analysis of well-characterized matrices, it is less useful for molecularanalyses and examination of poorly-characterized samples (such as those encounteredin real-world environmental applications, for example). As a result, studies combiningLIBS and Raman or LIBS and IR spectroscopy are expected to become more commonin the coming years, as is work combining LIBS and chemometric data analysis. Lastly,although the cost, complexity, and power and operator skill requirements for high-energyps and fs lasers currently render them unsuitable for routine use outside the laboratory,continuing instrumental and fiber optic improvements may eventually allow researchersto take advantage of the ablative and emissive improvements [53–56] seen in ultra-shortLIBS, further enabling the technique to fulfill its potential as a means of rapid, remote,online, on-site, and in situ multielemental analysis.

ACKNOWLEDGMENTS

The authors acknowledge the many researchers working to advance both single- anddual-pulse LIBS as viable means of multielemental analysis, and specifically thankDr. Mohammad Sabsabi and Dr. Louis St-Onge for making their raw data available tothe authors during the writing of this review. We also acknowledge the support of ourown work by the National Science Foundation (CHE-0316069 and OCE-0352242), theDepartment of Energy (DEFG0796ER62305), the Office of Naval Research (N0014-97-1-0806), Lawrence Livermore National Laboratory, Los Alamos National Laboratory,the National Aeronautics and Space Administration, and the Jet Propulsion Laboratory.Lastly, we thank the organizers of the biennial international LIBS conference series fortheir time and dedication to ensuring the continued existence of a forum for efficientexchange of information and ideas among LIBS researchers.

REFERENCES

[1] M. Tran, Q. Sun, B. Smith and J.D. Winefordner, Anal. Chim. Acta, 419 (2000) 153.[2] Q. Sun, Q, M. Tran, B.W. Smith and J.D. Winefordner, Anal. Chim. Acta, 413 (2000) 187.[3] C. Aragón F. Brech and L. Cross, Appl. Spectrosc., 16 (1962) 59.[4] J.A. Aguilera and F. Peñalba, Appl. Spectrosc., 53 (1999) 1259.[5] P. Lucena and J.J. Laserna, Spectrochim. Acta B, 56 (2001) 177.[6] J. Amador-Hernández, J.M. Fernández-Romero and M.D. Luque de Castro, Anal. Chim.

Acta, 435 (2001) 227.[7] H.H. Telle, D.C.S. Beddows, G.W. Morris and O. Samek, Spectrochim. Acta B, 56

(2001) 947.



Dual-Pulse LIBS 149

[8] R. Barbini, F. Colao, R. Fantoni, A. Palucci and F. Capitelli, Appl. Phys., A 69 [Suppl.](1999) S175.

[9] R.T. Wainner, R.S. Harmon, A.W. Miziolek, K.L. McNesby, and P.D. French, Spectrochim.Acta B, 56 (2001) 777.

[10] V. Lazic, R. Barbini, F. Colao, R. Fantoni and A. Palucci, Spectrochim. Acta B, 56 (2001) 807.[11] M.J. Lawrence-Snyder, J. Scaffidi, S.M. Angel, A.P.M. Michel and A.D. Chave, “Laser-

induced breakdown spectroscopy in high-pressure bulk aqueous solutions—part I: Com-parisons of single- and sequential-pulse excitation,” submitted to Appl. Spectrosc.November 2005.

[12] J.D. Blacic, D.R. Pettit and D.A. Cremers. “Laser-Induced Breakdown Spectroscopy forRemote Elemental Analysis of Planetary Surfaces,” published in the Proceedings of theInternational Symposium on Spectral Sensing Research, Maui, HI, November 15–20, 1992.

[13] A.K. Knight, N.L. Scherbarth, D.A. Cremers and M.J. Ferris, Appl. Spectrosc., 54 (2000) 331.[14] S.K. Sharma, P.G. Lucey, M. Ghosh, H.W. Hubble and K.A. Horton, Spectrochim. Acta A

59 (2003) 2391.[15] F. Colao, R. Fantoni, V. Lazic, A. Paolini, F. Fabbri, G.G. Ori, L. Marinangeli, and A. Baliva,

Planetary and Space Sci., 52 (2004) 117.[16] F. Colao, R. Fantoni, V. Lazic and A. Paolini, Appl. Phys. A, 79 (2004) 143.[17] Z.A. Arp, D.A. Cremers, R.D. Harris, D.M. Oschwald, G.R. Parker and D.M. Wayne,

Spectrochim. Acta B, 59 (2004) 987.[18] Z.A. Arp, D.A. Cremers, R.C. Wiens, D.M. Wayne, B.A. Salle and S. Maurice, Appl.

Spectrosc. 58 (2004) 897.[19] D.A. Cremers, L.J. Radziemski and T.R. Loree, Appl. Spectrosc., 38 (1984) 721.[20] R. Sattmann, V. Sturm and R. Noll., J. Phys. D, 28 (1995) 2181.[21] J. Uebbing, J. Brust, W. Sdorra, F. Leis and K. Niemax., Appl. Spectrosc., 45 (1991) 1419.[22] F. Colao, S. Pershin, V. Lazic and R. Fantoni, Appl. Surf. Sci., 197 (2002) 207.[23] L. St-Onge, V. Detalle and M. Sabsabi, Spectrochim. Acta B, 57 (2002) 121.[24] F. Colao, V. Lazic, R. Fantoni and S. Pershin, Spectrochim. Acta B, 57 (2002) 1167.[25] V. Sturm, L. Peter and R. Noll, Appl. Spectrosc., 54 (2000) 1275.[26] D.N. Stratis, K.L. Eland and S.M. Angel. “Dual-pulse LIBS: why are two lasers better

than one?” in Environmental Monitoring and Remediation Technologies II, T. Vo-Dinh andR.L. Spellicy, eds., Proc. SPIE 3853 (1999) 385.

[27] D.N. Stratis, K.L. Eland and S.M. Angel. Appl. Spectrosc., 54 (2000) 1270.[28] D.N. Stratis, K.L. Eland and S.M. Angel, Appl. Spectrosc., 54 (2000) 1719.[29] D.N. Stratis K.L. Eland and S.M. Angel, Appl. Spectrosc. 55 (2001) 1297.[30] S.M. Angel, D.N. Stratis, K.L. Eland, T. Lai, M.A. Berg, and D.M. Gold, Fres. J. Anal.

Chem., 369 (2001) 320.[31] J.R. Wachter and D.A. Cremers, Appl. Spectrosc., 41 (1987) 1042.[32] R.T. Wainner, R.S. Harmon, A.W. Miziolek, K.L. McNesby and P.D. French, Spectrochim.

Acta B, 56 (2001) 777.[33] S. Koch, W. Garen, M. Muller and W. Neu, Appl. Phys. A 79 (2004) 1071.[34] M.L. Barrette and S. Turmel, Spectrochim. Acta B, 56 (2001) 715.[35] L. Peter, V. Sturm and R. Noll, Appl. Optics 42 (2003) 6199.[36] R. Noll, H. Bette, A. Brysch, M. Kraushaar, I. Monch, L. Peter and V. Sturm, Spectrochim.

Acta B, 56 (2001) 637.[37] V. Sturm, L. Peter and R. Noll, Appl. Spectrosc., 56 (2000) 1275.[38] M. Stepputat and R. Noll, Appl. Opt., 42 (2003) 6210.[39] V.N. Rai, F-Y. Yueh and J.P. Singh, Appl. Opt., 42 (2003) 2094.[40] X.Y. Pu, and N.H. Cheung, Appl. Spectrosc., 57 (2003) 588.[41] X.Y. Pu, W.Y. Ma and N.H. Cheung, Appl. Phys. Lett., 83 (2003) 3416.[42] S.K. Ho and N.H. Cheung, Anal. Chem., 77 (2005) 193.




[43] A. Kuwako, Y. Uchida and K. Maeda, Appl. Optics, 42 (2003) 6052.[44] W. Pearman, J. Scaffidi and S.M. Angel, Appl. Optics, 42 (2003) 6085.[45] A.E. Pichahchy, D.A. Cremers and M.J. Ferris, Spectrochim. Acta B 52 (1997) 25.[46] L. St-Onge, E. Kwong, M. Sabsabi and E.B. Vadas, J. Pharmaceutical and Biomedical

Analysis, 36 (2004) 277.[47] J. Scaffidi, W. Pearman, M. Lawrence, J.C. Carter, B.W. Colston and S.M. Angel, Appl.

Optics, 43 (2004) 5243.[48] J. Scaffidi, W. Pearman, J.C. Carter, B.W. Colston and S.M. Angel, Appl. Optics 43

(2004) 6492.[49] J. Scaffidi, J. Pender, W. Pearman, S.R. Goode, B.W. Colston, J.C. Carter and S.M. Angel,

Appl. Optics 42 (2003) 6099.[50] M. Corsi, G. Cristoforetti, M. Giuffrida, M. Hidalgo, S. Legnaioli, V. Palleschi, A. Salvetti,

E. Tognoni and C. Vallebona, Spectrochim. Acta B 59 (2004) 723.[51] C. Gautier, P. Fichet, D. Menut, J.L. Lacour, D. L’Hermite and J. Dubessy, Spectrochim.

Acta B 59 (2004) 975.[52] J. Scaffidi, W. Pearman, J.C. Carter and S.M. Angel, “Observations in collinear femtosecond-

nanosecond dual-pulse laser-induced breakdown spectroscopy,” Appl. Spectrosc., in press.[53] V. Margetic, A. Pakulev, A. Stockhaus, M. Bolshov, K. Niemax and R. Hergenroder, Spec-

trochim. Acta B,55 (2000) 1771.[54] K.L. Eland, D.N. Stratis, T.S. Lai, M.A. Berg, S.R. Goode and S.M. Angel, Appl. Spectrosc.,

55 (2001) 279.[55] K.L. Eland, D.N. Stratis, D.M. Gold, S.R. Goode and S.M. Angel, Appl. Spectrosc., 55

(2001) 286.[56] Ph. Rohwetter, J. Yu, G. Mejean, K. Stelmaszczyk, E. Salmon, J. Kasparian, J-P. Wolf and

L. Woste, J. Anal. Atom. Spectrom., 19 (2004) 437.



Chapter 7

Femtosecond LIBS

Mohamad Sabsabi

Industrial Materials Institute, National Research Council of Canada,75 Boul. De Mortagne, Boucherville, Québec, J4B 6Y4, CANADA

1. CHAPTER ORGANIZATION

The purpose of this chapter is to provide the reader an overview of several aspects of theuse of femtosecond lasers for LIBS applications. These aspects include plasma dynamicsand characterization, ablation threshold, modeling of the laser-induced plasma (LIP),and LIBS spectrochemical analysis in terms of precision and sensitivity. In particular,we will highlight the advantages and the drawbacks of using ultrashort laser pulses forLIBS analysis. After an introduction to the subject, Sec. 3 will present a study on theplasma induced by ultra-short laser pulses which includes basic processes during laserablation, material removal and plasma expansion, and the influence of the pulse durationon the plasma properties. An evaluation of the influence of the laser pulse duration on thespectrochemical analysis by LIBS is given in Sec. 4. Sec. 5 will discuss a comparison ofgated and non-gated analysis by using ultrashort pulses. Finally, a summary is presentedin Sec. 6.

2. INTRODUCTION

Laser-induced plasmas are finding increasing interest as sources of materials resultingfrom the ablation of the target (called laser ablation or LA) and also of radiation. LA canbe used as a solid sampling technique where the ablated material is transported to a secondexcitation source and analyzed by alternative methods, i.e., atomic emission spectrometry(AES) and mass spectrometry (MS). Alternatively, a direct analysis can be made fromthe photons emitted by the plasma generated from the sample. In this chapter, we will befocusing on the second approach, called laser-induced plasma spectroscopy (LIPS) alsoknown as laser-induced breakdown spectroscopy (LIBS), by using ultrashort laser pulses.The first approach was the object of the Chapter 3 of this book and for complementaryinformation we refer the reader to two excellent review papers on the subject [1,2].

The interaction between a pulsed laser beam and any substance is extremely com-plex [3]. It is a non-linear process, dependent upon laser characteristics (fluence, pulse




152 M. Sabsabi

rise-time and duration, wavelength, beam quality), substrate composition and surfacecharacter, and the environment in which the plasma forms (pressure and composition).We counted more than 1500 publications in the last 5 years on laser-induced plasmasrelated to LIBS, many having to do with studies of the influence of laser wavelength,the effect of the surrounding atmosphere, time-resolved plasma imaging, the effect ofparticles and aerosols on plasma production, temperature and electron density measure-ments, ionization and plasma formation processes, modeling of the laser-induced plasmaand particularly the application of plasma formation on different materials for atomicemission spectroscopy. Most of these investigations were performed at long laser pulseduration in the nanosecond regime. To the best of our knowledge, very few papers weredevoted to the influence of laser pulse duration, and particularly to the ultrashort laserpulses on the LIBS performances (less than 50 papers devoted to LIBS with ultrashortlaser pulses out of 1500 papers in our databank).

Laser technology began to move into the subpicosecond time regime (this regimeis referred in the literature as ultrafast or femtosecond) in the early seventies [4], onedecade after the first invention of a commercial laser in 1960 [5]. In the followingyears subpicosecond laser pulses were primarily applied for the study of a broad varietyof ultrafast processes in different scientific fields [6,7], including time-resolved spec-troscopy of solids. More recently, femtosecond laser pulses have been considered foruse in laser processing of materials [8]. In the early nineties, scientists at the Universityof Michigan discovered that the transfer of heat from the laser beam to the work piececould be eliminated using ultrafast laser pulses instead of standard long-pulse lasers.Essentially, machining with laser pulses of very short duration eliminates heat flow tosurrounding materials. This discovery opened the way for fine laser micromachining.Furthemore, the rapid development of femtosecond lasers has opened up a wide range ofnew applications in industry, medicine, material science, military and X-ray lasers. Oneimportant application of femtosecond laser pulses is material removal or ablation. Laserablation with femtosecond pulses can be used for the deposition of droplet-free thinfilms, including semi-conductors, superconductor, magneto-resistive materials, and thecreation of new alloys. They can also be used for micro-machining, for the fabricationof nanomaterials, and even in the arts for picture restoration and cleaning. Femtosecondlaser ablation has an important advantage in such applications compared with ablationusing nanosecond pulses because there is little or no collateral damage due to shockwaves and heat conduction produced in the material being processed.

The general conclusions to be drawn from the literature concerning the benefits ofusing femtosecond pulses include:

• Ultrafast excitation can improve the material interaction with laser.• Ultrafast absorption of energy reduces post ejection interactions with laser.• Heat affected zone is confined to a smaller region – less vaporization of substrate.

It is then natural to suppose that the femtosecond regime could present many advan-tages over the nanosecond one for LIBS performance. For example, the improvedcontrol of the material removal (no melting and no mixing) should be an importantadvantage for using fs lasers for rapid surface layer analysis (in-depth profiling of thinlayered structures). The reduced thermal diffusion could enable a depth resolution in



Femtosecond LIBS 153

lower sub-�m range in combination with a good lateral resolution which is of impor-tance for microanalysis. Laser ablation in-depth profilling could be used for all kindsof solid materials. Investigations employing single-shot multielemental detection tech-niques could be carried out on conducting as well as non-conducting materials. Whileuse of long pulse laser in LIBS analysis has matured over many years, femtosecondLIBS is still in its infancy. In this chapter, we will present an overview on the use offemtosecond laser for LIBS plasmas, including plasma dynamics and characterization,ablation threshold, modeling of the laser induced plasma (LIP), and also a review on theuse of femtosecond laser for LIBS spectrochemical analysis in terms of precision andsensitivity. In particular, we will highlight the advantages and the drawbacks of usingultrashort laser pulses for LIBS analysis. I apologize for not citing numerous excellentpapers related to the studies presented herein; however, this chapter is not a literaturereview, but rather serves as an introduction to the field with useful tools for the LIBSresearcher.

3. PLASMA PRODUCED BY ULTRA-SHORT LASER PULSES

The interaction of ultrashort laser pulses with materials involves a number of spe-cial features that are different from laser-matter interaction for longer pulse durations.Depending on the time scales of the physical processes such as energy deposition, freeelectron heating and thermalisation (in the order of sub-picosecond), hot electron gascooling, energy transfer (few ps), thermal diffusion in the bulk (10 ps), thermal melting(100 ps), ablation (of the order of 1–10 ps) and plasma formation, we can distinguish dif-ferent regimes of laser pulse-matter interaction (see Figs. 1 and 2). For laser pulses witha duration longer than a few ps, the laser beam interacts with different transient statessuch as the plasma evaporated material and buffer gas above the sample surface. Forsub-picosecond pulses, the laser beam interacts only with the electron sub-system of thematerial before it undergoes any changes in thermodynamic state and material removaloccurs after the laser pulse. At such pulse durations, the physical mechanisms involvedduring the ablation process noticeably differ from those taking place with nanosecondlasers. Because of its very short pulse duration, the laser beam does not interact withthe resulting plasma. The part of laser energy absorbed is thus fully deposited intothe material at the solid density with only a little thermal diffusion while the pulseis on. The shortening of the laser pulse duration thus yields a shrinking of the heat-zone, which prevents an uncontrollable and often undesirable material modification andremoval.

0.1 101 100 1000 10000

Laser pulse duration (ps)100000

fs ps ns

Fig. 1. Scale of laser pulse duration used in LIBS applications.



154 M. Sabsabi

Melting, vaporization,plasma ignition

AblationAblation

Melting

Excitation

ns

ps

fs

Energy transfer fromelectron to atoms and ions

MetalsSemiconductors and dielectrics

Creation of free electronby multiphoton ionizationInverse

Bremsstrahlung

Fig. 2. General time scale of the various physical processes involved in femtosecond laser exci-tation from [10].

3.1. Basic Processes during Laser Ablation

Several mechanisms have been proposed to describe ultrafast laser ablation includingthermal evaporation, phase explosion, electrostatic ablation, and Coulomb explosion.However little is known about the importance of each mechanism and its dependence onthe laser fluence and material properties. Most experimental studies of subpicosecondlaser ablation deal with micromachining and a lot of work was dedicated to analyze theirradiated material surfaces. In particular, the ablation depth as a function of laser fluencehas been extensively studied. Only a few studies deal with the characterization of theplasma plume generated by ultrashort laser pulses. We will present briefly the funda-mental physical processes involved during laser ablation to make it easier to distinguishbetween the regimes of laser matter interaction based on the duration of the laser pulse.We will be focusing on the regime 1013–1014 W/cm2. For more information, we referthe reader to the excellent papers on the subject from different research groups [8–18].

When an intense femtosecond pulse (in the order of 1013–1014 W/cm2) interactswith a solid metallic target, electrons in the conduction band absorb photons and gainhigher energy through the inverse Bremsstrahlung mechanism. In semi-conductors anddielectrics electrons are excited from the occupied valence bands to empty conductionbands through photoionization [8,10,11,13]. Following ionization, the laser energy isabsorbed by free electrons due to inverse Bremsstrahlung and resonance absorptionmecanisms for specific wavelength. The absorption process is followed by thermalizationwithin the electron subsystem, energy transfer to the lattice and heat transport into thetarget due to the electron thermal diffusion. The time scale of this chain of events canbe classified as shown in Fig. 2. In order to meet the ablation conditions the averageelectron energy should increase from the initial room temperature to up to the Fermienergy, i.e., up to several eV. The electron–electron equilibration time is of the order ofmagnitude of the reciprocal electron plasma frequency, i.e., �−1

pe ∼10−2 fs which is much




shorter than the pulse duration. A distinctive feature of the ultrashort interaction modeis that the energy transfer time from the electrons to phonons or to ions by Coulombcollisions is significantly longer (∼hundreds of fs to picoseconds) than the laser pulseduration (100 fs) [8]. Therefore, the ions remain cold during the laser pulse interactionwith both metals and dielectrics. In addition, the conventional hydrodynamic motiondoes not occur during that time. Ablation takes place at pressure or temperature for whichthe electrostatic forces between electrons and ions are high enough to breakdown thematerial and to eject the ionized species. The minimum laser fluence for which ablationcan be initiated is defined as the ablation threshold or optical breakdown threshold.Expansion and ablation of the laser excited materials is a relatively slow process simplybecause it involves transport of heavy particles. Finally, the pulse duration in the sub-picosecond laser interaction with a solid target appears to be shorter than all characteristicrelaxation times: the electron-ion energy transfer time, the electron heat conduction time,and therefore the hydrodynamic or, the expansion time. Thus, the femtosecond laserpulse interacts with a solid target with a density remaining almost constant during thelaser pulse since there is no or little matter displacement during such a short pulse. Theheated matter then expands nearly adiabatically. For longer pulse durations, significantheat conduction takes place inside the target while the matter is ablated and laser energyabsorption occurs within the expanding plasma.

3.2. Material Removal and Plasma Expansion

The mechanisms governing the material removal during ultrashort laser ablation are stillsubject to intensive investigations. For typical intensities used in sub-picosecond laserablation �1012–1014 W/cm2�, maximum temperatures of several eV are achieved in thesurface layer, whose thickness is of the order of magnitude of the optical skin depth.Several processes are being discussed:

1. normal vaporization2. normal boiling3. phase explosion (explosive boiling)4. critical-point phase separation.

The first three are considered for longer-pulse laser ablation. The critical-point phaseseparation [19] is suggested as the possible mechanism for droplets formations observedduring ablation with ultrashort pulses only. All of them are termed thermal processessince they occur after the electron relaxation with the phonons, when the system isconsidered to be in a state of local equilibrium and the temperature has been established.

Fig. 3 shows a temperature-density phase diagram with several possible trajectoriesof the matter heated by an ultrashort laser pulse [19]. The material is heated isochoricallywith a rates of up to 1015 K/s into a hot, high-pressure (several GPa), fluid state witha temperature close to or well above the critical temperature (the vertical solid line ABin Fig. 3). Some of the particles in the uppermost layer have enough energy to transferinto the vapor phase directly (curve AC). The pressure gradient causes an expansion ofthe layer away from the target (and the formation of a shock wave that propagates intothe solid). At this point, the ambient pressure has little influence on the processes since



156 M. Sabsabi

10

1

Tem

pera

ture

(eV

)

Density (g/cm3)

0.1

0.010.01

Supercooledvapour

Unstablezone

Solid

VapourCritical point

Liquid

D

ESH

A

L

f g

B1

BC

0.1 1 10

Fig. 3. Temperature-density phase diagram typical for metals. Dashed line is the boundary betweenthe one-phase and two-phase regimes (binodal). Dotted line is the spinodal curve- the boundary ofabsolute instability. SHL = superheated liquid. Full lines schematically represent several possibletrajectories of heated matter during the fs-laser ablation.

it is much lower than the pressure within the hot layer. The expansion is accompaniedby adiabatic cooling (line BD). The parts of the layer material that pass through thestates close to the critical point can either be vaporized or remain close to the soliddensity. Mixing of the trajectories in the unstable zone lead to the formation of bubblesand droplets.

In the unstable zone, the smallest fluctuation will cause the rapid evolution of thethermodynamic state of the matter towards one of the two extrema of the isotherm on thespinodal curve (line fg) [19], so that either bubbles or droplets are formed. The bubblesundergo expansion through which the surrounding droplets are eventually pushed backin an explosive phenomenon. If the maximal lattice temperature is not above the criticaltemperature, the material remains in the unablated target (line B1E) [19].

The onset of the material removal described above takes place within a very shorttime after the pulse (1–100 ps): on the time scale of the plasma expansion into theambient medium (microseconds), this complete series of events can be regarded as avery brief release of energy. The plasma is initially much smaller than the distancesat which the expansion is observed. Under such conditions, several models have beendeveloped to better understand the ablation process and predict the plasma profilesand other parameters [13,19,20]. These models are based on the laws of conservationof mass, energy and momentum. The one-dimensional fluid model developed by ourgroup [12,19] includes treatment of:

(i) hydrodynamics,(ii) absorption of the laser energy,

(iii) electron thermal conduction,(iv) electron-ion energy coupling,(v) shock wave generation in ambient air and

(vi) radiative transfer and losses at the interface between plasma and air.




Fluid equations for the conservation of mass, momentum and energy are solvedthrough a one-dimensional Lagrangian scheme. The model also includes a realisticequation-of-state in order to describe thermodynamic properties (pressure and internalenergy as a function of density and temperature) over the various states of the matterranging from solid to plasma.

We show here a sample of the simulated plasma profiles for aluminum target. In theexample shown in Fig. 4a, and 4b the Gaussian laser pulse has a fluence of 20 J/cm2

and a full width at half maximum (FWHM) duration � = 100 fs and in Fig. 4c the pulseduration width was � = 100 ps.

Mas

s de

nsity

(g/

cm3 )Mass density

Ele

ctro

n te

mpe

ratu

re (

K)

107

106

105

104

1000

100–400 –200

Position (nm)

(a)100 fs – 20 J/cm2

t = 200 fs

Tc

t = 300 fs

t = 1 ps

00

1

2

3

(b)

Abs

orbe

d po

wer

(10

25 W

/m3 )

Ele

ctro

n de

nsity

(10

20 c

m–3

)

–50 0

Position (nm)

100 fs – 20 J/cm2

t = 300 fs

t = 200 fs

0

2

4

6

1000

100

10

Electron density

Critical density

Absorbed power

Fig. 4. Electron temperature (a) and absorbed power per unit volume (b) and mass/electron densityas a function of position. The profiles are taken at the time t = 200 fs, 300 fs and 1 ps after theonset at t = 0. The solid surface was initially at z = 0. The laser pulse is characterized by durationof 100 fs and a fluence of 20 J/cm2.



158 M. Sabsabi

Abs

orbe

d po

wer

(10

14 W

/cm

3 )

Ele

ctro

n de

nsity

(10

20 c

m–3

)

Position (μm)

(c)

0 1 2 3

ns

–1

2

0

4

61000

100

10

Electron density

Absorbed power

Critical density

Fig. 4c. Electron density and absorbed power per unit volume as a function of position. Theprofiles are taken at the time t = 250 ps. The solid surface was initially at z = 0. The laser pulseis characterized by a duration of 100 ps and a fluence of 20 J/cm2.

The plasma profiles are taken at 200 fs, 300 fs and 1 ps after the sudden onset att = 0. Fig. 4a and 4b show the electron density profile, the absorbed power, the electrontemperature and the mass density as function of the position z (direction of the plumeexpansion is in the ambient air for z > 0 and for z < 0 is into the target). The electrondensity profile decreases continuously from the solid density value �∼1�4 × 1023 cm3�(see Fig. 4b) near the surface at z = 0 to a value of ns = 1�8×1021 cm3 at z = 2�82 �m.The absorbed power profile showed that most of laser energy is absorbed before ablationbegins. The aluminum density is less affected by the laser pulse at 300 fs but at 1psthe matter starts to expand into the ambient air and reach few nanometers. The electrontemperature profile at 1 ps is larger than the 300 fs one’s. This broadening in the profile isdue to the electron thermal conduction, which transmit the laser energy near the surfaceto the bulk of the material.

Fig. 4c shows similarly the electron density profile and the absorbed power for longerpulse (100 ps). Fig. 4c shows a maximum near the plasma critical density, which isnc = 1�8×1021/cm3 for the (Ti:Sapphire) laser wavelength 0�8 �m of interest here. Theabsorbed power per unit volume shows oscillations with a period of approximately �/2(since the index of refraction is close to 1) due to interferences between the incidentand the reflected laser fields. In most simulation results, more energy is absorbed withinthe bell-shaped absorption profile near the critical density than in the remaining underdense plasma.

3.3. Influence of the Pulse duration on the Plasma Properties

The chirped pulse amplification (CPA) technique [21], which enables the productionof laser pulses from the sub-picosecond regime to the sub-nanosecond regime, makes




possible the investigation of the influence of the pulse duration on the plasma properties.Therefore, as demonstrated by several publications [22–25], lower ablation thresholdsand larger efficiency of material ablation can be obtained with high precision andminimal damage by using ultrashort laser pulses. In the following we will present brieflyan analysis of the results published in the literature on the ablation threshold, the ablatedmass and the diagnostics of the ultrashort laser-induced-plasma under ambient or reducedatmosphere.

3.3.1. Ablation Threshold

An essential feature of the laser ablation process is the existence of a threshold fluence(which depends on the material, wavelength, and pulse duration) below which no ablationis possible. Fig. 5 shows simulation results for the threshold fluence for ablation ofaluminum as a function of the pulse duration �. Matter is determined to be ablatedwhen a sharp jump appears in the plasma density profile, as explained in Ref. [19]. Oneobserves in Fig. 5 two distinct regimes, with a transition occurring between 1 and 10 ps.For subpicosecond pulses, the threshold fluence takes the constant value of 0�4 J/cm2,while for pulses longer than 10 ps, the threshold fluence rises as �1/2. Fig. 5 is inqualitative agreement with experimental results obtained for gold [26], fused silica, andcalcium fluoride [27]. The physical interpretation of these two regimes follows a rationalesimilar to that used for the damage threshold (interpreted as a melting point threshold),investigated in [28], that shows the same qualitative behavior as in Fig. 5. It is wellknown that the �1/2 behavior of the damage threshold for long pulse duration is due to thethermal conduction inside the target which drains the heat from the target surface [29].Comprehensive ablation experiments of metals by using solid-state femtosecond lasershave been reported [17,29], where laser fluences from 0.1 to 10 J/cm2 and pulse widthsfrom 150 fs to nanoseconds have been used.

Thr

esho

ld fl

uenc

e (J

/cm

2 )

0.10.01 0.1 1 10

Pulse duration (ps)100 1000 104

10

1 No ablation

Ablation

~τ1/2

Fig. 5. Threshold fluence for ablation as a function of the pulse duration.



160 M. Sabsabi

3.3.2. Ablated Mass

For a given laser fluence (laser energy per unit area) and wavelength, the ablationdynamics and results vary considerably with laser pulse duration [30]. Although under-standing the influence of pulse duration on ablation efficiency (i.e., on the volume ofmatter ablated per laser pulse) for a given wavelength and fluence is fundamental formost laser applications, this issue has been discussed in only a few papers to date.One of the most significant investigations was presented by Semerok et al. [31] whocarried out measurements of a variety of metals and investigated the dependence of theablation efficiency on pulse duration for only three specific values of pulse duration(150 fs, 25 ps, and 4 ns). However, various fluences and various laser wavelengths werecompared for each pulse duration. The main conclusion of that study was that the ablatedvolume per pulse and per unit laser energy in several metals, displayed as a function ofthe pulse duration, seems to show a minimum in the picosecond regime. Additionally,the same team presented ablation experiments with copper at a fluence of 21 J/cm2

with the pulse duration in the range 0.1–10 ps [32]. The results clearly showed a roughplateau in the ablated depth for subpicosecond pulses and a monotonic decrease forpulse durations longer than ∼1 ps.

The influence of the laser pulse duration on the ablation efficiency for the alum-nium was investigated theoretically using an one-dimensional code [12] and experimen-tally [33] as a function of pulse duration for a given wavelength and fluence. The ablatedvolume per pulse measured in an aluminum target for an average fluence of the order of100 J/cm2 (See Fig. 6a) shows a nearly flat plateau for subpicosecond pulses, then a dropby a factor of 1.8 with a minimum near 6 ps, and a modest increase up to subnanosecondpulse durations. In contrast, the ablated depth per pulse decreases monotonically as afunction of pulse duration for � > 1 ps. The crater diameter is constant up to 6 ps and

(a)

0.0

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

00.1 1 10 100 1000

60

80

100

120

140

160

180

200

220

240

260

Abl

atio

n de

pth

δ (µ

m)

Abl

ated

vol

ume

(µm

3 )

Laser pulse duration (ps)

(b)

0.1 1 10 100 100020

25

30

35

40

CBA

Cra

ter

diam

eter

(µm

)

Laser pulse duration (ps)

Fig. 6. (a) Measured ablated volume per pulse (filled squares) and ablated depth per pulse (opencircles) as functions of pulse duration. The scales were chosen such that the shortest pulses wouldcoincide. (b) Crater diameter as a function of the pulse duration.




increases rapidly beyond that value (See Fig. 6b). The physical interpretation given inRef. [33] is the following:

As for the ablation threshold, ablation characteristics as a function of the laser pulsewidth also show that there are two regimes, which are determined by the ratio r betweenthe thermal diffusion length of the electrons at the end of the pulse and the penetrationdepth of the laser light. For r < 1 �� < 1 ps�, the ablation depth is practically independentof the pulse duration while for r > 1 �� > 1 ps�, the ablation depth decreases as aconsequence of the energy losses due to thermal conduction, resulting in less energyremaining to accomplish ablation. For longer pulses �� > 5 ps�, an important fractionof the incoming laser energy is absorbed in the expanding plasma formed in front ofthe target surface, and this extra reservoir of energy assists ablation at later times. Theobserved behavior of the measured ablated volume shown in Fig. 6 is in qualitativeagreement with the prediction of our one-dimensional fluid model [12].

In addition, Fig. 6b clearly shows that, if one wishes to perform efficient ablationwith lateral precision, pulses of 1 ps or less are essential, whereas pulse durations largerthan 10 ps should be avoided. This advantage of subpicosecond laser pulses is now wellestablished for micromachining of a large variety of materials.

The results presented above were carried out in air at atmospheric pressure. Oneshould ask what is the influence of the ambient pressure on the ablation efficiency withultrashort laser?. The results published in the literature [12,34] indicate no dependenceof the ablated mass with pressure [34] or a slight decrease [12] in the ablation rateof copper (fluence of 8�4 J/cm2) in argon atmosphere for pressures of 0.05, 40 and700 mbar.

3.3.3. Plasma Characterization

An important part of the laser ablation process is the creation of small transient plasmasabove the sample surface that can be studied by spectroscopic methods. The plasmadynamics depends on the ambient conditions, the sample properties, and the laser param-eters: wavelength, energy and pulse duration. The ejected matter contains atoms, ionsand electrons as well as droplets and clusters. While the nanosecond laser-inducedplasma has been a subject of many investigations in the past (see the review papers onLIBS [35–39]), even six years ago the femtosecond laser induced plasmas (in the inten-sity range 1012–1014 W/cm2) were practically unexplored from the spectroscopic pointof view. Laser-induced fluorescence, plasma emission spectroscopy and direct time-of-flight (TOF) measurements were among the methods undertaken to study differentaspects of the plasma dynamics. From the obtained results, the conditions for reliable ana-lytical applications were deduced. Comparison of the LIBS spectra of plasmas induced byfemtosecond pulses with those of more familiar nanosecond ones helped in understandingthe underlying processes. Unless vacuum conditions are required for a given analyticalapplication, a noble gas in some cases should be chosen as the ambient gas in order toreduce the probability of chemical reactions with the sampled species. This choice willinfluence the plasma expansion dynamics and the diffusion properties of ablated parti-cles. While a reasonable amount of work has been carried out on the ablated mass (dueto the laser processing application) or to laser ablation coupled to other techniques ofanalysis (laser-induced fluorescence, plasma emission spectroscopy, time-of-flight), fewpapers have been published on the ultra-short laser-induced plasma properties [40–53].



162 M. Sabsabi

The main factors that influence the light emitted by the plasma are its temperature,the number density of the emitting species, and the electron density. The number densityof the emitting species (e.g. atoms, ions, molecules) depends on the total mass ablatedby the laser, the plasma temperature, and the degree of the excitation and/or ionizationof the plasma. The amount of ablated matter, in turn, depends on the absorption ofthe incident laser radiation by the surface, the plasma shielding, which is related to theelectron density of the plasma, and the laser fluence. Therefore, the knowledge of theplasma temperature and the density of plasma species are vital for the understanding ofthe dissociation–atomization, excitation, and ionization processes occurring in the plasmaand to determine the optimum experimental conditions for the quantitative use of LIBS.

The spectroscopic methods used in the determination of the temperature and the elec-tron density were described in earlier chapters of this book and will be only mentionedhere very briefly. The temperature is determined using the Boltzmann’s law assumingthe plasma is in local thermodynamic equilibrium (LTE) and the plasma is optically thinfor the lines used. The electron density has been deduced from the Stark broadening ofsuitable spectral lines.

The influence of the laser pulse duration on laser produced aluminum plasma prop-erties within the framework of LIBS applications (i.e., mostly in air at atmosphericpressure) was investigated for pulses varying from 100 fs to 270 ps [40,45], with somecomparison with data obtained at 8 ns as reported in the literature [45]. The comparisonwas carried out at constant energy density. Fig. 7 shows the temporal evolution of thetemperature and electron density of aluminum plasma produced in air at atmosphericpressure. One observes that the plasma temperature slightly increases with the pulse dura-tion (see Fig. 7a), while the electron density remains relatively constant for the typicaldelays considered in LIBS experiments (see Fig. 7b). For � > 5 ps, the laser–plasma

00 5 10 15 20 25 30 35 40

3000

4000

5000

6000

7000

8000

9000

10000

(a) (b)

τ = 100 fsτ = 500 fsτ = 5 psτ = 270 ps

Exc

itatio

n te

mpe

ratu

re (

K)

Delay (µs)0 1 2 3 4 65 7 8

0.1

1

10

Ele

ctro

n de

nsity

* 1

017 (

cm–3

)

Delay (µs)

τ = 500 fsτ = 5 psτ = 270 ps

Fig. 7. Time-resolved evolution of the excitation temperature (a) and electron density (b) ofaluminum plasmas produced in air at atmospheric pressure using various laser pulse durations (theexperimental data come from [40,45].




interaction results in plasma heating. In fact, as shown in Ref. [12], in the long pulseregime �� > 5 ps�, a significant fraction of the laser energy absorbed by the plasmaexpanding in front of the target should contribute to increasing the excitation temper-ature. This higher initial temperature lengthens the plasma cooling phase as comparedwith the sub-picosecond regime. In the latter case, the absorbed laser energy is fullydeposited in matter at the solid density and no further plasma heating takes place. Inboth cases (short and long pulses), at the end of the laser pulse, the plasma cools downby the same mechanisms, namely: (i) thermal conduction with the ambient air and theunablated target, (ii) the work done by the expanding plasma against the ambient air and(iii) radiative losses [41].

Fig. 8 shows the time evolution of the Mg (I) 285.2 nm, Al (II) 281.6 nm and thecontinuum for three different laser pulse durations. One observes significant differencesin the temporal evolution of the line emission for the various laser pulse durations

0.01 0.1 1 10 1001E–5

1E–4

1E–3

0.01

0.1

1

Continuum Mg (I) 285.21 nm Al (II) 281.60 nm

a: 500 fs

Nor

mal

ized

inte

nsity

(ar

b. u

nits

)

Delay (µs)

1E–5

1E–4

1E–3

0.01

0.1

1b: 5 ps


Nor

mal

ized

inte

nsity

(ar

b. u

nits

)

0.01 0.1 1 10 100

Delay (µs)


0.01 0.1 1 10 1001E–5

1E–4

1E–3

0.01

0.1

1

Nor

mal

ized

inte

nsity

(ar

b. u

nits

)

Delay (µs)

c: 270 ps

Fig. 8. Time evolution of the line intensity of Mg (I) 285.2, Al (II) 281.6 nm and the continuumfor three different laser pulse durations (500 fs, 5 ps, 270 ps).



164 M. Sabsabi

considered, while the rate of decrease of the continuum emission is constant. As thepulse duration increases the plasma takes longer to decay so that the radiation emissionlasts longer and the emergence of the spectral lines above continuum occurs later.Nevertheless, it is likely that laser-produced plasmas evolve through similar transientstates, the only difference being that these states, close to LTE, are reached at differentdelays after the laser shot. Therefore, temporal gating parameters appear to be keyparameters for LIBS performance optimization and must be appropriately chosen foreach laser pulse duration. As we will see later, once this optimization of the delay andintegration time is made, the laser pulse duration itself seems not to appear as a criticalpoint in LIBS science.

These results mentioned here for the temperature and electron density in the ultrashortplasma produced in air at atmospheric pressure are in agreement and in line with othervalues published in the literature [40–53] in similar conditions. It should be mentionedthat the values of temperature and electron density presented here are space-averagedvalues where no Abel inversion was used. The emission signal can be taken from the sideof the plasma or from the top or over the complete plasma volume from different views.

4. SPECTROCHEMICAL ANALYSIS BY ULTRA-SHORTLASER-INDUCED PLASMA

For spectrochemical analysis by laser-induced plasma, or any other sources of excitation,there are two important parameters considered by the chemist in the evaluation of aspectroscopic technique from an analytical point of view: the limit of detection and theprecision. In emission spectroscopy, the spectral line intensities, which are related tothe species concentrations, are strongly influenced by various parameters that cannotbe usefully controlled. Among these parameters, are the quantity of vaporized matter,the degree of ionization, which depend on the laser pulse parameters (pulse duration,wavelength, energy, beam quality, focusing conditions) and on the target characteristics(thermal conductivity, reflectivity, melting and vaporization temperature, etc.). Anotherimportant parameter is the surrounding atmosphere (pressure and composition). In thissection, we will compare the LIBS spectrochemical analysis performances by usingultra-short pulse and long pulse laser-induced plasma based on the recent literature in thefield. (Here we will deal only with ambient air as a surrounding atmosphere.). Althoughpulse duration is known to strongly affect the laser ablation dynamics as shown inSection 3.3, only a few studies have compared the characteristics of plasmas generatedwith nanosecond and sub-picosecond or picosecond pulses and discussed to what extentpulse duration will affect the analytical performances of the LIBS independently fromthe tools used to get the analytical signal [45–53].

It is natural to examine and evaluate to what extent for a given fluence the analyticalperformances of LIBS could be significantly improved by a suitable choice of laser pulseduration. In the last Section 3.3.3, we discussed the influence of the laser pulse durationon the plasma characteristics, in this section, we will be focusing on the analytical aspectsof LIBS based on the literature on the subject. For example, Le Drogoff et al. [45,50]studied the quantitative analysis of minor elements in aluminum and copper alloys forthree typical laser pulse durations, namely 90 fs, 2 ps and 270 ps, chosen to represent thethree regimes of laser-matter ablation. The 90 fs pulse represents the regime of pulses so




short that there is negligible conduction of heat into the target or movement of heatedmaterial out of the target during the laser pulse. The 2 ps pulse is representative of thesituation where, during the laser pulse, significantly more energy is conducted awayfrom the deposition (skin-depth) area, but where outward motion during the laser pulseis still negligible. Finally the 270 ps pulse, although much shorter than usual LIBS laserpulses, still represents the same situation, namely, that during the laser pulse there isconsiderable heat transfer to the interior and excessive laser-heating of the expandingplasma. Le Drogoff et al. used experimental conditions similar to Sabsabi et al. [54]with the more conventional Q-switched Nd:YAG laser operating at the fundamentalwavelength of 1064 nm and a pulse width of 8 ns.

Fig. 9 presents the calibration curves for silver in copper alloys for the two extremelaser pulse durations considered here (90 fs and 270 ps). One clearly sees that the calibra-tion curves do not vary linearly with the Ag concentration in the matrix, the nonlinearityproblem being much worse for the shortest pulse. It is only at very low Ag concentration,typically below 75 ppm, that the signal intensity varies linearly with the concentrationand falls to zero with the element concentration. This deviation at higher concentrationof the calibration curves from the linear relationship results from self-absorption due tothe resonant character of the lines considered here. Similar behaviors were observed forany of the temporal windows considered for other pulse durations.

The self-absorption mechanism depends on the concentration of the minor elementstudied in the lower level of the transition (most often the ground level) as well as onthe plasma thickness and on the oscillator strength of the transition. In the examplesof Fig. 9, self-absorption is so strong that a self-reversed effect of the line occurs, thisphenomenon being considerably more pronounced as the laser pulse duration is reduced.

3.0 × 104

2.5 × 104

2.0 × 104

1.5 × 104

1.0 × 104

5.0 × 103

0.00 50 100 150 200 250 300 350 400 450

20 J/cm2

τ = 90 fsτ = 270 ps

Inte

nsity

(cp

s)

Ag in copper alloys (ppm)

Fig. 9. Calibration curve of silver vs. its concentration in copper alloys for the two extreme laserpulse durations (90 fs and 270 ps). The temporal gating parameters were: delay time td = 2 �s andgate width tw = 5 �s. Note the severe nonlinear behavior for the short pulse and the significantlynonlinear behavior for the long pulse.



166 M. Sabsabi

This latter observation is consistent with the fact that when � decreases, the discrepancybetween the ideal linear calibration curves and the leveling of the experimental calibrationcurves increases (see Fig. 9). In addition, it is worth mentioning that using 8 ns laserpulses with similar samples, Sabsabi et al. [55] did not observe self-absorption for thisAg line �� = 338 nm�. The increase of self-absorption for sub-picosecond pulses maybe understood as a consequence of both higher ablation efficiency [13–33] and lowerplasma temperature [40–45], which tend to make the population density of the groundstate species higher thus enhancing the absorption of the resonant lines.

The use of non-resonant lines could help to minimize self-absorption and to increasethe dynamic range of the detection method to higher concentrations. Again this is the factthat non-resonant lines are usually much less intense than resonant ones (as one shouldexpect since excitation to the non-resonant lower state is likely to be rare). Clearly,when the concentration is very low, resonant lines should be used for analyte detectionsince nonlinearity is not then a problem. Since these lines are the most intense, they alsonaturally yield the lowest limits of detection (LOD).

As far as LOD is concerned, the results in the literature show the critical importance ofoptimizing the observation delay and the integration time [45]. Despite this, tremendousvariation in LODs obtained as gate windows are changed, in general, for a given element,comparable LOD values can be obtained for each of the different pulse durations providedthe best temporal window is chosen. However, provided the best gate is used for eachpulse duration, the results do not show significant evidence of an optimum pulse durationas far as LODs are concerned

Le Drogoff et al. [45] showed that, gate-optimized LOD values as low as ∼2, 14, 2,and 10 ppm were found for Cu, Si, Ag and Ni, respectively. Their values of LOD areconsistent with those obtained by Sabsabi et al. [54] of 10, 14, 1, and 10 for Cu, Si, Agand Ni respectively, for an optimal time window with td = 10 �s and tg = 10 �s, usinga 8 ns pulse duration and an energy density similar to that used in the present work.

From all these observations, it appears that, for a given pulse duration and element,the optimum gate and integration time need to be found for optimizing the LOD (seeFig. 10). Once this window is determined, there is no evidence of a particular pulseduration that would optimize the LOD, so the choice of pulse duration can be made onother grounds. Recently, similar finding were obtained by Stavropoulos et al. [52] incomparing the LOD of Al, Fe, and Si in metallic samples under nanosecond (6 ns) andpicosecond laser excitation (35 ps).

5. NON-GATED ANALYSIS BY ULTRA-SHORT LASER PULSES

Considering the lower continuum emission of plasmas produced by sub-picosecond ratherthan nanosecond laser pulses, it might be possible to perform LIBS analysis without anydetector gating (i.e. no delay and very long integration time), the protocol to be referred tohere as “non-gated”. Recent studies have shown that the non-gated spectra obtained withpicosecond and sub-picosecond laser produced plasmas show relatively low backgroundand better line thinness in comparison with nanosecond lasers [13,45,50–53]. However,according to the authors of those works, the signal-to-background ratio of the non-gatedspectrum emission is significantly poorer.




0.5–2.52–5

8–10 100 fs 500 fs 1 ps 5 ps 200 ps

0.00

20.00

40.00

60.00

80.00

100.00

120.00

Lim

it of

det

ectio

n (p

pm)

Delay - Int. time (µs)Laser pulse duration

Si(I) 288.16 nm

Fig. 10. Limit of detection obtained for silicon in aluminum alloys by using different laser pulseduration and different temporal conditions.

In order to test the analytical performances of LIBS in absence of gating, Le Drogoffet al. [45] examined this approach for three laser pulse durations representing the threeregimes (sub-pico, pico and nanosecond). For this purpose, the plasma emission wasintegrated over 200 �s from the laser trigger signal. Their results showed, as expected,that the continuum and line emission increase with �. This results from the longer plasmalifetime.

The non-gated detection limits obtained for Si and Cu in aluminum alloy and Ag incopper alloy are summarized in Table 1 for each of the three pulse durations and for the

Table 1. Non-gated limits of detection (LOD) and calibration plot data for Cu andSi in aluminum alloy samples and silver in copper alloy samples, together withbest-gated LOD values

�� Minor Element Best gated LOD (ppm) LOD (ppm)

80 fs 3�2 10�6 ± 2�92 ps Cu 1�7 8�0 ± 1�9270 ps 2�0 8�3 ± 2�5

80 fs 30�5 351�2 ± 53�82 ps Si 14�1 182�6 ± 25�4270 ps 17�2 194�9 ± 31�8

80 fs 1�4 3�7 ± 1�12 ps Ag 1�3 4�3 ± 0�9270 ps 2�2 8�6 ± 2�2



168 M. Sabsabi

best gate. The best non-gated LOD results (i.e., ∼182, 8, and 3 ppm for Si, Cu and Ag,respectively) are typically obtained for � = 2 ps, which presents the best compromisebetween a high signal-to-background ratio and a relatively low noise level. However, acomparison with the best-gate results indicates that in the case of Cu and Ag, the valuesfor limits of detection (LODs) are higher (i.e. worse) by a factor of 3–4 than thoseobtained by optimally gating the signal. In the case of Si, gated measurements allow amuch greater improvement (by about one order of magnitude).

6. CONCLUSIONS

While use of long pulse laser in LIBS analysis has matured over many years, femtosecondLIBS is still in its infancy. The general conclusions to be drawn from the literature inthe benefits of using femtosecond pulses include:

• Ultrafast excitation can improve the material interaction• Ultrafast absorption of energy reduces post ejection interactions• Heat affected zone is confined to smaller region – less vaporization of substrate• Potential for highly selective desorption-ionization

It is too early for a sound general assessment of the potential of femtosecond laser forLIBS analysis. This chapter is a first step for investigating whether these specific featuresof ultra-short laser–matter interaction may offer advantages with respect to longer pulseswhen LIBS applications are concerned. To this regard, we tried to identify the specificpotential advantages of the use of ultrshort laser pulses in LIBS analysis based on thefinding of the few works published on the subject.

We have discussed the effect of laser pulse duration on the ablation rate, ablationthreshold, plasma characteristics and analytical performance of the laser-induced plasmafor spectrochemical analysis. We have also discussed the sensitivity of LIBS analysis topulse duration for a few minor elements embedded in aluminum and copper alloys. Itappears that the dependence of the time decay of the plasma emission on the laser pulseduration requires optimizing the temporal gating parameters. The results indicated that,providing the best gate window is chosen, LIBS performance for most applications isalmost independent of the laser pulse duration (at least for the representative values usedof 80 fs, 2 ps, 270 ps). In this context, the choice of the laser system should be dictatedby considerations such as cost and robustness. To this regard, the ns laser is more robustfor field application and it is much cheaper than sub-picosecond laser.

Considering the lower continuum emission of plasmas produced by sub-picosecondrather than nanosecond laser pulses, we have examined the possibility of performingLIBS analysis without any detector gating. The results indicated that gated LIBS spectrausing picosecond or sub-picosecond lasers provide better LOD for the elements studiedin this chapter than the non-gated spectra. The degree of improvement of the LIBS sen-sitivity achieved with optimal gating rather than non-gated spectra is element-dependentfor all three laser pulse durations (80 fs, 2 ps, 270 ps).

We believe that LIBS could, however, benefit from using ultra-short laser pulses inthe framework of microanalysis and profilometry, but for reasons other than sensitivity orlimits of detection. There is a growing demand for the development of new microanalysis




techniques compatible with industrial requirements and applicable to routine analysis orindustrial applications. Femtosecond laser pulses, which provide high lateral and depthresolution of the ablation process could be advantageous for microanalysis, since thelateral and transverse thermal effects induced by nanosecond laser pulses can be avoided.Furthermore, since there is no thermal heating and no plasma shielding are involved webelieve that the accuracy of LIBS could be improved. This approach has just begun tobe investigated and further studies are needed to determine just how important theseadvantages may actually be.

Finally, it seems to me too early to establish a general assessment on the comparisonof the ns and ultra-short laser-pulse duration for LIBS analysis, however the resultsobtained by several authors in the field can be summarized as the following:

• The short pulses provide better ablation efficiency and lower threshold.• The temperature increases slightly with laser pulse duration while the electron

density is independent of it.• The optimal integration time varies with the laser pulse duration.• The performances of the LIBS in terms of sensitivity are almost independent from

the laser pulse duration if we chose appropriate optimal time conditions.• The spatial resolution obtained by fs pulses is better than ns pulses.• LOD are worse for non-gated than gated arrangements.

The accuracy is expected to be better, however, the cost of the fs laser is 10 timeshigher than ns laser.

ACKNOWLEDGMENTS

The author wishes to thank his colleagues S. Laville, F. Vidal, M. Chaker, J. Margot andL. Radziemski for their help in reading the manuscript and useful discussion. Supportfrom the NRC is also gratefully acknowledged.

REFERENCES

[1] D. Gunther, S.E. Jackson, and H.P. Longerich, Spectrochim. Acta Part B 54 (1999) 381.[2] R.E. Russo, X. Oleg, V. Borisov, and H. Liu, Laser ablation in atomic spectrometry. Encyclo-

pedia of Analytical Chemistry: Instrumentation and Applications, Wiley, Chichester, (2000).[3] N. Bloembergen. IEEE J. Quant. Elec.,Vol. QE-10, (1974) 375.[4] A. Penzkofer, D. von der Linde, A. Laubereau and W. Kaiser, Appl. Phys. Lett. 20 (1972) 351.[5] L. Radziemski, Spectrochim. Acta Part B 57 (2002) 1109.[6] W. Kaiser, Ed., Ultrashort Laser Pulses: Generation and Applications, Topics in Applied

Physics, Vol. 60, Springer, Berlin, (1993).[7] S.L. Shapiro, Ed., Ultrashort Light Pulses: Picosecond Techniques and Applications, Topics

in Applied Physics, Vol. 18, Springer, Berlin, (1977).[8] D. von der Linde, K. Sokolowski-Tinten and J. Bialkowski, Appl. Surf. Sci. 109/110 (1997) 1.[9] R. Russo, Appl. Spectrosc. 49 (1995) 14A

[10] B. Rethfeld, K. Sokolowski-Tinten, D. von der Linde, S.I. Ansimov, Appl. Phys. A 79(2004) 767.



170 M. Sabsabi

[11] E.G. Gamaly, A.V. Rode, B. Luther-Davies and V.T. Tikhonchuk, Phys. Plasmas, 9(2002) 949.

[12] S. Laville, F. Vidal, T.W. Johnston, O. Barthélemy, M. Chaker, B. Le Drogoff, J. Margot,and M. Sabsabi, Physical Review E, 66 (2002) 066415 1–7.

[13] Vanja Margetic, Ph.D thesis, Dortmund University, (2002).[14] D. von der Linde and H. Schüler, J. Opt. Soc. Am. B 13 (1996) 216.[15] S. Preuss, A. Demchuk, and M. Stuke, Appl. Phys. A 61 (1997) 33.[16] K. Furusawa, K. Takahashi, H. Kumagai, K. Midorikawa, and M. Obara, Appl. Phys. A

69 [Suppl.] (1999) 359.[17] S. Nolte, C. Momma, H. Jacobs, A. Tünnermann, B.N. Chichkov, B. Wellegehausen, and

H. Welling, J. Opt. Soc. Am. B. 14 (1997) 2716.[18] D. Günther, I. Horn, and B. Hattendorf, Fresenius J. Anal. Chem. 368 (2000) 4.[19] F. Vidal, S. Laville, M. Chaker, T.W. Johnston, B. Ledrogoff, J. Margot, and M. Sabsabi,

Phys. Rev. Lett. 86 (2001) 2573.[20] N. Arnold, J. Gruber, and J. Heitz, Appl. Phys. A 69 [suppl.] (1999) 87.[21] D. Strickland and G. Mourou, Opt. Commun. 56 (1985) 219.[22] B. Le Drogoff, F. Vidal, Y. von Kaenel, M. Chaker, M. Sabsabi, T.W. Johnston, S. Laville,

J. Margot, J. Appl. Phys. 89 (2001) 8247.[23] B.N. Chikov, C. Momma, S. Nolte, F. von Alvensleben and A. Tünnermann, Appl. Phys. A

63 (1996) 109.[24] A. Semerok, C. Chaléard, V. Detalle, J.-L. Lacour, P. Mauchien, P. Meynadier, C. Nouvellon,

B. Sallé, P. Palianov, M. Perdrix and G. Petite, Appl. Surf. Sci. 138/139 (1999) 311.[25] V. Margetic, A. Pakulev, A. Stockhaus, M. Bolshov, K. Niemax and R. Hergenröder, Spec-

trochim. Acta B55 (2000) 1771.[26] P.P. Pronko, S.K. Dutta, D. Du, and R.K. Singh, J. Appl. Phys 78 (1995) 6233.[27] B.C. Stuart, M.D. Feit, A.M. Rubenchik, B.W. Shore, and M.D. Perry, Phys. Rev. Lett. 74

(1995) 2248.[28] P.B. Corkum, F. Brunel, N.K. Sherman, and T. Srinivasan-Rao, Phys. Rev. Lett. 61

(1988) 2886.[29] H. Jacobs, B.N. Chichkov, C. Momma, S. Nolte, F. von Alvensleben and A. Tunnermann,

Appl. Phys. A 63 (1996) 109.[30] C. Momma, S. Nolte, B.N. Chichkov, F. von Alvensleben, and A. Tünnermann, Appl. Surf.

Sci. 110 (1997) 15.[31] A. Semerok, C. Chaléard, V. Detalle, J.-L. Lacour, P. Mauchien, P. Meynadier, C. Nouvellon,

B. Sallé, P. Palianov, M. Perdrix, and G. Petite, Appl. Surf. Sci. 139 (1999) 311.[32] B. Sallé, O. Gobert, P. Meynadier, M. Perdrix, G. Petite, and A.Semerok, Appl. Phys. A 69

(1999) S381.[33] B. Le Drogoff, F. Vidal, S. Laville, M. Chaker, T. Johnston, O. Barthélemy, J. Margot, and

M. Sabsabi, Appl. Opt. 44 (2005) 278.[34] S. Yalcin, Y. Y. Tsui and R. Fedosejevs, J. Anal. At. Spectrom. 19 (2004) 1295.[35] D.A. Rusak, B.C. Castle, B.W. Smith and J.D. Winefordner, Crit. Rev. Anal. Chem. 27

(1997) 257.[36] E. Tognoni, V. Palleschi, M. Corsi and G. Cristoforetti, Spectrochim. Acta B57 (2002) 1115.[37] E.H. Evans, J.A. Day, W.J. Price, C.M.M. Smith, K. Sutton and J.F. Tyson, J. Anal. At.

Spectrom. 18 (2003) 808.[38] W.-B. Lee, J. Wu, and Y-I. Lee, J. Sneddon, Appl. Spectrosc. Rev. 39 (2003) 27.[39] J. Vadillo, and J.J. Laserna, Spectrochim. Acta Part B. 59 (2004) 147.[40] B. Le Drogoff, J. Margot, F. Vidal, S. Laville, M. Chaker, M.Sabsabi, T.W. Johnston, and

O. Barthélemy, Plasma Sources Science & Technologies. 13 (2004) 223.[41] S. Laville, F. Vidal, T.W. Johnston, M. Chaker, B. Le Drogoff, O. Barthélemy, J. Margot,

and M. Sabsabi, Phys. of Plasmas. 11 (2004) 2182.




[42] R.E. Russo, X.L Mao, C. Liu and J. Gonzalez, J. Anal. At. Spectrom. 19 (2004) 1084.[43] V. Margetic, K. Niemax, and R. Hergenröder, Spectrochim. Acta B56 (2001) 1003.[44] F. Vidal, S. Laville, T.W. Johnston, O. Barthélemy, M. Chaker, B. Le Drogoff, J. Margot

and M. Sabsabi, Spectrochim. Acta B56 (2001) 973.[45] B. Le Drogoff, J. Margot, M. Chaker, M. Sabsabi, O. Barthélemy, T.W. Johnston, S. Laville,

and F. Vidal, Spectrochim. Acta B56 (2001) 987.[46] V. Margetic, T. Ban, F. Leis, K. Niemax and R. Hergenröder, Spectrochim. Acta B58

(2003) 415.[47] S.M. Angel, D.N. Stratis, K.L. Eland, T. Lai, M.A. Berg, and D.M. Gold, Fresenius J. Anal.

Chem. 369 (2001) 320.[48] K.L. Eland, D.N. Stratis, D.M. Gold, S.R. Goode, and S.M. Angel, Appl. Spectrosc. 55

(2001) 286.[49] G.W. Rieger, M. Taschuk, Y.Y. Tsui, and R. Fedosejevs, Spectrochim. Acta B58 (2003) 497.[50] Le Drogoff, M. Sabsabi, M. Chaker, J. Margot, O. Barthélemy, T.W. Johnston, S. Laville,

and F. Vidal, Appl. Spectrosc. 58 (2004) 122.[51] J.-B. Sirven, B. Bousquet, L. Canioni and L. Sarger, Spectrochim. Acta B59 (2004) 1033.[52] P. Stavropoulos, C. Palagas, G.N. Angelopoulos, D.N. Papamantellos and S. Couris, Spec-

trochim. Acta B59 (2004) 1885.[53] D. Grojo, J. Hermann and A. Perrone. J. Appl. Phys. 97 (2005) 063306 (1–9).[54] M. Sabsabi and P. Cielo, Appl. Spectrosc. 49 (1995) 499.



Chapter 8

Micro-LIBS

M. T. Taschuk, I. V. Cravetchi, Y. Y. Tsui and R. Fedosejevs

Department of Electrical and Computer Engineering, University of Alberta,Edmonton, Alberta T6G2V4, CANADA

1. INTRODUCTION

MicroLIBS (�LIBS) is a new growing area of Laser Induced Breakdown Spectroscopywhich employs �J energy laser pulses for excitation of plasma emission. Such �Jenergy pulses are required to carry out 1D, 2D or 3D microanalysis of material surfaceswith spatial resolutions approaching micron scale sizes laterally and nm scale sizesin depth. These pulses allow sampling of very small volumes �∼10–1000 �m3� andmasses �∼10 pg−ng�. �LIBS is also applicable where modest limits of detection, lowcost or portable systems are required. The use of femtosecond pulses for LIBS overthe past decade has in some cases also employed �J laser pulses [1–4] and many ofthe advantages of �LIBS are also observed using such ultrashort pulses as described inChapter 7 on femtosecond LIBS. In this chapter we will review the capabilities of LIBSas one scales to microjoule laser pulse energies and progress to date in the applicationof such systems.

The development of �LIBS has been driven by two factors:

1) the desire to obtain higher spatial resolution when carrying out 2D scans of materialsurfaces and

2) the development of high repetition rate compact microchip lasers leading to idealsources for very low energy LIBS applications.

It has been found that the plasma and continuum emission decreases significantlywith lower pulse energies and thus one can obtain reasonable performance without usingtemporal gating. The term �LIBS has also been used in the context of applications withablation spots of micron scale size [5,6]. In most cases, the definitions based on �Jenergy and micron resolution spot sizes are equivalent. This chapter will focus on thosestudies which have used pulse energies less than 1 mJ.

In the early 1990s Zayhowski developed the microchip laser [7–9] and he and otherresearchers started applying it to material analysis using both LIBS and laser-inducedfluorescence detection [9–11]. Bloch et al. were able to achieve limits of detection




174 M. T. Taschuk et al.

(LOD) of the order of 100 to 1000 ppm without temporal gating for Pb, Cu and Fe insoils using pulse energies of several �J [10]. More recently �J fiber laser sources withpulse lengths of 10 ps [12] to several nanoseconds [13] have been developed.

By the mid 1990s several groups had started to investigate the application of �Jpulses to LIBS in variety of studies [9,14–17]. In 1996, Geertsen et al. demonstratedthe use of 30–70 �J pulses for the microanalysis of aluminum alloys [15]. The authorsused 266 nm pulses for a variety of studies including LOD of minor elements downto 10 ppm, relative standard deviation (RSD) of ∼10% and lateral resolution to 6 �m.Sallé et al. [18,19] carried out studies of the crater diameters, ablation volumes andexpansion plumes for the interaction of 248 nm and 266 nm pulses with energies of65 to 130 �J and at 532 nm with energies of 10 �J to 4 mJ. Semerok et al. extendedthese studies to look at the scaling of ablation craters with pulse length for durationsof femtoseconds, picoseconds and nanoseconds and energies down to 10 �J [20,21].�LIBS has also been combined with scanning probe fiber tip microscopy to achievemicron scale size ablation spots by Kossakovski et al. [22]. However, in the later casethe emission was not strong enough to give good species identification for submicronablation spots using a non-gated and non-intensified camera. Rieger et al. [23] exploredthe scaling and optimization of measuring trace constituents in aluminum alloy for 50to 300 �J laser pulses at 248 nm as a function of gate time. They achieved LODs of 2to 450 ppm for elements of Mg to Fe for the case of optimized gate times and 200 �Jpulses. Further studies by the same group [24] compared LIBS signals for picosecondversus nanosecond 248 nm pulses. Scaling of line emission, continuum emission andemission lifetime with pulse energy were characterized with reported energy thresholdsfor observable line emission of 1�J for ns pulses and 0�1 �J for 50 ps pulses. Aboveenergies of 3 �J the characteristics of the LIBS emission were reported to be comparablefor both pulse lengths when identical laser and focusing conditions were used. In recentwork by Gornushkin et al. [25], studies were carried out with a 7�J 1064 nm microchiplaser using a non-gated and non-intensified detector. The authors highlight many of theadvantages of using microchip lasers, such as good mode quality of the beam, highshot-to-shot reproducibility, the high repetition rate, the low continuum emission andthe possibility of using ungated detectors. Observation of line reversal in the emissionspectra of Zn demonstrate that optically thick plasma conditions can exist for majorconstituents even for low energy microplasmas and observed signals and crater sizes forseveral metals were reported [25]. LODs of the order of a few percent were obtained formetallic samples but poorer sensitivity was observed for pelletized graphite samples.

Studies of surface mapping using �LIBS also began in the mid-1990s. Häkkänenet al. [14] used 200 �J 308 nm pulses to map Ca and Si concentrations in surfacecoatings of paper. They found good correlation with measurements of the surface usinglaser induced fluorescence. The group of Laserna et al. started their studies on surfacemapping using �LIBS with the investigation of depth profiling of a TiO2 antireflectioncoating on silicon [16] and 2D mapping of carbon impurities [26] using 400 �J pulsesat 337 nm. Further investigations indicated depth resolution of the order of 40 nm forcarbon impurity and demonstrated 3D scans of carbon contamination with 70 �m lateralresolution and 160 nm depth resolution [27]. Recently Menut et al. [5] demonstrated 2Dsurface scans with 3 �m spatial resolution using 5 �J pulses at 266 nm and LODs inthe percent range for mapping of the concentration of minor constituents on the surfaceof steel. They also reported that ablation probe spots down to 1 �m were possible but



Micro-LIBS 175

that there was not sufficient signal to allow for measurements of the trace elements atthe 1% level. Cravetchi et al. [28] studied crater size, line emission and RSD of signalsfrom trace elements in aluminum demonstrating that �LIBS can resolve different typesof micron scale size precipitates in aluminum. RSDs less than 10% were obtained with7 �J pulses at 266 nm. Full 2D scans mapping precipitate distribution in aluminum weresubsequently demonstrated by the same group with a lateral resolution of 10 �m [29].Redeposition of ablated material and cross contamination of a scanned surface has beennoted by a number of different groups [15,19,22,25,27,29,30].

In the late 1990s, the analysis of elemental chemical content of liquid samplesusing �J pulses with the goal of measuring the elemental contents of single cells wasdemonstrated by Ho and Cheung et al. [17,31]. Using 80–250 �J pulses at wavelengthsof 532 nm and 193 nm they demonstrated a LOD of 50 ppm for Na in water. An acousticnormalization which corrected for shot-to-shot variations in pulse energy and carefulspatial sampling of the expansion plume improved sensitivity to the few ppm range.

During the first decade of work in the �J energy regime many features of �LIBS havebeen identified and characterised as described in more detail below. In the followingsections, microjoule laser sources and their application to �LIBS are briefly reviewed inSection 2, the scaling of LIBS to �J pulse energies is discussed in Section 3, and finally,a review of the demonstrated applications of �LIBS to date is given in Section 4.

2. MICROJOULE LASER SOURCES

While traditional lasers can be operated in the microjoule range, one of the earliestsources specifically designed as a microjoule pulse source was the microchip laser deve-loped by the group at MIT [7–9,11,32–35]. Additional sources for the �J energy regimehave also been developed by other groups [13,36–40]. At the same time femtosecondlaser sources were developed, many of which also operate at microjoule energy levels.Studies of femtosecond LIBS are covered in Chapter 7 and thus femtosecond lasersources will not be discussed here. Recently, fiber optic oscillators and amplifiers havebeen developed to the point that �J output energies are obtainable in pulsed operationmode and offer a potential new option for robust sources which can be used in fieldportable LIBS systems.

2.1. Microchip Lasers

In 1989 Zayhowski et al. reported on the development of a single frequency microchiplaser in various different lasing materials [7]. Q-switched operation of the laser wasdeveloped using piezoelectric, electro-optic and passive techniques [8,33,34]. The outputat the fundamental wavelength is polarized and frequency conversion of the output andNd:YAG laser harmonics down to 213 nm have been demonstrated [9]. When using∼1 W pump power, output pulse energies of 8 �J at the fundamental wavelength, 3�5 �Jat 532 nm and 0�7 �J at 266 nm were reported. Higher pulse energies have since beenreported with 10 W of diode pump power resulting in pulse energies of up to 250 �Jand 310 ps pulsewidths at the fundamental wavelength of 1064 nm and 12 �J at 266 nmoutput with kHz repetition rates [11,35]. The layout of a low power harmonically




FiberNd:YAG

Cr:YAG

KTPBBO

Window

Fig. 1. Schematic of a UV harmonically converted passively Q-switched microchip laser. Entiredevice is about a cm across. (Reproduced with permission from Zayhowski [11]).

converted microchip laser is shown in Fig. 1. It is fabricated by bonding the gain mediumto a saturable absorber and harmonic conversion crystals. In a typical configuration a0.75 mm thick Nd:YAG gain medium is coupled to a 0.5 mm thick Cr:YAG saturableabsorber [34]. Diode pump laser light of ∼1 W at 808 nm is coupled to the gain mediumby a butt coupled fiber. The resonator is formed between a dichroic dielectric mirror at thefiber input face, with a high reflectivity at the laser wavelength and high transmissivityat the pump wavelength, and a partially transmitting mirror at the output face of thesaturable absorber. KTP and BBO crystals ∼5 mm long are butt coupled to the outputface to generate 2nd and 4th harmonic output respectively. The laser output is a singlefrequency TEM00 Gaussian mode with a diameter of the order of 50 �m. The shortcavity length ensures single longitudinal mode operation since only one axial mode hassufficient gain to exceed the lasing threshold within the laser bandwidth.

Other groups have also developed similar microchip lasers with various gain mediaand geometries. Fluck et al. [41] passively modelocked an Er-Yb:Glass gain mediumusing a semiconductor saturable absorber mirror (SESAM) to achieve 4 �J pulses at1535 nm with a repetition rate of 320 Hz. Spuhler et al. [42] applied a SESAM to aYb:YAG laser, producing 1�1 �J pulses at 1030 nm with a repetition rate of 12 kHz.Feldman et al. [43] used a 4 mm Nd:YAG microchip crystal bonded to a 2 mmCr:CaYAG saturable absorber crystal in a 31 mm external resonator to produce 50 �Jpulses at 1064 nm. Karlsson et al. [44] have produced 12 �J at 1535 nm with a 1 mmEr-Yb:Glass microchip laser with an external acousto-optic Q-switch and cavity mirror.Druon et al. [45] achieved ∼9 �J pulses at 1�06 �m and ∼0�8 �J pulses at 355 nm withpulse durations of 300 ps using a Nd:YAG microchip laser together with a double-passmicrochip amplifier. Higher repetition rate picosecond to subnanosecond pulsewidthmicrochip lasers with submicrojoule output energies have also been developed usingSESAMs at 1�06 �m [46] and 1�34 �m [47]. Hansson et al. used a low voltage multiplequantum well electro-absorption Q-switch system applied to an Er-Yb:Glass laser togenerate output pulses up to 470 nJ at a repetition rate of 10 kHz [48]. Further scalingin pump energy or addition of an amplifier chip should allow an increase in the outputenergy for some of these systems.



Micro-LIBS 177

Microchip lasers have several attractive features for LIBS as pointed out by a numberof authors [11,25]. They are compact, robust and relatively inexpensive. Because ofthe very short cavity length, the longitudinal mode spacing can be larger than thegain medium bandwidth and only a single narrow linewidth longitudinal mode will begenerated. High repetition rates of 1 to 20 kHz can be obtained by passive Q-switchingwhich can result in sub-nanosecond pulses making it easier to achieve the breakdownthreshold for materials compared to several nanosecond pulses with the same energy.Active Q-switching can be used to set exact repetition rates and synchronize to externalevents at the cost of somewhat longer pulse durations. The pulse to pulse stability ofmicrochip lasers is in the range of 0.05 to 0.5% [11,49]. Single transverse TEM00 modeoutput is readily achieved via gain guiding and M2 values of 1.0 to 1.3 have beenobtained [32,35]. This leads to low divergence output which can be focused to diffractionlimited spot diameters. Various output wavelengths in the range of 1030 nm to 1550 nmhave been demonstrated. With low energy pulses, 1550 nm pulses can fall in the eyesafe operation range which is an important advantage for system use in public areas.

There remain some disadvantages of microchip lasers with respect to their use for�LIBS. When converted to UV wavelengths microchip lasers still have limited energies,on the order of 1 to 10 �J per pulse. Further, when the simplest technique of passiveQ-switching is used the laser output is free running, making it difficult to synchronizegated detectors. Additionally, the repetition rates of passively modelocked microchiplasers may be too fast for some applications. It has been reported that in the case ofgraphite the damage from one pulse may modify the surface for the subsequent pulse,decreasing the reliability and sensitivity of the measurement [25]. Commercial versionsof microchip lasers are currently available with output energies of the order of tenmicrojoules at 1064 nm. It is expected that output energies from commercially availablemicrochip lasers will soon be sufficient to exploit the full capabilities of �LIBS. Suchlasers should lead to the design of compact LIBS units.

2.2. Microjoule Fiber Lasers

High power modelocked fibre lasers offer another potential laser excitation source for�LIBS. Fiber lasers have undergone intense development for applications in communi-cations and recently with the advent of cladding-pumped large mode area (LMA) fibersit is possible to achieve �J to mJ pulse energies. Erbium doped fibers at 1550 nm areof particular interest since they are eye safe at low microjoule pulse energies. Recently,acousto-optic Q-switching of LMA Er-doped [37] and Yb-doped fibers [38] have demon-strated close to diffraction limited transverse mode quality output pulses with pulsedurations of 100 ns, pulse energies of 500 �J and 700 �J, and repetition rates of 400 Hzand 2 kHz, with output wavelengths of approximately 1550 nm and 1060 nm respectively.Passive Q-switching of Er-Yb co-doped fiber has also been demonstrated yielding shorter3.5 ns, 60 �J output pulses at around 1550 nm with 0.6 to 6 kHz repetition rates [13].Active seeding with a pulsed CW diode laser injected into a multi stage erbium fiberamplifier has led to 118 �J 1550 nm pulses with a duration of few nanoseconds [36] andmore recently seeding with a thin disk laser source yielded longer but higher energydiffraction limited pulses of 4 mJ and 50 ns duration at 1060 nm from LMA Yb-dopedfiber [40]. In the latter case undoped plain fused silica end caps were fused onto the ends




of the fiber to allow the mode to expand before exiting to avoid damage on the fiberend faces. Harmonic conversion of the nanosecond output pulses from both acousto-optically modelocked and CW diode-laser-seeded Er-doped fiber systems has also beendemonstrated using periodically poled Lithium Niobate crystals yielding peak 2nd har-monic conversion efficiencies to 768 nm pulses of 83% and 62% respectively, a peak2nd harmonic energy of 80 �J in 45 ns pulses and 3rd harmonic conversion efficienciesof 15% [50]. Generally the acousto-optically Q-switched systems have pulse lengths oftens of nanoseconds which is longer than the optimum pulse length of picoseconds toa nanosecond for �LIBS applications. The alternative approaches of passively mode-locking and amplification of a short seed pulse allow much shorter pulses. However,the damage fluence levels of the fibers in the nanosecond regime scale with the 0.5power of pulse length given by heat diffusion scaling. Thus, shorter pulses are lim-ited to lower maximum energies. Even so, the amplification of 0.8 ns pulses to 1.2 mJhas been demonstrated in a high power chirped-pulse-amplification femtosecond lasersystem at 1055 nm [39] and 60 �J pulses have been generated by passive modelock-ing at 1550 nm [13], indicating that sources with nanosecond duration are possible atthe 100 �J level. Recently work has started on the development of high-pulse-energyhigh-repetition-rate picosecond fiber sources with 0�6 �J, 10 ps pulses at 1064 nm ampli-fied at an 80 MHz repetition rate in a Yb-doped holey fiber system. These pulses werealso frequency doubled with 50% efficiency to 532 nm. It is expected that by usinglower repetition rate seed sources the pulse energy should increase leading to 10 ps pulsesources with energies in the range of microjoules.

Recently a guide fiber has been employed for coupling light to micron scale sizespots onto a sample for �LIBS analysis [22]. While fiber laser systems have not yetbeen applied to �LIBS studies it is expected that they will soon become useful in LIBSmicroanalysis.

3. SCALING LIBS TO MICROJOULE ENERGIES

Over the past decade a basic understanding has been developed of the scaling of theperformance of LIBS systems to �J energies. It has been found that the duration ofthe line and continuum emission along with the relative amount of continuum radiationdecreases as one goes below 1 mJ excitation energy. In many cases the signal to noiseratio (SNR) is a weak function of energy and thus it is still possible to obtain goodsensitivity if care is taken in collecting the emission light. This means that working withungated detectors becomes possible which greatly simplifies the detector requirementsand reduces system cost. However, the highest sensitivities are achieved using gatedsystems. Due to the submillimeter size of the plasmas obtained in �LIBS a large fractionof the plasma emission can be coupled to the narrow input slit of grating spectrometersystems. As the pulse energy decreases, the craters produced in �LIBS decrease indiameter and depth. The smaller sample areas achieved allow the probing of muchsmaller features approaching a micron in size for microanalysis applications. However,there is a tradeoff between sensitivity and sample area that must be taken into accountfor any given application. In the following section the scaling of these properties isdiscussed in detail.



Micro-LIBS 179

3.1. Plasma Emission and Lifetime

Many materials have been examined using �LIBS, including Si photovoltaic cells, paperand paper coatings, and various metals. These studies have been performed across arange of energies from 400 �J [16] down to 0�1 �J [24]. A spectrum typical of what canbe obtained using �LIBS is given in Figure 2. This spectrum of aluminum was takenusing a single 8 �J pulse with zero gate delay and a gate width of 200 ns.

Emission scaling with energy has been studied using a photomultiplier with a bandpassfilter centered at 289 nm and a collection angle of f/6. Single line emission has beendetected from Si down to 1 �J with 10 ns 248 nm pulses and down to 0�1 �J with 50 ps248 nm pulses [24]. The resultant signal strengths are shown in Fig. 3 as a function ofpulse energy for these two pulse lengths. It is seen that above an energy of approximately3 �J the signals are of the same strength. Only as the breakdown threshold is approacheddoes one see a difference in the emission. Emission is observed for shorter pulses at lowerenergies while emission disappears for longer pulses because the intensity is no longersufficient to breakdown the target surface. Thus, above several microjoules the importantvariable appears to be energy fluence rather than intensity. The focal spot diameter wasapproximately 5 �m for these experiments leading to a fluence of approximately 5 J cm−2

for an energy of 1 �J. The vertical scale units for Fig. 3 correspond approximately tophotons per steradian except above ∼3×107 when the photomultiplier became weaklysaturated. A more efficient optical collection system and more sensitive photomultiplierdetector should be able to detect signals at even lower energies.

The scaling of peak emission time versus pulse energy has been studied by Häkkänenet al. [14] and Rieger et al. [24] for dielectric and metallic targets respectively. Thescaling for the former case is shown in Fig. 4 indicating that optimum measurement

0

1e+07

2e+07

3e+07

4e+07

5e+07

6e+07

7e+07

385 390 395 400 405

Tim

e in

tegr

ated

spe

ctra

l int

ensi

ty[P

hoto

ns S

r–1 n

m–1

]

Wavelength [nm]

Fig. 2. Background corrected spectrum of aluminum plasma emission using a single 50 ps, 8 �Jpulse at 248 nm, with a gate width of 200 ns and zero gate delay. The spectrum was obtainedusing a system for which an absolute calibration was performed. (Reproduced with permissionfrom Rieger et al. [24]).




108

107

106

105

1040.1 1 10

Energy [µJ]

PM

T s

igna

l [A

.U.]

50 ps10 ns

Fig. 3. Filtered photomultiplier detection of silicon line emission at 288 nm as a function of laserpulse energy for 10 ns and 50 ps pulses at 248 nm. The focal spot diameter was approximately5 �m, yielding a fluence of ∼5 J cm−2 for 1 �J pulse energies. The horizontal line represents thenoise floor of the PMT. (Reproduced with permission from Rieger et al. [24]).

0.0

0.2

0.4

0.6

0.8

1.0

0 200 400 600 800 1000

Nor

mal

ized

SB

R

Time [ns]

0.2 mJ

0.7 mJ

3 mJ

9 mJ

(a)

0

100

200

300

400

500

600

0.1 1 10

Dec

ay ti

me

[ns]

Energy [mJ]

CaSi

(b)

Fig. 4. (a) Smoothed and normalized time-resolved signal-to-background ratios of the silicon lineat 251 nm at various excitation energies using a 308 nm laser. (b) Decay time for calcium andsilicon signals at 422 nm and 251 nm respectively as a function of laser pulse energy. (Reproducedwith permission from Häkkänen et al. [14]).

times for peak signal to background ratio decreases to about 100 ns for 200 �J pulses inline with the decrease in plasma emission decay time with pulse energy. The results ofRieger et al. [24] shown in Fig. 5 indicate that the emission decay time reduces further toa few nanoseconds as the pulse energy is reduced below 2 �J for 10 ns pulses and below0�3 �J for 50 ps pulses. For energies above 3 �J an expanding spherical plasma with alifetime of tens of ns is formed for both ps and ns pulses leading to similar decay timeconstants for the emission. Similar observations have been reported by other authors withplasma emission decay times of 8 ns [25] to 15 ns [10] for 10 �J 1064 nm subnanosecondmicrochip laser pulses. Gornushkin et al. also observed a prompt emission which wasonly slightly longer than the subnanosecond laser pulse [25].



Micro-LIBS 181

1

10

100

0.1 1 10

Tim

e co

nsta

nt [n

s]

Energy [µJ]

50 ps

10 ns

Fig. 5. Decay time constant for silicon line emission at 288 nm as a function of pulse energy for10 ns and 50 ps pulses at 248 nm. The horizontal line represents the rise time of the PMT used.(Reproduced with permission from Rieger et al. [24]).

It has been observed that the decreasing decay time of the line emission is matched byan even faster decay of continuum emission. As a result, the ratio of line to continuumemission improves as one reduces pulse energy and thus one can detect the LIBS signaleven in the absence of a gated detector [16,24,25].

3.2. Crater Size – Lateral and Depth Resolution

One of the important advantages of �LIBS for microanalysis is the size of the ablationspot as compared to conventional LIBS. Several groups have studied the scaling ofcrater diameter or crater volume as a function of pulse energy [15,18,19,21,25,27,28,51].However, one must distinguish between the detection region, which is ionized sufficientlyto yield emission signals, and the total region, which is ablated by the laser pulse. Muchof the ablated material is removed after the laser pulse by the shock wave and melt wavepropagating into the target. The crater size will therefore represent an upper bound tothe actual region probed in composition measurements.

A careful test of the lateral resolution obtainable by �LIBS was performed by Geertsenet al. [15] using a specially fabricated test sample with a sharp Cu/Al interface. Usingpulse energies in the range of 35–40 �J at 266 nm, a series of shots were spaced at 2 �mintervals measured perpendicular to the interface. The sample was displaced parallel tothe interface by 15 �m between each shot to prevent previously ablated material frombeing resampled. The data from the experiment is shown in Fig. 6. The reported lateralresolution was ∼6 �m.

Kossakovski et al. [22] used 12�5 �J pulses at 337 nm coupled to an etched fiber probetip in a scanning probe microscope to investigate the surface of basalt and meteoritesamples. They were able to produce submicron ablation spot diameters but noted thatthe corresponding emission signals were too weak to obtain useful LIBS signals using a




0

0.2

0.4

0.6

0.8

1

1.2

0 1 2 3 4 5 6 7 8

Nor

mal

ized

inte

nsity

[A.U

.]

Relative position of the focal spot [µm]

Cu signal Al signal

Fig. 6. �LIBS 1D scan using pulses in the range of 35–40 �J at 266 nm across a Al-Cu interfacefor determination of lateral resolution. Special care was taken to prevent resampling of ablatedmaterial, and a lateral resolution of 6 �m was reported. (Reproduced with permission from Geertsenet al. [15]).

20X microscope objective and viewing the emission plasma from the side. Useful LIBSsignals were obtained when using spots greater than a micron in diameter.

The scaling of crater size and volume is an important variable in �LIBS applications.Aluminum is one of the most thoroughly studied materials in the �LIBS literature, andprobably represents the clearest dataset with which to investigate the scaling laws forsample volume. Measured crater diameters and volumes for aluminum using nanosecond�J pulses are shown in Fig. 7 and Fig. 8. Given the different methods of defining crater

0

5

10

15

20

25

30

35

0 50 100 150 200 250 300

Cra

ter

diam

eter

[µm

]

Pulse energy [µJ]

CravetchiGeertsen

GornushkinSalle

Fig. 7. Single-shot crater diameter as a function of energy for Al. Open circles represent shotsusing 248 nm pulses as described in [23]. Data from Cravetchi et al. [28], Geertsen et al. [15] andSallé et al. [18] taken with ∼10 ns pulses at 266 nm and data from Gornushkin et al. [25] takenwith 1064 nm are shown as solid points.



Micro-LIBS 183

0

2000

4000

6000

8000

10000

12000

0 50 100 150 200 250 300

Cra

ter

volu

me

[µm

3 ]

Pulse energy [µJ]

GeertsenSalle

Fig. 8. Single-shot crater volume as a function of energy for Al. Open circles represent shotsusing 248 nm pulses as described in [23]. Data from Geertsen et al. [15] and Sallé et al. [18] takenwith ∼10 ns pulses at 266 nm. The line is a linear regression to the 248 nm data points which areabove 25 �J and below 300 �J.

diameter and of measuring crater volume used in the literature the agreement betweengroups is quite good.

Ablation efficiency ��m3 �J−1� has been measured for a variety of metals underdifferent conditions. Results which used nanosecond pulses are given in Table 1, andpicosecond results are given in Table 2. There are significant variations which maybe due to the different focal geometries and intensities employed. While reasonableagreement for crater size scaling in the range applicable to �LIBS has been achievedin the literature between a number of groups, further work will be required to reach aconsensus on the scaling of ablation efficiency.

The volume which actually contributes to LIBS emission is expected to be smaller andshallower than the final ablation crater. These effects will depend on the focal spot pro-file as well as the material characteristics. Redeposition both immediately surrounding the

Table 1. Nanosecond Ablation Efficiencies ��m3 �J−1�

Author � Pulse-width Al Cu Fe Ni Pb Mo

Rieger [51] 248 nm 10 ns 30Geertsen [15] 266 nm 6 ns 9�8Semerok [20] 266 nm 4 ns 29�3 6�5 3�1 45�7Semerok [21] 266 nm 6 ns 6 2 1 3 9 1�4

Sallé [19] 532 nm 6 ns 4�9 1�93 6�11Semerok [20] 532 nm 4 ns 12�4 3�1 1�5 18�6Semerok [21] 532 nm 6 ns 5 2 6

Gornushkin [25] 1064 nm 0.55 ns 200Semerok [21] 1064 nm 6 ns 5 2 0�9 0�7 6 0�6




Table 2. Picosecond Ablation Efficiencies ��m3 �J−1�

Author � Pulse-width Al Cu Fe Ni Pb Mo

Semerok [21] 266 nm 25 ps 5�7 2�8 0�6 0�9 21�3 0.5Semerok [21] 532 nm 25 ps 4 0�9 0�4 0�7 12�5 0.7Gornushkin [25] 1064 nm 550 ps 200Semerok [21] 1064 nm 25 ps 1 0�4 0�45 0�6 2�0 0.3

crater and at further distances is an additional phenomenon which affects the effectivelateral resolution that may be achieved by �LIBS. Under some conditions ablated mate-rial may redeposit and contaminate unsampled areas as discussed in Section 4.2 below.Further work will be required to ascertain the actual source volume contributing to theLIBS signal observed, and the ultimate spatial resolutions that will be possible in �LIBS.

3.3. Limits of Detection

Only a few studies have reported LODs for elements using �LIBS. Geertsen et al. useda frequency-quadrupled Nd:YAG with pulse energies of ∼40 �J to measure LODs ofminor elements in aluminum [15]. Specially prepared homogeneous aluminum targetswere used for this work. In an alternative approach Rieger et al. [23] took advantageof the small probe spot size to probe only the matrix material in standard aluminumalloys for LOD measurements. The concentration of minor constituents in this matrixregion was calibrated using electron probe microanalysis of the matrix region of thealloys. In the latter measurement the SNR was obtained by taking the peak line emissioncompared to the 3 noise in nearby regions of the spectrum without line emission.For trace elements at concentrations below ∼1% it was assumed that the signal scaleslinearly with concentration. A comparison with the traditional technique for determiningLOD was performed and reasonable agreement between the techniques was obtained.The SNR was measured versus gate delay times and the optimum gate delay found forthe given plasma conditions for a number of trace elements.

An example of the 3 LOD for Cu in aluminum alloy as a function of pulse energyand gate delay is given in Fig. 9. It is seen that the LOD and optimum gate time areweak functions of pulse energy. The optimum LODs for a number of elements weredetermined and are presented in Table 3 together with values reported by Geertsenet al. [15]. A typical set of values for mJ energy LIBS measurements from Sabsabiet al. [52] is also given for comparison. To compare values taken with different numberof shots it was assumed that the LOD scales with the inverse square root of the numberof shots. The values presented have all been scaled to single shot values using thisscaling. It is seen that the optimized values are not greatly different from those reportedfor 60 mJ pulses, and single-shot LODs are mainly in the range of 20 to 400 ppm for40 to 200 �J pulses, depending on the element and line observed. Using more shotsimproves the LODs that are possible, as in the case of Geertsen et al. who report a LODof 3 ppm for Mg using an accumulation of 150 shots [15].

The detector used in Rieger et al. [23] and Sabsabi et al. [52] were both similargated intensified photodiode arrays, with similar spectrometer characteristics, including



Micro-LIBS 185

10

100

1000

0 100 200 300 400 500 600 700 800

LOD

[ppm

]

Delay [ns]

100 µJ

200 µJ

Fig. 9. LOD as function of gate delay for Cu emission at 324.8 nm in Al 7075 alloy for 100 �J(open squares) and 200 �J (solid circles) laser pulse energy. (Reproduced with permission fromRieger et al. [23]).

Table 3. LOD for minor elements in aluminum alloys. All values are scaled toequivalent single shot acquisitions values, using an N−1/2

shots scaling

Element Emission Geertsen [15] Rieger [23] Sabsabi [52]Wavelength 40 �J 200–240 �J 60 mJ

266 nm 248 nm 1064 nm

Cr 425.4 nm 204 ppmCu 324.8 nm 22 ppm

327.4 nm 245 ppm 71 ppmFe 438.4 nm 447 ppmMg 285.2 nm 37 ppm ≤2 ppm 3 ppmMn 279.5 nm 35 ppm

403.1 nm 67 ppm 14 ppmSi 251.6 nm 99 ppm

288.2 nm 141 ppmZn 334.5 nm 281 ppm

the slit widths. The only major difference in the experiments besides the energy is thelaser wavelength used: Sabsabi et al. used 1064 nm whereas Rieger used 248 nm. Theabsorption will be better at 248 nm for metals, and plasma shielding will be a greaterissue for the 1064 nm at higher energies which may affect the comparison somewhat. Atstill lower energies, Bloch et al. reported obtaining hundreds of ppm LODs for metalsin soil [10].

In the case of a high energy LIBS emission plasma, the entrance slit of the spec-trometer represents the limiting aperture for acceptance of light. The entrance slit andthe collection optics limit the spatial region of the plasma plume that may be observedat any given time. In the case of �LIBS, the spatial expansion of the plasma is much




104

103

102

101

101 102 103 104

Inte

nsity

[A.U

.]

Concentration [ppm]

Cu 324.8 nmCu 327.4 nm

Average

Fig. 10. Calibration curve of copper in aluminum matrix. Each point is an average of ten 240 �Jpulses at 248 nm. Gate width is 300 ns and gate delay is 200 ns. The 3 LOD is 12 ppm, andthe straight line is a linear fit to the averaged data points below 1000 ppm. (Reproduced withpermission from Rieger et al. [23]).

smaller with a size comparable to the slit widths used for spectral measurements. Thus, alarger fraction of the plasma emission can be collected by the spectrometer as comparedto traditional mJ pulse energy LIBS. The result is that the LODs reported in the �LIBSliterature are often comparable to those reported by more traditional LIBS systems usingmJ pulse energies.

3.4. Signal Linearity with Concentration

An important issue in the application of LIBS to analytical measurements is the scaling ofthe signal with concentration. For a small, optically thin plasma it is expected that the lineemission strength should scale linearly with concentration for minor constituents. Thislinear scaling is observed in the emission of Cu in aluminum targets for concentrationsbelow 0.1% as seen in Fig. 10. For the dominant species self reversal, indicating strongoptical opacity, can be observed at energies as low as 7 �J as reported by Gornushkinet al. [25]. Signal linearity depends on the characteristics of the line under observation,the focal conditions of the laser and the observation time. However, it appears that formany elements at concentrations less than ∼1000 ppm signal linearity can be assumed.

4. APPLICATIONS

To date, the main application areas of �LIBS have been in the analysis of very smallsample volumes �∼10–1000 �m3� and in the scanning microanalysis of surface com-position. Microanalysis can be carried out in 2D or 3D with depth resolution by usingrepeated scans over the surface. Initial reported results in these areas are presented below.



Micro-LIBS 187

4.1. Microanalysis of Small Volumes

Microanalysis of metallic samples has been reported by Geertsen et al. [15], Rieger et al.[23,24], Cravetchi et al. [28] and Gornushkin et al. [25]. Geertsen et al. demonstratedthe use of �J pulses for the microanalysis of aluminum alloys [15]. The authors used 30to 70 �J pulses at 266 nm focused onto the samples using a 25X reflective microscopeobjective. Due to the very small focal spot used in these experiments, the signals obtainedwere very sensitive to inhomogeneity on the �m scale size. In studies reported byCravetchi et al. [28,29] it was shown that the small probe spot could be positionedon individual precipitate crystals and used to analyze the composition of individualprecipitates. The placement of a probe spot either on the precipitate or in the surroundinghomogeneous matrix region is illustrated in Fig. 11.

It was shown that statistically significant determination of precipitate type couldbe made with single shot spectra by detecting emission lines which were more than3 higher than the same line for the homogeneous matrix [29]. It is essential for anyapplication on the micron scale that the analysis be obtained in a single shot since thefeatures being measured may be ablated in a single shot.

Broadband LIBS signals, covering a large spectral range using �J pulses have recentlybeen demonstrated by Gornushkin et al. [25]. The use of broadband LIBS is seen as amajor step forward for material analysis since one to two orders of magnitude more datacan be obtained on each laser shot, thereby making optimum use of the limited photonemission from a single laser shot. Using an ungated detector, Gornushkin et al. measuredthe �LIBS spectra of a number of metals with clear observation of emission lines butwith significant continuum for the 1064 nm 7�J probe pulses. The authors observedthat a moving target was necessary since the melting from previous laser shots left a

0

1000

2000

3000

4000

5000

320 340 360 380 400 420

Obs

erve

d si

gnal

[cou

nts]

Wavelength [nm]

Fe Mg

Al Mn

Cu

M1

P2

10 µm

M2

P1 Mn-Fe-Cu precipitateAl-alloy matrix

(a) (b)

Fig. 11. (a) Scanning electron microscope image of precipitates on the surface of aluminum alloyand individual single-shot �LIBS craters produced with 7 �J pulses at 266 nm. Matrix (dark area)shots are labeled M1 and M2, while �LIBS shots that sampled precipitates (bright areas) arelabeled P1 and P2. (b) Representative single-shot spectra from matrix and precipitate regions ofthe aluminum surface. Clear differences are observed with a pulse energy of 7 �J. (Reproducedwith permission from Cravetchi et al. [28]).




more reflective surface and the LIBS signal would disappear after the first shot. Thisindicates that the use of 1064 nm wavelength is not optimum for measurements of manymetals because of high reflectivity at this wavelength. For this reason many �LIBSinvestigations have been carried out using UV wavelength lasers to take advantage ofthe improved coupling.

Bloch et al. studied soil samples to detect Cu, Fe and Pb contamination using 10 �Jpulses at 1064 nm from a microchip laser [10]. The plasma emission was measuredusing an ungated and unintensified compact diode-array based spectrometer. The authorsnoted that the plasma continuum radiation decays quite rapidly with a time constant of∼15 ns. As a result, they were able to measure concentrations at the hundreds of ppmlevel without the need for temporal gating. Kossakovski et al. probed a meteorite samplecomparing probing with a focal spot from a 50 mm quartz lens with that through anetched fiber probe showing the signals were similar for similar power densities [22]. Inboth cases light was collected from the side with a 20X microscope objective and anungated unintensified spectrometer was used. Good signals were obtained when higherenergies per pulse were used leading to probe craters on the order of 2 �m in diameteror greater. Pelletized graphite targets impregnated with magnesium hydroxide powderwere studied by Gornushkin et al. using 7 �J pulses with limited success [25]. Theyattributed the lack of success to the fragility and roughness of the target surface whicheroded easily under the 5 kHz repetition rate laser. Due to the high repetition rate, thetarget was scanned in a spiral pattern in order to present a new spot to the samplinglaser for every pulse.

4.2. Scanning Microanalysis of Material Surfaces

One main application of �LIBS is in the scanning microanalysis of material surfaces. Toachieve high spatial resolution and small ablation depths, very small energies and smallfocal spots are desired. It has been reported that as the spot size approaches one micron,the signal becomes too weak for material composition analysis [5,22]. However, theseobservations were made for nanosecond pulses and without optical gating in one case.Using shorter picosecond or femtosecond pulses and better light collection efficiency itmay be possible to obtain LIBS signals in cases where the ablation crater is less than1 �m.

Häkkänen et al. have studied the application of �LIBS to the mapping of surfacecoatings on paper [14,30]. This work also represents one of the earliest uses of LIBS as asurface mapping tool. The LIBS results were compared with laser-induced fluorescence(LIF) and found to give good agreement. The results of the 2D �LIBS scan and 2D LIFscan are presented here in Fig. 12. 2 �J pulses at 308 nm were scanned over a 10 mmby 10 mm section of paperboard while monitoring the fluorescence signal at 422 nm.The same scan was performed after increasing the energy to 200 �J pulses at a fluenceof ∼109 W cm−2, leading to plasma emission. This fluence was sufficient to removethe paper coating, and generated craters 30 �m in diameter, and 2 �m deep. The Si I251 nm line was monitored using a PMT with a delayed boxcar integrator. 8 such shots,each displaced 32 �m, were averaged to generate a single data point corresponding toa pixel 30 �m ×250 �m. 40 such pixels were taken to make a single row, and 40 suchrows make up the entire image presented in Fig. 12b. The LIF and LIBS images are



Micro-LIBS 189

(a) (b)

Fig. 12. 2D scan of a 10 mm by 10 mm piece of coated paper board. (a) Laser-induced fluorescenceof underlying paper at 422 nm and (b) �LIBS scan at 251 nm for Si in paper coating. (Reproducedwith permission from Häkkänen et al. [14]).

expected to be negatives of each other since the Si line for the LIBS signal is sensitiveto the coated regions of the paper while the fluorescence measurements are sensitive toorganic compounds visible in the less coated regions.

Further improvements to the measurement technique were reported in a subsequentinvestigation [30]. Using a similar setup as previously the authors employed 80 �J pulsesat 308 nm and a 40 mm lens to generate focal spot sizes ∼100 �m in diameter. Theresulting craters were also 100 �m in diameter and 0�5 �m deep. Using a series of 40shots for each location on the target surface, the authors were able to measure the depthprofile distribution of the pigment layers that make up the smooth surface of modernpaper. A 2D depth resolved scan through the topcoat, precoat and into the base layer ofpaper, is shown in Fig. 13.

Scanning �LIBS of anti-reflection coated and uncoated silicon surfaces has beenstudied by Laserna’s group in several reports [16,26,27,53]. In the initial investigationof Hidalgo et al. TiO2 anti-reflection coatings for photovoltaic cells were studied usinglarge, low fluence spots in order to achieve better depth resolution [16]. Using pulseenergies of 400 �J delivered to the target and focal spots of 160 �m × 40 �m, depthprofiling of the TiO2 coating was performed, and the coating was distinguishable fromthe Si substrate. However, a depth resolution was not estimated by the authors. Oneinteresting feature noted by the authors was a dependence of the emission signal strengthon the coating thickness which also correlated with the film reflectivity. The peak field

10

20

0

0 10 20 30 40 50

Dep

th [µ

m]

Length (mm)

Fig. 13. �LIBS 2D depth profile scan of the composition of paper using 80 �J pulses at 308 nm.Gray indicates top coat, composed of a 50:50 mix of calcium carbonate:kaolin. Black indicatesprecoat, composed of a 80:20 mix of calcium carbonate:kaolin. White indicates the base paper.(Reproduced with permission from Häkkänen et al. [30]).




strength in the coating which causes breakdown and emission depends on the interferencebetween the reflected wave and incident wave and is a sensitive function of the layerthickness and thus a dependence on coating thickness is to be expected.

Using the same setup, Vadillo et al. applied �LIBS to a full 2D and 3D mapping ofphotovoltaic cell structures on silicon. The pulse energy was approximately 40 �J [26].Using these conditions, it was possible to produce a 2D surface map of carbon conta-mination distribution. By taking multiple shots to obtain depth profiling at each of thesurface map locations, a 3D map was also produced, giving carbon distribution not onlyat the surface, but at layers further down.

The mapping work was extended to simultaneous monitoring of multiple wavelengthsfor mapping of Si photovoltaic cells in Romero et al. [27]. The setup was similar to thatof Hidalgo et al. [16] where pulses of 100 to 400 �J were used. Using this setup, theauthors were able to generate a set of spectrally resolved images from their data, with alateral resolution of about 80 �m. Moving on to a full three dimensional analysis of thephotovoltaic cells Romero et al. studied the distribution of carbon in the solar cells, usinga series of 2D scans over the same area [53]. The resulting lateral resolution obtained byRomero et al. was 70 �m, and the depth resolution was approximately 0�16 �m. In thiswork the goal was to achieve good depth resolution and thus the focal spot size wasincreased to give the low fluences necessary.

Menut et al. combined a LIBS system with an optical microscope and generated a 2Dsurface scanning instrument with a lateral resolution of 3 �m using an Ar buffer gas [5].Crater sizes down to 1 �m are reported, though at such low energies the SNR wasinsufficient for analysis of minor constituents. The setup described by Menut et al. [5]detected signals at a pulse energy of 5�J, resulting in craters approximately 3 �min diameter for their steel sample. The system was able to acquire signals at 20 Hz,and has been used to map the surface composition of various samples. In Fig. 14 amulti-elemental map of a single inclusion in a steel alloy is shown.

Cravetchi et al. reported 2D mapping of aluminum surfaces and identification ofprecipitates using 8 �J pulses at 266 nm [29]. Particular attention was directed towardsimproving the statistical validity of the precipitate identification technique. A Gaussianfunction was fit to the signal intensity distribution of all shots in the mapped region toderive the average and standard deviation for signals corresponding to the backgroundmatrix. Only signals 3 above this level were deemed to be regions of precipitates.Correlations between various elements in a given type of precipitate can easily be

100 µm

Fe

100 µm

Mn

100 µm

Ti

100 µm

Ni

Lowconcentration

Highconcentration

Fig. 14. Scanning �LIBS image of a single inclusion on the surface of steel as seen in the emissionof elements Mn, Fe, Ti and Ni. (Reproduced with permission from Menut et al. [5]).



Micro-LIBS 191

0

1000

2000

3000

4000

5000

6000

7000

0 1000 2000 3000 4000 5000

Mn

peak

inte

nsity

[cou

nts]

Fe peak intensity [counts]

(a)

0

1000

2000

3000

4000

5000

6000

7000

0 5000 10000 15000

Mn

peak

inte

nsity

[cou

nts]

Mg peak intensity [counts]

(b)

Fig. 15. Correlation plots of peak intensity values for minor elements in aluminum alloys. (a) Mnvs Fe shows a positive correlation as they appear in the same precipitates, (b) Mn vs Mg yields anegative correlation in both lobes, as they do not appear in the same precipitates. Dashed lines are3 values from the nonlinear Gaussian fit to all available data. Solid lines are linear regressionswithin their respective quadrants. (Reproduced with permission from Cravetchi et al. [29]).

observed as shown in Fig. 15. The densely populated region in the lower left handcorner of the plots represents the matrix background. Based on the standard deviationof signals observed, it was possible to set detection thresholds for various trace ele-ments and map out the two dominant precipitates in Al 2024 alloy with 10 �m lateralresolution [29].

A few of the groups studying �LIBS have noted the issue of cross contamination frommaterial redeposited onto the surface from previous ablation spots. Romero et al. [27]and Häkkänen et al. [30] measured the single-shot contamination range from an ablatedlocation for silicon and paper targets using the LIBS signal itself, giving values of 80 �mand 200 �m respectively. Their results are shown in Fig. 16. Several other groups alsorefer to visible redeposition of target material on the sample surface [15,18,22,25].

Clear evidence of material redeposition was found in the 2D mapping of aluminumsurfaces experiment of Cravetchi et al. [29]. Redeposition of Al2O3 on the target surfacewas observed. The resulting coating of the target was quite pronounced when a largenumber of shots was taken, as can be seen in Fig. 17a. The left image is a SEM imagewhich shows a smooth coating over the original aluminum surface. However, as can beseen in Fig. 17b, the redeposited layer ceases abruptly as one approaches the mappedarea and around the isolated shots at the bottom of the images.

This can be understood by considering the blast wave in air and shock wave in thematerial created by the ablation plasma. As a LIBS plasma is created, it launches ashock wave that expands with a quasi-spherical symmetry and the force of this wavenear the ablation spot is sufficient to remove the deposited material from the surface.The radius of this cleaned area is larger than the distance to the subsequent shot in thescanning analysis and thus the original target surface is probed by the scanning �LIBSmeasurement. The dynamics of material deposition and cleaning will depend on thesample being scanned and the conditions being employed. Redeposition may be reducedif one carries out the scans in vacuum but detailed studies need to be carried out toquantify the reduction.




0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 20 40 60 80 100 120 140 160

Rel

ativ

e in

tens

ity [A

.U.]

Lateral resolution [µm]

3000

4000

5000

6000

7000

8000

100 200 300 400 500

Si i

nten

sity

[A.U

.]

Distance [µm]

Number of pulses/point3015

51

(a) (b)

Fig. 16. (a) Ratio of peak intensities of the Ti line measured at 626 nm from two adjacent pointson the surface of a TiO2 coated Si sample. A ratio of 1.0 indicates the second shot has sampled anundisturbed surface. (b) Silicon intensity of the first ablation layer of coated paper as a function ofdistance between sampling points and number of shots at each sampling point. In this case, Si is acontaminant from buried layers in a paper coating. (Reproduced with permission from (a) Romeroet al. [27], and (b) Häkkänen et al. [30]).

(a)

20 µm

10 um 1 um

5 µm

(b)

Fig. 17. (a) Scanning electron microscopic image of macroscopic redeposition of Al2O3 and shockcleaning near the perimeter of a 2D �LIBS scan area 300 �m by 900 �m in size with probespot separation of 10 �m. The edge of the scanned area is visible at the left edge of the image.(b) Isolated crater created using single 1�5 �J pulses at 266 nm. (Reproduced with permission fromCravetchi et al. [29]).

In order to compare the various mapping experiments, we define a surface mappingrate (SMR) as the sample area per shot multiplied by the sample rate. In this case, thesample area per shot is defined by the crater diameter. These are plotted for publishedreports of �LIBS surface analysis in Fig. 18. Included for comparison is the use of line-focused beam scans with milli-joule energy pulses, as applied by Mateo et al. [55,56]and Rodolfo et al. [57]. In such line focused beams, the irradiance applied to the targetcan be in the same range as that of �LIBS. This plot demonstrates the current capabilitiesof �LIBS scanning rates for 2D multi-elemental surface mapping.



Micro-LIBS 193

Sur

face

map

ping

rat

e [µ

m2 /

s]

Power [W]

1

2

3

4

5

6

78

9

10

Line Focused Beams

266 nm308 nm331 nm532 nm

1064 nm

107

106

105

104

103

102

10–5 10–4 10–3 10–2 10–1 100 101

Fig. 18. Surface scanning rate as a function of applied power. 1= Menut et al. [5], 2 =Cravetchi et al. [29], 3 = Vadillo et al. [26], 4 = Romero et al. [53], 5 = Romero et al. [27],6 = Häkkänen et al. [30], 7 = Romero et al. [54], 8 = Mateo et al. [55], 9 = Mateo et al. [56],10 = Rodolfo et al. [57]. For comparison, surface scans carried out using millijoule laser pulsesin a line focus geometry are also shown in the upper right area of figure [55–57].

4.3. Liquid Samples

In the early 1990s a series of experiments applying LIBS to the detection of elements inwater jet samples were performed by Cheung et al. [58,59]. This work has been extendedto the �LIBS regime in more recent work by Ho and Cheung et al. [17,31] for detectionof Na and K. One of the goals was to demonstrate sufficient sensitivity to measure thechemical content of single cells. 532 nm Nd-YAG and 193 nm ArF laser pulses wereused as excitation sources, with both a photomultiplier tube and ICCD detection. Toincrease the absorption of the liquid water, a solution of 12 mM methyl violet was used.Using 240 �J pulses at 532 nm a detection limit of 50 ppm was achieved. The reporteddetection limit using the ArF excitation beam was 230 ppb. In further work by the samegroup, Cheung et al. note that the plasma generated by the ArF beam is significantlycooler than that generated by the 532 nm beam at short delay times. Plasma temperatureand electron density were determined by line intensity ratios and line widths.

5. CONCLUSIONS

In the past decade there has been good initial progress in the development and under-standing of �LIBS. Pulsed microchip laser sources with energies of 1 to 240 �J havebeen demonstrated and are beginning to be commercially available. The primary sourcesdemonstrated to date are in the infrared region while the optimum wavelength for �LIBSis most likely in the UV region to give smaller, diffraction limited focal spots and bettertarget absorption. The energy of harmonically converted UV sources is still limited to




less than ∼10 �J. However, at 1550 nm (erbium based lasers) one also has the advan-tage of eye safe sources at microjoule energy levels making practical systems easier toimplement. More work needs to be done on the wavelength scaling issues to determinehow effective these mid-infrared sources could be for �LIBS.

In microanalysis applications the ability to achieve single shot LODs of 10 to 100 ppmhas been demonstrated with ∼100 �J energy pulses using gated detector systems and100 to 10,000 ppm with ∼10 �J energy pulses using ungated detectors. The ability toanalyze sample volumes of 10 to 1000 �m3 has been demonstrated. 2D surface scanshave been carried out with 3�m lateral resolution on steel and 10 �m lateral resolution onaluminum. However, issues of determining the exact region of LIBS emission sensitivitywithin the ablation volume and cross contamination remain to be addressed in detail.It is likely that cross contamination is very much material and laser parameter dependentand particular attention should be paid to this issue in any scanning microanalysissystem. Microanalysis of water samples has also been demonstrated achieving sub ppmsensitivities under optimized conditions.

While there is much that remains to be done in the study of �LIBS, particularly inthe area of wavelength and pulselength scaling of LODs achievable, the use of �LIBSalready appears as a promising new regime which should soon lead to cost effectiveportable systems.

REFERENCES

[1] V. Margetic, A. Pakulev, A. Stockhaus, M. Bolshov, K. Niemax and R. Hergenröder. Spec-trochim. Acta B55 (2000) 1771.

[2] V. Margetic, M. Bolshov, A. Stockhaus, K. Niemax and R. Hergenröder. J. Anal. At.Spectrom. 16 (2001) 616.

[3] K.L. Eland, D.N. Stratis, D.M. Gold, S.R. Goode and S.M. Angel. Appl. Spectrosc. 55(2001) 286.

[4] S. Yalçin, Y.Y. Tsui and R. Fedosejevs. J. Anal. At. Spectrom. 19 (2004) 1295.[5] D. Menut, P. Fichet, J.-L. Lacour, A. Riovallan and P. Mauchien. Appl. Opt. 42 (2003) 6063.[6] B. Al Ali, D. Bulajic, M. Corsi, G. Cristoforetti, S. Legnaioli, L. Masotti, V. Palleshi,

A. Salvetti and E. Tognoni. SPIE 4402 (2001) 25.[7] J.J. Zayhowski and A. Mooradian. Opt. Lett. 14 (1989) 24.[8] J.J. Zayhowski. Opt. Lett. 16 (1991) 575.[9] J.J. Zayhowski. Opt. Lett. 21 (1996) 588.

[10] J. Bloch, B. Johnson, N. Newbury, J. Germain, H. Hemond and J. Sinfield, Appl. Spectrosc.52 (1998) 1299.

[11] J.J. Zayhowski, J. Alloys Compd. 303–304 (2000) 393.[12] J. Limpert, A. Liem, M. Riech, T. Schreiber, S. Nolte, H. Zellmer, A. Tünnermann, J. Broeng,

A. Petersson and C. Jakobsen. Opt. Express 12 (2004) 1313.[13] M. Laroche, A.M. Chardon, J. Nilsson, D.P. Shepard and W.A. Clarkson. Opt. Lett. 27

(2002) 1980.[14] H. Häkkänen and J.E.I. Korppi-Tommola. Appl. Sectrosc. 49 (1995) 1721.[15] C. Geertsen, J.-L. Lacour, P. Mauchien and L. Pierrard. Spectrochim. Acta B51 (1996) 1403.[16] M. Hidalgo, F. Martin and J.J. Laserna. Anal. Chem. 68 (1996) 1095.[17] W.F. Ho, C.W. Ng and N.H. Cheung. Appl. Spectrosc. 51 (1997) 87.[18] B. Sallé, C. Chaléard, V. Detalle, J.-L. Lacour, P. Mauchien, C. Nouvellon and A. Semerok.

Appl. Surf. Sci. 138–139 (1999) 302.



Micro-LIBS 195

[19] B. Sallé, M.N. Libenson, P. Mauchien, G. Petite, A. Semerok and J.-F. Wagner. SPIE 3822(1999), 56–67.

[20] A. Semerok, C. Chaléard, V. Detalle, J.-L. Lacour, P. Mauchien, P. Meynadier, C. Nouvellen,B. Sallé, P. Palianov, M. Perdix and G. Petite. Appl. Surf. Sci. 138–139 (1999) 311.

[21] A. Semerok, B. Sallé, J.-F. Wagner, G. Petite, O. Gobert, P. Meynadier, M. Perdix. SPIE4423 (2001), 153–164.

[22] D. Kossakovski and J.L. Beauchamp. Anal. Chem. 72 (2000) 4731.[23] G.W. Rieger, M. Taschuk, Y.Y. Tsui and R. Fedosejevs. Appl. Spectrosc. 56 (2002) 689.[24] G.W. Rieger, M. Taschuk, Y.Y. Tsui and R. Fedosejevs. Spectrochim. Acta B58 (2003) 497.[25] I.B. Gornushkin, K. Amponsah-Manager, B.W. Smith, N. Omenetto and J.D. Winefordner.

Appl. Spectrosc. 58 (2004) 762.[26] J.M. Vadillo, S. Palanco, M.D. Romero and J.J. Laserna. Fresenius J. Anal. Chem. 355

(1996) 909.[27] D. Romero and J.J. Laserna. Anal. Chem. 69 (1997) 2871.[28] I.V. Cravetchi, M. Taschuk, G.W. Rieger, Y.Y. Tsui and R. Fedosejevs. Appl. Opt. 42

(2003) 6138.[29] I.V. Cravetchi, M. Taschuk, Y.Y. Tsui and R. Fedosejevs, Spectrochim. Acta B59

(2004) 1439.[30] H. Häkkänen, J. Houni, S. Kaski and J.E.I. Korppi-Tommola, Spectrochim. Acta B56

(2001) 737.[31] N.H. Cheung, C.W. Ng, W.F. Ho and E.S. Yeung, Appl. Surf. Sci. 127–12 (1998) 274.[32] J. Zayhowski. Lincoln Laboratory Journal, 3 (1990) 427.[33] J. Zayhowski and J. Keszenheimer. IEEE J. Quantum Electron. 28 (1992) 1118.[34] J.J. Zayhowski and C. Dill III. Opt. Lett. 19. (1994) 1427.[35] J.J. Zayhowski. Laser Rev. 26 (1998) 841.[36] D. Taverner, D.J. Richardson, L. Dong, J.E. Caplen, K. Williams, R.V. Penty. Opt. Lett. 22

(1997) 378.[37] H.L. Offerhaus, N.G. Broderick, D.J. Richardson, R. Sammut, J. Caplen and L. Dong. Opt.

Lett. 23 (1998) 1683.[38] C.C. Renaud, H.L. Offerhaus, J.A. Alvarez-Chavez, J. Nilsson. W.A. Clarkson, P.W. Turner,

D.J. Richardson and A.B. Grudinin. IEEE J. Quantum Electron. 37 (2001)199.[39] A. Galvanauskas. IEEE J. Sel. Top. Quantum Electron. 7 (2001) 504.[40] J. Limpert, S. Höffer, A. Liem, H. Zellmer, A. Tünnermann, S. Knoke and H. Hoelckel.

Appl. Phys. B: Lasers Opt. 75 (2002) 477.[41] R. Fluck, R. Häring, R. Paschotta, E. Gini, H. Melchior and U. Keller. Appl. Phys. Lett. 72

(1998) 3273.[42] G.J. Spühler, R. Paschotta, M.P. Kullberg, M. Graf, M. Moser, E. Mix, G. Huber, C. Harder

and U. Keller. Appl. Phys. B: Lasers Opt. 72 (2001) 285.[43] R. Feldman, Y. Shimony, Z. Burshtein. Opt. Mater. 24 (2003) 393.[44] G. Karlsson, V. Pasiskevicius, F. Laurell, J.A. Tellefsen. Opt. Commun. 217 (2003) 317.[45] F. Druon, F. Balembois, P. Georges and A. Brum. Opt. Lett. 24 (1999) 499.[46] B. Braun, F.X. Kärtner, U. Keller, J.-P. Meyn, G. Huber. Opt. Lett. 21 (1996) 405.[47] R. Fluck, B. Braun, E. Gini, H. Melchior, and U. Keller. Opt. Lett. 22 (1997) 991.[48] Björn T. Hansson and Ari T. Friberg. Opt. Lett. 26 (2001) 1057.[49] J.J. Zayhowski. J. Commun. Res. Lab. 46 (1999) 385.[50] D. Taverner, P. Britton, P.G.R. Smith, D.J. Richardson, G.W. Ross and D.C. Hanna. Opt.

Lett. 23 (1998) 162.[51] G.W. Rieger, M. Taschuk, Y.Y. Tsui and R. Fedosejevs. SPIE 4087 (2000), 1127.[52] M. Sabsabi and P. Cielo. Appl. Spectrosc. 49 (1995) 499.[53] D. Romero and J.J. Laserna. J. Anal. At. Spectrom. 13 (1998) 557.[54] D. Romero and J.J. Laserna. Spectrochim. Acta B55 (2000) 1241.




[55] M.P. Mateo, L.M. Cabalín and J.J. Laserna. Appl. Spectrosc. 57 (2003) 1461.[56] M.P. Mateo, L.M. Cabalín, J.M. Baena and J.J. Laserna. Spectrochim. Acta B57 (2002) 601.[57] K. Rodolfo and D. Cremers. Appl. Spectrosc. 58 (2004) 367.[58] N.-H. Cheung and E.S. Yeung. Appl. Spectrosc. 47 (1993) 882.[59] N.-H. Cheung and E.S. Yeung. Anal. Chem. 66 (1994) 929.



Chapter 9

LIBS Application to Off-Gas Measurement

F. Y. Yueh and J. P. Singh

Institute for Clean Energy Technology, Mississippi State University205 Research Boulevard, Starkville, MS 39759, USA

1. INTRODUCTION

Laser induced gas breakdown is the results of the interaction of high intensity laser beamwith a gas. Typically, an irradiance corresponding to an electric field strength on theorder of 105 volt/cm with a gas near atmospheric pressure can produce gas breakdownthrough multiphoton ionization or electron avalanche [1]. The gas breakdown thresholdsin the atmospheric pressure are proportional to the ionization potential of each gasdivided by the collision frequency. Due to the presence of micron-sized aerosols andimpurity particles, the observed breakdown threshold in gas samples is generally lowerthan that from the theoretical prediction. This is because the particles acting as seedscan significantly lower the breakdown threshold of clean gas. Laser-induced breakdownin gas has been studied extensively [2]. Beside the particle size, the laser-induced gasbreakdown thresholds also depend strongly on the gas pressure and the laser wavelength.Typically, laser-induced air breakdown has a plasma temperature of 20,000 K and anelectron density of 1017−1018 cm−3 after the plasma is formed [1].

The application of LIBS for gas analysis involved a focused high-energy pulsed laserto produce the breakdown in the gas medium. The high temperatures and electron densitylaser-induced plasma prepares and excites the sample in single step. The emission fromthe laser plasma can be used directly to measure the composition of gas, eliminatingthe need for sample preparation. Schmieder et al. were first to show that LIBS canbe applied as a combustion diagnostics for monitoring the elemental constituents ofa combustion product [3,4]. They used a time-integrated photographic technique anddiode array to detect N and O in gas mixtures and to measure the C/N ratio of theflame. Radziemski and Loree pioneered LIBS applications on gas measurements usingtime-resolved detection [5]. They used a time-gated optical multichannel analyzer or aPMT-boxcar detection system and found the detection limits for P and Cl in air as 15and 60 ppm, respectively. Cremers and Radziemski were later able to detect Cl and Fin air with a detection limit of 80 ng and 2,000 ng, respectively [6]. They also foundthat the absolute detection limit for Cl and F can be improved in a He atmosphere.Radziemski et al. have used LIBS to detect Be, Na, P, As, and Hg in air [7]. Due to the




200 F. Y. Yueh and J. P. Singh

small sample volume and possible sample inhomogeneity, LIBS measurement precisionin a gas sample is generally poor. The various size particulates in the gas can causethe breakdown to be generated at different locations along the axis of the laser beamand lead to significant signal variations. The most common interference found in the airbreakdown is the CN emission. CN is produced from the reaction of C and N, whichare produced in the spark. The intensity of CN bands depends on the concentration ofa C-containing compound in the gas stream. The analyte lines in the CN band coveredspectral region have less sensitivity due to the spectral interference.

Toxic metals from thermal processing units, is a great concern for environmentprotection agency. More strict limits on toxic metals from many process streams will beput on future regulations. Since LIBS is able to complete vaporization of aerosol particlesup to roughly 2–10 �m in diameter, and atomizes molecular species, it has great potentialfor detection of Toxic metals especially in the forms of fine particles (i.e. PM 2.5). Hahnhas used LIBS for sizing and elemental analysis of sub-micrometer to micrometer-sizedaerosol particles [8]. Buckley has studied the effects of experimental configuration,potential interferences and oxygen quenching to LIBS application to toxic metal emissionmeasurements [9]. Biological warfare agent is identified as a great threat to general publicdue to the potential bioterrorism. The feasibility of using LIBS for rapid detection andidentification of various biological aerosols has been demonstrated [10–12]. Hybl et al.have used a broadband LIBS system for laboratory measurements on some commonbiological agent simulants and a narrow band LIBS system to detect single simulant(Bg) particles in the size range 1–5 microns [12].

The optical characteristics of LIBS in gas measurements have been discussed in detailand can be found elsewhere [13,14]. This chapter explores the calibration techniquesand various LIBS applications for gas samples using a mobile LIBS system which wasdeveloped at Institute for Clean Energy Technology (ICET), Mississippi State University,USA. This versatile mobile system was originally developed to monitor toxic metalconcentrations in the off-gas emission of a plasma hearth process system. It has been usedto conduct various laboratory studies and field measurements for different applications.

2. EXPERIMENTAL SETUP

The experimental arrangement of the LIBS system requires a laser system that candeliver high pulse energy (e.g. 100–300 mJ/pulse) to produce a spark in the gas medium.A frequency-doubled Nd:YAG laser is directed and focused on the desired gas samplewith a lens of proper focal length (generally 10–20 cm). The emission from the sparkwas collected with a UV optical fiber bundle and sent to the detection system. Usuallyone Czerny-Turner spectrometer that can cover a spectral region of 20–40 nm simultane-ously is used for gas measurements. In some cases, two detection systems were neededto monitor two spectral regions simultaneously or two different measurement locations.The detection systems employed in the present study include a SPEX 500M spectro-graph equipped with a 1024-element intensified diode array detector (Model IDAD-1024,Princeton Applied Research) and an optical spectrograph (Model HR 460, Instrument SA,Inc., Edison, NJ) equipped with a 1024×256 element intensified charge-coupled device(ICCD, Princeton Instrument Corporation, Princeton, NJ). A fiber bundle with the output



LIBS Application to Off-Gas Measurement 201

end of the optical fiber bundle splitting into two bundles was coupled to two spectro-graphs for the measurements of two spectral regions. The detectors were operated in gatedmode with the control of a high voltage pulse generator (PG-10, Princeton InstrumentsCorporation, Princeton, NJ) and was synchronized to the laser output. Data acquisitionand analysis were performed using a personal computer and a notebook computer (ModelT-4700CS, Toshiba). The gate delay time and gate width were adjusted to maximize thesignal-to-background (S/B) and signal-to-noise (S/N) ratios, which are dependent on theemission characteristics of the elements as well as the experimental configuration. A gatedelay of 5–10 �sec and a gate width of 10–20 �sec were used in most of the work.

To quantify the LIBS data of gas sample, LIBS instrument needs to be calibrated withthe samples of known concentration. The gas phase sampling generally uses a nebulizerto produce aerosol from the solution of standard reference materials. In the present casean ultrasonic nebulizer (USN, Cetac U-5000AT+) is used to produce the dry aerosolsof selected metals. Two possible setups that were used for LIBS calibration in openand closed system are shown in Fig. 3 of chapter 5. Calibration has been performed byinjecting known concentrations of dry aerosols from an ultrasonic nebulizer into either asample cell (closed system) or air (open system). Volumetrically diluted plasma emissionstandard solutions (Spex Industries) were injected into the USN with a peristaltic pumpat a rate of 1.9 ml/min. A 0.8-ml/min flow rate of air was used as a carrier gas flowto transport the aerosol through the USN. The aerosol in the USN was first dried by aheated (140�C) tube and then passed through a chilled (3�C) condenser to remove water.In the open system, the dry aerosol from the USN was sent to a stainless steel sampleinjection tube, and the laser beam was aligned 2 mm above the end of the tube andfocused on the center of the tube to achieve reliable calibration. The sample injectiontube was enclosed in a Pyrex cylinder to reduce interference from the surrounding air.In the closed system, the metal aerosol was injected continuously to the sample cellthat is made of Polyvinyl Chloride (PVC). LIBS calibration data were collected afterthe composition equilibrium in the cell was reached. The waste solution was collectedduring the nebulization procedure. The USN system was later operated with the collectedwaste solution. A comparison of the LIBS signal from the stock and waste solution canbe used to determine the efficiency of the nebulizer [15].

In the laboratory, both the open and closed systems can be used to calibrate the LIBSsystem, but in the field measurement only the open system is more suitable for on-sitecalibration. The on-site calibration should be performed before the field test started andafter the test ended each day to verify the system response. The on-site calibrations havebeen carried out by injecting metal aerosol generated from a USN into the gas streamwith a probe. The sample injection probe was mounted on the opposing port across thegas stream. Each day, the LIBS spectra need to be recorded before the metal injectionfor zero check.

3. CALIBRATION

LIBS is an atomic emission spectroscopy. For quantitative analysis with LIBS, eitherinternal standard calibration or external calibration method is needed. However,calibration is the most difficult issue in the development of LIBS, especially forthe field measurement. This is because the calibration procedure should keep the




same experimental conditions for the known sample used in the calibration and theunknown sample. The parameters, which can affect the characteristics of the LIBSspectra in gas, include particle size, gas pressure, temperature, and laser energy. Dueto shot-to-shot laser fluctuations, it is hard to maintain the excitation condition for thecalibration data and data from the unknown sample in laboratory environment. LIBS isbeing considered as a non-sampling technique for on-site measurement, this implies anextra difficulty for calibration.

An extensive LIBS calibration study has been performed by Singh et al. [16–18].They have compared two calibration methods using a hydride generator and a USN inLIBS experiments. LIBS spectra have been recorded using a hydride generator (Fig. 1)and a USN with a mixture solution of As, Sb, and Sn in N2 and He to study interferenceeffects among different metals. Since HCl concentration plays an important role onhydride generation efficiency, different HCl concentrations in the mixture solution havealso been used in this experiment. The spectral interferences were not significant in thisstudy. However, the results from the hydride generation were quite sensitive to the acidconcentration in the mixture. Comparison of metal generation from metal oxide particlesproduced by an ultrasonic nebulizer shows that, the actual gas stream metal distributionis close to that from the USN. Efficient metal hydride generation requires different acidconcentrations for different metals. A USN, on the other hand, is easy to use, and worksfor all resources conservation and recovery act (RCRA) metals. Therefore, the calibrationcurves for every RCRA element have been obtained using a USN. Based on the datacollected from the USN and, after averaging 50 laser pulses, the precision for most of theRCRA metals was ∼10% or better, and the accuracy was ∼5−10%. Studies of relativecalibration were also performed to implement on-line calibration in field measurements.

From the experimental results, it is recommended to use the USN to conduct theLIBS calibration for gas samples. The calibration data used here is obtained by injectingknown concentrations of dry aerosols from the USN into air. Generally, LIBS data fromfour or more concentrations of an element were used to obtain the calibration curve. Thecalibration curve is based on either peak height or peak area of each analyte line. Theslope of the calibration curve is used as the calibration factor to infer metal concentration.

Rubberseptum

Reactor

To sample cellGas inlet

Reagents

NaBH4 + 3H2O + HCl H3BO3 + NaCl + 8H +EHn H2(excess)Em+

Fig. 1. Schematic diagram of the hydride generator. (Reproduced with permission from Ref. [16]).




The peak area (or peak height) of an analyte line from a demonstration test on LIBSspectra was normalized using its calibration factor to obtain the metal concentration. Ingeneral, peak height calibration and peak area calibration give about the same result foran interference-free line. For different types of spectral interferences, either peak heightor peak area must be selected for best results. From the obtained experimental resultsit was found that the peak area analysis yielded better results than did the peak heightanalysis for the self-absorbed spectral line, and the peak height analysis yielded betterresults than did the peak area analysis for a line overlapped with other lines.

The limits of detection (LOD) of selected analyte lines of seven RCRA metals deter-mined in the laboratory just before the CEM test are listed in Table 1. The precision ofthese measurements estimated from the calibration data are also listed in Table 1. The pre-cision and accuracy greatly depends on pulse-to-pulse laser fluence fluctuations at focalvolume and the concentration variation in the aerosol flow from the ultrasonic nebulizer(USN). The accuracy and precision of LIBS measurements can be improved by increasingthe signal integration time because some lines have spectral interferences and the actualfield detection limits may be slightly higher than the reported detection limit, dependingon the concentration of the interfering elements. The LODs depend on the experimentalconditions and can be reduced by improving the optical design and detection system.

In some cases, if the absolute concentration calibration is too difficult to obtain due tothe variation of the environmental conditions, relative concentrations may be considered.One can either use the calibration based on the intensity ratio of the analyte line andreference line or fit the observed spectra with a theoretical model. Analysis of LIBSdata using spectral fitting requires the knowledge of spectroscopic constants such asplasma temperature, and degree of ionization. These two parameters, however, are noteasy to be determined accurately. Alternately, Ottesen et al. used reference line intensity

Table 1. Limit of detection of some selected metals in gas [23]

Element Analyte Line (nm) Relative STD (%) LOD ��g/acm�∗

As 278.02 9 600Be 234.80 3 1Co 345.35 8 24Cr 425.44 5 7�8Cr 359.30 5 12Zn 330.30 15 570Cd 228.80 5 45Hg 253.65 13 680Sb 259.81 9 120Sn 283.99 10 190Mn 257.61 4 4Mn 403.08, 8 7�5

403.31,403.45

Ni 341.48 9 30Pb 405.78 6 90Fe 404.58 6 140

∗ acm = Actual cubic meter




information from the NIST collections to perform spectral fitting to obtain relativeconcentration [19]. However, the excitation condition needs to be verified for this simplemethod they used, since the intensities in these reference collections are obtained undercertain conditions, which may be very different from those in laser induced plasma.

Ciucci et al. have developed an algorithm for calibration-free quantitative LIBSanalysis and seem to have had great success with laboratory data [20]. However, suchan approach relies on some basic assumptions, such as laser-induced plasma (LIP) is inlocal thermodynamic equilibrium (LTE); LIP is an optically thin; plasma composition isrepresentative of the actual sample composition. It also requires measuring the emissionlines of all the elements presented in the sample. The selected analyte lines shouldbe free from the spectral interferences such as spectral overlapping, saturation or self-absorption. To measure all the analytes simultaneously, a broadband spectrometers ormulti-spectrometers are required for the calibration-free analysis. This technique stillneeds to be evaluated with more practical data to be widely accepted for most LIBS work.

The practical environments are quite different from those in a laboratory. The trans-fer of the LIBS calibration obtained in a laboratory to field measurements is a great

(a)

20140 160 180 200 220 240 260 280 300

25

30

35

40

45

50

55

60

Laser energy (mJ)

Slo

pe fo

r C

d 22

8 nm

120

100

80

60

40

2020000 60000 100000 140000 180000

Slo

pe fo

r C

d 22

8 nm

Background (counts)

(20, 30) (15, 30)

(b)

Fig. 2. (a) Cd calibration slope versus laser energy (b) Cd calibration slope versus LIBSbackground. (Reproduced with permission from Ref. [21]).




challenge. To establish a calibration scheme for quantitative measurement in practicalenvironments, a series of studies were made to correlate LIBS backgrounds with changesin excitation conditions [21]. A linear relationship between the LIBS calibration slopeand the backgrounds for Cd and Be was found. These data were obtained from spectrarecorded in the 230-nm spectral regions with different laser energies, gate windows,and test cells. Figure 2 shows the linear relation between the laser energy and the Cdcalibration slope. The LIBS background was also found to be linearly proportional tothe calibration slope (see Fig. 2). The data were recorded with gate delays of 20 �s and15 �s with a fixed gate width of 30 �s. These results imply that the background can beused to correct the changes in plasma conditions. However, the same experiment in the415-nm spectral region shows a linear relationship between background and calibrationslope only when laser energy is below a certain limit (see Fig. 3). At higher laser energy,the CN interference is dominant in this spectral region, and the intensities of the analytelines of Pb and Cr are possibly saturated. The results of the background study showthat background normalization can be used to correct the calibration factor due to minorchanges in the plasma condition. However, this approach demands great care due to itslimitations.

140 160 180 200 220 240 260 280 300

1800

1600

1400

1200

1000

800

600

Laser energy (mJ)

Slo

pe fo

r C

r 42

5 nm

(a)

2200

2000

1800

1600

1400

1200

1000

800150000 250000 350000 450000

Slo

pe fo

r C

r 42

5 nm

Background (counts)

(40, 40) (50/40)

(b)

Fig. 3. (a) Cr calibration versus laser energy (b) Cr calibration slope versus LIBS background.(Reproduced with permission from Ref. [21]).




4. APPLICATIONS

LIBS measurements of off-gas have a wide range of applications from the process controlof a production plant to thermal waste treatment process. The detection of trace metalsin the off-gas of various industrial plants such as coal-fired power plants, cement kilns,incinerators are a great public-health and environmental concern. Conventional analyticaltechniques involved getting sample from sites and sent to laboratory for analysis. It had todeal with sampling problems in off-gas system. LIBS can perform off-gas measurementsby focusing the laser beam on the gas stream through a window and collecting thesignal through an optical fiber. It offers a technique to perform remote and in-situmeasurements. Radziemski and Cremers have applied LIBS to analyze effluent gasesfrom a prototype fixed-bed coal-gasifier at the DOE Morgantown Energy TechnologyCenter [1]. They demonstrated that LIBS has the capability for near real-time monitoringof the concentrations of major and minor species in the off-gas emission. Neuhauseret al. tested their on-line Pb aerosol detection system with aerosol diameters between10 and 800 nm. A detection limit of 155 �g/m3 was found [22]. Singh et al. have usedLIBS as a process monitor and control tool for waste remediation [23]. They monitoredthe toxic metals from three plasma torch test facilities and proved that LIBS can beintegrated with torch control systems to minimize the toxic metal emission during plasmatorch waste remediation. Ferioli et al. have used LIBS to measure the equivalence ratioof a spark-ignited engine in a laboratory setting. They used C/N and C/O peak ratiosto quantify the equivalence ratio over a range from � = 0�8 to � = 1�2 [24]. Ballet al. investigated the feasibility to apply LIBS as hydrogen-sensing technique to detecthydrogen leak for real-time monitoring. Using hydrogen 656.28-nm line, they obtain alimit of detection of about 20 ppm (mass) [25].

4.1. Analysis of Air-Sampling Filters with LIBS

Filter collection with a sampling pump is widely used in environmental monitoring andpersonal protection. This technique can detect very low species concentrations throughtime accumulation. Conventionally, the filter is analyzed via chemical laboratory workthat includes two main steps: chemical washing of the filter to produce a solutionand, thereafter, performing a solution analysis. This is a time-consuming procedure andcan require several hours. With LIBS, the collected species mass on the filter can bedetermined rapidly by laser sparks across the filter surface. Since the laser inducedsparks vaporize and excite the sample without any sample preparation, the analysis timeis reduced to a few minutes. The direct detection of trace elements in air with LIBS isvery difficult due to its insufficient sensitivity. For trace metals below the LIBS detectionlimits, an air sample can be collected on a filter, which is then analyzed by LIBS. Thismethod results in a lower detection limit and provides a quasi-on-line measurement.Arnold and Cremers have used this technique to determine metal particles on an airdamping filter [26]. They used a cylindrical lens to form a long spark on the filter toincrease the sample volume and reduce the filter damage. Using the calibration curvefor Tl line at 535.05 nm, a LOD of 40 ng/cm2 for Tl in filter paper was obtained. Later,Yamamoto et al. determined LODs of 21 ng/cm2 and 5.6 �g/cm2 for Be and Pb on afilter [27]. They also noticed that particle size can affect the detection limit for filter




analysis. Cremers and Radziemski have measured Be on a filter surface and found thatBe particles greater than 10 �m on the filter were not completely vaporized [28]. Theparticle size dependence of the LIBS signal restricts LIBS applications to air samplingthrough a filter [28].

The method has been tested using filter sampling and LIBS analysis to monitorparticulate emission in a mini test stand. Two types of filters, PVC membrane filtersand glass fiber filters, were evaluated for this application. The PVC membrane filterswere burned by a couple of laser-induced sparks. Therefore, it is not suitable for LIBSapplication. Glass fiber filters made of borosilicate glass fiber have a maximum operatingtemperature of 500�C in air. Since it does not burn under the laser-induced spark, itis ideal for LIBS applications. This type of filter was then used in all the experimentsconducted at the ICET mini test stand.

The mini-test stand was operated at a 10-lb/hr air flow during the study. The drymetal aerosol of desired concentration produced by an ultrasonic nebulizer was injectedinto the test stand. A 1.5-inch diameter, 1-�m pore size glass fiber filter was put ina closed-face air-monitoring cassette. The sampling device with the filter cassette wasinstalled on a sampling port, which is 1.2 m downstream from the aerosol injection port,to collect the aerosol sample in-line. In most of the experiments, the sampling flowwas set to 1.5 L/min to match the velocity of the sampling flow with the velocity ofthe gas stream. Each filter was used to collect samples for 10 to 20 minutes beforeanalysis by LIBS. Since there was no absolute concentration measurement available atthe sampling position, LIBS measurements in the gas stream were also performed ina port 15 cm upstream from the sampling port to provide a reference. For LIBS filteranalysis, a filter was placed on a platform that rotated around the vertical axis with aspeed of 300 rpm. During the LIBS measurements, the platform was translated to letthe focused laser beam scan at different radii on the filter surface. The experimentalparameters for filter analysis are: a laser energy of 10–15 mJ, a gate delay of 1.5–2 �s,and a gate width of 5–10 �s. The experimental parameters for off-gas analysis are: alaser energy of 120–130 mJ, a gate delay of 10–20 �s, and a gate width of 10–30 �s.

To verify the performance of the filter, some sampled filters were sent to a laboratoryfor conventional chemical analysis to obtain the Be mass collected on the filter. Theabsolute Be concentration in the gas stream was then calculated from the sampling rateand Be mass deposited on the filter. The Be concentration in the gas stream was alsocalculated based on the concentration of the solution injected, nebulizer operation para-meters, and mini test stand operation parameters. A comparison of metal concentrationin the gas stream inferred from these two methods is shown in Table 2. The relativedifferences between these two methods are 4 to 15%. It is noted that the filter’s samplingability may be decreased as the aerosol concentration increased. This is because thistype of filter is initially designed to work with a very low concentration.

To evaluate the analytical performance of LIBS on a filter, some sampled filterswere analyzed by LIBS and compared with the on-line LIBS analysis. Fig. 4 showsthe LIBS signals obtained from filters and from the on-line measurements versus theBe concentration of the solution injected. The on-line LIBS measurements are used tomonitor the performances of the solution injection system and mini test stand. Since theintensities of LIBS obtained from the gas stream are linear to the solution concentration,a linear relationship is expected against the solution concentration from the LIBS filterdata. At low Be concentrations (1, 2, and 4 �g/ml), the intensities of LIBS on filters




Table 2. Comparison of the concentration of Be aerosol calculated from filter analysis andfrom the solution concentration

Be concentration insolution (�g/ml)

Be collectedon filter (�g)

Be concentration in gas stream (�g/m3� Relativedifference

calculationbased on thefilter analysis

calculated from thesolution concentration,gas flow rate, etc.

6 0�8 31�37 32�68 4�0%8 0�9 35�29 39�53 10�7%10 1�0 39�22 46�21 15�1%

00

2 4 6 8 10 12

10

20

30

40

50

60

70

Be solution concentration (µg/ml)

Be

(313

.04

nm)

inte

nsity

(th

ousa

nds) LIBS in flow

LIBS on filter

Fig. 4. LIBS signals obtained from filter and off-gas measurements.

do appear fairly linear with solution concentrations. The data taken at Be concentrationof 6 �g/ml shows a reduced intensity. It could be due to self-absorption in the plasma,incomplete vaporization of the particles, or lower sampling efficiency at high metalconcentration. However, it can be found that the reproducibility of LIBS on filters isfairly good (10–20% RSD), as compared to LIBS in flow (20–30% RSD). Since thefilter moves during data collection, a small standard deviation suggests that the massdistributed on the filter is fairly uniform.

Similar experiments for Cr and Pb have been conducted. The results are very similarto those shown in Fig. 4. The signal from low concentration filter data increases as thesolution concentration increases. The behavior of LIBS filter data at higher concentrationsis more complicated to explain. More systematic experimental study is required. Toevaluate the sampling performance at different sampling times, the gas stream wascontinuously monitored while a Cr solution of 30 �g/ml was injected into the mini teststand. The results of these measurement are shown in Table 3. Here, the intensitiesfrom LIBS filter data are normalized by the sampling time. The intensity ratio of theLIBS filter data and concurrent off-gas data is about the same for a sampling timeof 25 minutes. A higher intensity ratio was found in shorter sampling times (e.g., an




Table 3. LIBS signals from filters at different sampling times

Filter LIBS Measurement Concurrent Off-gas LIBS Measurement I1/I2

Sampling Time(Minutes)

Time averagedIntensity �I1�

Averaged LIBS signal �I2�

25.0 353 8366 0�04225.0 309 7460 0�041

8.5 468 6638 0�070

8.5-minute sampling). This shows that the longer sampling time for this Cr concentrationlevel reduced the Cr detection sensitivity. Therefore, sampling time needs to be adjustedfor different concentration levels.

Although this quasi-continuous emission monitor method is very useful for lowemission environments, it is difficult to achieve reliable quantitative results due to itssensitivity to the particle size, sampling time (concentration dependent), and collectionefficiencies for different elements. A great amount of chemical and LIBS analysis isneeded to establish the optimum conditions (filter, sampling time, LIBS setup, etc.) foreach type of sample (concentration level, particle size, etc.).

4.2. Continuous Emission Monitor

The amount of toxic metal added to the atmosphere is restricted and controlled byvarious U.S. Environmental Protection Agency (EPA) rules and permits. The EPA ismodifying regulations to further reduce metal emissions. Thus, the measurement of toxicmetals is very important for compliance with the existing EPA rules and also for theproposed Maximum Achievable Control Technology (MACT) rule in the future. Severalpotential techniques that have been evaluated for this application include inductivelycoupled plasma atomic emission spectrometry (ICP-AES), LIBS and X-ray florescence.Among these methods, LIBS is the only technique which provides real-time, in-situanalysis which is important for a continuous emission monitor (CEM) [29,30]. A CEMsystem needs to provide immediate warning as the level of the toxic metal in off-gasresults in a dangerous level of toxic metal released into the atmosphere. LIBS capabilityfor continuous, real-time analysis makes it an ideal technique for a CEM for thermaltreatment plants. The only problem with LIBS is that the sensitivity for some toxic metalmight not be enough.

A LIBS system developed in the laboratory has been tested in two U.S. Departmentof Energy (DOE)/EPA CEM tests [29,31]. The CEM test was designed to measure theperformance of multi-metal CEMs for regulatory compliance applications. It was con-ducted at the EPA’s Rotary Kiln Incinerator Simulator (RKIS) facility, which consists ofa primary combustion chamber, a transition section, and an afterburner in the secondarycombustion chamber [26]. The kiln and secondary combustion chamber were operatedwith natural gas during the tests. Metals were introduced into the fuel gas by injecting anaqueous metal solution directly into the secondary flame of the incinerator to achieve thetarget fuel gas concentrations. To simulate actual flue gas conditions, fly ash particleswere also injected into the incinerator. The LIBS system used at a port located 5.7 m




downstream from the air dilution damper. The EPA RM sample port is 1.4 m upstreamfrom the LIBS port.

The first test demonstrated LIBS’ rapid sampling rate and potential for metal CEM.It also helped us pinpoint various problems associated with LIBS field measurements.These problems include higher limits of detection (LOD), a need for on-line calibration,degradation of optical components, and the need for simultaneous monitoring of all theRCRA metals. These problems have been extensively studied in the laboratory since thefirst CEM test. Calibration techniques have been tested in the laboratory. The LOD hasbeen reduced by a factor of eight or more for most of the metals. A method to correctthe signal loss due to the degradation of optical components during the field test wasdeveloped. The various improvements made to the ICET-LIBS system were evaluated inthe second DOE/EPA CEM test [31]. The results of the LIBS calibration study, and theresults of LIBS measurements during the second DOE/EPA CEM test, are given here.

The CEM test focused on As, Be, Cd, Cr, Pb, and Hg, which are the RCRA metalsregulated in the EPA’s MACT rules. The test program consisted of a high- and low-metaltest. The target concentrations were 75 �g/dscm in high-metal tests and 15 �g/dscmin low-metal tests. The EPA’s Reference Method (RM) and CEM measurements wereperformed concurrently for each test condition. The number of RM measurements per-formed for each test depended on the target metal concentration. The RM sampling timewas one hour for the high-target-metal test and two hours for the low-metal tests. Therewere in total twenty RM samplings during the entire test, ten for low-metal tests, andten for high-metal tests. Due to the difficulty in injecting a known amount of sampleinto a practical gas stream, LIBS calibrations were performed in the laboratory beforethe field test. The two LIBS detection systems used in the field test were calibrated forall the RCRA metals. The peak area of an analyte line from the calibration LIBS datawas used to construct the calibration curves. Linear regression was used to obtain thecalibration factor.

On-site calibrations for Cr, Pb, Cd, and Be were performed at RKIS during theshakedown test with a calibration setup similar to that shown in Fig. 3a of chapter 5. Theon-site calibrations were done by injecting metal aerosol into the RKIS gas stream witha probe. The sample injection probe was mounted across the gas stream on the oppositeport. Since the gas flow quickly diluted the injected sample in the gas stream, the metalconcentration near the focal volume could not be accurately estimated. Therefore theon-site calibrations were mainly used to check system response. The temperature, fluegas flow rate, and particle loading in the test environment were ∼232�C, 3.4 scm/min,and 25–50 mg/dscm, respectively. The effects of these gas-stream parameters on LIBScalibration had not been systematically studied before.

The concentrations of Be, Cd, Cr, and Pb were monitored simultaneously in nearreal-time during the four-day test. Analyte lines of Cd and Be were monitored inthe 220-260-nm spectral region with a 1200-line/mm grating, while analyte lines ofPb 405.8 nm and Cr 425.44 nm were monitored simultaneously in the 400-429-nmspectral region with an 1800-l/mm grating. During the test, it was found that the high-quality optics used in the LIBS system degraded quickly, causing the LIBS signal todrop significantly. The dichroic mirrors used in LIBS have high-damage thresholds�>GW/cm2� under normal operating conditions. However, the properties of the opticalcoating changed in the humid and hot test environment, resulting in a lower damagethreshold than the specification and damage occurred rapidly in the field test. Fig. 5a




150

100

50

009:21 09:36 09:50 10:04 10:19 10:33 10:48

Be

conc

entr

atio

n (µ

g/ac

m)

Time

(c)

150

100

50

009:21 09:36 09:50 10:04 10:19 10:33 10:48

Be

conc

entr

atio

n (µ

g/ac

m)

(b)

30000

25000

20000

1500009:21 09:36 09:50 10:04 10:19 10:33 10:48

Bac

kgro

und

@ 2

30 n

m

(a)

Fig. 5. (a) LIBS background, (b) Raw CEM data, and (c) Corrected CEM data during RMsampling period. (Reproduced with permission from Ref. [21]).




shows the LIBS background recorded during one RM sampling period. It clearly showsthe LIBS signal falling as the optics gradually degraded. The background normalizationtechnique mentioned above was used to correct LIBS’ raw data. Figs. 5b and 5c showthe raw CEM data and background corrected CEM data during the RM sampling period,respectively. A more steady-inferred metal concentration was obtained after correction.This indicates that this technique is effective for the problems of optics damage. Thistechnique was then used to correct all the LIBS data collected during this field test. Theeffects of fly ash and temperature were taken into account in recalling the laboratorycalibration factor for CEM test.

This LIBS system was successfully used to simultaneously monitor concentrationsof Cd, Be, Cr, and Pb in near real-time during both the high- and low-metal tests.The system response time mainly depends on the sampling rate of the system. In thisCEM test, the LIBS system response time was 10–20 seconds. The measured metalconcentrations have been compared with the results from EPA’s RM. A comparisonof the time-averaged LIBS data (over the RM sampling period) along with the dataobtained with RM is shown in Fig. 6. The relative accuracy of LIBS for four elementsbased on the RM results was found to be in the range 19–78%. The expected accuracy inthese measurements was 20 or 50%, which is much higher than expected in an analyticallaboratory measurement. The LIBS data taken during the four test days roughly followedthe trend of the RM data. It was found that LIBS data was more consistent with RMdata for the last test day. This is because the experimental setup was more stable onthat test day due to a cooler probe and a new dichroic mirror. During the first three testdays, more technical problems were encountered such as optics damage and laser powerdropping due to the sensitivity of the frequency doubler affected by the environmentaltemperature. The rough correction with the background used in this test has shownpromising results. However, a more refined correction taking into account the effects ofgas-stream parameters should improve the accuracy of LIBS.

–1.5

–1

–0.5

0

0.5

1

1.5

(LIB

S-R

M)/

RM

0 5 10 15 20

RM #

Be Cr Cd Pb

High metal test Low metal test

Fig. 6. Comparison of CEM and RM data.




LIBS has shown its capability as a multi-metal CEM for Cd, Be, Cr, and Pb in theabove field tests. Background normalization technique has also proved to be a usefulmethod to correct the signal variation due to optics damage during the field test. However,currently LIBS’ sensitivity, precision, and accuracy for certain toxic metals, have stillnot reached EPA’s requirements to be accepted as a metal CEM. Future developmentin this area may include improving the detection sensitivity of all the RCRA metals.A calibration routine for automatically compensating plasma-condition changes due tovariations of gas-stream conditions or pulse-to-pulse laser fluctuations is also needed.

4.3. Process Control

A mobile instrument was used in the advanced analytical instrumentation demonstra-tion (AAID) test at the Science Applications International Corporation (SAIC)’s STARCenter, Idaho Falls, Idaho, to demonstrate LIBS’ capability in process control. TheSTAR Center’s plasma system consists of the following components: a plasma chamber,a secondary combustion chamber, a HEPA filter, a stack, and instrumentation and systemcontrols. The details of the test facility are given in Reference [32]. Fig. 7 shows theexperimental setup of the LIBS system. LIBS measurements were performed continu-ously at a port between the baghouse and the HEPA filter. The port was purged withnitrogen to keep the window clean and cool and the same port was also used to collectthe LIBS signal. A beam dump mounted on the opposite port across the gas streamwas used to dump the laser energy. The third port in the direction normal to the laserbeam was used to monitor the spark in the gas stream and also to align the spark withthe sample injection probe for calibration. The emission from the spark was collectedwith a UV optical fiber bundle coupled to a spectrograph. An intensified diode arraydetector (IDAD) was attached to the spectrograph to record the LIBS spectrum. A laptopcomputer interfaced to the detector controller with a PCMCIA-GPIB card was usedfor data acquisition and analysis. The EG&G OMA2000 software was used to collect

LIBS dataacquisition/analysissystem

Computermonitor

TTL signalreceiver/interface box

Facilitycomputer

LIBSprobe

Lens Lens

Window

Beamblock

Alarm/indicator

Fig. 7. Schematic of LIBS system and the SAIC’s STAR Center alarm/interface system.




021:07 21:21 21:36 21:50 21:04 21:19 21:33

200000

400000

600000

800000

1000000

1200000

1400000

Time

Hg

conc

entr

atio

n (m

g/sc

m)

Alarm concentration limits for Hg

Fig. 8. Variation of Hg concentration with time in SAIC’s STAR Center off-gas during theAAID test.

LIBS data. A user-written macro-program was used to analyze and display the datain near real-time. The real-time elemental concentrations of Pb, Ce, Cr, and Hg weredisplayed on the computer monitor during data acquisition. TTL signals were sent to thealarm/interface system, and a warning message was also shown at the bottom of the com-puter screen whenever the concentration of the measured species was above the alarmlevel during the LIBS measurements. This allows the operator to modify operationalparameters of the plasma system to prevent emission that exceeds the pre-establishedfacility. During the test, the metal emission did approach the alarm levels several times(see Fig. 8). The TTL signals were successfully sent to the alarm/interference systemwhen the concentrations exceeded the alarm limit. Details of this measurement can befound in Ref. [32]. The LIBS response was determined mainly by the sampling time ofthe measurement. In this test, LIBS had a response of 50 seconds, which is sufficient toprovide critical information for process control.

4.4. Filter Efficiency

The ceramic filters used in waste processing with the plasma torch play an important rolefor removing toxic metals. LIBS has been used to evaluate metal-removal efficiency.Two LIBS systems were used to record the spectra at the inlet and outlet of a ceramicfilter during the Plasma Arc Centrifugal Treatment Pact-6 Slip Stream Test Bed (SSTB),a 100-hour duration demonstration test at Mountain State Energy (MSE) [32]. Theelements that appeared in the filter inlet were Fe, Cr, Pb, Ca, Si, Cu, Mn, Mo, C, Mg,K, Na, Sn, Zn, and Cd. The metal identified from the filter outlet spectra were Cr, Pb,Fe, K, and Mn.




Table 4. The calculated minimum metal removal efficiency in WETO/MSE test [23]

Element Maximum concentration at theCeramic filter inlet (mg/acm)

LOD (�g/acm) Metal removalefficiency (%)

Cr 245 12 99�99Zn 240 570 99�7Cd 119 45 99�96Sn 200 190 99�9Mn 106 7�5 99�99Pb 888 90 99�9Fe 400 140 99�96

During most of the test, the concentrations of the target metals in the filter outletwere below our detection limit. Therefore, only the minimum metal-removal efficiencycould be calculated by assuming that the target-metal concentration in the ceramic filteroutlet equaled its detection limit. The metal-removal efficiency was calculated by usingthe ratio of the detection limit of the metal and the highest metal concentration foundin the ceramic filter inlet. The results of this calculation are shown in Table 4, whichclearly shows that the efficiency of the ceramic filter for most of target metal wasbetter than 99.9%. The estimated efficiencies for Zn, Sn, and Pb were found to belower than for other metals, and this is probably due to relatively higher LODs forthese three metals. Cr, Fe, and Pb were detected occasionally beyond the ceramic filterwhen the metal concentration levels were momentarily above our detection limit (thismight be related to some problems in the ceramic filter operation or torch processingsystem).

The metal partitioning is the ratio of the metal in the off-gas and the metal in thefeed. It can be used to monitor the facility operation. The major factors, which can affectthe metal partitioning, are the feed rate, the feed composition, and the plasma torch’soperating condition. The metal partitioning was calculated using the time-averaged metalconcentrations measured before the ceramic filter during the actual feed and the metalfeed rate. The calculated metal partitioning for Cr, Fe, and Pb were found to be 0.17%,0.066%, and 2.3%, respectively. Since Pb is more volatile, it has a higher partition thanCr and Fe, as expected.

To study the plasma torch vitrification process and the performance of various off-gascomponents, Mn and Cd were selected as tracers during the test, one with a high meltingpoint and the other with a low melting point. The tracers were injected into the plasmatorch vessel. The time lag between the metal addition to the plasma torch vessel andthe observation of the metal in the gas stream can be defined as the residence, and is animportant parameter for evaluating particular waste treatment processes. Mn was foundto be the best tracer for the present LIBS system. The concentration spikes of Cd athigh concentrations are not as well-defined as Mn although Cd can still be a good tracerwhen operated at a low injection amounts and for a longer injection interval.

Based on the LIBS data obtained in this test, some important facility operationinformation was obtained such as the minimum metal-removal efficiency of a ceramicfilter, suitable tracers for the residence time measurement, and metal partitioning.




4.5. Combustion Diagnostic

LIBS can also be used as a diagnostic tool for combustion systems by measurements offuel-air ratios, fuel composition, and temperature. In combustion, the hydrocarbon mix-ture and the equivalence ratio (i.e. fuel-to-air ratio) are two improtant parameters. LIBScan simutaneously detect varous elemental abundance in the combustion environment.It was first applied to the hydrocarbon mixture by Schmieder in 1981 [33]. Phuoc andWhite have determined the equivalence ratio of a CH4/Air mixture of a jet diffusionflame with LIBS by measuring the ratio of oxygen-to-carbon, nitrogen-to-carbon andhydrogen-to-oxygen in flame [34]. Sturm and Noll have performed measurements of C,H, O, and N in gas mixtures of air, CO2, N2, and C3H8 to establish the calibration curvesfor LIBS detection of these elements (i.e. LIBS signal versus the partial pressure) [35].Stavropoulos et al. have used LIBS in a methane/air premixed flame to demonstrate thatthe ratio of hydrogen atom and oxygen atom varies linearly with equivalence ratio [36].Blevins et al. have applied LIBS to high temperature industrial boilers and furnaces [37].They used a novel LIBS probes designed for these high temperatures and high particleloadings environments. Multi-elements were simultaneously detected with an Echellespectrometers coupled to intensified CCD cameras. It shows that LIBS can be used as asensitive, on-line process diagnostic for equivalence ratio monitoring in flame reactors.

The feasibility to apply LIBS to practical combustion environment was first evaluatedby Singh et al [38]. It has been used to characterize the upstream region of a largemagnetohydrodynamic (MHD) coal-fired flow facility (CFFF). The relative concentra-tions of several species were inferred by fitting the observed CFFF LIBS spectra withcomputer-simulated spectra. This was the first LIBS experiments in a harsh, turbulent,and highly luminous coal-fired MHD combustion environment.

Lee et al. also use LIBS to combustion and other thermal systems for simultane-ous measurements of a number of important thermo-chemical parameters, includingtemperature [39]. They have compared the results of LIBS flame temperature measure-ments with other methods such as thermocouple and Rayleigh scattering and have foundexcellent agreement even in sooting flames.

4.6. Rocket Engine Health Monitor

Detection and characterization of metallic species in the exhaust plume of hydrocarbonfueled rocket engines can indicate the presence of wear and/or corrosion of metal inthe rocket engine. This information on engine wear obtained during engine operation isvery useful, allowing the possibility of engine shutdown before any catastrophic failure.It has been observed that a catastrophic engine failure is generally preceded by a brightoptical emission, which results from the erosion of metal from the engine parts. This isbecause of high temperature in the rocket plume �∼2000 K�, which partially vaporizesand atomizes the metal species, leading to atomic emission in the near ultraviolet andvisible spectral range (300 nm–760 nm). A traditional method for monitoring the engineplume during a test is atomic emission spectroscopy in the near ultraviolet and visiblespectral regions. The hydrocarbon-fueled engine contains various species such as atomiccarbon, C+

2 , and other carbon-free radicals, which will increase the background emissioncomparatively more than the main OH band in the oxygen- and hydrogen-fueled engines.




Even the scattering from the unburned carbon will also produce a strong background,which has proved to be a disadvantage for the atomic emission spectroscopic techniqueto detect the presence of metal corrosion and engine wear in a hydrocarbon-fueled rocketengine.

LIBS provides an alternative technique to plume diagnostics. It uses a gated detec-tion, which discriminates against strong background emission. It can provide spatiallyresolved, real-time measurement of several metallic species critical for monitoring thehealth of a rocket engine.

The performance of LIBS was evaluated in detecting the trace of elements in the fuelplume of a hybrid rocket engine simulator at Stennis Space Centre (SSC), Mississippi,USA. The hybrid rocket engine simulator used Plexiglas as its fuel [40]. An adequateflow of oxygen was maintained for proper burning of plexiglass. Initially, spark wasstarted in a small chamber between the electrode and body of the ignition chamberusing an electric current from a 12 V battery. This initial spark started the ignition ofplexiglass as a main fuel, which generated a high-speed, luminous plume of ∼2 inchesfrom a 3–4-mm-diameter exit nozzle. The laser was focused at various locations of theplume to record the LIBS spectra at different spatial locations with a lens of 10 cm focallength. Copper and stainless steel wires were used as the seeding samples by keepingthem axially inside the ignition chamber extending up to the exit nozzle. The samplemetals melt and vaporize due to the high temperature of the fuel and then exit with theburnt fuel as a plume.

LIBS spectra of the rocket engine simulator plume were recorded when a 316Lstainless steel wire of diameter 1.76 mm was inserted into the ignition chamber. Thelaser was focused 3 inches away from the exit of the nozzle. Fig. 9a shows the strongatomic lines of chromium in the LIBS spectra. The stainless steel 316L contains ∼17%of Cr. LIBS spectra of the plume have shown a significant amount of Cr present in theplume. No Fe lines were found in the spectra at this location, which is probably due tothe low concentrations near the exit channel where the plume speed is very high. Thepresence of chromium lines seems to be due to their high transition probabilities. Fig. 9bshows the LIBS spectrum of the plume ranging from 305 to 350 nm. This spectrumshows the presence of strong OH and NH bands, as well as the two strong lines fromatomic copper. OH and NH bands appear as a result of the reaction of hydrogen andoxygen as well as atmospheric nitrogen and hydrogen, respectively. The copper linespectrum is from the sample of copper wire kept inside the ignition chamber. Strongcopper lines were recorded in the emission from the plume when copper wire was keptout of contact with the plume near the exit channel. However, it could be detectedfor a fraction of second as it vaporized and escaped with the high-speed burnt-fuelplume.

The LIBS spectra of the plume were recorded-at different spatial locations from theexit nozzle. Fig. 10 shows the LIBS spectra at different spatial locations of the plumewhen copper wire was used as the seeded sample in the ignition chamber. Spectra wererecorded at 13/8", 2", 3", and 5" from the exit nozzle. The presence of copper wasdetected strongly near the nozzle exit during an initial fraction of a second when theburnt-fuel plume started building up. The Cu signal decreased with time when the plumeattained its full length �>2"�, high temperature, and high speed.

It was noted that as the measurement point moved away from the plume exit channel(luminous zone), the copper lines appeared throughout the plume. The background




Cu

324.

7

Cu

327.

4

NHOH

Inte

nsity

Wavelength (nm)

80000

70000

60000

50000

40000

30000

20000

10000

0305 310 315 320 325 330 335 340 345 350

12000

10000

8000

6000

4000

Inte

nsity

Wavelength (nm)

2000

0345 350 355 360 365 370 375 380 385 390

Cr

357.

87

Cr

359.

35

Cr 360.53

Fig. 9. LIBS spectra of the rocket engine simulator. (Reproduced with permission from Ref. [40]).

emission from the plume also decreased in this measurement location. This is likelydue to the better mixing of metal vapor in the plume away from the exit channel. Inthis location, the plume has a lower speed and a lower gas temperature, as comparedwith the plume near the exit channel. The data from the preliminary test show that




40000

30000

20000

10000

00 5 10 15

Time (second)20

D = 1-3/8"

25

Inte

nsity

(C

u 32

4.7

nm)

40000

30000

20000

10000

00 5 10 15

Time (second)20

D = 2"

25

Inte

nsity

(C

u 32

4.7

nm)

40000

30000

20000

10000

00

5 10 15

Time (second)20

D = 3"

25

Inte

nsity

(C

u 32

4.7

nm)

40000

30000

20000

10000

00

5 10 15

Time (second)20

D = 5"

25

Inte

nsity

(C

u 32

4.7

nm)

Fig. 10. Variation of LIBS signal with time at different spatial locations. (Reproduced withpermission from Ref. [40]).

the measurements made away from the luminous part of the plume can provide moremeaningful information about the health of rocket engine.

This test demonstrated that LIBS is capable of being used as an engine health monitorto detect a trace amount of metal emerging from any part of a rocket’s engine. Theproper calibration and metal seeding techniques are now under investigation.




5. CONCLUSION

Laser-induced breakdown spectroscopy is now a very active field in analytical science.Considerable progress in the area of basic and applied research of LIBS has been madeduring the last two decades. Many research laboratories all over the world are workingin this field. However, the achievements are not always made known or implemented.A better international collaboration is needed for unification and application of currentLIBS techniques. This chapter describes the analytical analysis of gaseous samples usingLIBS. It demonstrates that LIBS may be utilized to detect trace metals in the off-gas ofvarious industrial plants and also used for combustion diagnostics. The results presentedhere reveal that glass-fiber filters can be used to collect air samples and are suitablefor LIBS analysis. The potential of using LIBS to detect metallic species in the exhaustplumes of rocket engines has also been demonstrated.

In the last decade, the LIBS technique has progressed due to improvement in seve-ral experimental parameters to detect trace elements in various types of samples. Tocommercialize the LIBS technique for industrial and environmental applications, itssensitivity and precision need to be further improved. Also more work is required toimprove the calibration methods especially an on-line calibration method for CEM.

REFERENCES

[1] L. J. Radziemski and D. A. Cremers, Laser Induced Plasma and Applications, Marcel Dekker,New York (1989) Chapter-7.

[2] D. C. Smith and R. G. Meyerand, Jr., Principles of Laser Plasma, G. Bekefi, Ed., Wiley,New York (1976) p. 457.

[3] R. W. Schmeider, Combustion applications of laser-induced breakdown spectroscopy, 13thAnn. Electro-Opt./Laser Conf. (1981) Anaheim, CA.

[4] R. W. Schmeider, and A. Kerstein, Appl. Opt. 19 (1980) 4210.[5] L. J. Radziemski and T. R. Loree, J. Plasma Chem. Plasma Proc. 1 (1981) 281.[6] D. A. Cremers and L. J. Radziemski, Anal. Chem. 55 (1983) 1252.[7] L. J. Radziemski, T. R. Loree, D. A. Cremers and N. M. Hoffman, Anal. Chem. 55

(1983) 1246.[8] D. W. Hahn, Appl. Phys. Lett. 72 (1998) 2960.[9] S. G. Buckley, Environmental Engineering Science, 22 (2005) 195.

[10] A. C. Samuels, F. C. Delucia Jr., K. L. McNesby, A. W. Miziolek, Appl. Opt. 42 (2003) 6205.[11] C. A. Munson, F. C. De Lucia, T. Piehler, K. L. McNesby, A. W. Miziolek, Spectrochim.

Acta B60 (2005) 1217.[12] J. D. Hybl, G. A. Lithgow, S. G. Buckley, Appl. Spectrosc. 57 (2003) 1207.[13] G. Bekefi, Principles of Laser Plasmas, Wiley, New York (1976) p. 457.[14] R. S. Adrain and J. J. Watson, Phys. D: Appl. Phys. 17 (1984) 1915.[15] D. P. Balwin, D. S. Zamzow and A. P. J. D’Silva, J. Air & Waste Management Association,

45 (1995) 789.[16] J. P. Singh, H. Zhang, F. Y. Yueh and K. P. Carney, Appl. Spectrosc. 50 (1996) 764.[17] W. S. Shepard, et al., “Application of modern diagnostic methods to environmental improve-

ment”, Mississippi State University: Diagnostic Instrumentation and Analysis Laboratory.(1996) 10575 FY 96 Annual.

[18] H. Zhang, J. P. Singh, F. Y. Yueh and R. L. Cook, Appl. Spectrosc. 49 (1995) 92.[19] D. K. Ottesen, J. C. F. Wang and L. J. Radziemski, Appl. Spectrosc. 43 (1989) 967.




[20] M. Corsi, G. Cristoforetti, M. Hidalgo, S. Legnaioli, V. Palleschi, A. Salvetti, E. Tognoniand C. Vallebona, Appl. Opt. 42 (2003) 6133.

[21] H. Zhang, F. Y. Yueh and Jagdish P. Singh, J. Air & Waste Manage. Assoc. 51 (2001) 681.[22] R. E. Neuhauser, U. Panne, R. Niessner, G. A. Petrucci, P. Vavalli and N. Omenetto, Anal.

Chim. Acta 346 (1997) 37.[23] J. P. Singh, F. Y. Yueh, H. Zhang and R. L. Cook, Process Control and Quality, 10 (1997) 247.[24] F. Ferioli, P. V. Puzinauskas and S. G. Buckley, Appl. Spectrosc. 57 (2003) 1183.[25] A. J. Ball, V. Hohreiter, D. W. Hahn, Appl. Spectrosc. 59 (2005) 348.[26] S. D. Arnold and D. A. Cremers, AIHA Journal, 56 (1995) 1180.[27] K. Y. Yamamoto, D. A. Cremers, M. J. Ferris and L. E. Foster, Appl. Spectrosc. 50 (1996) 22.[28] D. A. Cremers and L. J. Radziemski, Appl. Spectrosc. 39 (1985) 57.[29] H. Zhang, F. Y. Yueh and J. P. Singh, Appl. Opt, 38 (1999) 1459.[30] D. W. Hahn, W. L. Flower, and K. R. Hencken, Appl. Spectrosc. 51 (1997) 1836.[31] J. P. Singh, H. Zhang and F. Y. Yueh, Technique report for continuous emission monitor

(CEM) test at the Rotary Kiln Incinerator Simulator (RKIS) at the EPA EnvironmentalResearch Center, Research Triangle Park, Raleigh, NC, September (1997).

[32] A. L. Kielpinski, J. C. Marra, R. F. Schumacher, J. Congdon, J. Etheridge and R. Kirkland,“Testing of Refractory Materials for Plasma Vitrification of Low-Level Mixed Wastes”,in Proceedings of Waste Management Symposium, Tucson, Arizona, February 26-March2 (1995).

[33] R. W. Schmieder, “Combustion applications of laser-induced breakdown spectroscopy” inProceedings of the Electro-Optics Laser Conference (Cahners, Chicago, ILL. (1981) p. 17.

[34] T. X. Phuoc and F. P. White, Fuel 81 (2002) 1761.[35] V. Sturm, R. Noll, Appl. Opt. 42 (2003) 6221.[36] P. Stavropoulos, A. Michalakou, G. Skevis and S. Couris, Spectrochim. Acta B60 (2005)

1092.[37] L. G. Blevins, C. R. Shaddix, S. M. Sickafoose and P. M. Walsh, Appl. Opt. 42 (2003) 6107.[38] J. P. Singh. H. Zhang, F.-Y. Yueh, and R. L. Cook, Laser-Induced Breakdown Spectroscopy

in a Metal-Seeded Flame; 28th Intersociety Energy Conversion Engineering ConferenceProceedings (IECEC); August 8–13, Vol. 1 (1993) 995.

[39] T.-W. Lee, N. Hegde and I. Han, Laser-Induced Breakdown Spectroscopy for In-SituDiagnostics of Combustion Parameters Including Temperature, UKC2005: Aerospace Sci-ence & Technology Symposium (ASTS), University of California, Irvine (UCI), Irvine,California, USA. August 11–13 (2005).

[40] V. N. Rai, J. P. Singh, C. Winstead, F.-Y. Yueh and R. L. Cook, AIAA Journal Vol. 41(2003) 2192.



Chapter 10

Laser-Induced Breakdown Spectroscopyof Liquid Samples

V. N. Raia, F. Y. Yuehb and J. P. Singhb

aLaser Plasma Division, Raja Ramanna Centre for Advanced TechnologyP.O. CAT, Indore 452 013, INDIA

bInstitute for Clean Energy Technology, Mississippi State University,205 Research Boulevard, Starkville MS 39759, USA

1. INTRODUCTION

Laser-induced breakdown spectroscopy (LIBS) has been used for qualitative andquantitative analysis of elemental compositions from many different type of samples [1].LIBS uses a focused high intensity pulsed laser beam to produce laser induced spark onthe sample surface. In the resulting high-temperature plasma, the components of sampleare basically reduced to atoms and ions. The excited atoms and ions decay to lowerenergy states by emitting the radiation. Recording the atomic emission spectrum thusenables the identification and quantification of the elemental components in the sample.The main advantage of LIBS technique over conventional methods is the capability ofan online and real time analysis of almost all types of materials without any (or with alittle) sample preparation [2–6].

LIBS has generally been applied to the analysis of solid samples and comparativelyless attention has been paid to LIBS analysis of liquids [4,7–8], suspension in liquids[9–10] and samples submerged in liquids [11]. Production of a viable system for the on-line LIBS analysis of liquids requires solutions of some general problems encounteredwith plasmas generated from liquids in addition to a number of technical issues. Frequentcleaning of exposed optical components (focusing lens or window) has to be minimizedto remove accumulated matter ejected and splashed from the liquid sample by incidentlaser pulses. The miniature shock waves associated with vaporization of liquid samplescreate aerosols above the liquid surface and disrupt both the incident laser beam and theemitted light returning to the spectrometer. Shock waves also tend to induce waves onthe liquid surface, which increase shot-to-shot signal variation and lower the precisionof spectral measurements. The laser pulses also generate bubbles inside liquids that aretransparent at the laser wavelength. These bubbles may reach the liquid surface andchange the characteristics of the laser-induced plasma, thereby affecting reproducibilityof measurement. When the bubbles created inside the liquid by the laser pulse burst at




224 V. N. Rai et al.

the surface, or the waves induced on the surface by the laser pulse are not dissipated,they change the angle of incidence between the laser beam and the liquid surface. This,in turn, can change the fluence of the laser, and hence the emission intensity. Theaerosols created by the laser-liquid interaction also absorb the laser beam, and partiallyprevent the laser light from reaching the sample surface. This absorption can change thereproducibility of the measurement by affecting the energy delivered to the sample.

To overcome these problems a variety of experimental LIBS configurations have beenemployed for studies of liquid surfaces [12–13], bulk liquids [7] and liquid jets [14].Main aim of the present article is to discuss various techniques used for recording theLIBS of liquid samples with increased sensitivity.

2. LIBS OF LIQUID SAMPLES

2.1. Elemental Analysis in Liquids

The detection and quantification of light and heavy elements in liquid samples are impor-tant from application points of view, particularly in industrial processing, environmentalmonitoring, and the treatment of waste material [1–6]. Golovlyov and Letokhov [15],Esenaliev et al. [16], and Oraevsky et al. [17] have studied the physical mechanisms ofthe ablation and breakdown on liquid samples. Initially, liquid samples were studied byfocusing the laser on the surface of the liquid, which caused heavy splashing as well asshock waves [7–9]. These effects changed the position of the liquid surface with respectto the laser focus and adversely affected the analytical results. Laser induced plasma inthe bulk of the liquid prevented splashing, but presented a drawback in terms of decreasein the duration of plasma emission. The duration of light emission from the bulk plasmais extremely short, usually of the order of 1 �s or less. Haisch et al. [9] reported a fastplasma decay time of just a few hundred nanoseconds in their bulk liquid experiments.Cremers et al. [7] found that plasma parameters could not be derived for delay timesbeyond 1.5 �s.

The major disadvantage of bulk analysis is the severely reduced plasma emissionintensity in comparison with that obtained from the liquid’s surface. Watcher andCremers [18] overcame this problem by using a surface excitation scheme in which theliquid solutions were placed in cylindrical glass vials and the plasma was then boundedon one side by the rigid glass body of the vial. The light emission from the plasmadisplayed much longer durations of the order of several microseconds with an enhancedemission intensity.

Cremers et al. [7] described a method where an initial laser pulse produced a gasbubble within the water bulk, and a time-delayed second laser pulse analyzed the gaspresent inside the bubble. This approach resulted in an enhancement in the line intensitiesby a dramatic factor of 50 for oxygen line at 777.44 nm and by a moderate factor of 3 to 4for the calcium and magnesium resonance lines, increasing the analytical sensitivityand making the bulk analysis a reasonable option. This double-pulse plasma-generationapproach has also been used by Pichahchy et al. [11] and, Nyga and Neu [10] in theirstudies on the metal composition of specimens submerged in water. Despite the evidentproblems of splashing in the case of surface excitation configuration, some researchershave used this approach. Berman and Wolf [13] and Arca et al. [12] focused the laser



LIBS of Liquid Samples 225

pulse on the surface of liquid solutions and reported a minimum delay of 1–3 �s fordetecting trace concentrations of nickel, magnesium, calcium, and chromium.

LIBS analysis of liquid samples using the laminar flows of liquid jets have beenreported [4,19]. This approach was first used by Ito et al. [19] for the detection of colloidaliron in a turbid solution of FeO in water. The authors observed iron emission lines toabout 3.5 �s after the laser pulse. The limit of detection (LOD) for iron was estimatedas 0.6 ppm. After some time Nakamura et al. [14] used a similar technique along withdouble-pulse excitation in the presence of purged gas and reported an improved limit ofdetection (LOD) of 16 ppb. Ng et al. [20] and Ho et al. [21] performed spectroscopicstudies of plasma generated from a stable water jet by using two different excitationmethods; they derived plasma excitation temperature and electron density for a delaytime up to 1 �s.

The sensitivity of LIBS for quantitative analysis of liquid samples is often poorerthan that of other analytical techniques, such as atomic emission spectroscopy ICP-AESand ICP-MS. However, the importance of LIBS comes into prominence if a remoteonline analysis is required. Remote online analysis is preferred when the measurementsare to be carried out under hazardous or difficult environmental conditions, which is notpossible by any analysis technique other than LIBS.

The work described in the following sections stems from a need in the nuclear industryto conduct real-time, on-line analyses of radioactive waste in liquid specimens. Theseare encountered during the reprocessing of nuclear fuel and in monitoring of nuclearwaste storage tanks. Particularly, technetium (Tc) is a radioactive element and a productof the nuclear power cycle. The most stable Tc isotope has a half-life of 2�1×105 yearsand decays via beta emission. Due to the long half-life and the relatively high yieldfrom uranium decay, it is desirable to separate Tc from non-radioactive and short-lifecomponents found in the tank waste. It is important to isolate it with other long-liferadionuclides in geologically stable waste for long-term safe storage. Similar problemsare also encountered in other industries where toxic liquid effluent and/or waste arepresent, and a real need exists for proper regulation of these materials. This requires areal-time, remote, on-line LIBS analysis system.

Remote LIBS analysis was conducted with a fiber-optic probe that focused a laserbeam onto the fiber, whereas the output of the optical fiber was focused on the surfaceof the sample. This technique proved suitable for remote LIBS analysis of solid sam-ples [22–23]. A modified version of the optical fiber system, along with other telescopictechniques was used for remote analysis of liquid samples [4]. The laboratory basedexperiments for recording the LIBS of liquid samples generally use a simple system ofconvex lenses for focusing the beam, either on the liquid surface or on the jet.

For quick analysis of a liquid sample (typical laser pulse repetition rates 10 to 20 Hz),the laser beam is focused on the smooth vertical surface of a laminar jet stream of theliquid, which produces plasma by surface excitation. The analyte in the liquid jet sampleis vaporized into ambient air above the liquid surface, as in the LIBS of solid surfaces.In this case, the problem of splashing is largely minimized because only very minuteamounts of solution are vaporized. Furthermore, the vaporized material is extremely welldefined (the volume roughly equals the laser spot size multiplied by the jet thickness).As a consequence of the normally unchanged laser spot size and water jet thickness,the experimental repeatability is greatly enhanced (due to low scattering of data). In thiscase, plasma largely evolves in air with minimal interaction with the rest of the liquid,




the luminous phase of the plasma is prolonged, and many spectral lines can be observedwell beyond 10 �s. Low pulse repetition rates (less than 1 Hz) are required to avoidthe splashing and oscillations on the liquid surface due to the generation of resonantshock waves.

A reliable and quantitative LIBS analysis requires knowledge of some plasma param-eters like electron density, excitation temperature, line shapes, and the time-evolution ofthe spectral line emissions. These parameters are then utilized for generating operatingconditions optimal for trace element detection. It is important to identify and optimize theparameters in order to guarantee a reliable analysis. Since LIBS is a relative measurementtechnique, calibration curves of observed signal response versus element concentrationmust be available. Such calibration curves have been generated for the elements ofinterest, which were selected as dictated by some of the intended applications. Finally,enhancing the sensitivity of LIBS using other techniques is required for decreasing theLOD of elements.

2.2. LIBS of Molten Metal

The metal-producing industry faces the major challenge of increasing productivity toreduce cost and maximize the benefits from existing equipment. During refining, it iscritical that operating parameters be adjusted and controlled so that the chemistry of themelt is within predetermined limits. The current analytical approaches to the determi-nation of the chemical composition of the melt by spark optical emission spectroscopy,atomic absorption spectroscopy (AAS), X-ray fluorescence (XRF), inductively coupledplasma (ICP) spectroscopy, and ICP mass spectrometry (MS) are limited in practice bytheir off-line character. Furthermore, these methods are either based on analysis of thecold materials, or on laborious manual sampling from the melt at elevated temperaturesbetween 500–1600 �C, which results in insufficient turn-around time, and increased pro-cess and personnel costs. Motivated by potential savings in time, energy, and materials,as well as improved quality assurance, several LIBS groups are investigating the realtime analysis of molten metals [22–29]. However, LIBS analysis of high temperaturemolten metals in processing vessels often presents major difficulties and analytical chal-lenges. For a reliable and accurate LIBS sensor, many requirements should be metsuch as:

(1) The vaporized volume should be truly representative of the liquid bulk. Thisforbids interrogating the same surface for an extended period of time since a hotliquid metal surface can quickly get enriched with elements having higher affinityfor oxygen or nitrogen, or become poorer in elements with a lower vaporizationthreshold.

(2) Perturbations from aerosols and ejected particles should be eliminated since theirplasma emission is not representative of the melt, and they cause variations in thelaser power reaching the liquid surface and available for ablation.

(3) The sensor should be sufficiently rugged for use in the harsh environment of theindustrial plant.




The determination of the composition of molten phase samples by LIBS in a furnacehas been the subject of numerous studies in laboratory and several trials in indus-try [22–29]. St. Onge et al. [29–30] employed a patented approach based on the use ofa lance without optical components in which gas under pressure is introduced, therebyproducing bubbles inside the molten metal. In this approach, a new surface truly repre-sentative of the melt is continuously exposed to the laser beam. The problem of analyzinga non stationary surface and insuring high quality data representative of the bulk wassolved by selectively processing all acquired data in the presence of bubble motionand classifying spectra from molten or solid phases. The probe has been successfullytested in many industrial facilities for the production and processing of molten materials(zinc, zinc alloys, copper, magnesium, copper matte, electrolyte bath, etc.). For zinc bathanalysis the above technique permitted rapid identification and treatment of data frommultiple species and/or phases [30]. The probe was also subjected to the harsh conditionsof the copper smelting industry at 1200 �C where it was introduced through a tuyere intoa thousand-ton molten matte vessel to monitor Fe, Bi, and Ag content [29–30]. Proberobustness was established over many days of intense experimentation. The use of asimilar probe also successfully demonstrated in-situ analysis of molten electrolyte usedfor magnesium production at 700 �C. Measurements were performed in both a pilot plantand also on-line under hostile conditions in an operating plant. In all these conditions,the patented probe overcame problems related to non-representative melt surfaces dueto oxidation, contamination, and surface migration or depletion. Consequently, excel-lent measurement reproducibility and accuracy were obtained compared to conventionalLIBS measurements on stable and stationary liquid surfaces. To the best of our knowl-edge, this probe has for the first time demonstrated LIBS measurement reproducibilityof 1% for molten metal [30].

3. INSTRUMENTATION FOR LIQUID SAMPLES

3.1. Experimental Setup for Surface Excitation

Standard LIBS analysis systems consist of three typical major blocks, namely (a) thelaser source, (b) the laser light delivery and plasma emission collection system, and(c) the system for spectral analysis. The choice of light transfer arrangement dependsmainly on target exposure procedures, which may be either a direct surface excitationfrom the liquid surface, a laminar jet stream of the liquid or inside the bulk liquid ora sample submerged in liquid. The schematic diagram of the experimental setup forrecording the laser-induced breakdown emission on the bulk liquid surface as well asin the case of a laminar jet (see Fig.1, Chapter 5) has been reported by Rai et al. [31].They used a Q-switched frequency doubled Nd: YAG laser (Continuum Surelite III)that delivers energy of 400 mJ in a 5-ns pulse duration. The laser was operated at 10 Hzduring this experiment and was focused on the target (in the center of the liquid jetor on the surface of the bulk liquid, depending on the experiment) using an UV gradequartz lens with a focal length of 20 cm. The same focusing lens was used to collectthe optical emission from the laser-induced plasma. Two UV-grade quartz lenses withfocal lengths of 100 mm and 50 mm were used to couple the LIBS signal to an opticalfiber bundle. The fiber bundle consisted of a collection of 80 single fibers with a core




diameter of 0.01 mm. The rectangular exit end of the optical fiber was coupled with anoptical spectrograph (Model HR 460, Instrument SA Inc., Edison, NJ.) as an entranceslit. The spectrograph was equipped with an 1800 and 3600 l/mm diffraction grating,with dimensions of 75 mm × 75 mm. A 1024 × 256 element intensified charge coupleddevice (ICCD) (Princeton Instrument Corporation, Princeton, NJ) with a pixel width of0,022 mm was attached to the exit focal plane of the spectrograph and used to detectthe dispersed light from the laser-produced plasma. The detector was operated in gatedmode with the control of a high-voltage pulse generator (PG-10, Princeton InstrumentsCorporation, Princeton, NJ) and was synchronized to the laser output. Data acquisitionand analysis were performed using a personal computer. The gate delay time and gatewidth were adjusted to maximize the signal-to-background (S/B) and signal-to-noise(S/N) ratios, which are dependent on the emission characteristics of the elements aswell as the target matrix. In order to increase the sensitivity of the system, around 100spectra were accumulated to obtain one averaged spectrum. For liquid jet experimentsa Teflon nozzle of diameter ��∼1 mm was used with a Peristaltic pump (Cole-ParmerInstrument Co.) to form laminar liquid jet. The laser was focused on the jet such thatthe direction of laser propagation was perpendicular to the direction of the liquid jet.The laser was focused ∼15 mm below the jet exit, where the liquid flow was laminar.However, the extent of laminar flow was found dependent on the speed of the pump.The liquid jet was aligned in a vertically downward direction.

3.2. Experimental Set up for Bulk/Molten Liquid

Liquids have also been analysed by generating plasma in the bulk of liquid [32]. Thissetup has some advantages as well as disadvantages but this technique is of great impor-tance for qualitative chemical analysis in marine environment or in the case of moltenmetals. Normally optical fiber probes are suitable for this type of experiments [22–23,32].Beddows et al. [32] have used a single large core optical fiber, both for delivering thelaser radiation to the target and collecting the plasma emission for subsequent analysis.The fiber end-faces were prepared by a cleaving process, which provided fault freeoptical surfaces. The fiber was guided in a flexible tube and held at the end of the tubein a short glass capillary. A suitable buffer gas (ordinary air, dry N2 or argon) was bledthrough this tube/capillary assembly. However the input energy of laser was optimizedon the basis of threshold for the damage of optical fiber and the energy loss during itstransmission through a long distance. The irradiance on target was kept well above thethreshold for plasma generation. The fiber end was held within the capillary tube at asuitable distance away from the sample surface to create the luminous plasma withoutdamaging the fiber during the process. The buffer gas was blown down through theannular passage between the fiber cladding and the inside of the capillary tube, resultingin a bleed stream of gas displacing the water at the position of plasma generation. Dueto close proximity of the fiber end (D ∼ 1.5–2 mm) to the sample surface sufficientlight could be collected from the plasma by optical fiber end without a lens, whichmade alignment issue quite easy. The plasma radiation was finally delivered to thespectrometer with the help of a reflecting mirror.

Similar arrangements of fiber probes have been used by many researchers for thestudy of molten metal [22–30]. A ceramic guiding tube is generally used in the case of




molten materials so that it could sustain the high temperature of the bath. Some of theresearchers have used, lens system to focus the laser, whereas others have used the fiberdirectly to produce plasma.

3.3. Liquid Configuration for Plasma Formation

Trace elements in different types of liquid matrices without any sample preparation arelikely to be detected using LIBS. Matrices such as colloids, turbid, liquids, sludge, oils,etc., require the production of plasma at the liquid surface [33]. Bulk-liquid analysis isalso possible by optical fiber probe in the liquid samples having a turbid nature, whichwould prevent the laser beam from reaching the bulk liquid.

As discussed above the general configuration used for LIBS of liquid surfaces consistsof a laser beam perpendicular to the surface. This arrangement, however, leads to splash-ing, because the plasma expansion at atmospheric pressure is directed perpendicular tothe liquid surface. A tilted laser beam configuration with respect to the liquid surfacecan minimize this phenomenon [33]. The use of low laser repetition rate of ∼1 Hz canminimize the perturbation that takes place at the liquid surface following the laser pulse.It was shown that measurements with a l-Hz laser repetition rate were more reproducible.Although the droplet and jet configurations of the liquid sample demand a little samplepreparation, the use of a pump-backed jet has several advantages:

(1) The volume evaporated in the plasma formation process is extremely well defined,being equal to the laser spot diameter multiplied by the thickness of the jet; littleor no interaction with the residual material takes place, since nearly all of thesample volume is vaporized.

(2) A suitable flexible tube system on the entrance side of the pump can be used, inarbitrary location within a large volume of a liquid, and can be probed, both inlateral direction and in depth; this, therefore, provides the possibility of probingin real time the spatial distribution of concentrations in a liquid tank specimen.

(3) One can add known amounts of elements to the flow of the sample in order toget a standard element for normalization, which will help in providing a bettercalibration curve.

St. Onge et al. [34] have evaluated three different configurations for the analysis ofliquid formulations using LIBS: analysis in closed (transparent) bottles, on the surfacesof a horizontally flowing liquid stream and in open containers (on the non flowing liquidsurfaces). They used sodium chloride in solution form as a model compound. It wasfound that analysis of a non-flowing surface provided the best compromise in terms ofease of implementation and precision. This approach is also the one most easily adaptedto the configuration of an existing commercial LIBS instrument.

The choice of a given configuration may, however, be dependent on other practicalissues. However a simple comparison of bulk and liquid-jet experiment from the spec-troscopic point of view is discussed in the following section. On the basis of previousdiscussion for minimizing the splashing and surface distortion as well as considering itsapplication for most of the liquid samples a liquid-jet system was found more suitable.Most of the data presented here have been obtained using this method.




3.4. Optimization of Experimental Parameters

The most sensitive emission lines from the possible trace elements are used to studythe effects of various experimental parameters on the sensitivity of the system. Theexperimental parameters that can most affect the limits of detection are the laser energy,lens-to-surface distance (LTSD), gate delay time, gate width and the physical propertiesof the sample. The effects of these parameters on the emission characteristics werecarefully studied using targets, as bulk liquid and as liquid jet.

3.4.1. Effects of laser energy and lens-to-sample distance (LTSD)

A high intensity laser beam was used to produce the plasma, where the required excitedelements present in the liquid sample emitted the characteristic radiation. The LIBSspectra of all the elements were recorded at different laser energies. The emissionintensity (signal) was found proportional to the laser energy, when the laser-producedplasma was in the optically thin region. Initially this increase was the result of moreablation from the sample. The plasma temperature remains highest near the criticaldensity surface, where continuum emission is dominant. The atomic emission from theplasma decreased as the laser energy was increased to still higher values, which maybe either due to a decrease in the coupling of laser energy to the plasma as a result ofshielding by critical density, or due to the generation of instability in the plasma. Theoptimized laser-pulse energy for the jet and bulk-liquid targets was found to lie between150 and 250 mJ.

Any change of a few millimeters in the lens-to-sample distance (LTSD) affects theintensity of the atomic lines from the trace elements [35]. Therefore, keeping the LTSDconstant during the measurements was very important for accuracy and precision of thesystem. The LTSD was more critical in the case of liquid-jet experiments due to thesmaller surface area. It was noted that a shorter focal length lens produced a small beamwaist (tight focusing) and, therefore, a stronger breakdown. A smaller depth of focus,made it more sensitive to any change in the LTSD. It was also noticed that the movementof the target location by 1 mm away from the focal distance of the lens caused the LIBSsignal to drop by ∼25%. To improve the LIBS precision with a liquid-jet system, alonger focal length lens was preferred in order to increase the depth of focus.

3.4.2. Effects of gate delay

The lifetime of the laser-induced breakdown plasma plume has been found to be abouttwo times shorter in liquid than in air [1]. Since water has high, ionization potential(12.6 eV) and relatively high electro-negativity �−0�9 eV�, it produces fewer chargedparticles during laser-induced breakdown [4] leading to a much weaker laser-inducedplasma in water than in air. In our experiments, the variation in the intensity of chromiumatomic lines as well as that of the background emission, with the gate delay indicatedthat the continuum background emission was dominant in the first several microsec-onds but decayed much faster than the atomic line signal. The background emission,(Bremsstrahlung) is mainly dependent on the plasma temperature which decays fasteras a result of plasma expansion. Atomic line emission dominates as a result of radiativerecombination of the charged particles in plasma, which becomes prominent only at




lower temperatures after expansion of plasma. Since the background continuum andatomic emission decay at different rates, it was possible to obtain an optimum LIBSsignal with a properly selected detection window. In the liquid-jet target with Mn, theS/B ratio reached its peak between 5 and 10 �s duration while background decayed toits lowest value. The bulk-liquid data also indicated a similar trend. It was found thatLIBS data with an optimal signal-to-background ratio could be obtained by adjustingthe gate delay time. The LIBS spectra of magnesium (Mg), recorded at different delaytimes in bulk-liquid and liquid-jet targets were compared [35] and it was found that thebackground emission decreased and line emission increased in the case of the liquid-jet experiment in comparison to the bulk liquid experiment. Ultimately, the liquid-jetexperiment provided a better S/B ratio in comparison to the bulk experiment. A similartrend was found for the decay of background and line emission intensity in the case ofthe chromium-seeded liquid-jet experiment also.

3.4.3. Analytical measurements

For quantitative measurements, the recorded emission intensities should be related to theabsolute or relative elemental concentration in liquid. To obtain best sensitivity, LIBSsignals were optimized for different atomic and ionic lines by adjusting the gate delaytime and gate width of the detector, as well as the laser energy. The LIBS signals ofvarious trace elements (Cr, Mg, Mn, and Re) were recorded for different concentrationsto obtain a calibration curve under optimized experimental conditions for estimatingthe limit of detection. The linear calibration curves for rhenium (Re) obtained fromliquid-jet measurements [35] at a delay time of 8 �s and a gate width of 15 �s at twodifferent laser energies showed that an increase in excitation laser energy increases theLIBS sensitivity for each concentration. The detection limits for Cr, Mg, Mn, and Rewere calculated based on the calibration curves and were reported as 0.4, 0.1, 0.87 and10 �g/ml respectively [35]. The LOD of elements Pb, Si, Ca, Na, Zn, Sn, Al, Cu, Ni,Fe, Mg, and Cr obtained from bulk water and oil matrices (Table 1) has been reported

Table 1. Limit of detection of elements obtained from bulk liquidexperiments

Element Wavelength (nm) Detection Limit inwater (ppm)

Detection limit inoil (ppm)

Pb 405�87 100 90Si 288�15 25 20Ca 393�36 0�3 0�3Na 588�99 0�5 0�7Zn 334�50 120 130Sn 283�99 100 80Al 309�27 10 10Cu 324�75 7 5Ni 341�47 20 35Fe 371�99 35 20Mg 285�21 1 1Cr 425�43 10 20




Table 2. Limit of detection of elements recorded in liquid-jet experiment

Elements Wavelength (nm) Limit of detection(ppm){Ref# }

Limit of detection

Al 396�15 18Ca 422�67 0�6Cr 520�45 200

425�43 0�4Cu 324�75 5K 766�49 4Li 670�77 0�009Mg 285�21 3 0�1Mn 403�08 10 0�87Na 588�99 0�08Pb 405�78 40Tc 429�71 25U 409�02 450Re 346�05 10 10

in the literature [33]. Similarly, the LOD of elements Al, Ca, Cr, Cu, K. Li, Mg, Mn,Na, Tc, and U obtained using the liquid-jet system by Samek et al. [4] is reported inTable 2. The LOD of Cr, Mn, Mg, and Re reported by Rai et al. [35] were found to bebetter in comparison to the results described in the literature.

Various experimental parameters such as matrices, wavelength of emission, gatedelay, and the process of obtaining a calibration curve affect the estimation of thelimit of detection, which is why, an exact comparison of the LOD data from the tworesearch teams is difficult. The limit of detection reported for various elements in thisexperiment as well as in the literature has proved the LIBS technique suitable for findingpollutant trace elements at high and moderate concentrations. However, it is not possibleto detect them at very low concentrations. A serious effort is needed to make the systemversatile for very low concentration measurements. Recently, efforts were made byvarious research groups as well as by DIAL (now ICET, MSU, USA) to enhance thesensitivity of the LIBS system for different types of sample configuration. The techniqueincludes use of external magnetic field and double laser pulse excitation, which will bediscussed in the following sections.

4. ENHANCEMENT IN THE SENSITIVITY OF LIBS

4.1. Effects of a Magnetic Field on Plasma

Magnetic fields were utilized in the last decade for enhancing the analytical characteris-tics of various low-energy density plasma sources used for elemental analysis [36–46].The magnetic field ranged from a few hundred gauss to a few tens of kilogauss andmainly operated in the pulsed mode. Pulsed magnetic field was generated by discharginga capacitor through the pair of coils that produced a mirror-like structure of magnetic




field lines, which helped in confining the plasma. Various types of plasma sources suchas dc arc [36–37], low-pressure glow discharge [38–39], microwave plasma [40], sparkdischarge [41], and exploding conductor plasma [42–43] were used with these mag-netic fields. The magnetic field used to confine the plasma may also have an effect onits atomization, ionization and lifetime. The particle confinement time remained muchlonger in the presence of a magnetic field than the electron temperature decay time. Theplasma cooled mainly through the radiation losses. Thus energy given to the plasma maybe useful if its expansion is limited by applying the magnetic field. As reported previ-ously [43,47], magnetic confinement of the laser-produced plasma enhanced the emissionof radiation in the wavelength range from X-rays [47] to the visible region [43–46]. Theconfinement of a laser-produced plasma in a ∼100 kG pulsed magnetic field was founduseful in increasing the gain of the medium for the X-ray laser [48–49]. In anotherexperiment, the characterization of laser plasma with ∼80 kG pulsed magnetic fieldenhanced the visible emission and broadened the line spectra. Enhancement in the UVand visible emissions from beryllium plasma in the presence of a magnetic field has alsobeen reported. Instead of a pulsed high-intensity magnetic field, the application of a low(0.6 T) but steady magnetic field in laser-produced plasma caused a two to three-timesenhancement in X-ray emission [47]. However, the above magnetic field confinementexperiments were performed mainly with solid targets. The complication involved ingeneration of a high-intensity pulsed magnetic field as well as its synchronization withthe laser and the detection (measurement) system made the whole system difficult aswell as very critical. The effects of a low and steady magnetic field on the opticalemission characteristics of a laser-produced plasma from the trace elements present inthe liquid solution were studied systematically.

4.1.1. Emission from magnetically confined plasma

The emission from the laser-produced plasma under the effect of magnetic confinementcan be better understood by a simple analysis reported by Rai et al. [50]. It is well knownthat various types of radiations are emitted from plasmas, the nature of which dependsmainly on the density, temperature, and opacity of the plasma [1,51]. If the plasmais optically thick and has a high temperature, there will be a continuum black bodyradiation from it, whereas optically thin high-temperature plasma emits Bremsstrahlungradiation due to electron-ion collision (free-free transition), which also provides con-tinuum spectra. When a free electron recombines with the ion at a comparatively lowtemperature, it provides combination of continuum and line emission spectra. There willbe a three-body recombination process between electrons and ions resulting in the lineemission. The line emission can occur either from excited ions or atoms present in theplasma. The emission from the plasma ranges from X-rays to the visible spectral region,depending on the plasma parameters, which decide the dominant process of its emission.Emission can vary from Bremsstrahlung to line emission as the plasma temperatureand density decrease. Laser-induced plasma in the present experiment yielded all typesof emission, ranging from continuum (background) to line emission (signal) becausespatially integrated emission from the plasma plume was recorded in the direction oppo-site the laser propagation. Generally LIBS plasma was created at atmospheric pressure,thus X-rays will be absorbed in the air. For the simple analysis of the plasma emissioncondition, one can consider that for the same incident laser energy, the plasma plume




expanding in the absence and in the presence of a magnetic field will have nearly thesame plasma parameters, such as plasma density, temperature, and charge state Z. In thissituation all the emissions such as Bremsstrahlung, recombination, and line emission aredependent on the electron and ion density in the plasma. Therefore, the total amount ofplasma emission will be proportional to the square of the plasma density (∝ neni wherene = ni) as well as the volume of the emitting plasma plume. The laser-produced plasmais decelerated in the presence of the magnetic field, as no charged particle can cross themagnetic field lines [52]. However, cross-field diffusion of plasma particles is possibleonly when the plasma is either collisional or turbulent. This situation prevails either ata low plasma temperature (collisional plasma) or in the presence of instability in theplasma (turbulent). For the analysis of plasma emission in the presence of a magneticfield, we have assumed v1� t1 and v2� t2 as the asymptotic expansion velocity of plasmaand emission time in the absence and the presence of the magnetic field, respectively.Generally, the plasma plume expands in a hemispherical fashion, so the extent of plasmaexpansion after its formation can be given as v1tl and v2t2 and can be considered as theradius of the hemispherical coronal plasma in the absence and presence of the magneticfield, respectively. The mass ablated from the sample during the time duration of laserirradiation ��L� can be given as

M = dm

dt�r2��L (1)

Here, dmdt

is the mass ablation rate, and r is the focal spot radius. In this case, the density ofthe plasma can be calculated as the ratio of the total mass ablated (M) and the volume ofthe hemispherical-shaped plasma plume 2

3 �v1t1�3� for the case in which the magnetic

field is not present. Similarly, the density of plasma can be obtained in the presenceof the magnetic field. Considering optical emission from the plasma proportional to itsdensity square and its volume, one can write the ratio of the plasma emission in thepresence �I2� and in the absence �I1� of the magnetic field as [47]

I2

I1

=[

v1t1

v2t2

]3

(2)

This indicates that plasma emission intensity will be inversely proportional to the cubeof the size (product of expansion velocity and the emission time) of the plasma plume.Eq. (2) has already been verified experimentally in the case of X-ray emission byrecording the two-dimensional time-integrated image of the plasma plume in the absenceas well as in the presence of the magnetic field using an X-ray pinhole camera [53].The measured plume dimension provided more than two times enhancement in theX-ray emission, which was found in agreement with the observation of ∼2–3 timesenhancement in X-ray emission measured using X-ray vacuum photodiodes. The ratioof the plasma expansion velocities, explaining the plasma deceleration in the magneticfield, can be expressed in terms of the plasma � [54] as

v2

v1

=[

1− 1�

] 12

(3)




where � = 8nkTe/B2 is defined as the ratio of plasma pressure �nkTe� and mag-netic pressure �B2/8�. Finally, Eq. 2 for enhancement in the plasma emission can bewritten as

I2

I1

=(

1− 1�

)− 32(

t1

t2

)3

(4)

Eq. (4) indicates that the enhancement in plasma emission in the presence of the magneticfield is correlated with deceleration in plasma expansion velocity. Finally enhancementin the emission was mainly found dependent on the plasma �, which is the functionof plasma density and temperature. It was dependent on the ratio of the emission timeduration as well. v1 and v2 will be nearly equal (Eq. 2 & 3) for higher values of ,when the plasma is hot. The expansion velocity of the plasma �v2� in the presence of themagnetic field decreases (plasma confines) as the value of � goes down due to a decreasein plasma density and temperature after breakdown. This clearly indicates that plasmaconfinement will be effective only when the plasma � is low; that is, either the plasmatemperature and the density are low or the intensity of the magnetic field becomes high.Similarly, Eq. (4) indicates that no enhancement is possible if the plasma is hot and hasa high value of �. The variation of v2/vl and I2/I1 with plasma is shown in Figure 1,which clearly indicates the role of � in the enhancement of emission from plasma aswell as on change in the expansion velocity of plasma. I1 and I2 remains same for highervalues of �, which is possible in the case of Bremsstrahlung radiation only, which isdominant particularly in the high temperature plasma regime. Ultimately, enhancementseems possible mainly during the low � �<5� plasma when the plasma temperature aswell as its density decays to a lower value. The rate of the recombination of electronsand ions increases at low plasma temperatures due to the expansion and comparativelyhigh plasma density as a result of magnetic confinement. Probably an increase in the rateof recombination of electrons and ions increases the number of excited atoms playinga dominant role in enhancing the emission intensity. This analytical result is comparedwith the experimental observation in the following section.

0.00

0.50

1.00

1.50

2.00

2.50

3.00

0246810

Plasma beta

V2/

V1

and

I2/I1

V2/V1I2/I1

(a)

(b)

Fig. 1. Variation in the ratio of emission intensity �I2/I1� from plasma and the ratio of plasmaexpansion velocity �v2/v1� relative to changes in plasma .




4.1.2. Effects of magnetic fields on liquid-jet experiments

The optical emission from the laser-produced plasma of a liquid target having manganeseas a trace element was recorded in the absence and presence of a magnetic field [50].Plasma emission was collected in the direction opposite the laser propagation, whichprovided a spatially integrated intensity from the plasma plume. The characteristic lineemissions from the excited atoms and ions were observed only after the cooling downof the plasma during its expansion. This is the reason that all the line spectra frommanganese plasma (unless otherwise specified) were recorded after a 10 �s time delayfrom the laser pulse so as to prevent high-background emission from the hot plasmadue to Bremsstrahlung and black body radiation. The LIBS spectra of manganese (Mn)recorded in the absence and presence of the magnetic field had three strong peaks atthe wavelengths of 403.07, 403.30, and 403.44 nm, which were assigned as atomic lineemissions from the Mn atom. A similar spectrum was observed in the presence of themagnetic field but with an increase in the intensity of line emission by one and halftimes [50]. No other change was noted in the spectra in the presence of the magneticfield. However a different feature was noted when the experiment was performed atthe laser energy of 280 mJ in the presence of a magnetic field in comparison to thelow-energy (140 mJ) excitation. Both the background and the line emission intensitydecreased in the presence of the magnetic field at laser energy of 280 mJ. Figure 2 showsthe variation of the strongest line emission intensity with the laser excitation energy inthe absence and the presence of the magnetic field. It was found that the intensity of allthe three lines (only the intensity of the strongest line has been plotted) increases withlaser energy and follows a power law variation with a slope of 1.85 in the absence of amagnetic field. Similar power law variations having slopes ranging from 1.5 to 2.5 in thecase of x-ray emission from the laser-produced plasma has been documented [55–56].There is an indication of signal saturation towards the higher energy side in the presentexperiment, whereas no saturation was observed during the experiment performed in avacuum [47].

0.0E+000 50 100 150 200 250 300 350

2.0E+04

4.0E+04

6.0E+04

8.0E+04

1.0E+05

1.2E+05

1.4E+05

1.6E+05

1.8E+05

2.0E+05

Laser energy (mJ)

Inte

nsity

(ar

ea)

a

b

Fig. 2. Variation in the emission intensity of Mn �� = 403�07 nm� in a liquid sample relative tochanges in laser energy for a gate delay and a gate width of 10 �s (a) in the absence �B = 0 kG�and (b) in the presence �B = 5 kG� of a magnetic field.




It is likely that the presence of air at atmospheric pressure around the plasmaconfines the plasma and increases its effective density, which may be enhancing theself-absorption of emission from the plasma, leading to a saturation in the signal [57].Generation of instability in the plasma toward higher laser energy may also be possible.The presence of the magnetic field (Figure 2) enhances the LIBS intensity below 140-mJenergy, whereas it decreases toward the higher energy side. In fact the presence of themagnetic field clearly produced two slopes. Initially, during lower laser intensity, theslope was high, whereas it decreased toward the higher laser intensity side. The decreasein the slope or the presence of saturation in the plasma emission in the presence ofthe magnetic field indicated the loss of plasma energy. This probably results from theopening of a new channel of loss in the content of plasma energy. The generation ofinstability and high-energy particles in the plasma, as well as self-absorption of theemission by the plasma, may be the process responsible for the loss of plasma energy.

The temporal evolution of atomic emission at wavelength 403.07 nm was recorded inthe absence as well as in the presence of magnetic fields and is presented in Figure 3.The emission spectra were recorded at different gate delays, ranging from 5 to 45 �s.It was noted that emission intensity was high at a low gate delay, whereas it decayedexponentially with an increase in gate delay. Similar variations were noted, whetherspectra were recorded in the absence or in the presence of a magnetic field, although therewas an increase in the signal in the presence of a magnetic field [58]. This increase wasmaximal near the gate delay of 5 �s, whereas it showed a decrease below or above 5 �s.The emission from the plasma was dominated by Bremsstrahlung (continuum) spectradue to a high plasma temperature below 5 �s gate delay and the line emission was eitherfeeble or not observable. However, enhancement in plasma emission in the presence ofa magnetic field decreased quickly towards higher gate delay probably due to a decreasein the number of emitting atoms as a result of the diffusion process. It seems that the rateof recombination of electrons and ions increased as a result of an increase in effectiveplasma density due to magnetic confinement and a decrease in plasma temperature due

0.0E+000 10 20 30 40 50

2.0E+04

4.0E+04

6.0E+04

8.0E+04

1.0E+05

1.2E+05

1.4E+05

1.6E+05

1.8E+05

Gate delay (micro sec.)

Inte

nsity

(ar

ea)

abc

(a)

(c) (b)

Fig. 3. Variation in the emission intensity of Mn �� = 403�07nm� in a liquid sample relative tochanges in gate delay for a laser energy ∼140 mJ, a gate width of 10 �s (a) no magnetic field and(b) a linear magnetic field �B = 5 kG� (c) a cusp magnetic field.




to its expansion. This seems to be the main reason behind an enhancement in emission inthe presence of a magnetic field. However, the maximum enhancement will be decidedby the balance between the rate of recombination of plasma particles and the loss ofplasma particles or the neutral atoms from the emission region as a result of variousdiffusion processes. A small or negligible enhancement in emission below the 5 �s gatedelay is probably due to the plasma’s high density and temperature, making the plasma high in the presence of the magnetic field. These experimental observations were inqualitative agreement with the results presented in Figure 1, which clearly shows thatthere will not be any enhancement toward the high side. It will be high only belowthe plasma = 5. However, Figure 1 clearly indicates that, in principle, enhancementin emission intensity can be more than double by maintaining the plasma close to 1,which seems to be possible either by increasing the value of a steady magnetic fieldor by avoiding the loss of atoms due to various diffusion processes (minimization ofinstabilities in the plasma).

4.1.3. Effect of magnetic field on H� emission

An increase in plasma density caused by magnetic confinement was verified experimen-tally [50] with a liquid sample. H� spectra were recorded from an aqueous solution ofMg near 656 nm in the absence and presence of a magnetic field, which showed broad-ening in the spectral emission. The result indicated an increase in plasma density in thepresence of a magenetic field. Electron density of 5�47×1016 cm−3 and 9�54×1016 cm−3

were inferred in the absence and presence of magnetic field with the help of Stark broad-ening measurement (Fig. 4). This increase in density by nearly a factor of 2, confirmedthat the enhancement in the emission was due to an increase in plasma density as a resultof magnetic confinement. A variation in plasma density obtained using Stark broadeningtechnique in the absence as well as in the presence of magnetic field was reported by Raiet al. [50]. Plasma density was found changing from 1019 to below 1017 particles/cm3

with a change in gate delay from 0.5 to 10 �s. No significant change in the plasma

0

0.2

0.4

0.6

0.8

1

1.2

653 654 655 656 657 658 659

Wavelength (nm)

Nor

mal

ized

inte

nsity

B = 5 kGB = 0

Fig. 4. H� spectra recorded from a liquid jet with 200 mJ laser energy at a gate delay of ∼ 7 �sand a gate width of 150 ns.




density was noted for the gate delay of 2–3 �s. Plasma density increased in the presenceof magnetic field as the gate delay was increased beyond 3 �s [31]. These observationsconfirmed that the presence of the magnetic field confined the plasma and increasedthe density for gate delays higher than 3 �s. Confinement became effective only aftergate delay of 3 �s, which was in agreement with the conjecture that enhancement inemission intensity in the presence of magnetic field resulted from plasma confinement.It is possible mainly due to an increase in the rate of radiative recombination, whichis mainly dependent on the plasma density and its temperature. This observation wasfound similar as reported earlier in the case of enhancement in the X-ray emission inthe presence of a steady magnetic field. These observations were found in qualitativeagreement with the model discussed in Section 4.1.1.

The observations from this experiment have been compared with the experimentalresults reported earlier [47] to better understand the physical processes taking place inthe plasma in the presence of a magnetic field. It seems that in the lower laser energyregime the plasma expansion is nearly smooth (linear). Because the kinetic energy of theplasma is comparatively low, such that most of it is confined by the applied magneticfield (B = 5 KG). Usually, confinement of the plasma increases the effective number ofplasma particles in the confinement region. This is due to decreased plasma expansionvelocity as a result of its deceleration in the presence of the magnetic field. Initially,plasma has a high temperature, high density, and, as a result, high plasma pressure(nkTe) during or just after the laser pulse. The magnetic field applied in this initialtime regime is not sufficient to confine such high-energy-density plasma. The plasmacools after expanding for some time and its density decreases, resulting in a low plasmapressure. It is clear that the change in plasma expansion velocity is mainly dependenton of the plasma (Fig.1). In fact, the plasma expansion velocity may be zero in thepresence of the magnetic field, when it is completely stopped at a certain spatial location,where the plasma pressure (kinetic energy) and magnetic pressure (energy) becomeequal � = 1�. But in reality, plasma cannot be completely stopped by the magneticfield, because of its finite resistivity at low plasma temperature (collisional plasma) [59].Perfect confinement of plasma is possible only when the plasma is fully conducting,which is impossible for the present experimental condition. Finally, a part of the plasmawill escape from the confinement zone either due to cross-field diffusion as a result ofincreased collision at low plasma temperature, or due to the generation of instability inthe plasma. The presence of instability in the plasma also enhances the particle diffusionout of it. The decrease in the slope (LIBS intensity) in the presence of a magnetic fieldtowards the higher laser energy side indicates that either the laser energy absorbed inthe plasma is being lost through some channel other than radiation, or the increasedlaser energy has not been absorbed at all. In the former case, part of the absorbed laserenergy may be utilized in the generation of instability in the plasma, which is expectedin the present experimental condition [58]. Another possibility of energy loss may bethe generation of high-energy particles (ions and electrons), which can escape from theplasma with a certain amount of kinetic energy. The second possibility that part of thelaser energy is either not being absorbed or being scattered from the plasma may bepossible if some parametric instability is present in the plasma, which seems to be eitherimpossible or very unlikely at such a low laser intensity. Finally, it was concluded thatthe change in the slope (decrease in the signal intensity) toward higher laser energy is




dominated by the generation of instability in the plasma and the escaping of high-energyplasma particles from it. The generation of a laser-induced shock (that is, laser-induceddetonation wave and laser-induced ablation pressure) [1] also may distort the laminarjet of the liquid sample. All these factors were found important, making their owncontributions in decreasing the signal toward higher laser intensity in the presence of amagnetic field.

4.1.4. Limit of detection in magnetic field

An enhancement in the visible emission from the plasma in the presence of a steadymagnetic field has an important implication in the trace element analysis from solidand liquid samples using laser-induced breakdown spectroscopy [1–6]. The calibrationcurves for the manganese and magnesium were obtained in the absence and the presenceof magnetic fields. The limit of detection obtained from these calibration curves ispresented in Table 3. The comparison of data in Table 3 indicated that the LOD improvedin the presence of magnetic field. It was noted that the enhancement in emission signalwas correlated with the decrease in LOD. Finally, the LOD of Mn was improved to0.83 ppm in the presence of a magnetic field in comparison to 1.74 ppm in the absenceof a magnetic field [31,60], whereas it was 0.43 and 0.23 ppm for Mg in the absence andpresence of magnetic field respectively. Enhancement in the line emission observed inthe case of manganese and magnesium in the presence of a magnetic field was comparedfor other trace elements such as Cr and Ti also in the liquid solution. Emission from allthe elements showed enhancement in intensity in the presence of a magnetic field witha similar behavior as reported for manganese. The enhancement factor for Mn, Cr, Mgand Ti is presented in Table 4. An increase in emission was found for all these elements(Cr, Mg and Ti), even at 280-mJ energy, for which manganese emission showed adecrease. However, the saturation or decrease in the emission signal for these elementshas been noted at still higher laser energies. This indicates that the onset of saturationmay vary from element to element and is dependent on emission characteristics as wellas on elemental concentration.

Table 3. Limit of detection obtained for Mg and Mn lines forlaser energy of 140 mJ in the absence and presence of themagnetic field

Elements Wavelength (nm) LOD at 140 mJ

B = 0 kG B = 5 kG

Mga 279�55 0�43 0�23280�27 1�93 1�10285�20 12�20 7�15

Mnb 403�08 1�74 0�83403�31 2�47 1�53403�45 6�23 2�21

a gate delay of 4 �s; gate width of 2 �s,b gate delay of 10 �s; gate width of 10 �s




Table 4. Variation in the enhancement of LIBS intensity for Mn, Mg, Tiand Cr in solution in the presence of the magnetic field

Emission Wavelength (nm) Enhancement Factor (G)

At 140 mJ At 280 mJ

Mn(10 ppm)a 403�08 1�74 0�75403�31 1�86 0�85403�45 1�27 0�84

Mg (10ppm)b 279�55 1�71 1�14280�27 1�66 1�25285�29 1�56 1�49

Ti(100 ppm)c 363�55 1�26 1�52364�27 1�14 1�42365�35 1�29 1�40

Cr (10 ppm)d 425�44 1�56 1�81427�48 1�50 1�82428�97 1�55 1�99

a B = 5 kG; gate delay of 10 �s and gate delay of 10 �sb B = 6 kG; gate delay of 4 �s and gate width of 2 �sc B = 6 kG; gate delay of 4 �s and gate width of 8 �sd B = 6 kG; gate delay of 4 �s and gate width of 10 �s

4.2. Effects of Double Laser Pulse Excitation on LIBS Signal

The use of multiple laser pulses for producing more intense and sustained plasmaemission is one of the techniques for improving the sensitivity of LIBS, which has beenfound productive in various experiments in improving the signal to noise ratio [61–65].During the double pulse excitation experiments, the first pulse generates a gaseouscavity inside the liquid, which is then excited using the second pulse for analysis.Uebbing et al. [61] used a double pulse excitation scheme mainly for reheating preformedplasma. The second pulse, after a certain time delay, re-excited the gas present inthe form of ions and atoms in the plasma. Two separate lasers were used during thisexperiment in perpendicular configuration, where first laser beam was incident at rightangle on the target leading to an ablation of its surface and the formation of a plasmaplume. The second laser was directed parallel to the target surface and was incidenton the previously formed plume to re-excite it. Sattaman et al. [62] used a single Nd:YAG laser, which generated a double laser pulse separated by a desired time delay.It has been shown that the volume of plasma formed from the steel target after adouble pulse burst is about twice as large as that formed with a single laser pulsehaving energy equal to the sum of both the laser pulses. Nakamura et al. [14] haveused two pulses separated by 1 �s to analyze iron suspension in water flowing from anozzle, which provided a substantial decrease in limit of detection. Stratis et al. [63]reported that signal enhancement could be attributed to an increase in sample ablation.Enhancement in the emission was also found to be dependent on the geometry of thecollection optics. L. St-Onge, et al. [64–65] used different wavelengths for both lasersand reported that the mixed wavelength approach was better for enhancing the signal.




Since the initial study, most of the double pulse experiments were performed on thesolid samples. However, double laser pulse excitation technique in the study of liquidsample came into prominence due to its application in enhancing the sensitivity of LIBSfor lower concentration of trace elements in liquid and some important applicationsin the study of molten material in industry as well as in the study of under watermaterials.

Recently Rai et al. [66] used the double pulse LIBS to study liquid samples usingjet configuration and reported that delay between the laser for optimum enhancementis ∼2–3 �s. Kumar et al. [67] studied the double pulse laser-induced breakdown spec-troscopy with liquid jet of different thicknesses and reported that thick jet of diameter1 mm provided better sensitivity than the thin jet of diameter ∼0.3 mm or the mist ofthe liquid sample. Kuwako et al. [68] reported supersensitive detection of sodium inwater using dual pulse LIBS. In optimized condition this experiment provided limit ofdetection as 0.1 ppb for sodium (Na) in water. Pearman et al. and Scaffidi et al. [69–70]used dual pulse LIBS system for bulk aqueous solution with orthogonal beam geometry.They reported more than 250 fold enhancement in the emission in comparison to singlepulse experiment. Detection limits of Ca, Cr, and Zn were reported as 41.7 ppb, 1.04 and17 ppm respectvely, better than earlier reported results in the literature. Peter et al. [28]reported the study of molten steel in the furnace. They used LIBS with multiple pulseexcitations for multi elemental analysis of liquid steel. In this system a Q-switched Nd:YAG laser was used with a modification. The Q-switching electronics was modified togenerate as many as three separated laser pulses within a single flash-lamp pulse. Finallythree equal energy (110–125 mJ) pulses separated by 25 and 42 �s were used in thisexperiment and the limits of detection for C, P, S, Ni and Cr were obtained as 5, 21,11, 9 and 9 respectively. Giacomo et al. [71–72] used double pulse LIBS to study themetallic target in the sea water. A quantitative chemical analysis of Ti, Cu, Pb, Sn andZn submerged in sea-water was presented. They found that ablated matter was stronglyconfined by the water vapour inside the cavitation bubble, which led to higher valuesof excitation temperature and held the conditions suitable for chemical analysis for alonger time than in the gaseous case.

Normally two types of experimental setups have been used for the LIBS under doublelaser pulse excitation. In the first setup single laser is used, which provides two/or moresuccessive laser pulses separated by few tens of microseconds for the excitation ofplasma. In the other setup two separate synchronized laser systems are used that allowa broad variation in laser energy and the delay between the laser pulses to optimizethe optical emission. Experimental detail is presented in the next section. The use of asingle or two-lasers for double laser-pulse excitation experiment mainly depends on therequirement of the proposed experiment and can be decided by the user.

4.2.1. Double pulse LIBS using single laser

Many researchers [28,66–72] used a Nd:YAG laser for double or multiple successivelaser pulse excitation experiments, where pockel cell trigger was controlled by an externalpulse generator in order to extract two or more pulses by the same flashing lamp.Giacomo et al. [71–72] kept the delay time of laser from the pulse triggering the lamp as145 �s whereas the delay between two laser pulses was 45 �s. The delay was optimizedto obtain two stable laser pulses with nearly same output energy (∼100 mJ). The delay




between laser pulses was variable between 45–140 �s. The laser was focused on thesample kept in a cuvette or directly in the bulk water using a glass lens. The collectionoptics was mounted along the normal to the laser beam direction, which consists of acompact telescope coupled with an optical fiber bundle connected to the spectrometer.The detection system was a gated ICCD and controlling system was synchronized withthe pockel cell trigger.

The use of single laser for generating two successive pulses for LIBS has someadvantages as well as disadvantages. A lot of optical alignments and synchronizationsare avoided in the case of a single laser. On the other hand, there is some limitationon the use of inter laser-pulse delay below ∼40 �s and independent variation in laserenergy for optimization of emission from the plasma. Using two separate lasers hasthis flexibility, but there are experimental problems associated with additional alignmentoptics and synchronization requirements.

4.2.2. Double pulse LIBS using two lasers

The schematic diagram of the experimental set-up (see Fig. 4, Chapter 5) for makingthe two laser beams collinear for recording the laser-induced breakdown emission fromthe liquid sample under double pulse excitation has been reported earlier [66–67]. Itconsisted of two Q-switched, frequency-doubled Nd:YAG lasers (Continuum SureliteIII and Quanta-Ray DCR-2A-10) that deliver energy of ∼300 mJ at 532 nm in 5-ns pulseduration. Both the lasers were operated at 10 Hz during this experiment and were focusedon the target (in the center of the liquid jet). The first laser provided a p-polarized laserbeam, whereas, the second laser beam was s- polarized. Both the lasers were madecollinear using a thin film polarizer (CVI Lasers), which transmitted the p- polarizedlight, but reflected s- polarized light. For an optimum performance of the thin filmpolarizer (TFP) it was necessary that p- and s- polarized light beams keep an angleof incidence nearly 57 degree. Both the beams became collinear after TFP, which wasfocused on the target with the help of a dichroic mirror and a spherical, ultra violet, quartzlens of 20 cm focal length. The combination of dichroic mirror and the focusing lenswas also used to collect the optical emission from the laser-induced plasma. The lasersoperations were synchronized using a programmable trigger pulse generator (StanfordResearch System Inc. Model DG 535) that made possible the arrival of both the lasersat a certain time delay. The delay between lasers could be changed from nanosecondto microsecond time range. Two UV grade quartz lenses of focal length 100 mm and50 mm were used to couple the plasma emission to an optical fiber bundle. The fiberbundle was made up of a collection of 80 single fibers of 0.01-mm core diameter, whichwas coupled to an optical spectrograph (Model HR 460, Instrument SA, Inc., Edison,NJ) and used as an entrance slit.

4.2.3. Spectral emission under double laser pulse excitation

The enhancement in the sensitivity of LIBS system using double laser pulse excitationhas its own importance due to its wide application. Rai et al [66] selected aqueoussolution of Mg, Cr and Re elements for their experiments in laboratory. These elementshave triplet line emissions in UV-VIS region of the spectrum similar as technetium




except that they are not radioactive. This study was performed to evaluate the doublepulse LIBS for making technetium monitor [73–74]. Mg solution was used for mostof the study, because the behavior of line emission from ions and neutral atoms underthe effect of double pulse excitation can be studied simultaneously [66]. The spectra ofmagnesium were recorded in the single and double pulse excitation mode. The doublepulse excitation spectrum was recorded at 4 �s gate delay with 2 �s inter-pulse separationbetween the lasers, when emission was maximum. The single pulse excitation spectrumwas recorded at 1 �s gate delay where the emission was maximum. The spectra recordedby single laser pulse showed mainly two dominant line emission due to ion at 279.55and 280.27 nm, whereas spectrum under double pulse excitation showed more than 4times (peak to peak) enhancement in the line emission intensity. Background emissionalso had a higher level in double pulse than the single pulse excitation. Either no or verysmall neutral line emission was found at 285.20 nm, because the spectrum was recordedat lower gate delay, when the plasma was hot and most of the magnesium atoms presentin the plasma plume were in the form of ions. To verify whether the enhancement inthe emission was due to an increase in the laser intensity (simple addition of intensity ofboth the lasers), emission spectra were recorded by changing the delay between both thelasers. During this experiment, gate delay and gate width were fixed at 10 �s in orderto see the neutral line emission also.

Figure 5 shows the spectra recorded in double pulse mode when the inter-pulseinterval between both the lasers was zero and 3 �s. A very small enhancement was notedin the emission intensity, when the delay between both the lasers was zero in comparisonto single pulse experiment. In fact, this was the case of addition of intensity. However,emission intensity was enhanced by nearly ∼4 times (peak to peak) when the delaybetween both the lasers were increased to 3 �s. Similar enhancement (4–10 times) inemission was reported under double pulse excitation for aluminum line emission from thesolid target [64]. It seems that the first laser pulse induced the plasma, which expanded

0.0E+00

2.0E+05

4.0E+05

6.0E+05

8.0E+05

1.0E+06

1.2E+06

1.4E+06

270 275 280 285 290

Wavelength (nm)

Inte

nsity

(A

.U.)

Double pulse(delay 0 micro sec.)

Double pulse(delay 3 micro sec.)

Mg+

(27

9.55

)

Mg+

(28

0.27

)

Mg

(285

.2)

Fig. 5. Emission spectra of Mg (5 ppm) in double pulse excitation with a change in interpulsedelay (0 and 3 �s) between lasers (Laser 1: 100 mJ; Laser 2: 120 mJ; gate delay/gate width:10 �s/10 �s).




normal to the target surface, and the second laser interacted with preformed plasma.The interaction of the second pulse with expanding plasma increased the probabilityof absorption of the laser, which probably heated the plasma and finally excited morenumbers of plasma particles to its excited state. Even the second laser added up in thevaporization (ablation) of the material. This could be the reason behind the enhancementin the emission intensity of magnesium in the presence of a delayed second laser. Itclearly indicates that the enhancement in the signal was not due to the simple additionof intensity of lasers.

A strong emission from neutral magnesium line (285.20 nm) was also observed inthis experiment, because the plasma was comparatively cool at 10 �s gate delay. Theneutral line emission also showed similar enhancement in the emission, when the inter-pulse delay was increased from zero to 3 �s. At the same time, background emissiondecreased with an increase in inter-pulse separation. It seems that the lighter elements ofthe solution matrix (H, O and OH), which contributed mainly to background emission,diffused out faster in comparison to the heavier element Mg. This is because, initially,all the species of the plasma have similar kinetic energy. Thus when the second pulseinteracts with the expanding plasma (from first laser) after 3 �s delay, a comparativelysmall amount of matrix species (H, O and OH) are present in the plasma to contributeto the background emission.

4.2.4. Effect of delay between lasers

It was shown earlier that the delay between both the lasers played an important role inthe enhancement of emission from the plasma. For this purpose, the emission from theplasma was recorded in double pulse excitation mode by changing the delay betweenboth the lasers. In this experiment, delay between both the lasers was increased bypre-triggering the first laser before the second laser, whereas, the detector gate delaywas changed with respect to second laser. Figure 6 shows the variation in emissionintensity from magnesium solution (5 ppm), with an increase in delay between the lasers

0.E+00

1.E+06

2.E+06

3.E+06

4.E+06

5.E+06

6.E+06

7.E+06

8.E+06

9.E+06

1.E+07

0 5 10 15 20 25 30

Delay between lasers (microseconds)

Inte

nsity

(A

rea)

Mg+ (279.55 nm)Mg+ (280.27 nm)Mg (285.2 nm)

Fig. 6. Variation in emission intensity from Mg (5 ppm) ions and atoms with a change in delaybetween the lasers (Laser 1: 100 mJ; Laser 2: 120 mJ; gate delay/gate width: 10 �s/10 �s).




(laser delay), under double pulse excitation mode. The inter-pulse delay was increasedfrom 0 to 25 �s. Emission intensity was recorded at the gate delay and gate width of10 �s. A higher gate delay (10 �s) allowed us to observe the behavior of neutral lineemissions also, which remained nearly absent or present with a very low intensity,when the spectrum was recorded at a lower gate delay (≤ 4 �s). The emission intensityincreased from both the magnesium ions and neutral atoms and reached a maximumbetween 2–3 �s delay between the two laser pulses (Fig. 6). Maximum enhancementin the emission is more than six-fold for emission of magnesium ion (� = 279�55 nm).Similar variation was reported in signal to noise ratio for iron suspension in liquid byNakamura, et al. [14] with a maximum enhancement of more than 2.5 times at 1-�sinter pulse interval. Any further increase in the inter-pulse delay decreased the lineemission. Initially this decrease was fast and then slow. This indicated that if one hastoo long inter-pulse interval, the pre plasma would have expanded too much to effi-ciently absorb the second pulse. Finally, a condition reached slowly towards that ofcompletely distinct laser pulses, where the plasma from the first laser pulse completelydisappeared at the arrival of the second laser pulse. However, significant emission inten-sity was noted in the case of double pulse excitation even at an inter pulse separation of15–20 �s.

The neutral magnesium line showed lower intensity than that of the ion emissionline for lower inter-pulse delay. However, neutral line emission dominated the ion lineemission as the laser delay was increased beyond 10 �s. The neutral line emissionpeaked at slightly higher inter-pulse delay of ∼4 �s in comparison to ∼2�5 �s for ionlines. The neutral line emission showed (Fig. 6) a broader profile with inter-pulse delayin comparison to the ion line emission. This clearly indicated that more than six-foldenhancement in the emission in double pulse excitation was not due to an increase inthe total effective laser intensity. In this case emission was maximized, when the plasmagenerated by the first laser pulse expanded to such an extent as to absorb the maximumenergy from the second laser pulse. The study shows that it is possible to obtain anoptimum enhancement in the line emission for an element by controlling the inter-pulseinterval, which in fact determines the size and physical properties of the pre-plasma,when maximum absorption of the second laser pulse occurs. The decrease in the emissionintensity with an increase in inter-pulse delay indicated the interaction of second laserpulse with a rarified plasma plume generated by the first laser pulse. A depleted densityof ions and neutrals in plasma plume ultimately decreased the absorbing capacity of theplume for the second laser light. The flattening in the peak (Fig. 6) observed for neutralmagnesium line emission was because of the accumulation of neutral atoms in a smallvolume of the expanding plume, which is not possible for ions to sustain for a longertime in the plume during its expansion. The dominance of neutral emission on the ionline emission beyond a 10-�s delay showed that less ions were left in the plume of thefirst laser induced plasma after its expansion for 10 �s and only magnesium atoms weregetting excited by the second laser and providing line emission. However, in this timerange emission intensity was less due to the loss of numerous atoms because of diffusion.The interaction of the second laser with the preformed expanded plasma plume showedthat the volume of the emitting plasma was enhanced. The optimum size of the plume(from first laser) for maximum emission intensity after interaction with second laser wasalso determined by the plasma temperature (expansion velocity of the plasma), whichwas dependent on the energy of the first laser interacting with the target.




4.2.5. Effect of gate delay from first laser

The temporal evolution of plasma emission was recorded in double pulse excitation modeby changing the gate delay from the first laser, which enabled us to see the variation inthe emission intensity after the first, as well as, the second laser pulse. The delay betweenboth the lasers was kept ∼2 �s, whereas, gate delay was varied from 1 to 10 �s withrespect to the first laser pulse during this experiment. The gate width was set to ∼0.1 �sso as to get a better time resolution. At 1 �s gate delay the emission intensity was smallbecause emission from only first laser-induced plasma was contributing to detector, forthe inter-pulse delay of 2 �s. The effect of second laser on the plasma emission wasnoted when the gate delay was changed to 2 �s. This is the time when the second laserinteracted with the plasma formed by the first laser. The background emission increasedin such a way that detector was saturated and no data point was recorded at 2 �s gatedelay (Fig. 7). This point is shown by an arrow in Figure 7. An enhancement in thebackground emission was the indication of an increase in plasma temperature, when thesecond laser pulse interacted with the preformed plasma. A small increase �<10%� inplasma temperature has been reported for the case of solid aluminum sample [64–65].Further increase in the gate delay showed fast decay in the background emission up to4 �s followed by a slow decay. The fast decay of background emission indicated that theplasma heated by second laser pulse cooled down due to its expansion, which resulted inthe dominance of line emission from magnesium ions and the neutrals. Finally, the lineemission reached its maximum around 4 �s gate delay, when the background emissiondecayed to a much lower value. The line emission intensity from both the ions andneutrals of magnesium, once they reached maximum, started decaying with an increasein the gate delay. Since the plasma present at 1 �s gate delay was generated only from thefirst laser, the increase in background emission between 2 to 3-�s gate delay indicatedthat the second laser pulse heated up or excited the plasma to an extent such that onlybackground emission was observed. However, the background emission decayed fastbecause of plasma expansion. The relatively colder plasma helped in increasing the

0.E+00

1.E+06

2.E+06

3.E+06

4.E+06

5.E+06

6.E+06

7.E+06

0 2 4 6 8 10 12

Gate delay (micro sec.)

Inte

nsity

(A

rea)

A

B

CD

A - Mg+ 279.55 nmB - Mg+ 280.27 nmC - Mg 285.20 nmD - Background

Fig. 7. Variation in signal and background emission from Mg (5 ppm) with detector gate delay fromthe first laser (Laser 1: 130 mJ; Laser 2: 100 mJ; Delay between lasers/gate width: 2 �s/0�1 �s).




rate of recombination of the electrons and ions present in the plasma, which ultimatelyincreased the intensity of line emission to a maximum near 4-�s delay. The decayin the line emission with a further increase in gate delay started because of loss ofplasma particles (magnesium ions/atoms) out of imaging range due to diffusion. Thedecrease in plasma temperature also reduced the number of excited atoms in the plasma.This study clearly indicated that there is an optimum gate delay in the case of doublepulse excitation, which produced maximum line emission for a fixed inter-pulse delaybetween the lasers. The occurrence of maximum emission was decided mainly by thedelay between the two lasers, the plasma temperature and the dynamics of plasma. Itwas noted that as the delay of second laser increased with respect to the first laser(gate delay is fixed with respect to first laser) the occurrence of maximum line emissionshifted towards higher gate delay. However, the emission intensity decreased, as thedelay between both the lasers increased, because plasma density decreased fast withan increase in the gate delay. This clearly indicated that enhancement in the emissionin this experiment was mainly due to the excitation of more ions/atoms. However asignificant amount of material ablation due to the second laser pulse also may haveplayed important role in the enhancement of intensity. All the factors had a role inincreasing the signal under the double pulse excitation. A comparison of the variationin emission intensity from Mg 279.55 nm line showed that enhancement in the emissionintensity in the double pulse excitation was ∼20 times than the emission in single pulseexcitation at 4 �s gate delay.

Similar variation in the LIBS of chromium was also noted when excited using thedual laser pulses. It was noted that the emission from chromium decayed slowly andlasted for a longer time of 40 �s in both type of excitation (single or double pulse). Themaximum enhancement in the emission from the chromium was found around 15 �s gatedelay. Emission intensity towards lower gate delay was dominated by the backgroundas well as noise. However, signal to noise ratio was found to be better in double pulserather than the single pulse excitation. The different decay constants and total emissiontimes for Mg and Cr indicated different values of emission transition probabilities forthese elements. Magnesium ion line emission decayed fast, whereas, neutral Mg atomicemission decayed slowly and lasted for a longer time (Figs. 6 and 7). Similarly, emissionat the neutral chromium line decayed very slowly signifying its presence in the plasmaplume over an extended time.

4.2.6. Effect of laser energy

Two different laser pulses were used for double pulse excitation experiment, whichneeded an optimization of its energy for a maximum line emission from the plasma.Variation in the emission intensity from plasma under double pulse excitation for variousenergies of the first laser was investigated. During this experiment energy of the firstlaser was varied up to 180 mJ whereas the energy of second laser was kept constant at120 mJ. The emission was recorded at a gate delay and gate width of 10 �s. Figure 8shows the variation in emission intensity with change in inter-pulse delay between boththe lasers recorded for three different energy of the first laser. In this experiment firstlaser was pre triggered with respect to second laser and the gate delay was set fromsecond laser. Maximum emission intensity was obtained for ∼100-mJ energy of the firstlaser. However, emission decreased as the energy of the first laser was either decreased




0.E+00

1.E+06

2.E+06

3.E+06

4.E+06

5.E+06

6.E+06

7.E+06

8.E+06

9.E+06

0 5 10 15 20 25 30

Delay between lasers (micro sec.)

Inte

nsity

(A

rea)

L1 = 100 mJL1 = 140 mJL1 = 180 mJ

Fig. 8. Variation in emission from Mg (279.55 nm) with inter-pulse delay between the lasers fordifferent energies of Laser 1 (Laser 2: 120 mJ; gate delay/gate width: 10 �s/10 �s).

or increased from 100 mJ. The emission intensity was highest between 1.5–3.5 �s laserdelay for the 100-mJ energy but it shifted towards higher delay time with an increase inlaser energy. The lower energy from the first laser is expected to cause smaller emissionprobably due to comparatively less material ablation which may increase for higherenergy. The variations of intensity in the three curves of Fig. 8 indicate that along withan increase in the size of the plasma plume, the loss in the plasma particles was alsofast, when energy of the first laser was increased (>100 mJ). An increase in the size ofthe plasma plume and quicker loss of the plasma particles may cause the broadeningin the peak and a decrease in the signal, respectively, when the laser pulse-energy wasincreased to 140 and 180 mJ. The development of a small peak at 0.5 �s delay for140 and 180 mJ laser energy could not be understood. However, it seems that out ofmore ablated material at higher laser energy a particular number (type) of particleswere excited by the second laser pulse 0.5 �s after the first pulse. These particles couldhave been the cluster of particles, heavier than the other normal plasma particles, whichexpanded slowly.

The energy of second laser pulse was also equally important for the double pulseexcitation experiment. It was observed that the emission intensity in double pulse excita-tion mode increased as the second laser energy increased up to 120 mJ. Emission startedsaturating (and then decreasing) for further increase in the second laser pulse energy.The enhancement in the emission with an increase in the energy of second laser couldbe due to better absorption of laser in the pre-plasma but the energies higher than 120mJ could have created either self-absorption or generated instability in the plasma [1],thus decreasing the emission intensity. A higher energy for the second laser (∼140 mJ)was required for maximum emission when the spectrum was recorded at decreased gatedelay and gate width (5 �s/0.1 �s). It was concluded from this study that the laser energyrequired for an optimum emission in the double pulse excitation was also dependent onthe gate delay and gate width of the detector such that larger gate delay and gate widthinduced saturation at lower intensity of the second laser.




For explaining the above experimental observations, it is necessary to understandthe absorption of laser light in the pre-formed coronal plasma. The possible absorptionmechanism in the plasma plume is inverse Bremsstrahlung absorption [75], which takesplace as a result of electron-ion collision in low temperature plasma. When the plasmaelectrons are subjected to momentum changing collisions as they oscillate back and forthin the laser electric field, the laser light wave feels an effective damping. In this caseabsorption coefficient �kib� of plasma can be written as

kib ∝ Zn2e

T3/2e

(1− ne

nc

)1/2 (5)

Eq. (5) clearly shows that inverse Bremsstrahlung absorption is the strongest for lowplasma temperature �Te�, high density �ne� and high Z plasma. Here nc is the criticalplasma density. This qualitatively explains our observation: when the delay between thetwo lasers is very small, temperature of pre-formed plasma remains high giving rise tosmall probability of absorption of second laser-pulse by ions and atoms of the plasmaplume. In addition, the high temperature plasma emits Bremsstrahlung continuum, whichis observed experimentally in the form of background emission. Plasma temperaturedecreases as it expands away from the ablation surface resulting an increase in probabilityof absorption of second laser-pulse and corresponding increase in line emission. Nakanoet al. [76] have calculated the absorption of laser light in expanding pre-formed plasmaby solving the Helmholtz wave equation with a density gradient profile n (x, t) [77]given as the Reimann solution of the hydrodynamic equation [78] so that

n�x� t� = n0

[34

− x(4vexpt

)

]3

(6)

Here n0 is the solid or liquid state density and the plasma scale length L = vexp�t. Theircalculation shows that absorption of second laser starts as the plasma scale length L�vexp�t� exceeds the wavelength of laser light �, that is L ≥ �. In our experimentalcondition Te ∼1 eV corresponds to a plasma expansion velocity vexp = 5�5×106 cm/s.The time delay between two laser pulses for maximum emission was ∼2 �s. The scalelength of the plasma is obtained as L∼ vexp�t = 1�10 cm, which is very much largerthan the laser wavelength �∼0�53 �m. This indicated that in this case an efficientabsorption of second laser was possible in the pre-formed plasma produced by thefirst laser. However, any further increase in delay between the lasers will decrease theemission intensity due to drastic decrease in plasma density as a result of combined effectof electron ion recombination and plasma diffusion. According to Eq. (5) absorptioncoefficient is directly proportional to n2, which indicates that any decrease in plasmadensity will lead to a large decrease in the probability of laser absorption in the plasmaresulting in less emission even under the double laser pulse excitation. Finally, in thecase of very large delay between lasers, interaction (absorption) of second laser withpre-formed plasma would be either negligible or nonexistent and the plasma emissionwill be observed as produced by two separate lasers. The qualitative agreement ofthis analysis with our observations confirms that the main reason for enhancement inthe emission under double laser pulse excitation is due to better absorption of second




laser in the pre-formed plasma produced by first laser, where optimum absorption(and the resulting emission) occurs when the plasma scale length is L∼1�10 cm >>�∼0�53 �m. Further analysis of laser-plasma interaction considering all aspect of plasmaformation and its dynamics will provide a quantitative understanding of the experimentalresults.

4.2.7. Analytical measurements with double laser pulses

The variation in the emission intensity with concentration was recorded for Mg, Cr andRe solutions in the single and double laser pulse excitation in order to find the limitof detection (LOD) of these elements [66]. The calibration curve for magnesium ionemission (279.55 nm) was obtained using single and double laser pulse excitation. Theemission intensity from ions showed a linear variation in the concentration range of0.1 to 5 ppm in single pulse excitation mode, whereas increase of emission intensitywith concentration was nonlinear in the double laser pulse excitation mode. Two slopeswere observed in double pulse excitation mode,the first slope covered the 0 to 1 ppmconcentration range whereas the second slope (1–5 ppm) seemed to be due to saturationof emission as a result of self-absorption. The limit of detection was defined here asthe ratio of three times standard deviation with the slope of calibration curve. Thelimit of detection was calculated as 69 ppb in double laser pulse excitation in 0–1 ppmconcentration range, whereas, it was 230 ppb for the single pulse excitation mode. Thecalibration curve for neutral magnesium emission was recorded under single and doublepulse excitation, which showed only one slope between 0.1 to 5 ppm concentration rangefor both excitation modes. The limit of detection for neutral emission was estimated as370 ppb in double pulse in comparison to 970 ppb in the single pulse excitation. Thisshows that limit of detection obtained for magnesium ion as well as for the neutral atomimproved (decreased) in double laser pulse excitation mode. The limit of detection forchromium was also obtained as 120 ppb in double pulse mode in comparison to 1300 ppbin single pulse excitation mode. This shows that the double laser pulse excitation canimprove the limit of detection of Cr by an order of magnitude. Similar observationswere noted in the case of Re. Table 5 shows the LOD obtained for Mg, Cr and Re usingdifferent spectral lines.

Table 5. Comparison of Limit of Detection (LOD) for different elementsin single and double laser pulse excitation

Elements Wavelength (nm) Limit of Detection (LOD) (ppm)

Single Pulse Double Pulse

Mg 279�55 0�23 0�06285�20 0�97 0�57

Cr 425�44 1�30 0�12427�48 2�16 0�16428�97 1�84 0�18

Re 346�04 22�21 8�55346�47 14�42 8�85




5. CONCLUSION

In summary, it has been demonstrated that LIBS is a useful technique for the analysisof trace element present in the liquid matrices. Different techniques were used for theplasma formation depending on the type of the sample. Excitation of plasma on the liquidsurface in the bulk as well as in the molten state of metal was discussed in the lightof their different requirements. A simple comparison indicated that normally liquid jetsystem could be useful for most of the samples present in the liquid form. Optical fiberprobe was found to be more suitable for the measurements of molten metal and thesamples in bulk of liquid (under sea).

It has also been realized that inspite of its many advantages over the conventionaltechniques, LIBS is still lacking the sensitivity in the measurement of very low concen-tration of trace elements in samples. It was demonstrated that application of externalmagnetic field and double laser pulse excitation can be used for increasing the sensitivityof LIBS by a factor of two and six respectively. Analytical measurements also confirm asignificant change in limit of detection. It has been found that confinement of plasma inthe presence of magnetic field was the main reason for an increase in intensity of emis-sion from plasma. Our analysis shows that enhancement in intensity can be increasedeven more by keeping plasma close to one.During double laser pulse excitation, itwas found that the first pulse created an expanding plasma, which absorbed second laserpulse more efficiently and excited more number of plasma particles. A simple analysisshows that for optimum increase in the plasma emission the plasma scale length mustbe larger than the laser wavelength.

ACKNOWLEDGMENT

This work was supported by Savannah River Technology Center through Education,Research & Development Association of Georgia Universities, grant no. GA0046 andDepartment of Energy contract no. DE-FG02-93CH-10575.

REFERENCES

[1] L. J. Radziemski and D. A. Cremers, Laser Induced Plasma and Application Marcel Dekker,New York (1989).

[2] T. L. Thiem, Y. I. Lee and J. Sneddon, Appl. Spectrosc. Rev. 32 (1997) 183.[3] K. Song, Y. I. Lee and J. Sneddon, Microchem. J. 45 (1992) 1.[4] O. Samek, D. C. S. Beddows, J. Kaiser, S. V. Kukhlevsky, M. Liska, H. H. Telle and

J. Young, Opt. Engg. 38 (2000) 2248.[5] F. Y. Yueh, J. P. Singh and H. Zhang, Encyclopedia of Analytical Chemistry, John Wiley

& Sons Ltd. 3 (2000) 2066.[6] A. K. Rai, V. N. Rai, F. Y. Yueh and J. P. Singh, Trends in Appl. Spectrosc. 4 (2002) 165.[7] D. A. Cremers, L. J. Radziemski and T. R. Loree, Appl. Spectrosc. 38 (1984) 721.[8] R. Knopp, F. J. Scherbaum and J. I. Kim, J. Anal. Chem. 355 (1996) 16.[9] C. Haisch, J. Lierman, U. Panne, R. Niessner, Anal. Chim. Acta. 346 (1997) 23.

[10] R. Nyga, W. Neu, Opt. Lett. 18 (1993) 747.[11] A. E. Pichahchy, D. A. Cremers, M. J. Ferris, Spectrochim Acta B 52 (1997) 25.




[12] C. Arca, A. Ciucci, V. Palleschi, S. Rastelli and E. Tognoni, Appl. Spectrosc. 51 (1997) 1102.[13] L. M. Berman, P. J. Wolf, Appl. Spectrosc. 52 (1998) 438.[14] S. Nakamura, Y. Ito, K. Sone, H. Hiraga, K. I. Kanedo, Anal. Chem. 68 (1996) 2961.[15] V. V. Golovlyov and V. S. Letokhov Appl. Phys. B 57 (1993) 417.[16] R. O. Esenaliev, A. A. Karabutov, N. B. Podymova and V. S. Letokhov Appl. Phys.

B59 (1995) 73.[17] A. A. Oraevsky, S. L. Jacques and F. K. Tittel, J. Appl. Phys. 78 (1995) 1281.[18] J. R. Watcher and D. A. Cremers Appl. Spectrosc. 41 (1987) 1042.[19] Y. Ito, O. Ueki and S. Nakamura, Anal. Chim. Acta 299 (1995) 401.[20] C. W. Ng, W. F. Ho and N. H. Cheung Appl. Spectrosc. 51 (1997) 976.[21] F. W. Ho, C. W. Ng and N. H. Cheung Appl. Spectrosc. 51 (1997) 87.[22] A. K. Rai, F. Y. Yueh and J. P. Singh, Rev. Sci. Instrum. 73 (2000) 3589.[23] A. K. Rai, F. Y. Yueh and J. P. Singh, Appl. Opt. 42 (2003) 2078.[24] C. Aragon, J. A. Aguilera, J. Campos, Appl. Spectrosc. 47 (1993) 606.[25] L. Pasky, B. Nemet, A. Lengyel and L. Kozma, Spectrochim. Acta B 51 (1996) 27.[26] J. Gruber, J. Heitz, H. Strasser and N. Ramasedar, Spectrochim. Acta B 56 (1981) 685.[27] J. Gruber, J. Heitz, N. Arnold, N. Ramsedar, W. Meyer, F. Koch, Appl. Spectrosc. 58

(2004) 457.[28] L. Peter, V. Sturn and R. Noll, Appl. Opt. 42 (2003) 6199.[29] M. Sabsabi, R. Heon and J. Lucas US Pat. No. 700660 Issued Sept. (2003).[30] E. Baril, L. St. Onge, M. Sabsabi and J. Lucas (Galvatech 2004) Chicago (IL) April 4–7

(2004). Conf. Proceeding (Association for Iron & Steel Technology 2004) pp 1095.[31] V. N. Rai, H. Zhang, F. Y. Yueh and J. P. Singh, Appl. Opt. 42 (2003) 3662.[32] D. C. S. Beddows, O. Samek, M. Liska and H. Telle, Spectrochim. Acta B 57 (2002) 1461.[33] P. Fichet, P. Mauchien, J. F. Wagner and C. Maulin, Anal. Chemica 429 (2001) 269.[34] L. St. Onge, E. Kwong, M. Sabsabi and E. B. Vadas, J. Pharm. And Biomed. Anal. 36

(2004) 277.[35] F. Y. Yueh, V. N. Rai, J. P. Singh and H. Zhang, AIAA-2001-2933, 32nd AIAA Plasmady-

namics and Laser Conference, 11–14 June (2001) Anaheim, CA.[36] A. A. Fakhry, M. A. Eid and M. S. Hasem Appl. Spectrosc. 32 (1978) 272.[37] K. Trivedi, D. Coll and R. Sacks, Appl. Spectrosc. 42 (1988) 1025.[38] L. McCaig, R. Sacks and D. Lubman, Appl. Spectrosc. 43 (1989) 912.[39] S. Tanguay and R. Sacks, Appl. Spectrosc. 43 (1989) 918.[40] S. R. Goode and D. T. Pipes, Spectrochim. Acta B 36 (1981) 925.[41] V. Majidi and D. M. Colemen, Appl. Spectrosc. 41 (1987) 200.[42] D. Albers, M. Tisak and R. Sacks, Appl. Spectrosc. 41 (1987) 131.[43] E. T. Johnson and R. D. Sacks, Appl. Spectrosc. 42 (1988) 77.[44] K. J. Mason and J. M. Goldberg, Anal. Chem. 59 (1987) 1250.[45] K. J. Mason and J. M. Goldberg, Appl. Spectrosc. 45 (1991) 370.[46] K. J. Mason and J. M. Goldberg, Appl. Spectrosc. 45 (1991) 1444.[47] V. N. Rai, M. Shukla and H. C. Pant, Laser and Particle Beam 16 (1998) 431.[48] S. Suckwer and H. Fishman, J. Appl. Phys. 51 (1980) 1922.[49] S. Suckwer, H. Skinner, H. Milchberg, C. Keane and D. Noorhees, Phys. Rev. Lett. 55

(1985) 1753[50] V. N. Rai, A. K. Rai, F. Y. Yueh and J. P. Singh, Appl. Opt. 42 (2003) 2085.[51] R. H. Huddlestone and S. L. Leonard, Plasma Diagnostic Techniques, Academic Press, New

York (1965).[52] F. F. Chen, Introduction to Plasma Physics and Controlled Fusion, Plenum Press,

New York (1974).[53] V. N. Rai, M. Shukla and H. C. Pant, Pramana, J. Phys. 52 (1999) 49.[54] S. Yu. Guskov, T. Pisarczyk and V. B. Rozanov, Laser and Particle Beam 12 (1994) 371.




[55] R. D. Bleach and J. D. Nagel, J. Appl. Phys. 49 (1978) 3832.[56] V. N. Rai, M. Shukla and H. C. Pant, Proc. National Laser Symposium IRDE, Dehradun

10–14th Feb. (1995) pp 269.[57] W. J. Trestle, K. W. March, J. B. Orenberg, P. W. Carr, D. Miller and D. Glick, Anal.Chem.

43 (1971) 1452.[58] V. N. Rai, J. P. Singh, F. Y. Yueh and R. L. Cook, AIAA-2001–2812, 32nd AIAA Plasma-

dynamics and Laser Conference, 11–14th June (2001) Anaheim CA.[59] H. R. Griem, “Plasma Spectroscopy” Mc Graw Hill, New York (1964).[60] V. N. Rai, H. Zhang, F. Y. Yueh and J. P. Singh, Bulletin of Am. Phys, Soc. 46 (2001) 120.[61] J. Uebbing, J. Brust, W. Sdorra, F. Leis and K. Niemax, Appl. Spectrosc. 45 (1991) 1419.[62] R. Sattmann, V. Sturn and R. Noll, J. Phys. D 28 (1995) 2181.[63] D. N. Stratis, K. L. Eland and S. M. Angel, Appl. Spectrosc. 5 (2001) 1297.[64] L. St. Onge, M. Sabsabi and P. Cielo, Spectrochim Acta B 53 (1998) 407.[65] L. St. Onge, V. Detalles and M. Sabsabi, Spectrochim Acta B 57 (2000) 121.[66] V. N. Rai, F. Y. Yueh and J. P. Singh, Appl. Opt. 42 (2003) 2094.[67] A. Kumar, F. Y. Yueh, T. Millar and J. P. Singh, Appl. Opt. 42 (2003) 6047.[68] A. Kuwako, Y. Uchida and K. Maedo, Appl. Opt. 42 (2003) 6052.[69] W. Pearman, J. Scaffidi and S. Michael Angel, Appl. Opt. 42 (2003) 6085.[70] J. Scaffidi, J. Pender, W. Pearman, S. R. Goode, B. W. Colston Jr., J. C. Carter and S. M.

Angel, Appl. Opt. 42 (2003) 6099.[71] A. De. Giacomo, M. Dell’Aglio, F. Colao and R. Fantoni, Spectrochim Acta B 59

(2004) 1431.[72] A. De. Giacomo, M. Dell”Anglio, F. Colao, R. Fantoni and V. Lazic, Appl. Surf. Science

(2005) in Press.[73] W. A. Spencer, F. M. Pennebaker, N. M. Hassan, J. E. McCarty and C. W. Jenkins, Rep.

BNF-003-98-0199 Rev. (Savannah River Technology Center, Aiken SC. 2000).[74] F. Y. Yueh, R. C. Sharma, H. Zhang, J. P. Singh and W. A. Spenser, J. Air Waste Manage.

Assoc. 52 (2002) 747.[75] C. E. Max, “Physics of the coronal plasma in laser fusion targets” Ed. R. Balian and

J. C Adams Laser Plasma Interaction, North Holland Publishing Co. (1982).[76] H. Nakano, T. Nishikawa, H. Ahn and N. Uesugi, Appl. Phys. B 63 (1996) 107.[77] H. M. Milchberg, R. R. Freeman, J. Opt. Soc. Am. B 6 (1996) 1351.[78] R. M. Moore, K. H. Warren, D. A. Young, G. B. Zimermann, Phys. Fluid 31 (1988) 3059.



Chapter 11

Laser-Induced Breakdown Spectroscopy of Solidand Molten Material

A. K. Raia, F. Y. Yuehb, J. P. Singhb and D. K. Raic

aDepartment of Physics, University of Allahabad, Allahabad-211002, INDIAbInstitute for Clean Energy Technology, Mississippi State University,205 Research Boulevard, Starkville, MS 39759, USA.

cDepartment of Physics, Banaras Hindu University, Varanasi-210005, INDIA

1. INTRODUCTION

The current analytical techniques used for both qualitative and quantitative analysis oftoxic elements (e.g. chromium, lead, nickel, cadmium, copper, mercury etc) have sig-nificant limitations as regards their practical application. It involves sample collection,transportation, sample preparation and laboratory analysis which are labour intensive,costly, and require a considerable amount of time (a few days to a few weeks) for theresults to be available. Thus it is desirable to develop an analytical technique which isquick, sensitive, and is also capable of analyzing the material in-situ, especially in sit-uations involving hazardous materials. Laser Induced Breakdown Spectroscopy (LIBS),a powerful spectro-analytical technique rapidly making a transition from laboratories tofield use, has many advantages in this regard. It requires no special sample preparation,any type of material (solid, liquid, gas, slag) may be analyzed in situ and it is capableof detecting and analyzing several elements at the same time.

The present chapter aims at summarizing in some detail the diverse analytical methodsemploying LIBS that have been developed during the past two decades. The LIBStechnique makes use of a simple plasma spectrochemical approach. A high peak powerlaser pulse is focused on the sample (solid, liquid, or gas) to produce a spark whoseemission contains characteristic spectral signatures from excited atoms, radicals, and ionsin the plasma plume. The emitted radiation is collected by using optical fibers or lensesand passed through a monochromator where the spectrally resolved light is detected bya CCD/ICCD detector. The light intensity as a function of wavelength is recorded ina computer, and this digitalized data is analyzed using appropriate software. The endproduct of the analysis provides identification as well as concentration information aboutthe various elements present in the sample. Thus, LIBS is an advanced diagnostic toolfor rapid and remote analysis of target-composition [1–5].

LIBS can also provide on-line elemental analysis of compounds at the prepara-tion stage so that quality assurance and quality control decisions can be made during




256 A. K. Rai et al.

processing [6,7]. This facility may lead to enhanced control of product quality, savetime and improve the efficiency of processes involved in glass, metal and pharmaceuti-cal industries. Currently, no real-time measurement of melt constituents is available forthe glass, aluminum, and steel industries. Melt composition is determined by collectingmolten samples and taking them to a laboratory for analysis, making it a time- andenergy-consuming process. Moreover, the melt composition itself can change due tothe vaporization of the more volatile components during its transportation and hencecompositional fluctuations cannot be effectively monitored using the currently availablemethods. To improve production efficiency, such industries require a technique thatcan provide rapid, on-site melt composition measurement. This technique should alsoallow chemical additions to be made (if necessary) to the melt, so that an acceptableproduct composition is achieved prior to draining a melter/furnace. On-site, real-timemeasurements are expected to be more cost effective than separate sampling and off-siteanalyses. In the following sections we describe the experimental arrangements based onfiber optic (FO) LIBS sensor to measure on-line, in-situ elemental composition of solidand molten samples.

2. FO LIBS SENSOR FOR DETERMINATION OF ELEMENTALCOMPOSITION OF SOLID ALUMINUM ALLOYS

The laser breakdown threshold is known to be lower in solids than in a gas, so lower opticalenergy is needed for measurements on a solid sample. Analytical results of LIBS studieson solids are more frequent in the literature than on liquids or gaseous samples. Severalpublications [8], describe the determination of elemental composition in steel, Al alloys,soil, and paints, using the LIBS technique. LIBS has also been used for on-line qualitycontrol of rubber mixing and in the analysis of mining ores. A number of review articleson these topics have also appeared in scientific literature in recent years [4,9, and 10].

Gomba et al. [11] have determined the very low concentration of Li in an aluminum-lithium alloy by recording its LIBS spectra in a vacuum chamber in a controlled xenonatmosphere. Hemmerlin et al. [12] have demonstrated that LIBS is comparable to thespark technique for the quantitative determination of trace elements in steel. Femtosecondlaser pulses have been used by Drogoff et al. [13] to obtain detection limits in therange of a few ppm in Al alloys. Rosenwasser et al. [14] have used LIBS to identifythe metallic elements in ores, while Samek et al. [15] utilized it to measure traceelement concentration in hard biological tissue (e.g. teeth and bone). LIBS has also beenemployed to analyze wood [16], glass [17], concrete [18], limestone [19], and paint [20],soil and sand [21].

The early LIBS systems consisting of a number of lenses required elaborate alignmentfor recording the spectra [22,23]. Such an experimental set up is not well suited forindustrial/field use where minimum of on-site alignment is a great advantage. Recentadvances in fiber optic materials have opened up new areas of applications for the LIBStechnique. A beam delivery system is used to send the laser beam to the desired locationand the signal collected through optical fibers greatly facilitates remote measurement.One of the most difficult tasks in designing a FO-LIBS probe is to couple a high-energylaser beam into an optical fiber without damaging the fiber [24–26]. In the initial stagesfiber bundle replaced the lenses for collecting the emission from the laser-spark but in



LIBS of Solid and Molten Material 257

later experiments, two optical fibers; one for delivering the laser pulse and the otherfor collecting the emission from the spark were used [27,28]. In harsh and hazardousenvironments such as in aluminum, glass and steel industries the adjustment of twoseparate optical fibers is a delicate and difficult task, and therefore it is desirable to usethe same fiber for delivery as well as for collecting the optical energy [29–31].

The schematic diagram of the FO-LIBS probe is shown in Fig. 3 of chapter 5.The second harmonic (532 nm) of a pulsed Nd: YAG laser (Big Sky, Model CFR400) operating at 10 Hz, pulse duration 29 ns, beam diameter 7 mm and the full angledivergence 1.0 mrad, is directed towards the optical fiber by a 532/1064 nm beam splitterand a 532 nm dichroic mirror.

A 45� dichroic mirror (DM), with special coating that reflects at 532 nm and transmits180–510 nm and 550–1000 nm, is used for delivery of the laser energy and collection ofthe optical signal from the laser-spark. This simple design protects the detector from thepotential damage by the reflected laser light. To transmit sufficient laser energy throughthe fiber optic cable while keeping it below the damage threshold of the fiber, laser beamwas focused at a spot ∼3 mm in front of the fiber tip using a 10 cm focal length lens.A cap with a 0.8 mm pinhole was placed at the fiber input end to avoid the possibility ofany damage to the core and cladding of the fiber. The laser beam transmitted through theoptical fiber is collimated with a 10 cm focal length lens and then focused on the sampleby a 5 cm focal-length lens. The emission, from the laser produced plasma, is collectedby the same lenses and the optical fiber. The collimated radiation passes through thedichroic mirror and is focused onto an optical fiber bundle with a 20-cm focal lengthlens. The fiber bundle consists of 78 fibers each of 100 �m diameter and 0.16 numericalaperture (NA). The slit type output end of this fiber bundle delivers the emitted light to theentrance slit of a 0.5 m focal length spectrometer (Model HR 460 JOBIN YVON-SPEX)equipped with a 2400 lines/mm grating blazed at 300 nm. An intensified charge coupledetector (ICCD, Model ITE/CCD Princeton Instruments) with its controller (Model ST133, Princeton Instruments) was used as the detector. A programmable pulse delaygenerator (MODEL PG-200, Princeton Instruments) was used to gate the ICCD. Theentire experimental apparatus was controlled by a (Dell Dimension M 200a) computerrunning the WinSpec/32 (Princeton Instruments) software. Multiple (100) laser shotswere recorded and the resulting spectrum was stored in “accumulations” mode. Fiftyspectra were stored in one file for analysis to obtain average area/intensity value for thespectral line of interest.

2.1. Parametric Studies

To obtain optimum signal for the quantitative analysis of minor elements in the aluminumalloys, LIBS signals were recorded by changing the various experimental parameters(laser energy, sample surface, detector gain, gate delay and width etc).

2.1.1. Transmission of Laser energy through Optical Fiber

The fiber used in our experiment [31] was a silica core/silica cladding multimode fiber(FG-1.0-UAT from ThorLabs Inc.). The stability of silica cladding allows for high power-handling capability and correcting any laser mis-alignment. The silica cladding design




also provides superior UV transmission required to transfer the LIBS signal. The lengthof the fiber was 3 m and SMA 905 stainless steel fiber connectors (ThorLabs Inc.) wereused at both ends. The fiber was polished with 0.3-mm size aluminum oxide particles inthe final step. The core diameter is 1.0 mm, the cladding diameter 1.25 mm, the numericalaperture 0.16, and manufacturer’s suggested maximum power capability is 5 GW/cm2.The low numerical aperture provides for low beam divergence and a uniform spot sizethat facilitates focusing the beam after transmission through the fiber. The Nd: YAG laser(Big Sky Inc. CFR400) was operated at 10 Hz and its second harmonic (� = 532 nm)radiation has a pulse width (FWHM) 8 ns and maximum pulse energy 180 mJ. The laserhad a Gaussian beam profile and beam diameter was 6.5 mm. A spherical plano-convexfused silica lens of 10-cm focal length was used to couple the laser beam into the fiber.A 30-mJ-laser beam after passing through this lens can create breakdown in air, andhence this value (30 mJ) is the maximum laser energy that might be transferred throughthe fiber. A metal cover with a 0.8-mm pinhole at the center was placed just in front ofthe fiber end to avoid any damage to the core-cladding boundary during alignment. Thefiber was placed about 5 mm behind the focal point and it is estimated that only about0.6–0.7 mm of the core diameter was illuminated by the diverging laser beam. A simplecalculation indicates that a 30-mJ-pulse energy with a spot size of 0.5-mm diameter willproduce an energy density of 2 GW/cm2 in the fiber, which is lower than its damagethreshold. However, even at this energy level it is still possible that damage may occuron the input surface of the fiber due to randomly occurring hot spots in the laser profile.Fig. 1 shows that the energy transmission efficiency with our coupling setup is about88%, which is fairly high.

2.1.2. Influence of Laser Power on Fiber Damage

In order to improve signal-to-background (S/B) ratio, effects of various experimentalparameters were tested and during this process, the optical fiber was damaged severaltimes. In most of the cases damage occurred inside the fiber when the laser energyinput exceeded 20 mJ. It was soon realized that as long as the laser energy was kept

y = 0.8884x – 0.3631

R2 = 0.9972

0

2

4

6

8

10

12

14

16

18

20

0 2 4 6 8 10 12 14 16 18 20

Laser energy before fiber (mJ)

Lase

r en

ergy

afte

r fib

er (

mJ)

Fig. 1. Laser transmission from the optical fiber. (Reproduced with permission from Ref. [31]).




below 20 mJ at the fiber input end, no damage occurred and for all the experiments laserenergy was kept below this threshold. On several occasions the damage caused due tocore-cladding breakdown took place not at the input end but at a point 2–5 cm insidefrom the input face. This kind of damage is most likely at the location where the firstreflection of the laser beam inside the fiber takes place. In our first experiments, thefiber was clamped about 10 cm behind the input end face and several times the fiberdamage occurred just behind the clamped position, due to additional stress caused byclamping. In later experiments the fiber was kept straight and was clamped around 30 cmbehind the input end and no such damage occurred. We recorded the LIBS spectra byvarying the laser power up to the damage threshold, and it was observed that the signal-to-background ratio is best at laser-pulse energy of 13.6 mJ. All subsequent experimentswere performed at this laser power.

2.1.3. Effect of Laser Radiation on the Surface of the Sample

If the focused pulsed laser beam is directed at the same spot on the sample surface, theLIBS signal decreases with time. This decrease is believed to be due to the formationof an oxide layer and a crater, which modifies the optical properties of the target. If thelaser is focused continuously at the same location, the crater size changes and this resultsin a time varying LIBS signal. Therefore, to obtain reproducible signals, measurementswere made by slowly translating the sample with a stepping motor to ensure that thelaser strikes a fresh spot for each new measurement.

2.1.4. Influence of Gain of the Detector

In our initial experiments, the gain of the ICCD was kept high, but the S/B ratio wasfound to be very poor. These experiments were performed with a short gate delay andatomic lines were found to be buried in a strong background that caused saturation of thedetector. In order to improve the S/B ratio, a longer time delay was used which not onlyreduced the background but also caused the disappearance of some of the weak spectrallines. To reduce the scattered laser light, a notch filter was placed in front of the inputend of the receiver fiber, but no significant improvement in the S/B ratio was observed.This shows that scattered laser light is not the main cause of the strong background.Finally it was found that by keeping the detector gain at a moderately low level, discretespectral lines in Al alloys could be recorded with better S/B ratios. Fig. 2 shows LIBSspectra recorded at detector gains of 1 and 2 respectively and S/B ratio is higher forthe lower gain setting (see upper spectra of Fig. 2). At lower gain, one can record theLIBS spectra with good S/B ratio even by setting a shorter delay time and without losingthe weak lines (for example, the 404.136 nm line of Mn). We thus conclude that thegain-setting of the detector is an important parameter in the present experimental setup.To avoid saturation of some strong lines at short delay time and at low detector gain,the spectra were recorded by using neutral density filters.

2.1.5. Effect of Detection-time Window

The spectral line emission signal is always accompanied by a strong continuum from thelaser-produced plasma. The continuum background dominates during the first several




100000

80000

60000

40000

400000

300000

200000

100000

0

20000

0370

Pr

975.

02

Ha

900.

67 Pr

902.

04

Pr

900.

69

Hq

909.

29

Hq

909.

09

Pr

971.

95

370

Wavelength (nm)

Inte

nsity

(a)

(b)

Inte

nsity

375 380 385 390

Wavelength (nm)375 380 385 390

Fig. 2. LIBS Spectra of solid Al alloy at two different gain of the ICCD detector; upper spectra (a)recorded at low gain and lower one (b) at high gain. (Reproduced with permission from Ref. [31]).

microseconds after the laser pulse but decays faster than atomic emission. Therefore,one can use a time-resolved technique to discriminate against the continuum radiation.Fig. 3 shows S/B ratio for a spectrum recorded with a FO-LIBS system at various gatedelay times with gate width fixed at 2 �s. The best S/B is obtained with delay timesof 2–3 �s and hence in the present work, LIBS spectra for parametric studies wererecorded at 2 �s gate delay using 2 �s gate-width. The plasma temperature estimated

0

20

40

60

80

100

120

140

0 2 4 6 8 10 12 14 16

Delay time (μs)

S/B

rat

io

Fig. 3. LIBS spectra of solid Al alloy recorded at different gate delay. (Reproduced with permissionfrom A. K. Rai et al. [31]).




from the spectra recorded in this experimental condition was found to be 5570 K. Thecritical electron density for local thermal equilibrium (LTE) is found to be 5 ×1015cm−3,evaluated on the basis of the Griem equation [32]. Since the measured electron density�2×1016 cm−3� is higher than the calculated value, this indicates that LTE exists underthese conditions.

2.2. Effect of Angle of Incidence on LIBS Signal

We visited an aluminum factory at Syracuse, NY, USA to explore the possibility ofusing a LIBS probe inside their furnace. From an inspection of their facilities we cameto the conclusion that it was not possible to insert the probe from the top of furnace sothat it is perpendicular to the melt surface but the probe insertion was possible only atan angle with the melt surface.

In order to evaluate the performance of the FO-LIBS probe in a factory environment,laboratory studies were performed to assess the effect of angle of incidence of the laserbeam on the intensities of the analyte emission. The LIBS signals using fiber opticprobe as well as without such a probe were recorded for various angles of incidence(0�� 15�� 30�� 45� and 60�� where 0� corresponds to normal incidence. Great care wastaken to maintain the constancy of the lens-to-sample distance at each angle of incidence.For this, the axis of rotation of the sample was made coincident with the axis of theincident beam. Lenses of different focal lengths were used to focus the laser radiationon the sample and the results are summarized in the following sections.

2.2.1. Fiber with Lenses of Focal length 5 and 10 cm

We have recorded the LIBS spectra from neutral (Fe, Cr, Mg, Mn etc.), and ionic speciesat gate delays of 0.3, 0.5, 1, 2 and 3 �s and at various angles of incidence ranging from0� to 60�. Our results show that intensities of both line and continuum emission decreaseas the angle of incidence changes from 0� (normal incidence) to 60�. In the case oflines from neutral atoms, the decrease in intensity is steeper at higher time delay, butin the case of the background continuum the trend is opposite (Fig. 4 (a) and 4 (b)).This observation is in accordance with the fact that in the first microsecond after thelaser pulse the continuum emission is strong whereas the line emission appears strongonly after several microseconds. A similar experiment has been performed by Multariet al. [33] who noticed that emission intensities were the largest for incidence at 0�

and decreased as the incidence angle was increased upto 40�. An increase of intensityfor angles of incidence beyond 40� was also noted. This increase was greatest for theneutral emission, which became almost as intense as at normal incidence for angleof incidence of 60�. In contrast the intensity for ionized species and the backgroundcontinuum continued to decrease beyond 40� and become a minimum at 60�. In ourexperiments the intensity in all three types of emissions (neutral, ion and backgroundcontinuum) is largest at 0� and is smaller for all other angles.

As the sample is rotated with respect to the incident laser beam the mass of the ablatedmaterial as well as the temperature of the atomic material ejected from the surfacemay change, which may lead to changes in emission intensities. Multari et al. [33]




00

0.2

0.4

0.6

0.8

1

1.2

–100

(a) (b)

–60 –10 40–50 50 100

Angle Angle

Inte

nsity

Inte

nsity

Mn2

Cr2

Fe3

Backgr

Mn2Cr2Fe3Backgr

1

0

0.2

0.4

0.6

0.8

1.2

Fig. 4. Variation of line intensity and background continuum with angle of incident of laser beamon the sample surface in the LIBS spectra of solid Al alloy recorded using 5 cm focal length lenswith fiber and at gate delay; (a) 0�5 �s (b) 2 �s. (Reproduced with permission from A. K. Raiet al. [6]).

have reported that there is no variation in the mass of the ablated material as the angleof incidence is changed from 0� to 60�, thereby eliminating changes in total ablatedmass as the cause of the observed changes in the emission intensity. Thus, changein temperature of the ejected material may be the cause in the decrease of the LIBSsignal at higher incidence angle. Our measurements showed a monotonic decrease inthe plasma temperature as the sample was rotated from 0� to 60�, which would decreasethe intensity of emission from the neutral as well as the ionized species. Another causeof decrease in the measured intensity is probably the fact that the symmetric centralaxis of emissions (which is perpendicular to the surface for all sample orientation) nolonger remains aligned to the collecting optics. The third reason for the decrease in theLIBS signal may be the difference in the amount of laser light reflected from the samplesurface at different angles of incidence. In our experiments, we noticed an increase inreflection with increase in the angle of incidence which causes a reduction in the laserenergy available for producing the spark.

In fiber optic experimental setup also, the LIBS signal decreases with an increase inthe rotation angle of the sample, but there are differences in the trend of decrease in theatomic emission. The decrease in the intensity of atomic lines is steeper at lower delaytime (Fig. 5a) but for higher delay time the effect of rotation on the intensity of theatomic lines is small (Fig. 5b).

2.3. Calibration Curve

It is clear from the above parametric studies that the FO-LIBS probe is just like aflash light with analytical capability, so that if you shine this flashlight at any materialyou are directly able to see the various elements in that material. This probe/sensor isvery suitable for qualitative analysis or even for semi quantitative elemental analysisof the sample material. For quantitative analysis, however, some shortcomings must beovercome before one can use this technique. If one wants to perform the quantitative




0

0.2

0.4

0.6

0.8

1

1.2

0

0.2

0.4

0.6

0.8

1

1.2

Inte

nsity

Inte

nsity

–70

(a) (b)

–20 30 80 –100 –50 0 50 100

Angle Angle

Mn2Cr2Fe3Bkgr

Mn2Cr2Fe3Bkgr

Fig. 5. Variation of line intensity and background continuum with angle of incident of laser beamon the sample surface in the LIBS spectra of solid Al alloy recorded using 10 cm focal length lenswith fiber and at gate delay; (a) 0�5 �s, (b) 2 �s. (Reproduced with permission from A. K. Raiet al. [6]).

0

20

40

60

80

100

120

140

160

180

0 1 2 3 4

Hun

dred

s

Weight %

Inte

nsity

Fig. 6. Ideal Calibration curve for Quantitative Analysis of minor elements in Alloys.

analysis of minor elements (i.e. Mn, Mg, Cr, Cu etc.) in an Al alloy, one should firstprepare calibration curve between the concentration of the element in the alloy and theLIBS signal intensity of a spectral line of this element. A perfect calibration curve is one,which passes through the origin and also has a small standard deviation. (Fig. 6). In actualpractice it is difficult to get such an ideal calibration curve. To obtain calibration curvesfor Mn, Cr, Mg and Cu etc, which are most important minor elements in the Al alloys,we have obtained commercially available Al alloys, whose component concentrationsare given in Table 1 and their LIBS spectra in different spectral regions have beenrecorded using the experimental set-up shown in Fig. 3 of chapter 5.




Table 1. Concentration of different elements in weight % present in Al alloy

Sample Si Fe Cu Mn Cr Ni Zn Mg Al

70750�06∗ 0�018∗ 1�37∗ 0�02∗ 0�20∗ 0�00∗ 5�78∗ 2�46∗ 89�86∗

0�13+ 0�15+ 1�35+ 0�02+ 0�19+ 0�00+ 5�69+ 2�62+ 88�93+

20170�33∗ 0�28∗ 3�97∗ 0�45∗ 0�18∗ 0�01∗ 0�10∗ 0�73∗ 96�46∗

0�30+ 0�26+ 3�80+ 0�56+ 0�20+ 0�00+ 0�09+ 0�57+ 93�85+

60610�65∗ 0�29∗ 0�27∗ 0�073∗ 0�073∗ 0�073∗ 0�053∗ 0�85∗ 97�95∗

0�35+ 0�33+ 0�31+ 0�09+ 0�07+ 0�00+ 0�06+ 0�86+ 97�57+

62620�35∗ 0�48∗ 0�31∗ 0�01∗ 0�07∗ 0�00∗ 0�01∗ 1�00∗ 97�66∗

0�37+ 0�50+ 0�35+ 0�00+ 0�06+ 0�00+ 0�00+ 1�06+ 97�32+

20260�06∗ 0�10∗ 4�33∗ 0�48∗ 0�01∗ 0�00∗ 0�05∗ 1�39∗ 93�6∗

0�06+ 0�07+ 4�29+ 0�59+ 0�00+ 0�00+ 0�06+ 1�45+ 93�07+

20110�15∗ 0�039∗ 5�38∗ 0�02∗ 0�01∗ 0�00∗ 0�03∗ 0�09∗ 94�00∗

0�13+ 0�39+ 5�65+ 0�00+ 0�00+ 0�00+ 0�02+ 0�00+ 93�75+

60630�23∗ 0�18∗ 0�06∗ 0�00∗ 0�00∗ 0�00∗ 0�00∗ 1�60∗ 99�03∗

0�36+ 0�15+ 0�00+ 0�00+ 0�00+ 0�00+ 0�00+ 0�48+ 98�68+

∗ Analysis based on MSU chemical lab (atomic absorption).+ Analysis based on ICP.

In the atomic spectrum of Mn there is a group of four lines in the wavelength regionof ≈400 nm (Fig. 7) of which three lines (403.448, 403.306, 403.075 nm) are very strongand one line (404.135 nm) is weak. A calibration curve using one of the strong linesat 403.075 nm is shown in Fig. 8a. One can easily see that for this particular line ofMn the calibration curve is not a straight line. It seems that the LIBS signal for thisparticular line gets saturated due to self absorption in the case of samples with higherconcentration of Mn. The three lines (403.448, 403.306, 403.075 nm) are the resonantlines which means that the lower state of these lines is the ground state of the atom.Therefore, it is more likely that these lines would suffer from self absorption. The line at404.135 nm is not a resonant line. The lower state for this line is an excited state of theatom. Therefore, self absorption is not likely for this line. The calibration curve usingthis line is shown in Fig. 8b and is a straight line.

To reduce the influence of experimental parameters like laser power, sample to lensdistance and the nature of the matrix elements on the LIBS signal from different samples,one can use a ratio calibration curve. In other words, one uses the ratio of the intensity ofthe analyte atomic line and the intensity of a reference atomic line. Since Fe has atomiclines in almost every spectral region, we have divided the intensity of the analyte atomicline with the intensity of a Fe reference line. Fig. 9a shows the ratio calibration curve forthe strong Mn line (403.075 nm) and once again the calibration curve is nonlinear. Forthe nonresonant weak Mn line, however, the calibration curve is a straight line (Fig. 9b).

The next minor element in the Al alloy tested for calibration is Cr. As shown inFig. 10, there are two groups of Cr lines; one in the wavelength region ≈360 nm andthe other in the wavelength region ≈425 nm which may be utilized for drawing thecalibration curves.

Fig. 11a shows the nonlinear calibration curve using Cr line at 359.35 nm whereasthe calibration curve using Cr line at 428.97 nm is a straight line (Fig. 11b). This isagain because the Cr line at 359.35 nm is nearly ten times more intense than the Cr




403.448 nm403.306 nm403.075 nm

24802.25 cm–1

24788.05 cm–1

24779.32 cm–1

41789.48 cm–1

404.135 nm

Ground level

17052.29 cm–1

Mn

403.306

404.135

403.075120000

100000

80000

60000

40000

20000

0400 405

Wavelength [nm]410 415

403.448

Fig. 7. Atomic energy level diagram of Mn and its spectrum.

0 0

50

100

150

200

250

300

0 0.2 0.4 0.6 0.8 1

0.2

0.4

0.6

0.8

1

1.2

1.4

Mill

ions

Tho

usan

ds

Weight % of Mn

(a) (b)

0 0.2 0.4 0.6 0.8 1Weight % of Mn

Inte

nsity

of M

n lin

e

Inte

nsity

of M

n lin

e

♦ Mn(403.075) nm ♦ Mn(404.135) nm

Fig. 8. (a) Calibration curve using absolute intensity of Mn (403.075 nm) resonant line;(b) Calibration curve using absolute intensity of Mn (404.135) non-resonant line.




00 0.5 1 1.5

2

4

6

8

10

12

Weight % of Mn/Fe Weight % of Mn/Fe

Inte

nsity

of M

n/F

e

1

00 1 2 3 4

2

3

4

5

6

Inte

nsity

of M

n/F

e

♦ Mn(403.075)/Fe(406.39) nm ♦ Mn(404.135)/Fe(406.39) nm

(a) (b)

Fig. 9. (a) Calibration curve using ratio of Mn (403.075 nm) line with Fe (406.39 nm) line;(b) Calibration curve using ratio of Mn (404.135 nm) line with Fe (406.39 nm) line.

360.53 nm

23498.84 cm–1

23386.35 cm–1

23305.01 cm–1

27935.26 cm–1Cr27820.23 cm–1

27728.87 cm–1

425.43 nm

427.48 nm

428.97 nm

357.87 nm

359.35 nm

Ground level

428.97

427.48425.43

357.8712000

80000

40000

0

350 355 360 365 420

0

10000

20000

Inte

nsity

Inte

nsity

30000

425

Wavelength (nm)Wavelength (nm)430

359.35

360.53

Fig. 10. Atomic energy level diagram of Cr and its spectrum.




0

0.2

0.4

0.6

0.8

1

1.2

1.4

0 0.1 0.2 0.40.3 0.5

Weight % of Cr

Inte

nsity

of C

r lin

e

Cr(359.35 nm)

0

0.2

0.4

0.6

0.8

1

1.2

1.4

0.2 0.40.30 0.1 0.5

Weight % of Cr

Inte

nsity

of C

r lin

e

Cr(428.97 nm)

(a) (b) M

illio

ns

Mill

ions

Fig. 11. (a) Calibration Curve using absolute intensity of Cr (359.35 nm) Line, (b) CalibrationCurve using absolute intensity of Cr (428.97 nm) Line.

0

1

2

3

4

5

6

7

8

9

10

0.4 0.8

Weight % of Cr/Fe

Inte

nsity

of C

r/F

e

Cr(359.35 nm)/Fe(364.98 nm)

0

1

2

3

4

5

6

7

8

0.4 0 0.2 0.60 0.2 0.6 0.8

Weight % of Cr/Fe

Inte

nsity

of C

r/F

e

Cr(428.97 nm)/Fe(432.71 nm)

(a) (b)

Fig. 12. (a) Calibration curve using ratio of Cr (428.97 nm) line with Fe (432.71 nm) line,(b) Calibration curve using ratio of Cr (359.35 nm) line with Fe (364.98 nm) line.

line at 428.97 nm and the atomic line at 359.35 nm saturates the detector for higher Crconcentrations in the sample. The calibration curve based on intensity ratio at 428.97 nmis a straight line whereas the similar curve for the 359.35 nm emission is not linear(Figs. 12a, 12b). Since saturation of the detector may be avoided by reducing the incidentlaser power, we have shown in Fig. 13 the calibration curves for the 359.35 nm attwo-laser powers (13.6 mJ and 10.2 mJ) and it is clear that the calibration curve at lowerlaser power is nearly linear.

In the spectrum of Mg there are three close lying atomic lines (382.93, 383.22,383.82 nm) having a common upper level. The intensity of 383.82 nm line is the largestwhereas intensity of 382.93 nm line is the smallest (Fig. 14). The calibration curvecorresponding to 383.82 nm line is not linear because of the saturation of the detectorwhereas the calibration curve corresponding to 382.93 nm line is a straight line (Figs. 15a,15b). Similar behavior is seen in the ratio calibration curves for these two lines of Mg.(Figs. 16a, 16b).




0

0.1

0.2

0.3

0.4

0.5

0.6

0 0.05 0.1 0.15 0.2 0.25

Weight %

Line

inte

nsity

(ar

ea)

Cr(359.349 nm) at laserpower 13.6 mJCr(359.349 nm) at laserpower 10.2 mJ

Mill

ions

Fig. 13. Calibration curve using absolute line intensity of Cr 359.35 at two different laser powers.

380 385 3900

100000

200000

Inte

nsity

300000

Wavelength (nm)

382.

93 M

g

383.

22 M

g

383.

82 M

g

382.93 nm

47957.06 cm–1 Mg

383.22 nm383.82 nm

Ground level

21911.18 cm–1

21870.46 cm–1

21850.41 cm–1

Fig. 14. Atomic energy level diagram of Mg and its spectrum.




0

100

200

300

400

500

600

700

0.2 0.40 0.1 0.3 0.5 0.6

Tho

usan

ds

Tho

usan

ds

0

50

100

150

200

250

300

350

400

450

0.2 0.40 0.1 0.3 0.5 0.6

Mg(383.82 nm) Mg(382.93 nm)

Weight % of Mg Weight % of Mg

Inte

nsity

of M

g lin

e

Inte

nsity

of M

g lin

e

(a) (b)

Fig. 15. (a) Calibration curve using absolute intensity of Mg (383.82 nm) line, (b) CalibrationCurve using the absolute intensity of Mg (382.93 nm) line.

0

0.75

1.5

2.25

3

3.75

4.5

Weight % of Mg/Fe

0

0.5

1

1.5

2

2.5

3

0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8

Weight % of Mg/Fe

Mg(383.82 nm)/Fe(382.04 nm) Mg(382.93 nm)/Fe(382.04 nm)

Inte

nsity

of M

g/F

e lin

e

Inte

nsity

of M

g/F

e lin

e

(a) (b)

Fig. 16. (a) Calibration curve using the ratio of Mg (383.82 nm) line with Fe (382.04 nm) line,(b) Calibration curve using the ratio of Mg (382.93 nm) line with Fe (382.04 nm) line.

It is thus clear that by selecting the proper atomic line, one can get a linear calibrationcurve but the fluctuations about the straight line as measured by the standard deviation,have still to be tackled. For the calibration curves shown in the Figs. 8b, 12b, 15b,the standard deviation is quite large �∼20%�. The standard deviation may be reducedby excluding certain data points which differ from the average value by more than0.5 � (� is the original standard deviation) i.e. we have to exclude all points whichdeviate beyond A ±0�5� , where A is the average value of the intensity.

2.4. Effect of Sample-Lens Distance and Focal Length

It has been found that a change in the sample to lens distance causes a change in theintensity of the LIBS signal from one laser shot to another which ultimately increasesthe standard deviation and hence the error bar for the measurements. Since the confocal




parameter is short for a shorter focal length lens, we expect that the fluctuation inintensity of the LIBS signal would be larger for short focal length lens than for a largefocal length lens.

We have performed experiments to see the effect of sample-to-lens distance onLIBS signal by using lenses of focal lengths 20 cm and 5 cm (with and without fiber).The experimental results clearly demonstrate that even a slight (0.5 mm) change in thesample-to-lens distance may considerably reduce (by as much as 30%) the LIBS signalfor a lens having f = 5 cm (Fig. 17a). For a lens having f = 20 cm, similar reduction insignal strength takes place for a 2 mm change in the sample to lens distance (Fig. 17b).Therefore, it is advisable to use a longer focal length lens to focus the incident radiationon the surface of the sample. Further, we noticed that the rate of decrease in LIBS signalis slow when the focal point is situated in front of the sample surface and it decreasesmore rapidly when the focal point is beyond the sample (Fig. 17a). This observationis seen more clearly in the case of a longer focal length lens (Fig. 17b). If the sampleto lens distance is not kept constant during the translatory motion of the sample, theLIBS signal will fluctuate in intensity from one laser shot to another thereby affectingthe analysis.

Our experimental observations demonstrate that to reduce the standard deviation incalibration curve, one should use a longer focal length lens. To confirm this, we havecalculated the percentages of standard deviation (Table 2) for intensity of the Cr line fordifferent Al alloy samples by recording the LIBS spectra using lenses of focal length5 cm and 20 cm. It is seen from Table 2 that the standard deviations for the intensity ofthe analyte lines obtained from the LIBS spectra using a lens of focal length of 5 cmare larger in comparison to those from the LIBS spectra using a lens of focal length of20 cm. One can also notice that the standard deviation is large for the sample, whichhas lower Cr concentration (Table 2). We can now conclude that in calibration curve forquantitative analysis of an element, non-resonant spectral lines should be preferred andfocal length of the lens collecting emission from the laser spark should have a large value.

0

0.2

0.4

0.6

0.8

1

Position from focal point (cm)

Mn2Cr2Fe3

0

1

2

3

4

5

6

7

– 4 –2 0 2 4 –6 –4 –2 0 2 4 6

Position from focal point (cm)

SiMg2MnBkg

(a) (b)

Inte

nsity

Inte

nsity

Mill

ions

Mill

ions

Fig. 17. Variation of LIBS signal intensity with sample-to-lens distance using; (a) focusing lensof 5 cm focal length and gate delay of 0�7 �s (b) focusing lens of 20 cm focal length and gatedelay of 2 �s.




Table 2. Variation of the STD of the intensity of analyte line�Cr � = 357�869 nm�

Sample Cr concentration F = 20 cm F = 5 cm

% STD % STD

6262 0.07 5�7 21�111258 0.01 10�88 22�211259 0.17 11�59 15�247075 0.20 5�73 10�632017 0�018/0�02 9�35 27�92011 0.01 14�39 23�876061 0.073 8�62 9�442024 0.01 12�01 19�73

3. LIBS SPECTRA OF MOLTEN ALUMINUM ALLOYIN A LABORATORY FURNACE

The use of fiber optic (FO) LIBS technique for in-situ and on-line compositional anal-ysis/studies of the molten alloy inside a furnace is not yet a practical proposition andeven laboratory studies are rare. Recently, Paksy et al. [34] performed quantitative anal-ysis of metals in the molten phase. They focused the laser light on the surface of themolten alloy with the aid of a fixed optical system while the emission from the laserinduced plasma was collected using optical fiber in a direction perpendicular to the laserbeam. Gruber et al. [35] have used the LIBS technique for monitoring of Cr, Cu, Mnand Ni in steel by focusing the laser beam on the surface of the molten sample in thefurnace. Noll et al. [36] analyzed the top gas composition for monitoring the elementalcomposition of molten steel in the blast furnace. It is to be noted that the surface of themolten alloy as well as the top gas may not contain the actual elemental compositiondue to the formation of slag-like/oxide material on the surface of the molten alloy. Alsomeasurements on the surface will not provide any information about the uniformity ofmixing in the Al melt inside the furnace. Therefore, it is desirable to analyze the moltenalloys by recording the LIBS signal by probes inserted fairly inside the melt surface. Toachieve this goal, we have modified the FO-LIBS probe, which was developed for thecompositional analysis of solid Al alloy [31]. The laser beam is coupled with the opticalfiber in the same way as for the experiment on solid Al alloy. The main modification inthe FO-LIBS probe is after the exit point of the optical fiber. The laser beam at the exitof the optical fiber is collimated by a plano-convex lens �f ≈ 15 cm� and focused withthe help of another plano-convex lens �f ≈ 5 cm� which is kept at a distance of 75 cmfrom the collimating lens. Both these lenses are kept in a stainless steel (s. s.) holderwith an internal diameter ≈2.2 cm and outer diameter ≈3.0 cm (Fig. 18). At the bottomof the holder a cave is cut to hold the focusing lens, which sits on an iron ring. Thes.s. holder below the collimating lens contains an inlet designed for purging an inert gasthat cools the lens and applies pressure to the aluminum melt surface. The purging gascomes out through eight holes near the focusing lens and does not allow the Al meltto reach the lens surface. This s.s. holder is then inserted into a ceramic pipe with the




Optical fiberconnector

Collimating lens

Focusing lens

Iron ring

N2 gas out

Swagelok

Opticalfiber

Purge gas

Fig. 18. Stainless steel holder for collimating and focussing lens.

help of an s.s. flange and a thermocouple is also placed in the s.s. holder to monitor thetemperature near the optics.

The Al alloy melt was produced in a laboratory furnace (L-83102-56622, GS, LIND-BERG) placed in a crucible �8�13 cm ID×9�15 cm OD×16�5 cm high, AC 36265 Al2O3

crucible, Ozark Technical Ceramics, Inc.). To avoid breakage and thermal shock, thiscrucible was placed in another crucible �15�24 cm OD × 14 cm ID × 7�6 cm high� andthe temperature of the furnace was increased in steps of 60� every half hour until itreached 800�C. The schematic diagram of the experimental set up of FO-LIBS probefor measuring the elemental composition of molten Al alloy in laboratory is shown inFig. 19. Initially, the focal point was nearly 2.5 cm inside the ceramic pipe but in laterexperiments we found it necessary to insert the probe more than 2.5 cm below the meltsurface. The focusing lens was damaged due to splashing of the melt at low flow rates ofthe purging nitrogen gas. To avoid damage to the focusing lens the design of the probewas changed by focusing the laser beam at the circumference of the ceramic pipe. TheLIBS spectra of seven molten alloys were recorded without any damage to the focusinglens by adjusting the inlet flow rate of the purging gas between 1.5 and 3 l/min, andthe outlet flow rate between 100 and 600 ml/min [7]. At times, the LIBS signal strengthdecreased, but recovered once the flow rate of the purging gas was adjusted.

We also recorded the LIBS spectra by inserting the probe at different depths insidethe melt and it was found that at greater depths a higher inlet-flow rate was necessaryfor sufficient LIBS signal. These experiments demonstrated the success of the probeto record LIBS signals from inside the melt. The LIBS spectra of the melt could be




2XNd: YAG laser

Data acquisition /Analysissystem

Computer Controller

Spectrograph

Pulse generator

ICC

BH

LDLLFO

F

Collimating optics

Furnace

Al melt

BD – Beam Dump DM – Dichroic Mirror FO – Fiber Optics HS – Hormonic Separator L – Lens ICCD – Intensified Charge Coupled Device 2X – KDP Doubler

Fig. 19. Schematic diagram of the experimental set up of FO LIBS probe. (Reproduced withpermission from A. K. Rai et al. [7]).

recorded with laser pulse energy of 9.5 mJ whereas the minimum laser-pulse energyneeded for recording the spectra of solid Al alloy is 13.2 mJ [31].

3.1. Effect of the Surrounding Atmosphere on LIBS Signal

In order to obtain the optimum LIBS signal in the molten alloy, the effects of the gaseousatmosphere surrounding the sample on the emission characteristics of the laser-inducedplasma were also studied. The intensities of the different atomic lines, the continuumemission and noise were measured in nitrogen, argon and helium atmosphere in twospectral regions (� � 360 nm and 300 nm). As seen from Fig. 20 the most intenseemission signal is obtained in argon atmosphere. Kuzuya et al. [37] have performed asimilar study for solid samples and they also observed that maximum emission intensityis obtained in the argon atmosphere. Paksy et al. [34] also performed experiments tostudy the effect of air and argon atmospheres on the plasma emission from both solidand molten samples. These authors noted that the background intensity is larger in argonatmosphere if the plasma is generated from a solid sample, whereas it is smaller, if theplasma is generated from a molten sample. Our results are not directly comparable toothers because while they had focussed the laser on the surface of the molten aluminumalloy, we have measured the emission intensity after inserting the probe more than 2.5 cmbelow the melt surface. Further, they measured the emission intensity in a directionperpendicular to the laser beam, while our measurements are of the emission in thebackward direction. On the basis of the present experiments we conclude that for thesame experimental condition the background (BKG) continuum in the presence of Aris almost two times more intense than in the N2 atmosphere. The BKG intensity inhelium atmosphere is lower than both, but the intensity of the plasma emission in heliumatmosphere is too low to be detected when one uses a 2 �s gate delay for which N2 andAr measurements have been carried out. Therefore, for the case of helium the intensitydata has been recorded at 1 �s gate delay keeping the other experimental parameters the




N2 Purging gas

Fe

Fe

Si

Si

Cr

0

50000

100000

150000

200000

250000

287 289 291 293 295 297 299 301 303 305 307

Wavelength (nm)

Inte

nsity

Inte

nsity

20000

0

40000

60000

80000

100000

120000

140000

287 289 291 293 295 297 299 301 303 305 307

Wavelength (nm)

Ar Purging gas

Cr

Fig. 20. LIBS spectra of molten Al alloy in laboratory furnace taken with different purging gases.(Reproduced with permission from A. K. Rai et al. [7]).

same as in argon and nitrogen atmosphere. The shorter delay is necessitated since thebreakdown threshold of helium is higher than for the other two gases [37].

From the viewpoint of analytical performance, it is useful to evaluate the line-to-background ratio (LBR) of the spectrum. The values of LBR for the Si 288.158 nm,Fe 297.334 nm and Cr 301.757 nm lines were derived from the intensity data for differentatmospheres and are presented in Table 3 where LBR are found to be higher in argonatmosphere than in nitrogen atmosphere. LBR value for Si is nearly 2.1 times larger in

Table 3. Calculated S/B and S/N of different elements from the LIBSspectra of Al molten alloy in the presence of various atmospheric gases

Gas Elements

Cr Fe Si

S/N S/B S/N S/B S/N S/B

N2a 12�32 0�49 42�05 1�78 46�82 1�97

Ara 14�87 0�65 49�44 1�96 103�82 4�13Heb 4�2 0�85 7�49 1�57 16�68 3�34

a 2-�s gate delay; b1-�s gate delay.




Ar as compared to N2 whereas for Cr and Fe the increase is by a factor of 1.32 and 1.10respectively. For helium atmosphere the LBR value for Cr is larger than both for argonand nitrogen atmospheres but for Fe helium yields a lower LBR than argon and nitrogenatmosphere. For Si, the LBR in He is lower than in argon but is larger than in nitrogen.

The expansion of the laser-induced plasma is dependent on the pressure of thesurrounding gas and it is related to the mass density of the gas. Since the density ofargon is larger in comparison to helium and nitrogen, hence at the same pressure, theconfining effect on the plasma is stronger in the case of argon atmosphere, which resultsin an increase in the emission intensity for both BKG and line emission.

We have also calculated the line-to-noise ratio (LNR) for Si 288.158 nm, Fe297.334 nm and Cr 301.7569 nm (See Table 3). As in the case of the LBR value, itis seen from Table 3 that LNR value for Si is also 2.2 times larger in case of argonatmosphere than in the nitrogen atmosphere, whereas, for Cr and Fe the increase in LNRvalue in argon is only 1.20 and 1.17 times respectively. In contrast to LBR value, thevalue of LNR for Si, Fe and Cr for helium atmosphere is lower in comparison to argonand nitrogen atmosphere.

In summary, we can state that for the analysis of the molten phase an argon atmosphereis more appropriate, because:

1. it ensures higher LBR value,2. it ensures higher LNR value (favorable detection limit),3. and it helps avoid surface oxidation

3.2. Calibration Curves for Molten Aluminum Alloy

Calibration based on line intensity is a very straightforward method for elemental anal-ysis. However, calibration curves based on absolute intensity are only applicable forthe samples where data are taken under the same experimental conditions (laser power,detection duration and delay, sample to lens distance and for samples of similar mate-rial/matrix). To obtain the spectrum from a molten-phase sample, one has to heat thesample slowly up to 800�C which takes several hours (nearly 4 to 5). Therefore, one isable to record the LIBS spectra in desired spectral range for only one sample a day, andit takes a whole week to obtain the LIBS data for seven samples. Since, it is difficult tokeep all the experimental parameters the same for such an extended (in time) experiment,calibration curves using ratios of the intensity of the analyte line to the intensity of a ref-erence line of another element are considered more reliable. To obtain reliable calibrationdata, the reference element should have reasonably high concentration in each standardsample. The selected reference line should be interference-free and its upper energy levelshould be close in energy to that for the analyte line. Although Al is the major speciesin all Al alloys, the Al lines are present only in two spectral regions in our experimentsand most of them suffer from spectral interferences. Since iron lines are abundant in thewavelength range covered in the present experiment (300–420 nm) an interference freeFe line from each spectral region was selected as the reference line for ratio calibration.Calibration curves were obtained for seven different aluminum alloy samples. Figs. 21(a)–(e) show some typical calibration curves using this method. The calibration curves




y = 2.9245x + 4.3719R2

= 0.9475

0

10

20

30

40

50

60

70

80

0 2 4 6 8 10 12 14 16 18 20Weight ratio Mg/Fe

Inte

nsity

rat

io

Inte

nsity

rat

io

(Mg 383.82 nm /Fe 383.63 nm)

(c)

y = 79.56x + 0.0862

R2 = 0.9062

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 0.001 0.002 0.003 0.004 0.005 0.006Weight ratio Fe/Al

(Fe 297.344 nm /Al 305.468 nm)

(d)

Weight ratio Si/Al

Inte

nsity

rat

io

2024

70752011

2017

6063

6061

6262y = 66.892x – 0.0121

R2 = 0.9227

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0 0.001 0.002 0.003 0.004

(Si 72 /Al 901)

(e)

0

2

4

6

8

10

0 10 20 30 40 50

Weight ratio (Cu/Fe)

Inte

nsity

rat

io (

Cu/

Fe)

Cu 327.396 /Fe 344.06

0

2

4

6

8

0 5 10 15 20

Weight ratio (Cu /Fe)In

tens

ity r

atio

(Cu

/Fe)

202

2017

20117075

6061626

6063

(a)

0 1 2 3 4 5 6 7

Weight ratio (Mn/Fe)

Mn 404.136 /Fe 406.39

0

1

2

3

4

5

6

2024

2017

6066262

0

0.4

0.8

0 0.1 0.2 0.3Weight ratio (Mn/Fe)

60617075

6262201

1

6063

(b)

Fig. 21. (a)–(e): Calibration curves of the molten Al alloys. (Reproduced with permission fromA. K. Rai et al. [7]).

of the different atomic lines of the same element are reproducible. The calibration curvefor the Cu 327.396 nm line is linear up to a concentration of 3.8 wt% (Fig. 21a) but forlarger concentrations exhibits curvature. The nonlinear behavior in the calibration curveof Cu is believed to be due to self absorption as was already noticed in our previouswork [31] on solid samples. In contrast to the solid sample [31], the calibration curvefor Mn 404.36 nm is also showing curvature after 0.54 wt% (Fig. 21b). Calibration curvefor Mg 383.82 nm is, however, a straight-line (Fig. 21c).

To obtain the absolute concentration of analyte elements from the intensity ratio-basedcalibration we need to know the concentration of the reference element (i.e. Fe in thepresent case). The concentration of Fe may be obtained from the ratio calibration curveof Fe and the major element Al. The large intensity and interference-free location of Alline at 305.468 nm makes it possible to obtain the calibration curve for Fe 297.334 nm/Al305.468 nm as a straight-line (Fig. 21d). Since Al is the major constituent, by using thiscurve one can get the concentration of Fe and the concentrations of other elements may




be calculated from their ratio calibration curves with Fe. As Si (288.158 nm) line is alsoavailable in the region of Al 305.468 nm line, we have obtained the calibration curve forSi 288.158 nm/Al 305.468 nm, which again shows a linear variation with concentration(Fig. 21e).

Calibration curves obtained in the present work clearly demonstrate that the exper-imental set up with FO LIBS probe is quite suitable to monitor the concentration ofminor elements in the molten Al alloy in situ (furnace).

3.3. Comparison of LIBS Spectra of Molten and Solid Alloy Samples

To compare the LIBS spectra of molten alloy with that of the of solid sample, LIBSspectra of solid sample were also recorded with the same FO LIBS probe. By comparingthe melting points of Al, Cu, Cr, Mg, Mn, Si, Fe, and Zn, one can broadly divide theseelements into two categories: one having a melting point between 500 to 1000�C andthe other having a melting point between 1000 to 2000�C (Table 4). Al, Cu, Zn and Mgcome in the first category, whereas the rest of the elements fall in the second category.

Comparing the LIBS signal of these elements in molten and solid phase, we have thefollowing observations:

(i) Ratio of the line intensity of Al/Fe, Cu/Fe (Fig. 22a), and Mg/Fe (Fig. 22b) inmolten phase was found quite small in comparison to its value in solid phase(Table 5). These observations may be understood by considering the fact that theconcentration of elements having a lower melting point is higher above the meltsurface. Therefore, the intensity of the line of an element, having lower meltingpoint and having higher concentration, would often decrease in the molten phasedue to self-absorption. The observation already noted earlier strengthens the aboveexplanation that the spectra in melt show that the Cu, Al, and Mg lines havesuffered self-absorption. Fig. 23 shows LIBS spectra of the solid and of the meltin the spectral region of 380 nm where Mg line in the melt is found to be broaderthan in the solid. This observation indicates that the concentration of Mg is higherabove the melt surface due to its lower melting temperature.

Table 4. Melting points of analyte elements

Sample Temperature ��C� for vaporpressure of 1 Torr

Melting point ��C�

Al 1557 660Cu 1617 1084Cr 1737 1857Mg 605 649Mn 1217 1244Ni 1907 1453Si 2057 1410Fe 1857 1535




0

2

4

6

8

10

12

14

16

18

20(a) (b)

0 5 10 15 20

Weight %

Inte

nsity

rat

io

Inte

nsity

rat

io (

Mg2

/Fe4

)Solid with Fiber Cu (327.40 nm)/Fe (344.06 nm)

Molten Cu (327.40 nm)/Fe (344.06 nm)

0

100

200

300

400

500

0 5 10 15 20 25

Weight %

With Fiber Mg 688/Fe 938

Molten Mg 690/Fe 939

Fig. 22. (a) Comparison of intensity ratio (Cu/Fe) vs concentration ratio in LIBS spectra of solidand melt. (Reproduced with permission from A. K. Rai et al. [7]; (b) Comparison of intensity ratio(Mg/Fe) vs concentration ratio in LIBS spectra of solid and melt. (Reproduced with permissionfrom Ref. [6]).

Table 5. Intensity ratio of analyte lines of solid and molten aluminum alloy in the LIBS spectra

Ratio Mn/Fe Cu/Fe Mg1/Fe Mn2/Fe Fe/Al Cr/Al Si/Fe Cr/FeSample

6063 – – – ∗25�4198 0�1471 0�0244 2�2398 0�122045∗0�1199 ∗0�9951 ∗16�5333 – ∗0�1722 ∗0�0585 ∗1�460426 ∗0�422136

2011 0�1331 12�3104 0�0535 0�1087 0�3010 0�0275 0�3042 0�07999∗0�1846 ∗4�6166 ∗5�3482 ∗7�1848 ∗0�3623 ∗0�0310 ∗0�25966 ∗0�172954

2017 3�2945 15�5868 20�0832 36�6633 0�2026 0�0523 1�118637 0�304452∗3�3769 ∗6�2498 ∗6�8436 ∗10�0633 ∗0�3068 ∗0�2420 ∗0�589848 ∗0�805809

2024 5�4129 76�8489 186�3045 323�9095 0�0879 0�0146 0�515252 0�163362∗4�0906 ∗7�2346 ∗33�8356 ∗53�1594 ∗0�1629 ∗0�2099 ∗0�179353 ∗0�12970

6262 0�0763 3�7796 16�1861 26�3214 0�3862 0�1414 1�343422 0�369332∗0�1614 ∗1�3413 ∗9�8259 ∗15�1845 ∗0�4929 ∗0�2821 ∗0�839196 ∗0�816889

7075 0�3623 18�8995 175�4791 283�6476 0�1035 0�3152 0�813926 3�02612∗0�3749 ∗3�8857 ∗32�4723 ∗53�7989 ∗0�2335 ∗0�4872 ∗0�428902 ∗2�557745

6061 – – – – 0�26079 0�1495 2�015179 0�584828∗0�5006 ∗1�1343 ∗4�8786 ∗7�3426 ∗0�4184 ∗0�5094 ∗0�640486 ∗1�143162

∗ molten phase

(ii) In contrast to the above observation, the ratio of Mg/Fe (Fig. 22b) in the moltenphase was found to be larger than in the solid phase for one of the samples 2011(Table 5), whereas the concentration of Mg in this sample is very small (0.09%)in comparison to other samples (Table 1). The actual concentration of Mg at thesurface of the melt becomes larger than in the melt due to the lower melting pointof Mg. Since the effect of self-absorption is small for the low Mg concentrationsin sample 2011, the intensity ratio of Mg/Fe in this sample is larger in the moltenphase than in the solid phase.




0

0.2

0.4

0.6

Inte

nsity

0.8

1

1.2

380 381 382 383 384 385 386

Wavelength (nm)

SolidMelt

Fig. 23. Comparison of the LIBS spectra of an Al alloy recorded in solid and molten phases.(Reproduced with permission from A. K. Rai et al. [7]).

0

0.1

0.2

0.3

0.4

0.5

0.6

0 0.0005 0.001 0.0015 0.002 0.0025

Weight %

Inte

nsity

rat

io (

Cr/

Al)

Solid with Fiber Cr (301.756 nm)/Al (305.468 nm)Molten Cr (301.756 nm)/Al (305.468 nm)

Fig. 24. Comparison of intensity ratio (Cr/Al) vs concentration ratio in LIBS spectra of solid andmelt. (Reproduced with permission from A. K. Rai et al. [7]).

(iii) Ratio of Cr/Fe and Cr/Al (Fig. 24 and Table 5) for molten phase were found tobe larger than in the solid phase. This is due to the fact that Al and Fe have asmaller melting point in comparison to Cr, therefore, the line intensity of Al andFe is reduced in molten phase due to self-absorption.

(iv) The assumption that the intensity of the element having a lower melting pointdecreases in molten phase gets further support from our experimental ratio forSi/Fe. Ratio Si/Fe in the molten phase is quite small in comparison to its value inthe solid phase. Melting point of Si is small in comparison to that for Fe and theline intensity for Si in the molten phase is reduced by self absorption yielding asmall intensity ratio for Si/Fe.

(v) The comparisons of LIBS spectra from melt and solid samples show that theintensity ratios from melt data strongly depend on the concentration levels andmelting temperature of the analyte elements. The selective vaporization in melts




causes the elemental concentration in laser plasma to be different from the trueelemental composition in the melt. Therefore, the intensity of the analyte lines inthe molten phase is quite different from the intensity in the solid phase.

In conclusion, these results reveal that the FO LIBS probe is quite suitable for real-time in-situ monitoring of minor elements (Cr, Cu, Mg, Mn, Si, Zn etc) present in moltenAl alloys in a laboratory furnace. These results also show that Ar atmosphere yields ahigher line-to-background ratio and a higher line-to-noise ratio, and hence it is moresuitable for molten alloy measurements. Experimental observations clearly demonstratethat the melting points of elements can also affect the calibration curve for the moltenalloy. So, we cannot use the calibration curve obtained from studies on solid alloy to thecase of the molten alloy. To obtain accurate elemental concentrations in melt, one needscalibration data, properly obtained for the melt. The calibration curve for an elementhaving a lower melting temperature should cover a wider concentration range.

4. FO-LIBS PROBE FOR ALUMINUM ALLOY IN INDUSTRIALPILOT FURNACE

The results obtained in the laboratory furnace demonstrated the suitability of the presentFO LIBS probe to monitor the concentration of minor elements in the molten Al alloyeven in an industrial setting. The distance between the collimating lens (for the laserbeam coming from the fiber) and the focusing lens (to focus the laser beam in the moltenAl alloy) in the LIBS probe used in the laboratory furnace was nearly 75 cm. Use ofthe same FO LIBS probe for the measurement of elemental composition of the Al alloyin an industrial pilot furnace, would require a part of the fiber to be inside the furnacemaking the fiber, especially the fiber connector vulnerable to damage due to the high�800�C� temperature. To ensure that the fiber remains wholly outside the furnace, a smallmodification in the design of the FO LIBS probe was effected. Keeping in view the largevolume of the industrial furnace, the distance between the collimating lens and the focus-ing lens should be nearly 200 cm. Stainless steel (s.s.) holders were constructed whichcan house the collimating and the focusing lens without disturbing the optical alignmenteven at high temperature of about 800�C [38]. This holder protects the fiber and the col-limating and focusing lenses from damage when the probe is inserted inside the furnaceinto molten material. This part of the probe is shown in Fig. 25. It is constructed fromsix different pieces of stainless steel tubes each having an internal diameter �.2.2 cm,an outer diameter of �3.0 cm, and a length of �30 cm. These holders are connected toone another with the help of fine male and female threads as shown in Fig. 25.

At the top of the holder a provision for swagelok connection is made for the inletflow of the purge �N2� gas. As shown in Fig. 25 an aluminum flange of outer diameter�15 cm and internal diameter �3.1 cm is connected to the s.s. holder with lock screwand Teflon system. With the help of this flange the s.s. holder is tightened in the s.s. pipewhich is described in Ref. [38]. A provision for the insertion of a thermocouple, whichmeasures the temperature at the bottom end of this holder, is made in the aluminumflange. A flow meter is connected which controls the gas flow to cool the whole lensholder. At the top of stainless steel holder an aluminum holder having the same internaland outer diameters is connected using male and female threads. This aluminum holder




Opticalfiber

Optical fiberconnector

Collimating lens

Swagelok

O ring

Purge gas

Lock screw

Thermocouple

O ring

Male thread

Male thread

Female thread

Focusing lens

Snap ring

Aluminum holder

Teflon

Lockscrew

Aluminum flange

O ring

N2 gas out

N2 gas out

Fig. 25. Stainless steel holder for collimating and the focusing lens in the pilot furnace. (Repro-duced with permission from A. K. Rai et al. [6]).

houses the collimating lens with the help of a spiral lock ring. At the top of the aluminumholder a provision is made to connect the optical fiber through SMA 905 stainlesssteel fiber connectors. The aluminum holder also has a provision for a rotating ring,which permits fine adjustment of the distance between output end of the fiber and thecollimating lens. With the help of this adjustment procedure the laser beam passingthrough the stainless steel holder is collimated without any change of the circular spotof the laser beam. The bottom part of the holder houses the focusing lens. A circularcave is cut in the internal wall at the bottom end, to hold the internal snap ring (SAE1060–1090 steel). This snap ring prevents the movement of the focusing lens duringthe experiment. Provision is made by the side of the lens and the snap ring for the outflow of the purge (nitrogen) gas, which enters through the upper portion of this holder.This flow of purge gas helps to keep the lens and snap ring cool and also prevents thealuminum melt to reach the lens surface. In the present sensor we are able to adjust




the distance between the output end of the fiber and the collimating lens so that thecollimating beam illuminates the focusing lens without any loss of intensity of the laserbeam. Finally the stainless steel holder was kept inside a stainless steel pipe to preventits direct contact with the Al melt.

4.1. Testing the Long Stainless Steel Probe

First, the alignment of the laser beam was tested by checking the image of the laserbeam outside the stainless steel probe. By properly adjusting the lens holder with respectto the 180 cm pipe, one is able to get the proper circular spot of the laser beam, outsidethe probe. A special cap, which closes the bottom end of the probe, was designed toensure that the focal point of the laser beam remains at the periphery of the probe. An Alrod attached to the cap was kept in the center of the probe, and its tips remained at theperiphery of the probe. By adjusting the length of the spacer, we focused the laser on thetip of the Al rod. LIBS spectra of this Al rod were recorded to check the performance ofthe probe and after the successful test performance, we inserted the probe in the furnacemelt. After properly adjusting the inlet and outlet gas flow, we were able to record theLIBS spectra of molten Al alloy.

4.2. LIBS Measurements inside the Industrial Pilot Furnace

The LIBS assembly (which includes the FO LIBS probe, spectrometer, laser, computer,etc.) was packed and taken for field measurement. After testing the optical alignment,proper connections for water cooling and N2 gas flow were made and the probe wasslowly inserted into the pilot furnace containing the molten alloy. By adjusting the depthof insertion of the probe and the flow of gas in the inlet, one is able to get LIBS spectraof the molten alloy with good S/N ratio. During the experiment, it was observed that the

0

1

2

3

4

5

6

7

8

0 0.2 0.4 0.6 0.8

Cr/Fe Weight ratio

Inte

nsity

rat

io

Cr 359.35 nm /Fe 364.78 nm

Fig. 26. Variation of the line intensity ratio vs. the concentration ratio in the LIBS spectra ofthe molten Al alloy during metal feed tests in a pilot furnace. (Reproduced with permission fromA. K. Rai et al. [6]).




Zn

330.

23

Zn

334.

55

Ni 3

41.4

4

Fe

344.

05

Cu

327.

4

0

10000

20000

30000

40000

50000

60000

70000

80000

90000

326 328 330 332 334 336 338 340 342 344 346

Wavelength (nm)

Td = 0.5 us; Tw = 5 us

Inte

nsity

0

50000

100000

150000

200000

250000

300000

326 328 330 332 334Z

n 33

0.23

Zn

334.

55

Ni 3

41.4

4 Fe

344.

05

Cu

327.

4

336 338 340 342 344 346Wavelength (nm)

Td = 0.3 us; Tw = 5 us

Inte

nsity

Zn

330.

23

Zn

334.

55

Ni 3

41.4

4

Fe

344.

05

Cu

327.

4

0

10000

20000

30000

40000

50000

60000

70000

80000

90000

326 328 330 332 334 336 338 340 342 344 346

Wavelength (nm)

Inte

nsity

Td = 1.0 us; Tw = 5 us

Fig. 27. LIBS spectra recorded with different gate delay time. (Reproduced with permission fromA. K. Rai et al. [6]).

signals are quite sensitive to the depth of this probe in molten alloy, as well as to theflow rate of the purging gas.

The data were taken by varying the experimental parameters for the best signal-to-noise ratio. Fig. 26 shows the variation in the intensity of Cr emission line for thethree different tests which shows a significant increase in the line intensity of the seededmetal. Fig. 27 shows the LIBS spectra recorded in the spectral region ≈336 nm withthree different gate delay times. It is clearly seen that the intensities of Fe and Zn linesare enhanced at shorter gate-delay times, as compared to that of Cu 327.4 nm line which




may suffer self-absorption at the shorter delay times. After successfully recording thespectra of the original melt in all spectral regions of interest, known amounts of Cr,Mn, Mg, and Cu metals were added into the melt. After waiting for one hour to letthe metals get mixed in the Al alloy, the LIBS probe was again inserted into the meltto detect the change in the melt concentration. These observations demonstrate that thepresent FO LIBS sensor would be useful for on-line, in-situ monitoring of minor metalconcentrations in pilot furnaces.

5. CONCLUSIONS

In conclusion, this chapter describes a complete optical fiber (OF) LIBS system thatwas developed for on-line, in-situ elemental composition measurements of solid andmolten samples. The critical issues associated with the coupling of the pulsed laser beamwith an optical fiber are described. The parametric study was performed to optimize theperformance of the OF-LIBS system. Application of the OF LIBS system to solid andmolten aluminum measurements has been demonstrated and the test results show thatOF-LIBS system can be used for on-line process monitoring and control of industrialfurnaces.

ACKNOWLEDGMENTS

This work was supported by the U. S. Department of Energy, Office of Industrial Tech-nology (OIT) grant number DE-SC02-99 CH-10974, through a subcontract from theenergy Research Company, and DOE Cooperative Agreement DE-FC 26–98 FT-40395.We are also thankful to Shiwani Pandhija, Junior Research Fellow for help in the prepa-ration of the manuscript. During preparation of the manuscript, the financial assistancefrom DRDO project (No ERIP/ER/04303481/M/01/787) is fully acknowledged.

REFERENCES

[1] M. Sabsabi and P. Cielo, Appl. Spectrosc., 49 (1995) 499.[2] D. E. Kim, K. J. Joo, H. K. Park, K. J. Oh, and D. W. Kim, Appl. Spectrosc. 51 (1997) 22.[3] R. Q. Auccolio, B. C. Castle, B. W. Smith and J. D. Winefordner, Appl. Spectrosc. 54

(2000) 832.[4] F. Y. Yueh, J. P. Singh and H. Zhang, Encyclopedia of Analytical Chemistry, Wiley, New

York (2000).[5] C. F. Su, S. Fang, J. P. Singh, F. Y. Yueh, J. T. Rigsby III, D. L. Monts, and R. L. Cook,

Glass Technol. 41 (2000) 16.[6] A. K. Rai, F. Y. Yueh and J. P. Singh, Rev. Sci. Instrum., 73 (2002) 3589.[7] A. K. Rai, F. Y. Yueh and J. P. Singh, Appl. Opt., 42 (2003) 2078.[8] V. Sturm, L. Peter and R. Noll, Appl. Spectrosc, 54 (2000) 1275.[9] D. A. Rusak, B. C. Castle, B. W. Smith and J. D. Winefordner, Anal. Chem., 27 (1997) 257.

[10] A. K. Rai, V. N. Rai, F. Y. Yueh and J. P. Singh, Trends in Appl. Spectrosc., 4 (2002) 165.[11] J. M. Gumba, C. D. Angelo, D. Bertuccelli, and G. Bertuccelli, Spectrochim. Acta B56

(2001) 695.




[12] M. Hemmerlin, R. Meil, R. Hazlett, J. Martin, T. Pearee, and A. Zigler, Spectrochim. ActaB56 (2001) 707.

[13] B. L. Drogoff, J. Margot, M. Chaker, M. Sabsabi, O. Barthelemy, T. W. Johnston, S. Laville,F. Vidal, and Y. V. Kaenel, Spectrochim. Acta B56 (2001) 987.

[14] S. Rosenwasser, G. Asimellis, B. Bromley, J. D. Caceres, and A. Gonzales Urena, Spec-trochim. Acta B56 (2001) 865.

[15] O. Samek, D. C. S. Beddows, H. H. Tella, J. Kaiser, M. Liska land, H. Falk, P. Wintjens,and L. Paulard, Spectrochim. Acta B56 (2001) 661.

[16] A. Uhl, K. Loebe, and L. Kreuchwig, Spectrochim. Acta B56 (2001) 795.[17] G. Zikratov, R. Vasudev, F. Y. Yueh, J. P. Singh, and J. C. Mara Glass Technol, 40 (1999) 84.[18] A. V. Pakhomov, W. Nichols & J. Borysow Appl Spectrosc, 50(1996) 880.[19] I. Gobernado-Mitre, A. C. Prietro, V. Zafiropoulos, Y. Apetsidou, and C. Futakis, Appl.

Spectrosc., 51 (1997) 1125.[20] R. Salimbeni, R. Pini, S. Siano, Spectrochim. Acta B56 (2001) 877.[21] R. Krasniker, V. Bulatov, and I. T. R. Lorce, Appl. Spectrosc., 38 (1984) 721.[22] V. Majidi and M. R. Joseph, Spectroscopic Sehechter, Spectrochim. Acta B56 (2001) 609.[23] D. A. Cremers, L. J. Radziemski, Applications of Laser Induce Plasma, Critical Reviews in

Analytical Chemistry, 23 (1992) 143.[24] K. Y. Yamamoto, D. A. Cremers, M. J. Ferris and L. E. Foster, Appl. Spectrosc, 50

(1996) 222.[25] R. Barbini, F. Colao, R. Fantoni, A. Palucci, S. Ribezzo, H. J. L. Van der Steen and

M. Angelone, Appl. Phys., 65 (1997) 101.[26] B. J. Marquardt, D. N. Stratis, D. A. Cremers and S. M. Angel, Appl. Spectrosc.,

52 (1998)1148.[27] R. E. Neuhauser, U. Panne and R. Niessner, Analytical Chemica Acta 392 (1999) 47.[28] R. E Neuhauser, U. Panne and R. Niessner, Appl Spectrosc, 54 (2000) 923.[29] D. A. Cremers, J. E. Barefield II and A. C. Koskelo, Appl. Spectrosc. 49 (1995) 857.[30] A. I. Whitehouse, J. Young, I. M. Botheroyd, S. Lawson, C. P. Evans and J. Wright,

Spectrochim. Acta B56 (2001) 821.[31] A. K. Rai, H. Zhang, F. Y. Yueh, J. P. Singh and A. Wiseburg, Spectrochim. Acta B56

(2001) 2371.[32] H. R. Griem, Plasma Spectroscopy, McGraw Hill, New York (1964).[33] R. A. Multari, L. E. Foster, D. H. Cremers and M. J. Ferris, Appl. Spectrosc. 50 (1996) 1483.[34] L. Paksy, B. Nemet, A. Lengyel, L. Kozma and J. Czevkel, Spectrochim. Acta B51

(1996) 279.[35] J. Gruber, J. Heitz, H. Strasse, D. Bauerle and N. Ramaseder, Spectrochim. Acta B56

(2001) 685.[36] R. Noll, H. Bette, A. Brysh, M. Kraushaar, I. Monch, L. Peter and V. Sturm, Spectrochim.

Acta B56 (2001) 637.[37] M. Kuzuya, H. Matsumoto, H. Takechi and O. Milkami, Appl. Spectrosc, 47 (1993) 1659.[38] H. Zhang, A. K. Rai, J. P. Singh and F. Y. Yueh, Fiber optic laser-induced breakdown

spectroscopy probe for molten material analysis. Patent No. 6762835 (2004).



Chapter 12

LIBS Technique for Powder Materials

Bansi Lala, L. St-Ongeb, Fang-Yu Yuehc and Jagdish P. Singhc

aCentre for Laser Technology, Indian Institute of Technology Kanpur, Kanpur 208016, INDIAbNational Research Council Canada, Industrial Materials Institute, 75 de Mortagne Blvd,Boucherville, Québec J4B 6Y4, CANADA

cInstitute for Clean Energy Technology, Mississippi State University, Starkville,Mississippi 39759, USA

1. INTRODUCTION

Powder materials both granular as well as fine powder represent the most commonform of raw material in the industry world-wide. Industries like chemical, pharmaceu-tical, glass, ceramic, food, mining, metallurgy, construction and many others use thepowder material continuously in their applications and processes. Most of the time thepowder material used in an industrial application is a mixture of various pure chemi-cals and the quality of the end-product invariably depends on the composition of themixture being used necessitating the on-line/in-situ monitoring of the elemental com-position of the powder material before it is fed into a process. This on-line/in-situmonitoring of elemental composition can be also helpful in resolving the environmen-tal issues by identifying the pollutants before starting of the process through whichthe powder material has to undergo. A large number of analytical techniques like wetchemistry, infrared/visible/ultraviolet absorption/fluorescence spectrometry, light scat-tering, chromatography, continuous/pulsed NMR, mass spectrometry and X-ray diffrac-tion/fluorescence can be used to monitor the elemental composition of the powdermaterial. State of the art instruments based on these techniques are available commer-cially with enough speed and sophistication of data collection required to meet the everincreasing demand for higher sensitivity, selectivity, precision, accuracy and number ofsamples to be processed. However, almost all these techniques need sample preparationand most of the operating costs and work activity are spent in sample preparation forinjection into a measurement device. The operating costs are further escalated due towaste storage, segregation and disposal of the chemicals/solvents used for sample prepa-ration. Hence, there is need for a better on- line/in-situ analytical technique which doesnot need any sample preparation; products from the various stages of an assembly linecan be directly checked. Laser induced breakdown spectroscopy (LIBS) is almost anideal technique for such applications as it needs minimal sample preparation, results are




288 B. Lal et al.

obtained in a few seconds (results in principle are available with a single laser pulse) andseveral elements can be monitored simultaneously. Also, the data collection techniquesof LIBS are sophisticated enough to be used for the control of the process in whichpowder material is used. No doubt the speed and the data collection sophistication ofLIBS is comparable/even better than the techniques listed above, yet the precision ofLIBS in general and more-so with powder materials, is less than the other availableanalytical techniques. This chapter summarizes the optimization of various experimentalconditions of the LIBS technique so as to obtain the best possible reproducible datain case of powder materials. Also, application of LIBS to pharmaceutical and glassindustry are discussed as both these industries use raw material in the form of powdersextensively.

2. LIBS TECHNIQUE FOR POWDER MATERIALS

In LIBS technique a high energy laser pulse when focused on the sample of interestresults in the formation of micro-plasma, characteristic of the sample composition. Theemission from this micro-plasma is analyzed for the quantitative determination of theelemental composition of the sample. This micro-plasma formation is accompanied bythe generation of a high-pressure absorption wave (shock wave) whose propagation isa function of the incident laser energy [1]. This shock wave is the cause of highlyinaccurate data in the case of powder samples because:

1. the surface of the powder sample is disturbed so that the focal spot of the laser isnot same for the subsequent laser pulses and

2. powder is ejected out due to shock wave (aerosol production) so that a varyingpart of laser pulse is absorbed in front of the sample.

Both these factors cause pulse-to-pulse fluctuations in the irradiance level resulting inthe poor reproducibility of the LIBS data. Wisburn et al [2] investigated the quantity ofheavy metals in soils, sand and sewage sludge using powder samples. The LIBS datathey obtained has a relative standard deviation (RSTD = 100×standard deviation/mean)of about 25% which they explained in terms of persistent aerosols and relatively biggercrater formation. Pascal et al. [3] used an echelle spectrometer based portable LIBSinstrument for the analysis of powder soil samples but could not quantitatively interprettheir results. Similar observations have been reported by Lal et al. [4] while applyingLIBS for the determination of the elemental composition of glass batch. On the otherhand, dramatic change in the reproducibility of the LIBS data by using pellets insteadof powder samples has been reported by several workers. Martin et al. [5] employedLIBS to determine the concentration of carbon and nitrogen in a variety of soil samplesin pellet form. Rosenwasser et al. [6] used LIBS for quantitative analysis of phosphateores and they reported RSTD of about 3.79% with pellet samples. Krasniker et al. [7]used polyvinyl alcohol as binder to prepare soil and sand pellets for the investigation ofmatrix effects in LIBS. Lal et al. [8] have investigated the effect of various parameterson the accuracy of the LIBS data recorded with pellet samples. The various experimentalparameters which are to be optimized so as to obtain reproducible data for powdermaterial are discussed in the following sections.



LIBS Technique for Powder Materials 289

2.1. Preparation of the Pellets of the Powder Samples

The dramatic improvement, as discussed earlier, in the RSTD of LIBS data when thepellets of the powder materials are used has prompted the investigation of the effects ofvarious factors involved in pellet making on the reproducibility of LIBS data. Typicallyfinely ground powder material after thorough mixing with a binder is pressed into apellet by a die-hydraulic press combination. The degree of roughness of the powdermaterial, nature and amount of the binder, pressure used to pelletize the powder andheat-treatment of the pellets are the experimental parameters affecting the RSTD of theLIBS data. The conclusions of the various studies are summarized as follows:

(i) The RSTD of LIBS data improves by finer grounding of the powder material [6].(ii) Improvement in the RSTD of the LIBS data on drying the pellet for about 15

minutes at about 90 �C has been reported [8] in the literature. LIBS data collectedfrom one side of the freshly prepared pellet (0.8ml of 2 wt% PVA with 5g oflime) has RSTD of about 17%. After drying the pellet the RSTD reduces to about5% when data is collected from the other side of the pellet. Similar observationshave been reported for mineral ores [6].

(iii) Variation of the RSTD of the LIBS data with the amount and nature of the binderhas been investigated in detail in Ref. 8. In this study three types of binders(polyvinyl alcohol, sucrose and starch) have been investigated using industriallyimportant powder materials namely, silica, alumina and lime. The variation inthe intensity of the Ca 395.7 nm spectral line as a function of the amount ofthe polyvinyl alcohol (PVA) is shown in Fig. 1. The relative standard deviation(RSTD) of the same emission spectral line is also plotted as a function of the PVAamount added to the powder as binder. As shown in Fig. 1, the emission intensityobserved in case of pellet with 0.8ml binder is more than 1.5 times of that of

2.3

4.3

6.3

8.3

10.3

12.3

14.3

16.3

18.3

20.3

22.3

Amount of PVA in ml

105 ×

inte

nsity

(AU

) of

Ca

395

nm

00 0.2 0.4 0.6 0.8 1

0.5

1

1.5

2

2.5

3

3.5

4

4.5

RS

TD

of t

he in

tens

ity o

f Ca

395

nm

RSTDIntensity

Fig. 1. The intensity of Ca 395.7 nm line and the relative standard deviation of the data as afunction of the amount of polyvinyl alcohol (PVA) added as a binder to 5g of lime. 24 MPa hasbeen used to make the pellets.



290 B. Lal et al.

the pellet with no binder. On the other hand, there is decrease in RSTD with theincrease in the amount of binder used to make the pellets. The RSTD of the Ca395.7 nm emission line is about 21% when pellets of the lime prepared with nobinder are used to record LIBS spectra. It is about 5% for the same emissionline when the pellets are made from 5g of lime mixed thoroughly with 0.8ml ofPVA. Any further increase in the amount of binder used to make a pellet doesnot decrease RSTD in any significant manner.

Similar results have been obtained with silica and alumina powders with PVA (2 wt%)as binder. However, alumina and silica, unlike lime, cannot be pressed into pelletswithout binder. For alumina the minimum amount of PVA binder required to press 5g ofpowder into a pellet is 0.2ml and the LIBS data recorded with such a pellet has RSTD ofabout 47% in the emission line of Al 394.4 nm. The RSTD for the same line decreasesto about 5% when LIBS data is recorded from a pellet made from 5g of alumina powdermixed with 0.8ml of PVA. As is the case with lime, no significant decrease in RSTDis observed on further increasing the amount of PVA. On the other hand, the minimumamount of PVA needed to press 5g of silica in a pellet is 0.4ml. The RSTD of Si390.5 nm emission line recorded from this pellet is as high as 67% which reduces toabout 6.5% when a pellet made with 0.8ml PVA is used. Again no significant decreasein RSTD is observed when pellets with PVA more than 0.8ml are used to record LIBSspectrum. In this study 2% by weight of PVA (CAS#9002-89-5, Alfa Aesar) dissolvedin distilled water has been used. No significant decrease in RSTD or increase in emissionintensity has been observed by further increasing �>2%� the concentration of PVA.

The general trend of increase in the intensity of emission lines and decrease in theRSTD of the data with the increase in the amount of binder up to certain limit beyondwhich there is no significant increase (decrease) in intensity (RSTD) has been observedfor sucrose (CAS# 57-50-1, Alfa Aesar) as well as starch (CAS # 9005-84-9, AlfaAesar). However, in the case of sucrose (2% wt concentration) the minimum RSTD ofCa 395.7nm emission line observed with 0.8ml added to 5g of lime powder is about 6.5%which does not change significantly by further increase in the amount of sucrose added tomake pellets. In the case of alumina the minimum RSTD observed in Al 394.4nm spectralline is about 7% while for Si 390.5nm (from silica) it is about 9%. Both these figures donot change significantly by further increasing the concentration of sucrose from 2 wt%.

The general observations about the role of the amount/nature of binder used in pelletmaking are:

(a) The LIBS spectra of the pellets made with 2 wt% solution of PVA in distill waterhas the lowest RSTD.

(b) The lowest RSTD is observed when an optimum amount of PVA is added asbinder. This optimum amount depends on the nature of the powder material andhas to be found-out experimentally.

The reduction in RSTD is mainly because the pellets are more rigid than the powderso that the position of the focal spot is almost unchanged for all the laser pulsesarriving at the pellet surface during a data acquisition cycle By increasing the amountof binder, pellets with higher rigidity are obtained. The LIBS data acquired from suchpellets is more precise. This dependence of LIBS data accuracy on pellet rigidity is




0

1

2

3

4

5

6

7

8

9

10

1250 1750 2250 2750 3250 3750

Pressure (lb/inch2)

RS

TD

of t

he in

tens

ity o

f Ca

395

nm

Fig. 2. Variation of the RSTD of the intensity of Ca 395.7nm with the amount of pressure usedto make pellets.

further illustrated in Fig. 2 where RSTD of the acquired data has been plotted as afunction of the pressure used to make a pellet. The RSTD changes from about 8.5%to about 5% when the pressure used for the pellet-making is changed from 10MPa toabout 24MPa.

2.2. Apparatus

The experimental setup used for recording the LIBS spectra of pellet samples is similarto the one used for solid material. A typical setup is shown in Fig. 3. Laser pulses

Harmonicseparator

Beamdump

Nd: YAG LASER

Optical fiber Dichroicmirror FL

(300 mm)

Prism

Spectrograph IDAD or ICCD

Controller PC

Pulse generator

Rotary-translation stagemounted sample holder

LL

Fig. 3. Schematic diagram of the apparatus for recording the LIBS spectra.



292 B. Lal et al.

generally from a frequency doubled Nd: YAG laser are focused on the sample usinga fused silica convex lens of proper f-number. This focusing lens can also be used tocollect the plasma emission which is fed to a spectrometer, generally through a UVgrade fiber. The spectrograph could be a Czerny-Turner 0.5m spectrometer fitted witha gated intensified diode array detector (IDAD) or an echelle spectrometer having agated intensified charge coupled device (ICCD) as optical detector. Both gate delayand gate width are controlled by a pulse generator which is synchronized with thelaser. A number of PC software are available for data acquisition and processing. WithCzerny-Turner spectrometer, a spectral region covering about 20nm is recorded in asingle run while with echelle spectrometer LIBS spectra in a broader region (typically200–800nm) can be recorded simultaneously. The pellet samples are mounted on arotating platform.

2.3. Position of the Focal Spot

Studies have shown [8] that the RSTD of the LIBS data depends on the position of thefocal spot on the pellet. The dependence of the intensity of 395.7 nm emission spectralline of Ca from a lime pellet, on the position of the focal spot on the pellet is shownin Fig. 4 alongwith the variation of the RSTD of the same data. The “0” position onthe x-axis of both these figures corresponds to the position of focal spot on the surfaceof the pellet, +ve value to a focal spot above the surface while −ve value correspondsto a focal spot inside the surface of the pellet. These figures show that the intensity ofthe emission line is maximum for the “0” position of the focal spot while RSTD of the

0.0–8 –6 –4 –2 0 2 4 6 8

0.5

1.0

1.5

2.0

2.5

3.0

3.5

Distance from focal point (mm)

RS

TD

of t

he in

tens

ity o

f Ca

395

nm

105 ×

Inte

nsity

(A

U)

of C

a 39

5 nm

0

2

4

6

8

10

12

14

Intensity

RSTD

Fig. 4. Variation of the intensity and RSTD of Ca 395.7nm emission spectral line with the changein the position of the focal spot. The “0” position corresponds to the focal spot on the surfaceof the pellet, +ve value to the position of focal spot above the surface while −ve value to theposition of the focal spot inside the surface of the sample.




LIBS data is minimum �<5%� when the focal spot is ∼1.5mm inside the surface of thepellet. There are two reasons for this observation:

(i) when the position of the laser focal spot is at or slightly above the surface of thepellet, part of the laser energy is used to excite the air in contact with the pelletleading to poor shot to shot reproducibility while with the laser focused slightlyinside the surface, a smaller volume of the air is excited resulting in the reductionof RSTD,

(ii) the shot to shot fluctuations in emission intensity due to non-uniform distributionof the various species in the sample, are less as the area covered by the focusedlaser spot on the sample in this case is more than that of the position when focalspot is on the surface of the pellet.

Similar observations have been reported for steel [9] and steel alloy [10]. The increasein RSTD as the focal spot is moved further inside the sample could be due to variationin the position of the focal spot on the pellet resulting from the material removal by theshock-wave accompanying the breakdown.

2.4. Delay Time

In all LIBS experiments the delay time, time interval between the arrival of the laser pulseon the sample and activation of the detector, is to be optimized so as to (i) minimize thebackground noise due to continuum emission and (ii) maximize the emission intensityof the spectral line of interest. Generally the background noise is negligible after thedelay time of about 0�1�s or less [11] while the intensity of most of the neutral atomspectral lines is maximum in 0�1–2�0 �s range [12,13]. Lal et al. [8] have observed thatthe RSTD is almost constant in 0�1–2 �s range while it increases from 5.5% to 14%when delay time is changed from 2 �s to 4 �s. This increase in RSTD at higher delaytimes is due to decrease in signal to noise ratio resulting from the decrease in emissionintensity with increase in delay time.

2.5. Sample Rotating Speed

The sample needs to be rotated to ensure that laser pulse is not incident on an alreadyexposed sample area to avoid defocusing of the laser pulse. The rotation speed shouldbe such that LIBS data are acquired from all parts of the sample to take care of thenon-uniformity of the sample. The effect of sample rotation on the signal intensity [14]is shown in Fig. 5

The sample (pellet made from lime) was rotating with speed of 1 rotation in 270s forinitial ∼50s of the data acquisition cycle when the rotation is stopped. As seen in Fig. 5there is dramatic change in the emission intensity of Ca336 nm spectral line as soon asthe rotation is stopped. The intensity is almost zero after about 130s. This decrease inthe intensity, as mentioned earlier, is due to defocusing of the laser beam in a stationarysample as the crater produced by the laser pulses on the sample becomes deeper anddeeper with time.



294 B. Lal et al.

00 20 40 60

Time (second)80 100 120 140

200

400

600

800

1000

Inte

nsity

× 1

03 (c

a 33

6 nm

) Ca 336 nm

Stop rotation

Fig. 5. Effect of the sample rotation on the intensity of the emission spectral line of Ca 336 nm.

3. APPLICATION TO PHARMACEUTICAL INDUSTRY

3.1. Introduction

The LIBS technique for powder materials discussed in earlier sections are very relevantto the pharmaceutical industry as it is estimated [15] that about 70% of the marketed drugproducts are in solid dosage forms and a bulk of them are in tablet (pellet) form. Therole of pharmaceutical products in health care is gaining more and more prominence.In accordance with strict governmental regulations, pharmaceutical manufacturers needto ensure the safety, purity, and conformity (correct potency) of their products. Inthis context, analytical technologies play a vital role in the testing of raw materials,formulation development, process optimization, impurity testing, dissolution testing, andproduct release.

Solid dosage forms are small and convenient vehicles for drugs, but the apparent sim-plicity of a tablet often hides a complex formulation (comprising several ingredients) anda complex manufacturing process. The drug, or active pharmaceutical ingredient (API),generally accounts for less than half (and often much less) of the solid dosage form mass,the latter ranging from 50 to 1000 mg [15]. Other ingredients include substances which,although physiologically inactive, nevertheless serve a function in the formulation. Theseinclude the lubricant, which helps the tabletting process, and the disintegrant, which facil-itates the liberation of the drug once the solid dosage form is ingested. The remainder ofthe solid dosage form, and often the main component, is an inactive and non-functionalfiller material, such as cellulose or lactose, or a mixture of the two. Finally, a significantnumber of solid dosage form products possess one or more coatings, which may alsohave several functions (taste masking, opacifying, modulation of drug release).

The very small volume of powder that forms a tablet (a few mm3) in fact comesfrom a much larger �m3� powder blend. Acknowledging the fact that such a blendcomprises several ingredients, one readily sees the challenge inherent in ensuring theconsistency of composition from tablet to tablet. This is especially important in the caseof the API, which is generally required to fall between 95% and 105% of its nominal




concentration. Assuming that LIBS can provide a specific spectral signature for the API,two opportunities appear for the use of LIBS:

(i) as a quality-control tool to verify the consistency of drug content in the finishedtabletted product, or

(ii) as an upstream indicator of uniformity in the powder blend, prior to tabletting.

The latter case is the closest to the use of LIBS described in Section 2, since it generallyrequires that the powder is first pressed into a pellet for presentation to the LIBSinstrument. For a more complete description of the analytical needs of the pharmaceuticalindustry, and of the adequacy of LIBS compared to other analytical tools, the reader isreferred to Ref. [15].

In common with the analysis of inorganic powder, the analysis of pharmaceuticalmaterials involves relatively inhomogeneous samples, compared to, say, metal alloysor liquids. On the 1–100 �m scale, relatively large variations of composition may beencountered from site to site on a solid sample. This dictates certain requirementsin terms of sampling (number of sites and depths analysed) to ensure that a givenanalysis is representative of a given sample, especially if the desired information is thecomposition at the scale of the tablet. In contrast to the analysis of inorganic powders,where elemental analysis is an obvious choice, the analysis of pharmaceutical materialshas traditionally involved techniques that provide information specific to a molecule,such as high-performance liquid chromatography (HPLC) or, more recently, vibrationalspectroscopies. In the following sections it will be shown that, in several cases, LIBS cannevertheless provide valuable compositional information about pharmaceutical materials,especially when spatial resolution is beneficial. Although not as molecule-specific asother analytical techniques, LIBS can often be competitive on the basis of its simplicity,rapidity, and freedom from a sample preparation step. These features of LIBS havespawned many applications in the pharmaceutical field since the late 1990s.

3.2. Analysis of Organic Materials

3.2.1. Element-specific Analysis

When a laser is focused on a solid organic sample, molecular compounds of the targetare vaporized and the ensuing fragments are further dissociated in the accompanyingplasma. Ultimately, these molecular compounds may be identified through emission oftheir elemental components. An unambiguous identification (and quantification) of agiven compound is possible if it possesses a specific element, i.e. an element absentfrom other molecular compounds in the sample. In the case of solid dosage forms,this analytical principle applies to all ingredients of the sample. Examples of specificelements are: chlorine, sulfur, fluorine or bromine as part of the drug; magnesiumas part of the lubricant (i.e. magnesium stearate); sodium as part of the disintegrant(i.e. croscarmellose sodium); or titanium in the form of titanium oxide in the coating.In contrast, carbon, oxygen, hydrogen and nitrogen cannot be considered specific sincethey are generally present in several components of the same formulation, including theinactive matrix (e.g. cellulose, lactose). Of course, this principle also makes LIBS suitable



296 B. Lal et al.

Wavelength (nm)240 250 260 270 280 290

(a)In

tens

ity (

coun

ts)

0

1e+5

2e+5

3e+5

4e+5

5e+5

6e+5

7e+5

C

P

Mg II

Mg I

0 1 2 3 4 65

%API (w/w)

0.00

0.05

0.10

0.15

0.20

P/C

inte

nsity

rat

io

(b)

0.25

0.30

Fig. 6. LIBS spectrum (a) and response curve (b) for a phosphorus-containing drug.

for the analysis of mineral-supplement products and metallic elements (generally presentas oxides) having particularly rich emission spectra. Finally, as with other elementalanalysis techniques (e.g. based on ICP), LIBS also enables the detection of metalliccontaminants (see Sect. 3.4.4).

Fig. 6a provides an example of element-specific analysis of solid dosage forms.In this particular case, phosphorus lines originate solely from a phosphorus-containingAPI, magnesium lines originate solely from magnesium stearate (the lubricant), while acarbon line originates non-specifically from all components of the sample. The capacityof elemental analysis to provide relevant quantitative information about an organiccompound is illustrated in Fig. 6b, which shows the LIBS response curve for phosphorusas a function of the API concentration. In this particular case, it was possible to usecarbon as an internal standard in order to increase analytical precision. The issue ofinternal standardization will be discussed below in greater detail.

The detection limits attainable by LIBS are element-dependent, with order of magni-tude ranging from 0.1 ppm (e.g. Na) to 0.1% (e.g. F). Unfortunately, elements specificto APIs are generally halogens or other non-metals (sulfur), which inherently havehigh excitation energies, leading to low emission efficiencies. A simple approach forenhancing sensitivity to halogens will be discussed in Section 3.3.

3.2.2. Structure-specific Analysis

Up to now, quantification of APIs in pharmaceutical preparations by LIBS has requiredthe API to contain a specific element, as explained above. As a consequence, not allAPIs have been amenable to quantification by LIBS.

As an alternative to elemental analysis, small diatomic fragments, present in theplasma, emit light and can provide valuable information about the chemical structure ofthe target compound. Previous work, notably by Locke et al. for gaseous compounds [16]and by Berman and Wolf for liquids [17], has shown that the emission from C2 pro-vides such structure-specific information. The C2 signal is stronger when double-bondedcarbon is present in the compound. Similarly, work on solid pharmaceutical materialsrevealed a clear relationship between C2 emission and the presence of aromatic rings(containing C==C double bonds) in the starting compound [18].




Inte

nsity

(co

unts

)

5000

Wavelength (nm)504 506 508 510 512 514 516 518 520

10000

15000

20000

MgC2(0,0)

C2(1,1)

C2(2,2)C

10%

0%

N

ClO

O

N

CH

H2

H2CH3

CH3

HO

HO

CCH

C

CC

CH

Fig. 7. LIBS spectra for pellets containing 0% (blank) and 10% chlorpheniramine maleate (CM)in a 50/50 lactose/cellulose matrix, with 0.5% magnesium stearate added. The inset shows thestructure of CM.

Given that most drugs possess an aromatic structure, and that excipients generally donot, C2 emission was recently evaluated for the quantitative analysis of drugs [19]. Pelletscontaining varying concentrations of chlorpheniramine maleate (CM), a model drug,were submitted to LIBS analysis, and a response curve was built. The spectra shownin Fig. 7 include three band heads of the C2 Swan system, as well as three magnesiumlines originating from magnesium stearate. There is some C2 emission from the blank,but the C2 signal is clearly stronger with 10% CM. The C2 signal from the blank samplecontributes a non-zero intercept in the response curve. This is a non-structure-specificsignal that presumably comes from atom-atom recombination in the plasma.

Given that CM also contains a chlorine atom, it has been possible to show that anAPI can be quantified using C2 emission with comparable linearity and sensitivity aswhen using an elemental signal (chlorine in this case). The structure-specific approachwas also extended to commercial formulations of APIs having an aromatic structurebut lacking a specific element, and which could not be analysed by LIBS in the past.Again, a capacity for quantitative analysis was demonstrated. C2 emission was also foundto provide useful information in depth-profile analysis of coated tablets. C2 emissiontherefore enables the determination of compounds that contain double-bonded carbon,and for which a specific element (hetero-atom) is not necessarily present. This extendsthe range of APIs for which LIBS is applicable.

3.3. Experimental Approaches

For similar laser parameters, the ablation of solid dosage-forms results in much greatermaterial removal than the ablation of metals or of other dense, opaque materials. Thelaser radiation generally penetrates much deeper in pharmaceutical materials, resultingin the distribution of laser energy in a larger volume. Furthermore, the shock wave



298 B. Lal et al.

accompanying laser ablation can cause powder particles surrounding the vaporized regionto be dislodged, thereby contributing to more material removal. Whereas for metals theablation efficiency is on the order of 10 nm per pulse, with pharmaceutical materials itis on the order of 10 �m per pulse.

Up to now most applications of LIBS, to pharmaceutical materials, have used a solid-state Nd:YAG laser operating at its fundamental frequency (1064 nm). This is justifiedby this laser’s stability, robustness, compactness, and minimal maintenance require-ments. Other lasers that produce higher absorption by pharmaceutical materials havealso been used. Lam and Salin report the use of a frequency-quadrupled Nd:YAG laser,emitting radiation at 266 nm, for the laser ablation of pharmaceutical tablets followed byinductively-coupled plasma atomic emission spectrometry and mass spectrometry [20].The choice of an ultraviolet laser was dictated by their use of a commercial laser ablationsystem. Another LIBS analysis using a KrF excimer laser suggested that, for a particulardrug product, only the active agent in the analysed tablets significantly absorbed thelaser radiation at 248 nm [21].

A distinctive feature of LIBS spectra obtained from pharmaceutical materials is theirrelative simplicity, compared to the spectra of metals. Although emission bands fromdiatomic species (C2, CN, NH, etc.) appear, atomic and ionic lines are much fewerthan for metals. In addition, the most useful spectral information is generally found inthe visible and near-infrared, in contrast to the abundance of lines in the ultraviolet formetals. These considerations naturally dictate the choice of spectrograph and detector.

As already mentioned, the elements most often used as specific markers for APIs (S,Cl, F and Br) have low emissivities. Helium is known to enhance the signal-to-noiseratio for halogens. This can be achieved simply by blowing a narrow helium streamto displace the air above the sample surface. In this way, significant improvements insensitivity (by more than seven-fold) have been demonstrated for fluorine and chlorine-containing APIs [22]. Improvements in sensitivity have also been obtained for bromine.Figure 8 shows the impact of using helium on the detection of fluorine. In this case,using helium leads to a much weaker background signal, and to increased line intensities.

Wavelength

680 682 684 686 688 690 6920

20000

40000

60000

80000

F

with He

without He

Fig. 8. Comparison of spectra obtained with and without a helium gas flow, for tablets containing5% m/m of a fluorine-containing API (corresponding to 0.264% of fluorine).




Finally, it is worth mentioning that, as with several analytical applications, perform-ing internal standardization can significantly improve analytical figures of merit, suchas precision. As was shown in Ref. [22], using carbon as internal standard can also helpeliminate or minimize some matrix effects during the analysis of pharmaceutical mate-rials. In this case, two very different formulations of the same chlorine-containing drugdid not produce the same response of the chlorine signal. However, using the chlorine-to-carbon signal ratio, a calibration curve valid for both formulations was obtained. Thebenefits of internal standardization are however not universal, and it should be evaluatedon a case-by-case basis.

3.4. Main Applications

3.4.1. Global Analysis of Solid Dosage Forms

Although LIBS would allow a spatially-resolved analysis of solid dosage forms (asdescribed further), many applications call only for a reliable and representative analysisat the scale of a whole tablet.

For HPLC, the global composition is obtained following dissolution of the wholetablet, and its homogenization once in solution. With LIBS, this sample preparation stepis avoided, but a strategy must nevertheless be developed so that analysing only partof the tablet will still provide a reliable value for the global composition. In general,significant variation of composition within a tablet occurs at the scale of particles�10–200 �m�, but larger-scale heterogeneity can also occur. When several laser pulsesare fired at the same position on a tablet, the ablated depth for each successive shot ison the order of tens of microns. This spatial exploration of the tablet is on the samelength scale as the particle size, and therefore the LIBS signal variation from pulse topulse is relatively high (RSTD = 10–20%).

When one averages the signal for several pulses (say, ten) at the same site, oneobtains an analysis for a volume much larger than the particle size, and therefore thesignal variation between different sites is much lower (RSTD = 5–10%). Averagingthe results from several sites on a tablet, further increases precision. The variation fromtablet-average to tablet-average falls typically to the level of 1–4% RSTD. This is suf-ficient precision for most applications, with precision below 5% often required. Withinternal standardization, and a uniform sample set, repeatabilities can fall in the range0.5–2% [23]. It should be noted that when a tablet is analysed with LIBS at several posi-tions (using several laser pulses), there are three possible sources of variation: variabilityof pulse energy (generally below 1%), variability of laser-sample interaction (whichcomes from changes in absorption properties or irradiated area), and real compositionalheterogeneity of the sample. Internal standardization can be used to compensate for thefirst two variation sources, so that remaining variations can be attributed principally tocomposition heterogeneity.

In order to infer the composition of a tablet from a limited amount of material, ithas been determined that increasing the number of sites across a tablet is preferableto increasing the number of laser pulses at a given site [15]. For a given number ofsites, firing more than ten pulses per site does not further improve the tablet-to-tabletRSTDs. In contrast, analysing more than 10 or 20 sites across a tablet proves beneficial.



300 B. Lal et al.

At a given site, heterogeneity only on the 100-�m scale is accounted for, while analysingseveral sites with spacings on the mm-scale accounts for larger-scale heterogeneity inthe tablet, due for example to non-optimal blending.

3.4.2. Chemical Mapping in Solid Dosage Forms

Instead of averaging spectral data gathered using LIBS at different depths and sitesacross a tablet surface, it may in some cases be useful to treat this data separately andbuild two- or three-dimensional compositional maps. Mouget et al. have reported onthe capabilities of LIBS for this purpose [15,24]. The three-dimensional distribution forvarious compounds containing elements such as Mg, Ti, Ca, S and Cl was obtained.

Recently, Heuser and Walker have compared LIBS to scanning electron microscopy-energy dispersive X-ray emission (SEM-EDX), another analytical technique requiringno sample preparation and offering spatial resolution [25]. Although the SEM offereda better lateral resolution, LIBS was found to be more sensitive, especially for fluorineand lighter elements (the authors inaccurately state that LIBS is limited to boron andhigher-mass elements). SEM-EDX was also found to be inadequate for coated tabletswhen depth resolution was required, as both the core and coating generated a signal. Incontrast, by sequentially ablating a small amount of mass at the same lateral position,it is possible by LIBS to obtain a depth-resolved analysis. In particular, this approachcan be utilized for determining the coating thickness at a given point, and the thicknessuniformity across a tablet.

The capabilities of LIBS for the characterization of coatings had previously beenexplored in detail by Mowery et al. [26]. Calcium signal (from the tablet core) and Mg, Siand Ti signals (from the tablet coating and sub-coating) were monitored simultaneouslyas a function of laser pulse (shot) number. The resulting non-calibrated depth profilesare shown in Fig. 9. The appearance of calcium signal was chosen as representingpenetration of the coating into the core. The number of laser pulses required to exceed agiven threshold value for the calcium signal was used as a measure of coating thickness

0

0.25

0.5

0.75

1

1 3 5 7 9 11 13 15 17 19 21 23

Shot number

Rel

ativ

e in

tens

ity

Ca

inte

nsity

Ca Mg Ti Si

0

50000

100000

150000

200000

250000

300000

350000

400000

1 4 7 10 13 16 19 22

Shot number

5% 6% 7.5% 10%

12.5% 15% 17% 21%

Fig. 9. (left) Depth profiles of Ca, Mg, Si and Ti signal from a coated tablet; (right) depth profilesof Ca signal for different coating thicknesses (expressed as the percentage of the sub-coated tabletmass). Taken from Ref. 26 (with permission).




and this number was found to increase linearly with the coating weight (thickness). Thisscheme was used to investigate coating thickness uniformity across a tablet surface andits edges. It was also possible to assess the uniformity of coating thickness from tabletto tablet.

When compositional information only from the core of a coated tablet is required,the laser ablation function of LIBS offers a convenient means for coating removal. Thissample preparation step is performed first, followed by the actual analysis of the core.Clearly, the analysis of coated tablets and the measurement of coating thickness are areaswhere LIBS possesses definite advantages over other analytical techniques, includingSEM-EDX and near-infrared spectroscopy.

3.4.3. Blend Uniformity Studies

One key attribute of the LIBS technique is its rapidity, due in part to its avoidanceof a sample preparation (digestion) step. In a pharmaceutical context, this can greatlyfacilitate the optimization of a process, when a large number of samples taken in differentconditions need to be analysed. This is the case for instance for the optimization ofblending processes.

As a proof-of-concept, the uniformity of a powder blend containing lactose, cellulose,magnesium stearate (the lubricant), croscarmellose sodium (the disintegrant) and chlor-pheniramine maleate (the API), was assessed as a function of the number of rotationsof a low-shear blender [15]. At different stages of blending, powder samples were col-lected at five different locations, and pressed into pellets for analysis by a commercialPharmaLIBS™ instrument. The %RSTD of Mg, Na and Cl signal between the five pel-lets was plotted as a function of the number of rotations of the blender. In this particularcase, the %RSTDs for the three components decreased below 4% after 200 rotations,indicating the appropriate endpoint for the blending process. This study required theanalysis of 40 pellets (with tens of laser pulses per pellet), but at a fast throughput ofabout one pellet per minute.

The at-line determination of the lubricant distribution in powder blends using LIBShas been reported by Good et al. [27]. The lubricant, magnesium stearate, was analysedquantitatively. For two different formulations, a linear calibration of the Mg signal wasobtained for 0–3% magnesium stearate, based on in-house standards prepared by spikingunlubed powder blends. The slope of the calibration was slightly different for the twoformulations. Powder samples were then taken from 12 different locations in a full-scaleribbon blender after a pre-determined blending time. In one process, the blend was foundto be highly uneven, because of the location where the lubricant was added to the blend,and of an insufficient blending time. In another process, the lubricant was found to beevenly distributed.

The lubricant concentration at 12 blender locations could be determined in approx-imately 15 minutes. The LIBS technique could therefore provide reliable informationabout lubricant uniformity, in a time frame adequate for making real-time process deci-sions. In relation to this application, Green et al. reported recently a thorough comparisonof LIBS with near-infrared (NIR) spectroscopy [28]. Although LIBS is destructive andhas less molecular selectivity, it was found to be more sensitive, and its calibration lessspecific to the formulation and manufacturing process, i.e. more universal, than NIRspectroscopy.



302 B. Lal et al.

3.4.4. Analysis of Metallic Contaminants

The applications described above call upon the detection of elements which are compo-nents of larger organic molecules. There is another class of applications in the pharma-ceutical industry, the detection and quantification of contaminants, which consists moresimply of detecting elements in their elemental (metallic) form. This constitutes in factthe principal use of LIBS in most industrial sectors.

Metal contamination of drug substances (APIs) or drug products (API + excipients)has several possible sources. It can originate from raw materials entering the manu-facturing process, from reagents and solvents, and from various equipments (vessels,plumbing, etc.) used in the synthesis of the drug [29]. Another important source ofcontamination is from the metal catalysts used in drug synthesis. The catalysts mostoften used to synthesise APIs include: ethyl magnesium bromide, NaOBr, nickel, pal-ladium, zinc, cesium, and numerous complexes with chromium, aluminum, cobalt ormanganese [30].

Traditional tests for the presence of metal contaminants, as provided by the UnitedStates Pharmacopoeia (USP) or European Pharmacopoeia (EP), rely on time-consumingsample preparation, followed by non-specific and insensitive colorimetric detection.Inductively-coupled plasma mass spectrometry (ICP-MS) has been proposed as an alter-native to the traditional tests [29]. A new general chapter on plasma spectrochemistryhas been published recently in the USP [31], providing an introduction to the use of theICP for detecting inorganic compounds such as contaminants. In order to be inclusiveof other plasma spectrochemistry techniques, this general chapter also includes a briefdescription of LIBS, stating in particular that “while LIBS is not currently in wide useby the pharmaceutical industry, LIBS is suited for at-line or on-line measurements ina production setting as well as in the laboratory. Because of its potential, it shouldbe considered a viable technique for plasma spectrochemistry in the pharmaceuticallaboratory.”

As a proof of concept, LIBS was used for the detection of palladium traces inpharmaceutical material [32]. A controlled experiment was carried out using palladiumnitrate �Pd�NO3�2� diluted at different concentrations in a 50/50 lactose/cellulose mixture,with 0.5% m/m magnesium stearate added. This powder was pressed into pellets for theLIBS measurements. Fig. 10 shows spectra obtained with a PharmaLIBS™ instrument.Seven palladium lines were clearly observed at 0.001% and 0.01% of palladium nitrate.Using additional samples, a �3�� limit of detection of 0.3 ppm for elemental palladiumwas determined. Given that the relevant analytical range for palladium contaminants isof 0.5–20 ppm, LIBS is found in this case to be readily applicable.

3.4.5. Direct Analysis of Powder

Although up to now most applications of LIBS for the analysis of pharmaceuticalpowders have included a sample preparation step where the powder is pressed intoa pellet, there is no real conceptual hurdle to the analysis of loose powder. Associ-ated with laser ablation is a laser-produced shock wave, which would tend to scatterloose powder and produce an uneven surface for subsequent pulses. A simple approachfor mitigating this problem consists in fixing the loose powder onto an adhesivesurface.




Wavelength (nm)

Inte

nsity

(co

unts

)

1500

340 342 344 346 348 350

2000

2500

blank0.001% Pd nitrate0.01% Pd nitrate

Pd spectral lines

0.001%

0.01%

blank

Fig. 10. LIBS spectra containing Pd lines for two samples with different concentrations of Palla-dium nitrate and one blank sample [32].

Tran et al. [33] report the use of such a scheme for the analysis of organic powders.This same scheme had been developed previously for the analysis of mineral ore samples.A given mass of powder was spread onto double-sided adhesive tape, itself fixed to aglass microscope slide. Excess powder was shaken off, leaving 2–5 mg of powder onthe tape. One laser pulse was fired at each of 100 separate sites on the sample, and sixslides were analysed. The carbon signal (at 247.86 nm) from the 100 sites on each slidewas averaged. The relative standard deviation (RSTD) of the carbon signal among thesix slides was 2%. Within each group of 100 sites, the RSTD was 10–17%. Incidentally,these repeatability results are comparable to the sample-to-sample RSTD and site-to-siteRSTD obtained when powder is pressed into pellets (Sect. 3.4.1).

The advantages of the adhesive-tape approach are its simplicity, and the fact that pos-sible matrix effects (coming from the compression of pellets) are avoided. The drawbackis that some interference may come from ablating part of the underlying tape. However,Tran et al. evaluate that, at least in the case of carbon, such an interference amounts toless than 1% of the signal from any organic powder sample present on the tape.

3.4.6. Other Types of Sample

In principle, LIBS could be applied to pharmaceutical samples other than solid dosageforms or powders, including liquids, gels, lotions, pastes, etc. These sample types fall out-side the scope of this chapter. The reader is referred to Ref. [34] for a proof-of-concept ofthe analysis of pharmaceutical liquids (injectables, nasal solutions) using LIBS. Anotherstudy worth mentioning is by Lademann et al., wherein LIBS was utilized for the anal-ysis of sunscreen lotions containing coated titanium dioxide microparticles [35]. Thetitanium-to-aluminum signal ratio (aluminum being present in the particle coating) wasused to check the stability of the coating during the manufacturing of the lotion or itspenetration in skin.



304 B. Lal et al.

4. PROSPECTS: LIBS AS A PROCESS ANALYTICALTECHNOLOGY

It would be difficult to conceive of a LIBS instrument which would be able to tackle allpossible applications. The analysis of pharmaceutical materials is one of very few appli-cations for which there exists a dedicated commercial instrument, the PharmaLIBS™(Pharma Laser, Boucherville, QC, Canada). This instrument, although designed forat-line analysis (i.e. on the production floor), has been utilized up to now mainly in phar-maceutical R&D laboratories. The transfer of LIBS from the laboratory to the productionsite necessitates extensive validation of this new technique, as required by regulatorybodies.

This acceptance process is likely to accelerate in the near future, thanks to a newinitiative surrounding the notion of Process Analytical Technologies (PAT), launchedin 2002 by the U.S. Food and Drug Administration (FDA). PAT is presented as anew philosophy for pharmaceutical process control. It is based on the observation that“quality cannot be tested into products; it should be built-in or should be by design”(from the FDA’s Guidance for Industry on PAT [36]). In other words, instead of relyingon quality control of the final product emerging from the production line, sensor devicesand knowledge management tools should be included in the process loop, in order tobetter understand each step of a process, to characterize materials at different points,and to possibly influence the process itself in real time. This approach, routinely usedin other industries, is relatively new in the pharmaceutical sector, and calls for novelsensors that can be implemented in-process at the production floor.

It is clear that LIBS has all the capabilities for the rapid chemical characterizationof pharmaceutical ingredients and mixtures, in tablet, powder, or even liquid form. Thecommercial LIBS instrument currently on the market has already been designed forat-line analysis of solid dosage forms and powder pellets, and as such fits nicely inthe PAT framework. The prospects for the inclusion of LIBS in pharmacopeias, for itsacceptance by analytical chemists, and for its transition to the production floor appeartherefore very bright.

5. APPLICATION TO GLASS INDUSTRY

The raw material for glass industry is in powder form which is a mixture of manychemicals. This mixture is called glass batch. Various steps involved in glass making are:

(i) preparation of the glass batch of the composition needed for making the intendedglass product,

(ii) the glass batch is fed into furnace where it is melted and(iii) cooling of the melt into stable glass.

Most of energy used in glass making is consumed in step (ii). To make the glass-making energy efficient it is necessary to ensure that repeatable glass batch entersthe glass melting furnace so that huge energy is saved by optimization of the furnaceparameters for a particular glass batch.




LIBS is one technique which can be employed for in-situ monitoring of the com-position of glass batch before it is fed to furnace. LIBS has been applied to determinethe composition of finished glass as well as that of glass melt. Kurniawan et al. [37,38]investigated glass for a number of elements and reported a detection limit of the orderof 10 ppm. However, the experiments were carried on samples placed in 1–10 torr pres-sure. Panne et al [39,40] used LIBS for the analysis of glass and glass-melts duringvitrification process of fly and bottom ashes. Su et al. [41] used LIBS to determine theconcentration of about 10 metallic elements in vitrified glass samples produced by Jouleheated glass melters. Lal et al. [42] investigated the application of LIBS to determinethe elemental composition of the glass batch. They experimented upon two types ofglass batches, namely, (i) the glass batch used as surrogate for the batch employed fortrapping of radioactive waste by vitrification and (ii) the batch used to manufacturethe most common type of soda-lime glass (flat glass). Surrogate glass batches wereinvestigated because disposal by vitrification is one of the best available waste disposaltechniques for radioactive waste. On the other hand in-situ monitoring of flat glassbatch is expected to result in (i) the energy saving, as mentioned earlier, due to opti-mization of the furnace parameters for a particular glass batch by ensuring that everytime identical glass batch enters the furnace, (ii) resolution of environmental issues byidentifying the pollutants before their entry into the furnace and (iii) increase in thedegree of uniformity in the final glass products by monitoring the uniformity of theglass batch.

The composition of the surrogate glass batch is 24–40 wt% of silica �SiO2�, 16 wt%ferric-oxide �Fe2O3�, 11 wt% alumina �Al2O3�, 6 wt% zinc-oxide (ZnO), 1–3 wt% boronoxide �B2O3� besides very small quantities �<1 wt%� of oxides of sodium, potassium,magnesium, manganese, tin, nickel etc. The flat glass batch consists of SiO2 (58.7 wt%),CaCO3 (23.6–17.7 wt%), Na2CO3 (22.9–16.9 wt%) and small amounts (0.8 wt%) ofAl2O3. The glass batch is prepared by thoroughly hand mixing (in a mortar) the carefullyweighed quantities of chemical compounds. The LIBS spectra are recorded using pelletsprepared from the powder glass batch. 5g of the powder are hand grounded in a mortartill no resistance is felt. To this 0.8ml polyvinyl alcohol (2 wt% in distilled water) isadded as binder. The mixture is thoroughly hand mixed and this powder is pressedinto pellet by putting it in a 25mm bore die which is subjected to about 24 MPapressure. The pellets thus obtained are dried in an oven at 90 �C for about fifteenminutes.

The apparatus used to record the LIBS spectra is same as shown in Fig. 3.However, in-addition to the detection system based on Spex 500M Czerny-Turner0.5m spectrometer fitted with a diode array detector (IDAD), the LIBS spectra havealso been recorded using a ESA 3000 echelle spectrometer with ICCD camera. TheSpex 500M spectrometer equipped with a 2400l/mm can only cover a spectral regionof 20 nm. The echelle spectrometer has a spectral coverage of 200–780 nm, therefore,almost all the elements in the sample can be simultaneously detected. The detectionsystem with the echelle spectrometer also provides a spectral resolution about 4 timesbetter in the UV region than is obtainable by the detection system with the Spex 500Mspectrometer system. However, the data acquisition time with the echelle spectrometersystem is much longer for the same exposure time and accumulation. The longersystem response time for this system is because it needs more time to transfer all data�1024×1024 pixel� from camera to frame grabber and then to the computer.



306 B. Lal et al.

To record the LIBS spectra of glass batch in the 250–450 nm region, it requiressequential scanning of the spectrometer using Spex 500M system while spectral range200–780nm has been recorded simultaneously using the echelle spectrometer system.Fig. 11 shows typical LIBS spectra recorded with both systems around 390 nm region.It is clear from Fig. 11 that the spectrum recorded with the echelle spectrometer systemhas better resolution in this spectral region. The elements identified from the spectra areSi, Al, B, Ca, Mg, Mn, Na, K, Cr, Cu, Fe, Ti, Sr, Zr, Zn, Pb, and Ni. For quantitativeanalysis we have to choose intense spectral lines, which have minimal interference fromother emission lines, and which do not involve ground state so that self-absorption isabsent. The LIBS spectra recorded with the Spex 500M system has RSTD of about 4.5%with the position of laser focal spot ∼1.5mm inside the sample and gate delay of 1�s.The calibration curves for the major constituents (Si, Fe, Ca and Al) of the surrogate glassbatch have been prepared by plotting the emission intensity as a function of elementalconcentration in the region of 390nm (Fig. 12). Calibration curve (Fig. 12) for Si preparedby plotting the intensity of the Si 390.5 nm emission line as a function of the wt% ofSi is a straight line with correlation coefficient �R2 = 0�9927� close to unity. Similar

0

40

80

120

160

200

240

280

385 386.5 388 389.5 391 392.5 394 395.5 397 398.5 400

11961

9003

6045

3087

129

Ca

Ca

Si 3

90.5

nm

Fe

387.

8 nm

Inte

nsity

Inte

nsity

Ca

393.

366

nm

Ca

396.

84 n

m

Al 3

96.1

6 nm

Al 3

94.4

nm

AlAl

Si

Fe

Wavelength (nm)

Fig. 11. LIBS spectra of a surrogate glass batch around 390nm. Top: recorded by 0.5m Czerny-Turner spectrometer fitted with IDAD. Bottom: recorded with ICCD fitted ESA 3000 echellespectrometer.




0.8

1

1.2

1.4

1.6

10 12 14 16 18 20

Si 390.5 nm

R2 = 0.9927

wt% of Si

104 ×

inte

nsity

4

5

6

7

wt% of Fe

R2 = 0.9937

104 ×

inte

nsity

Fe 387.8 nm

9 11 13 15 17

wt% of Al

1.2

1.3

1.4

1.5

1.6

1.7Al 394.4 nm

R2 = 0.993610

5 × in

tens

ity

1.8

5 6 7 8 9

wt% of Ca

2.5

3.5

4.5

5.5

6.5

7.5

104 ×

inte

nsity

Ca 336.19 nm

6 8 10 12 14

R2 = 0.9964

Fig. 12. Calibration curves obtained with 0.5m Czerny-Turner spectrometer detection systembased on line intensity.

calibration curves have been obtained for Fe (emission line 387.8 nm, R2 = 0�9937), Ca(emission line 336.19 nm, R2 = 0�9964) and Al (emission line 394.4 nm, R2 = 0�9936).

However, when same procedure is followed for minor constituents like B and Ni, thecalibration curves thus obtained have correlation coefficients of the order of 0.6. Onthe other hand, the intensity-ratio calibration curves have correlation coefficient veryclose to unity. The calibration curve obtained by plotting the ratio of the intensity ofthe B 249.6 nm emission line to that of the Fe 250.1 nm as a function of the B to Feconcentration is a straight line with R2 = 0�9979. Similar results have been obtained forNi/Fe intensity ratio calibration. Data recorded from this detection system has a RSTDof better than 5% for all the measured elements either using the absolute intensity orintensity-ratio.

Since the entire wavelength range 200–780nm is recorded simultaneously with theechelle spectrometer detection system there are more than one emission line for eachconstituent which can be selected for the quantitative analysis. The spectral lines withthe least interference for each element were selected for the analysis. On analysis ofthe LIBS spectra recorded with this system it is found that (i) RSTD of the intensitydata is ∼8% and (ii) the intensity of an emission line of a constituent shows very poorcorrelation with its concentration. This might relate with longer system response time(18 times higher for 60 shots averaged measurement).



308 B. Lal et al.

Fe 440.48 nm

6000

8000

10000

12000

14000

16000

10 12 14 16 1850

55

60

65

70

75

IntensityNormalized intensity

Fe (wt%)

Nor

mal

ized

inte

nsity

Inte

nsity

Si 288.16 nm

5000

10000

15000

20000

25000

30000

10 12 14 16 18 20

Si (wt%)

Nor

mal

ized

inte

nsity

Inte

nsity

40

50

60

70

80

90

100

110

Intensity

Normalized intensity

Fig. 13. Calibration data obtained with the echelle spectrometer detection system with and withoutplasma emission normalization.

Alternately, calibration curves have been prepared by normalizing the absolute lineintensity with the total plasma emission. Fig. 13 shows the calibration data with andwithout the normalization. The dramatic improvement in the calibration curve after thesignal normalization can be seen clearly in this Figure. Also the RSTD has improvedsignificantly. For example, the signal normalization reduces RSTD from 8.4% to 2.6%for Fe 440 nm line and from 9.1% to 1.8% for Si 288 nm line of one set of glass batchdata. This improvement in the RSTD of the data and in the linearity of the calibrationcurve by normalization technique can be explained in terms of the elimination of theeffects of the fluctuating plasma conditions during the measurement.

Although the normalization technique can improve RSTD and accuracy for mostelements yet the calibration based on the line intensity ratio still gives equal or betteranalytical figure of merit. Using the line intensity ratio with respect to Si 288 nm linefor echelle spectrometer detection system we obtained the linear calibration curvesfor Ca, Pb, Cr, Ba, Zr, Al, Fe and Ti shown in Fig. 14. Data recorded with theechelle spectrometer detection system has a RSTD of better than 5% for almost all themeasured elements except Si and B. The RSTD for these two elements are ∼9% and6%, respectively.

The accuracy and precision for the glass batch data obtained from two detectionsystems are compared. Both detection systems have comparable analytical figure ofmerit. The accuracy and precision of the glass batch measurement with Czerny-TurnerSpectrometer fitted with diode array detector as well as with ICCD fitted broadbandechelle spectrometer are better than 5% for the major elements and better than 10% formost of the minor elements. This investigation [42] clearly demonstrates the potential

ElsevierAMS

Jobcode:

LRN

Ch12-N51734

24-7-2007

8:05p.m.

Page:309

Trim:165 ×

240MM

TS:Integra

Font:

Times

Size:10/12pt

Margins:Top:13MM

Gutter:20MM

T.Area:125mmX198mm

4Color

COP:Recto

Floats:

Inline

LIB

ST

echniquefor

Powder

Materials

309

0

0.5

1

1.5

2

2.5

3

3.5

4

0 0.01 0.02 0.03 0.04 0.05 0.06

Weight ratio

I(C

a 39

3.37

)/ I(

Si 2

88.5

)

I(P

b 40

5.78

)/ I(

Si 2

88.5

)

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

Ca /SiPb /Si

I(F

e 40

4.58

)/ I(

Si 2

88.5

)

I(T

i 499

.11)

/ I(S

i 288

.5)

0

1

2

3

4

5

6

0 0.5 1 1.5

Weight ratio

00.020.040.060.080.10.120.140.160.180.2

Fe /SiTi /Si

I(A

l 308

.22)

/ I(2

88.5

)

I(Z

r 360

.12)

/ I(S

i 288

.5)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0 0.2 0.4 0.6 0.8

Weight ratio

0

0.5

1

1.5

2

2.5

3

Zr /SiAl /Si

I(C

r 425

.43)

/ I(S

i 288

.5)

I(B

a 45

5.4)

/ I(2

88.5

)

Cr /Si

Ba /Si

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

0 0.002 0.004 0.006 0.008

Weight ratio

0

0.2

0.4

0.6

0.8

1

1.2

1.4

Fig. 14. Some typical calibration curves based on line ratio with echelle spectrometer detection system.



310 B. Lal et al.

of LIBS for in-situ monitoring of glass batch in glass industry. Its shows that LIBSis capable of directly measuring the composition of glass batch (in pellet form) withreasonable accuracy.

REFERENCES

[1] S. Palanco and J. Laserna, Appl. Opt., 42 (2003) 6078.[2] R. Wisbrun, I. Schechter, R. Niessner, H. Schroder and K. L. Kompa, Anal. Chem, 66

(1994) 2964.[3] P. Fichet, D. Menut, R. Brennetot, E. Vors and A. Rivoallan, Appl. Opt., 42 (2003) 6029.[4] B. Lal, F-Y. Yueh, J. P. Singh and W. G. Ramsay, Proc 105th Annual Meeting & Exposition,

April 27–30 (2003) Nashville, Tennessee, USA.[5] M. Martin, S. Wullschleger and C. Garten Jr, Chemical and Biological Early Warning

Monitoring for Water, Food and Ground, J. L. Jensen and L.W. Burggraf ed. SPIE ProcNo.4576 (2002) 188.

[6] Rosenwasser, G. Asimellis, B. Bromley, R. Hazlett, J. Martin, T. Pearce and A. Zigler,Spectrochim Acta B 56 (2001) 707.

[7] R. Krasniker, V. Bulatov and I. Schechter, Spectrochim Acta B56 (2001) 609.[8] B. Lal, H. Zheng, F-Y. Yueh and J. P. Singh, Appl. Opt., 43, (2004) 2792.[9] R. Krasniker, V. Bulatov and I. Schechter, Spectrochim Acta B56 (2001) 609.

[10] M. A. Khater, John T. Costello and T. Kennedy, Appl. Spectrosco., 56 (2002) 970.[11] I. Bassiotis, A. Diamantopoulou, A. Giannoudakos, F. Roubani-Kalantzopoulou and

M. Kompitsas, Spectrochim Acta B 56 (2001) 683.[12] B. T. Fisher, H. A. Johnsen, S. G. Buckley and D. W. Hahn, Appl. Spectrosco., 55

(2001) 1312.[13] B. Le Drogoff, J. Margot, M. Sabsaabi, O. Barthelemy, T. W. Johnston, S. Laville, F. Vidal

and Y. von Kaenel, Spectrochim Acta art B, 56 (2001) 987.[14] Private Communication.[15] S. Béchard, Y. Mouget, in Laser Induced Breakdown Spectroscopy (LIBS): Fundamen-

tals and Applications, A. Miziolek, V. Palleschi, I. Schechter (eds.), Cambridge UniversityPress (2006).

[16] R. J. Locke, J. B. Morris, B. E. Forch, A. W. Miziolek, Appl. Opt., 29 (1990) 4987.[17] L. M. Berman, P. J. Wolf, Appl. Spectrosc., 52 (1998) 438.[18] L. St-Onge, R. Sing, S. Béchard, M. Sabsabi, Appl. Phys. A, 69 (1999) S913.[19] L. St-Onge, M. Tourigny, M. Sabsabi, AAPS Journal. 6(4) (2004), abstract T3040. Available

from http://www.aapspharmsci.org/.[20] R. Lam, E. D. Salin, J. Anal. At. Spectrom., 19 (2004) 938.[21] L. St-Onge, E. Kwong, M. Sabsabi, E. B. Vadas, AAPS PharmSci. Suppl., 3(S1) (2001),

abstract M2156. Available from http://www.aapspharmsci.org/.[22] L. St-Onge, E. Kwong, M. Sabsabi, E. B. Vadas, Spectrochim. Acta B, 57 (2002) 1131.[23] L. St-Onge, P. Faustino, M. Tourigny, M. Sabsabi, AAPS Journal, 6(4) (2004), abstract

T3053. Available from http://www.aapspharmsci.org/.[24] Y. Mouget, M. Tourigny, S. Béchard, AAPS PharmSci. Suppl., 3(S1) (2001), abstract M2331.

Available from http://www.aapspharmsci.org/.[25] D. Heuser, D. S. Walker, J. Anal. At. Spectrom. 19 (2004) 929.[26] M. D. Mowery, R. Sing, J. Kirsch, A. Razaghi, S. Béchard, R. A. Reed, J. Pharm. Biomed.

Anal., 28 (2002) 935.[27] J. A. Good, M. D. Mowery, R. A. Reed, AAPS PharmSci. Suppl. 3(S1) (2001), abstract

M2344. Available from http://www.aapspharmsci.org/.




[28] R. L. Green, M. D. Mowery, J. Good, J. P. Higgins, S. M. Arrivo, K. McColough, A. Mateos,R. A. Reed, Appl Spectrosc., 59 (2005) 340.

[29] T. Wang, J. Wu, R. Hartman, X. Jia and R. S. Egan,[30] J. Pharm. Biomed. Anal., 23 (2000) 867.[31] E. B. Vadas, personal communication. <730> Inductively Coupled Plasma [new] (1st Supp

to USP 28), Pharmacopeial Forum. 30(3) (2004) 1022.[32] L. St-Onge, M. Tourigny, M. Sabsabi, Pittcon 2005, abstract 670–6 (2005).[33] M. Tran, Q. Sun, B. W. Smith, J. D. Winefordner, Appl. Spectrosc. 55 (2001) 739.[34] L. St-Onge, E. Kwong, M. Sabsabi, E. B. Vadas, J. Pharm. Biomed. Anal. 36 (2004)[35] J. Lademann, H.-J. Weigmann, H. Schäfer, G. Müller, W. Sterry, Skin Pharmacol. Appl.

Skin Physiol. 13 (2000) 258.[36] Guidance for Industry: PAT – A Framework for Innovative Pharmaceutical Development,

Manufacturing, and Quality Assurance (Sept. 2004). Available from http://www.fda.gov/cder/OPS/PAT.htm.

[37] H. Kurniawan, S. Nakajima, J. E. Batubara, M. Marpaung, M. Okamoto and K. Kagawa.Appl. Spectrosc. 49 (1995) 1067.

[38] H. Kurniawan, S. Nakajima, J. E. Batubara, M. Marpaung, M. Okamoto and K. Kagawa.Appl. Spectrosc. 50 (1996) 299.

[39] U. Panne, C. Haisch, M. Clara and R. Niessner, Spectrochimica Acta B53 (1998) 1957.[40] U. Panne, C. Haisch, M. Clara and R. Niessner, Spectrochimica Acta B53 (1998) 1969.[41] C. Fu Su, S. Feng, J. P. Singh, F-Y. Yueh, J. T. Rigsby III, D. L. Monts and R. L.Cook,

Glass Technol. 41 (2000) 16.[42] B. Lal, F-Y. Yueh and J. P. Singh, Appl. Opt. 44 (2005) 3668.



Chapter 13

LIBS for the Analysis of Chemical and Biological Hazards

Steven G. Buckley

Department of Mechanical and Aerospace Engineering,University of California San Diego, La Jolla, CA 92093-0411and Photon Machines, Inc.; San Diego, CA, USA

1. INTRODUCTION

Much of the re-invigoration of LIBS that occurred in the United States in the late 1980sand 1990s occurred in pursuit of a diagnostic method for toxic and hazardous sub-stances. In particular, work of David Cremers and Lee Radziemski in New Mexico [1–3],winner of several R&D 100 awards from R&D Magazine, was in part motivated by theneed for monitors for toxic substances such as Be. Researchers at Sandia National Labo-ratories worked in a similar vein, on the application of LIBS as a diagnostic for airbornemetals from the incineration of toxic mixed and municipal wastes [4,5], while researchersat the Army Research Laboratory investigated LIBS for detection of halon replacementcompounds [6]. This previous work focused on LIBS-based elemental analysis for safetymonitoring and environmental applications.

Recent work has attempted to extend the role of LIBS in environmental and hazardmonitoring to include diagnostics for chemical and biological agents. Such agents areobviously complex molecules and intricate living structures, and it is not immediatelyobvious that an elemental analysis technique such as LIBS should be useful in diagnosticsfor chemical and biological agents, as molecular information is not preserved in LIBSspectra. It was not until the relatively recent application of broadband spectrometersto LIBS, either linear silicon array spectrometers or echelle spectrometers, that thepossibility of LIBS determination of stoichiometry of a broad range of compoundsthrough analysis of elemental ratios became feasible. The ability to accurately determineelemental ratios in a gas stream, in particles suspended in a stream, or for gases adsorbedor particles captured on a substrate, provides the diagnostic potential for LIBS analysis ofchemical and biological agents. In addition, so-called “hyphenated” techniques, e.g. LIF-LIBS or Raman-LIBS, may take advantage of the particular strengths of two or moremethods.

LIBS thus provides a broadly applicable elemental analysis platform, requiring virtu-ally no sample preparation, applicable in multiple media. Due to the rigorous nature ofchemical and biological agent sampling requirements, in particular high sensitivity, low




314 S. G. Buckley

false alarms, and standoff detection, an understanding of matrix effects, potential inter-ferences, and high-efficiency collection optics becomes crucial for these applications.This chapter outlines recent research and developments in LIBS analysis of chemicaland biological hazards.

2. APPLICATION TO CHEMICAL HAZARD ANALYSIS

The application of LIBS to the identification of chemical hazards has primarily focusedon the identification of explosive materials, with a small amount of work on chemicalnerve agents, but this work has strong general parallels with work to determine chemicalstoichiometry in general using LIBS, e.g. for chemical identification [7,8] and for fuel/airratio identification [7,9–11]. Typically this work relies on line ratios to determine relativemolecular concentrations. For example, Sattmann et al. [8] used the C/H line ratio andthe chlorine line to identify polymers, while groups working on hydrocarbon/air mixtureshave used combinations of C, H, O, and N lines (and ratios) to determine stoichiometry.

De Lucia and co-workers have performed LIBS on a number of different energeticmaterials, including PETN, HMX, RDX, and TNT, as well as on propellants and militaryexplosives [12]. They used a broadband (200–980 nm) LIBS spectrometer with 1�5 �sdelay and an open gate of 2 ms to collect spectra following excitation with a 30 mJ pulsefrom a 10 ns Q-switched Nd:YAG laser. Single-shot spectra of pure energetic materials,shown in Fig.1, revealed characteristic C, H, O, and N atomic lines. In addition, one ofthe strongest spectral features was Na, and additional Mg, and K impurities were alsoobserved. The authors observe that, based on the N:O atomic ratio in air measured withLIBS, the N:O atomic ratio measured with an argon purge on an RDX sample surface

200150

550

950

1350 C

TNT

HMX

NC

PETN

RDX

NO

O NN

HNa

300 400 500 600

Wavelength (nm)

Inte

nsity

(ar

b. u

nits

)

700 800 700 1000

Fig. 1. Single-shot LIBS spectra of pure energetic materials, Ref [12].



LIBS for the Analysis of Chemical and Biological Hazards 315

conforms to the expected value for RDX, and suggest that LIBS peak ratios could be usedfor energetic materials analysis, as previous work in polymer analysis might suggest.Using the same system, the group has also showed that the measured C:P atomic ratiocan be used to determine the stoichiometry of C:P in chemical agent simulants [13],both neat and (more weakly) when the agent simulants are on soil samples.

Portnov, Rosenwaks, and Bar [14] have shown that molecular emission may alsobe used to infer material composition. In particular, they observed LIBS spectra ofnitroaromatic and polycyclic aromatic hydrocarbons in laboratory air. They found thatintensity ratio of the C2 Swan bands to the CN violet system was useful in determinationof the number of carbon-carbon double bonds in the analyte, and rises with increasingnumbers of aromatic rings. This is related to the finding of Sabsabi’s group [15] whichfound similar results with samples mixed in a cellulosic matrix. Portnov et al. [14]combined the molecular band ratio with the O/N ratio, which carries information aboutthe number of nitro groups in the sample. Together, it was postulated that these piecesof information could help infer the presence of polycyclic aromatics and nitroaromaticsin a particular sample.

The previous work was accomplished at close distances in laboratory settings.Recently López-Moreno et al. [16] have demonstrated stand-off LIBS detection ofexplosive residues on solid surfaces. Their experiment used a Herschelian telescope toproject and focus 350 mJ pulses of 1064 nm laser light onto a solid surface 30 metersdistant, upon which small quantities �∼5 �g, for example) of energetic material hadbeen deposited. Light was dispersed by a 1/8-meter spectrometer and detected on anICCD camera. Known samples were tested in the field, and unknown samples were alsointerrogated in a blind test. Molecular emission from the C2 Swan bands, from N, H,and O emission lines, and from atomic intensity ratios, including H/C2, O/H, and O/N,but also including O/K and Na/C2, was detected and used in a flow-chart analysis todetermine whether a sample is explosive or not. The latter two ratios, O/K and Na/C2,are dependent on contaminants K and Na commonly found in explosives. Of 15 knownsamples the test generated two false positives, while for 6 unknown samples all of thesamples were classified correctly.

3. APPLICATION TO BIO-AEROSOL & BIO-AGENT DETECTION

Initial LIBS-related work on biological aerosol detection was published in October,2003, by four groups. Three papers appeared in Applied Optics [17–19], while oneappeared in Applied Spectroscopy [20], each is discussed below. These papers werelikely motivated in some manner by the events that unfolded in late 2001 in the U.S.;following the September 11 airplane hijackings there was a highly-publicized series ofletters contaminated with aerosolized anthrax spores. These letters sickened a total oftwenty-two people and caused five deaths from inhalation of anthrax; thousands morewere put on a prophylactic course of the antibiotic Cipro to protect against infection.This disturbing event mobilized researchers to investigate the discrimination potentialof LIBS for biological aerosol. Previous work on biological material had been primarilylimited to investigations of teeth and bones, such as in Samek et al. [21] and the workreferenced therein.



316 S. G. Buckley

Table 1. Elemental analysis of Bg and fungal spores, Ref [20]

Elementa Bg Fungal spores (smuts)

1 2 3 Oat Wheat Corn

Ca 1�16 1�08 10�21 0�16 0�0147 0�12Mg 0�30 0�37 2�80 0�20 0�0937 0�19Na 0�45 0�38 5�82 0�0132 0�0110 0�0171K 0�49 0�49 0�68 1�60 2�24 1�63Fe 0�67 0�57 0�0088 0�0253 0�0032 0�0081P 2�30 2�32 8�52 0�44 0�41 0�58Mn 0�0081 0�0122 0�10 0�006 0�0024 0�0037

a Elemental concentrations as percent by weight.

The potential for elemental analysis to discriminate between various types ofbioaerosol is observed in Table 1, from Hybl et al. [20], which shows the elementalconcentration as a percent by weight of a number of elements detectable using LIBSin fungal spores (smuts) and three strains of Bacillus subtilis var. niger, also known asBacillus globigii (Bg). Bg is a common bacterium found in soil and decomposing organicmatter that is widely used as a simulant for anthrax and other biohazardous bacterialspores. In this case, each strain of Bg has been differently prepared, with variations ingrowth media and conditions. From Table 1 it is apparent that ratios of the elements varywidely between the bacterium and pollen samples, in general more widely than withinan individual sample type. For example, the ratio of Mg/Na in the three Bg samples isbetween 0.5 and 1, while in the pollen samples it is between 8 and 16. Such elementalratios, or combinations of these ratios, could be used to group classes of biologicalaerosols. This could be useful because common ultraviolet excitation/broadband fluo-rescence techniques, which largely rely on amino acid fluorescence, are known to havecross-sensitivities to many forms of biological aerosols, as well as other contaminantsin the atmosphere such as diesel soot. Hence LIBS could be an effective discriminationtool to minimize cross-sensitivities.

The work on LIBS detection of bioaerosols can be classified into work that has beendone with aerosols captured on solid surfaces, and work that has been done with airborneaerosols. In both cases, work to date has shown promise but has not been conclusiveabout the potential of LIBS detection of bioaerosols, suggesting that further work in thisvein is warranted.

3.1. Deposited or Pelletized Samples

Samuels et al. [17] used LIBS to study bacterial spores, pollens, molds, and proteins,which were prepared and deposited on porous silver substrates. This work thoroughlycovered sample preparation similar to other work discussed below, so it is discussed indetail here. Preparation of bacterial samples included streaking, suspension in phosphate-buffered saline, and spreading onto a solid nutrient sporulating medium; incubation(up to 7 days at 30 �C) was terminated when phase-contrast microscopy determined that




more than 95% of cells from each colony had sporulated. These spores were suspendedin buffered saline and centrifuged at 10,000 rpm, after which water was poured off andspores were resuspended in ultrapure water and vortexed. This procedure was repeateda total of four times, after which the bacteria were suspended in water and freeze-dried.For preparation of thin films for LIBS analysis, bacteria, pollen, proteins, and mold wereall resuspended in ultrafiltered water, vortexed, pipetted onto a 0�45 �m silver membranefilter (Millipore) and underwent vacuum filtration to form thin films on the surface ofthe filter.

The plasma was formed on the surface of the filter by a 30 mJ pulse of ∼10 nslaser light at the Nd:YAG fundamental wavelength focused onto the sample surfaceusing a 5 cm focal length convex lens. Light was collected into a 6-channel broadbandspectrometer system utilizing linear silicon array detectors with a resolution of 0.1 nmper pixel. Spectra were collected with a 1�5 �s delay and a 2 ms gate open width.Relatively low numbers of spectra were collected, each from fresh surface on the thinfilms: 30 individual spectra of each spore type, 15 of each mold, 5 of each pollen, 13of ovalbumin (protein), and 20 blank spectra from a silver disk membrane. Average andstandard deviations were calculated from this data.

Significant differences were observed in the averaged spectra from each class ofbiological material. These data were then treated with a principal components analysis(PCA) [22,23] based on 30 features observed in the 67 single-shot spectra. The firstprincipal component was observed to contain much of the shot-to-shot variability, while

Loadings for PC2

Loadings for PC2 versus PC1

Load

ings

for

PC

1

–0.2

–0.125

–0.124

–0.123

–0.122

–0.121

–0.12

–0.119

–0.118

–0.117

ov

ov

ovovovov

ov

Vo

Vo

Vo

VoVoVo

Dr

DrDr Dr

DrAlAlAl

Al Al

Al

cl

cl

BG

BGBGBGBG

BG

BGBGBG

BG

Bc

Bc

Bc

Bc

Bt

BtBt

Bt

Bt

Bt Bt

Bc

BcBt

BlBlBl

clcl

clcl

clcl

clclclcl

AlAl

Al

–0.116

–0.115

–0.15 –0.1 –0.05 0.05 0.1 0.15 0.20

Fig. 2. First two principal components from PCA of bacterial, mold, and spore samples, Ref [17].Al, Alternaria tenuis; cl, Cladosporum herbarum; BG, B. subtillis; Bc, B. cereus; Bt, B. thuringien-sis; Vo, Virginia oak pollen; Dr, desert ragweed pollen; ov, ovalbumin.



318 S. G. Buckley

the second component contained much of the desired information about the source ofthe spectra. Figure 2 shows the PCA spectra from these experiments, where the secondprincipal component is displayed on the x-axis. As observed, this treatment of the LIBSdata from spectra collected from the filter membranes shows good separation betweenvarious biological samples, as indicated by the legend. The authors concluded thatdiscrimination was possible and that based on these results, further study was warranted.

Morel et al. [18] investigated six biological samples, prepared in a similar way asSamuels et al. [17], except that they formed compressed pellets from the freeze-driedsamples, as opposed to vacuum deposition of the samples onto surfaces. Spectra froma plasma formed from 10-ns 100 mJ pulses of 1064 nm laser light focused at normalincidence is dispersed by a 1-m focal length Czerny-Turner spectrometer onto an opticalmultichannel analyzer/charge-coupled device detector.

These authors observed that there were several types of features detected in theirspectra, which were initially collected over the spectral region ∼245 nm to 930 nm. Theywere able to see elemental lines from mineral elements Mg, Na, Fe, K, and Ca, butsuggested that some of these (particularly Na and Ca) may suffer from background inter-ference. So-called “organic” elements (presumably from their association with commonorganic molecules) that they saw in their spectra were C, N, P, and H. In addition tothese elemental lines, they observed strong emission from excited CN formed duringrecombination of C2 and N2 in the cooling plasma [15,24]. They suggested that elim-ination of atmospheric nitrogen via detection of biological agents in a rare gas matrixsuch as argon might allow use of information contained in the CN bands for furtherdiscrimination.

Nitrogen is associated with numerous moieties in biological materials, includingamino acids, proteins, and enzymes, and thus may provide additional discrimination.However, since at nanosecond time scales the LIBS plasma essentially erases all molec-ular information, the intensity of the CN bands observed in LIBS with nanosecondlasers is dependent on the carbon concentration, the nitrogen concentration, the matrix(which influences recombination chemistry) and the plasma parameters, which influencethe cooling rate in the plasma. Hence these CN emissions are likely only a reliablemarker for the C/N elemental ratio in a particular plasma volume, but not for the originalproportion of C and N bonds.

Baudelet and colleagues [25] have shown in a preliminary study that femtosecondlaser pulses may be able to preserve information pertaining to C and N in the originalmaterial. Specifically, they observed prompt CN emission �t < 200 ns� from both abiological sample (E. coli) which has CN bonds in amino acids and from nitrocellulosefilters (which contain C and N, but few CN bonds) during LIBS measurements with a120 fs pulse of 810 nm laser light, while CN emission from graphite in air under thesame conditions was delayed by ∼200 ns or more from the laser pulse. This shows thatwhen both the C and N are part of the sample material, the emission is much faster thanwhen the excited N or N2 is contributed from the plasma plume. Further, they note thatthe CN emission from the E. coli sample has a decay time constant roughly half thedecay time of the nitrocellulose (94 versus 185 ns), and attribute this to the presence ofablated CN fragments in the case of the biological sample, as opposed to unassociatedC and N in the nitrocellulose sample. While this is not conclusive, as the effect couldbe due to matrix effects or other considerations, it is certainly possible that the presenceof the native CN bond is associated with a faster CN emission decay.




Morel et al. [18] consider the sum of the intensity ratio of particular lines in 10single-shot spectra, the “cumulative intensity ratio” (CIR), defined by

CIR =10∑

i=1

�Ia/Ib�i� (1)

where Ia and Ib are the integrals of individual emission lines measured for each shot.They found that of several combinations of lines, the Phosphorous line at 253.56 nmand the Carbon line at 247.86 nm were a highly reproducible combination in the CIR,with a low standard deviation. Figure 3 illustrates the CIRs for the several of the testedbacteria, including two strains of Bg, and also for two strains of pollen. The diagonally-banded bars on the top of each histogram illustrate the standard deviation of the P/Cratio calculated from the 10 shots. From this preliminary data set it seems possible todiscriminate this limited set of biological samples on the basis of the P/C ratio alone. Theauthors concluded on the basis of this initial study that LIBS provided some substantialbenefits, including in situ, real time operation and sensitivity to a wide range of elements.

A study in 2004 by Kim et al. [26] also used a similar variable to separate bacterialsamples of Bacillus subtilis, Bacillus thuringiensis, Bacillus megaterium, and E. coli.Emission attributed to phosphate functional groups at 588.1 nm was normalized by an Feline at 578.9 nm, and this was plotted against Calcium emission at 393.7 nm normalizedto an Mn line at 398.3 nm. Both the Fe and Mn lines were at least moderately strong inall of the bacterial samples. Data from 15 independent laser shots (5 areas in 3 differentcultured samples) provided good separation between nearly all of the samples, as shownin Fig. 4. The authors suggest that additional work in detailed spectral fingerprinting of

0

Bacillus g

lobigii 1

Bacillus t

huringiensis

Escheric

hia coli

Staphyloco

ccus a

ureus

Proteus mira

bilis

Poplar pollen

Elin pollen

Bacillus g

lobigii 2

0.2Cum

ulat

ive

inte

nsity

rat

io o

f P/C

0.4

0.6

0.8

1.2CIR of PIC Standard deviation

1

Fig. 3. CIR of P/C ratio for several bacteria, from [18]. Dark bars indicate standard deviation.



320 S. G. Buckley

Intensity ratio of 393.7 nm/398.3 nm

Inte

nsity

rat

io o

f 588

.1 n

m/5

78.9

nm

00

0.5

1

1.5

2

2.5

3

1 2 3 4 5 6

LB

B1551

PV361

B.sub

B.thur

E. coli

7

Fig. 4. Separation of several bacterial samples based on ratios of LIBS lines, from [26].

species of interest would allow chemometric-type software analysis [22] which couldbe applied to the development of spectral libraries to improve methods of classificationand sample discrimination.

Boyain-Goita et al. [19] did some notable work on LIBS and Raman detection ofindividual pollen spores attached to needle tips. In their LIBS experiments, where CN,C2, Ca, and several trace elements were observed, they found that there was a largedegree of variation between samples, and while it was possible to normalize spectraand minimize the pulse-to-pulse variation, they found it difficult to identify featuresor patterns in the spectra associated with particular pollen type. They concluded thatmultivariate and pattern-recognition techniques should be applied to LIBS analysis ofbioparticles to improve discrimination. They came to a similar conclusion regardingtheir Raman measurements, in which they were able to see particular vibrational featuresassociated with plant structure.

3.2. Airborne samples

The previous section has highlighted work with biological material that has beendeposited on a surface or pelletized. In 2003 Hybl et al. [20] published work on LIBSdiscrimination of airborne biological aerosol. These experiments eliminate matrix effectsexcept those due to the carrier gas, required no sample preparation, and are likely tobe closer to a real-world sampling scenario than deposited samples might be. Twogeneration methods and two detection systems were considered. The first method usedthe shock wave from a repetitive (5 Hz) LIBS plasma to disturb and entrain a pile of




sample located in a small chamber. As the sample was aerosolized, it entrained into theplasma volume. After a few shots of the laser, reproducible spectra were obtained insuccessive shots from the continuously refreshed sample in the plasma volume. Thesewere collected using an Ocean Optics spectrometer system covering the spectral region200–825 nm with a nominal resolution of 0.15 nm per pixel.

In addition, a Pitt generator was used to acoustically excite aerosol samples into aco-flowing air stream, in which spectra from a 200 mJ/pulse Nd:YAG operated at 532 nmare resolved by a 0.25 m spectrometer onto a intensified CCD camera. The excitation andair flow rates were adjusted such that bioaerosols were detected in the plasma emissionapproximately 10% of the time. Only a limited spectral range of approximately 50 nmcould be collected with each laser shot using this system.

The broadband measurements with the Ocean Optics spectrometer were performedon three samples of Bg, three different protein/growth media samples, three differentpollen samples, and three different fungal spore samples. As expected from Table 1,significant variation was observed in elemental emission signals from different airbornesamples. In particular, the spectral power associated with lines of Mg, Ca, Na, and Kwere observed to vary strongly and systematically between different classes of material.When the 30 strongest spectral features from averaged spectra were used as a trainingset for PCA analysis, individual spectra from various classes separated very well in a3-dimensional PCA plot, as shown in Fig. 5. Further, on two-dimensional PCA plots,individual members of each class (e.g. each of the three Bg samples, each of the threepollen samples) separated reasonably well in some cases, as shown in Fig. 6.

1.0

0.5

0.0

PC

A 3

–0.5

–1.0

PCA 1

Media

Bg

Pollen

Fungal spores

–1.0–0.5

0.00.5

1.0

PCA 2

1.0

0.5

0.0

–0.5

–1.0

Fig. 5. Three dimensional PCA showing discrimination between anthrax simulant and variouspossible classes of interferences, from [20].



322 S. G. Buckley

PCA 1

PC

A 2

1.0

0.5

0.0

Bg B

Bg A

Bg C

–0.5

–1.0 –0.5 0.0 0.5

OvalbuminLB broth

Brain-heart infusion

PCA 1

PC

A 2

1.0

0.5

0.0

–0.5

–1.0–1.0 –0.5 0.0 0.5

Penicillium

Corn smut

Oat smut

1.0

1.1

0.9

0.8

0.7

0.6

0.5

PCA 1

PC

A 2

–1.0 –0.5 0.0 0.5

Fig. 6. Two-dimensional PCA showing discrimination of samples within particular classes,from [20].

Measurements performed with the Pitt generator and the intensified CCD camerawere also instructive; while it is virtually certain that each of the “particle hits” was anagglomerate of several spores, it is likely that spores of a bioterrorism agent would beagglomerated following a release. Hence while the concentrations of inorganic elementsper spore from Table 1 would appear to be near the LIBS mass detection limits, thesemeasurements of individual agglomerates were encouraging in that single clumps ofairborne particles could be detected with LIBS. The authors concluded that LIBS mightbe considered a useful complement to existing techniques for bioaerosol detection, suchas fluorescence. Further research into spectral processing was recommended.

Dixon and Hahn [27] published an assessment of the feasibility of detection ofsingle bacterial spores using LIBS. They carefully generated single aerosolized sporesof Bacillus atrophaeous, Baccillus pumilus, and Bacillus stearothemophilus using apneumatic nebulizer, recording total mass loss of the nebulizer over time to calibrate thespore flow rate. The air stream, which was previously dried, was also carefully controlled.Spore concentration near the LIBS sample point was verified using an independent lightscattering measurement to be between 2–5 spores/cm3. Light from a 5 Hz, 275 mJ/pulse1064 nm Q-switched Nd:YAG laser was focused into the aerosol stream. Light wascollected on axis with the incident laser beam using a pierced mirror, dispersed usinga 0.275-m spectrometer, and imaged onto an intensified CCD camera. The hit rate ofspores was controlled to be approximately 1%, yielding a negligible number of “doublehits” (two spores in the same plasma volume) providing that spores are unagglomerated.Hence most spectra were considered to contain a single bacterial spore, and spectra werecarefully processed to avoid false hits.

In this study, Ca was the only trace element emission line visible in the single-sporespectra. 40 individual particle hits from B. atrophaeous were examined in detail, yieldingan average mass of 3.1 femtogram (fg) of Ca with a residual standard deviation of 39.6%.The average computed from single particles compared well with the ensemble averagemass of 2.6 fg. From measured spore dimensions and the LIBS-measured mass, a massfraction of 0.5% was calculated, in reasonable agreement with published values expectedfor several species. Further, the detection limit of Ca assessed in these measurementsagrees well with previously published results by the same group.

An intensive effort was made to detect Na and Mg in spectra from these single spores,with an evaluation of 54,000 laser spectra revealing not a single detectable Na or Mg




line. This “rather definitive” effort was judged by Dixon and Hahn to be significant, dueto the body of previous work relying on fingerprinting of bioaerosols using multiple traceelemental lines. These authors concluded that single-spore LIBS analysis is not feasible,and also made valid comments concerning the difficulty of detection of bioaerosolsagainst an ambient background with LIBS alone.

Additional work on detection of single submicron particles with LIBS has also shownthat absolute mass quantification can be problematic. For example, in 2005 Lithgow andBuckley showed that emission from individual nanoparticles engulfed in a plasma viewedfrom two different directions showed little correlation [28]. This study was followedby another by Lithgow and Buckley in the same year illustrating that single-particleemission is distributed locally (rather than homogeneously) in LIBS plasmas [29]. Thesemeasurements have been confirmed by images of particles in plasmas by Hohreiter andHahn that show emission from ablated particles expanding in time [30]. It is expectedthat these considerations, plus particle-plasma energy transfer limitations that imposean upper size limit for particle measurements [31], may also be important for themeasurement of supermicron bacterial spores.

4. CONCLUSION

This chapter has outlined the recent LIBS work related to explosive, chemical, andbiological hazard detection. From these studies it is clear that there is a surprising amountof useful information in LIBS spectra pertaining to measurement of these molecular andcellular moieties, but it is also clear that in these applications LIBS presently mightbe expected to play a supporting, rather than leading, diagnostic role. Improvementsin LIBS sensitivity from new, dedicated LIBS hardware, and improved employmentof statistical methods (e.g. chemometrics and PCA analysis) are both expected to playmajor roles in improvement in LIBS-based discrimination of chemical and biologicalsamples in the future.

REFERENCES

[1] L.J. Radziemski, T.R. Loree, D.A. Cremers, and N.M. Hoffman, Anal. Chem, 55 (1983) 1246.[2] L.J. Radziemski, D.A. Cremers, and T.R. Loree, Spectrochim. Acta B 38 (1983) 349.[3] D.A. Cremers and L.J. Radziemski, Anal. Chem. 55 (1983) 1252.[4] D.W. Hahn, W.L. Flower, and K.R. Hencken, Appl. Spectrosc. 51 (1997) 1836.[5] S.G. Buckley, H.A. Johnsen, K.R. Hencken, and D.W. Hahn, Waste Management 20

(2000) 455.[6] C.K. Williamson, R.G. Daniel, K.L. McNesby, and A.W. Miziolek, Anal. Chem. 70

(1998) 1186.[7] T.X. Phuoc and F.P. White, Fuel 81 (2002) 1761.[8] R. Sattmann, I. Monch, H. Krause, R. Noll, S. Couris, A. Hatziapostolou, A. Mavromanolakis,

C. Fotakis, E. Larrauri, and R. Miguel, Appl. Spectrosc. 52 (1998) 456.[9] F. Ferioli, P.V. Puzinauskas, and S.G. Buckley, Appl. Spectrosc. 57 (2003) 1183.

[10] V. Sturm and R. Noll, Appl. Opt. 42 (2003) 6221.[11] F. Ferioli and S.G. Buckley, Combustion and Flame 144 (2006) 435.



324 S. G. Buckley

[12] F.C. De Lucia, R.S. Harmon, K.L. McNesby, R.J. Winkel, and A.W. Miziolek, Appl. Opt. 42(2003) 6148.

[13] F.C. De Lucia, A.C. Samuels, R.S. Harmon, R.A. Walters, K.L. McNesby, A. LaPointe,R.J. Winkel, and A.W. Miziolek, IEEE Sensors J. 5 (2005) 681.

[14] A. Portnov, S. Rosenwaks, and I. Bar, Appl. Opt. 42 (2003) 2835.[15] L. St-Onge, R. Sing, S. Bechard, and M. Sabsabi, Appl. Phys. A-Mater. Sci. & Processing

69 (1999) S913.[16] C. López-Moreno, S. Palanco, J.J. Laserna, F. DeLucia, A.W. Miziolek, J. Rose, R.A. Walters,

and A.I. Whitehouse, J. Anal.l At. Spectrometry DOI: 10.1039/b508055j (2006).[17] A.C. Samuels, F.C. DeLucia Jr., K.L. McNesby, and A.W. Miziolek, Appl. Opt. 42

(2003) 6205.[18] S. Morel, N. Leone, P. Adam, and J. Amouroux, Appl. Opt. 42 (2003) 6184.[19] A. Boyain-Goitia, D.C.S. Beddows, B.C. Griffiths, and H.H. Telle, Appl. Opt. 42 (2003) 6119.[20] J. Hybl, G.A. Lithgow, and S.G. Buckley, Appl. Spectrosc. 57 (2003) 1207.[21] O. Samek, D.C.S. Beddows, H.H. Telle, J. Kaiser, M. Liska, J.O. Caceres, and A.G. Urena,

Spectrochim. Acta B 56 (2001) 865.[22] D.L. Massart, B.G.M. Vandeginste, S.N. Deming, Y. Michotte, and L. Kaufman, Chemo-

metrics: a textbook. Data Handling in Science and Technology. Vol. 2. 1988, Amsterdam:Elsevier Science Publishers.

[23] M. Meloun, J. Capek, P. Miksik, and R.G. Brereton, Anal. Chim. Acta 423 (2000) 51.[24] C. Vivien, J. Hermann, A. Perrone, C. Boulmer-Leborgne, and A. Luches, J. Phys.

D-Appl. Phy. 31 (1998) 1263[25] M. Baudelet, L. Guyon, J. Yu, J.P. Wolf, T. Amodeo, E. Frejafon, and P. Laloi, Appl.

Phys. Lett. 88 (2006) 063901.[26] T. Kim, Z.G. Specht, P.S. Vary, and C.T. Lin, J. Physical Chem. B 108 (2004) 5477.[27] P.B. Dixon and D.W. Hahn, Anal. Chem. 77 (2005) 631.[28] G.A. Lithgow and S.G. Buckley, Appl. Phys. Lett. 87 (2005) 011501.[29] G.A. Lithgow and S.G. Buckley, Spectrochim. Acta B 60 (2005) 1060.[30] V. Hohreiter and D.W. Hahn, Anal. Chem. 78 (2006) 1509.[31] J.E. Carranza and D.W. Hahn, Anal. Chem. 74 (2002) 5450.



Chapter 14

Life Science Applications of LIBS

F. Y. Yueha, A. Kumarb, and J. P. Singha

aInstitute for Clean Energy Technology (ICET), Mississippi State UniversityStarkville, MS 39759, USA

bDepartment of Physics, Tuskegee University, Tuskegee, AL 36088, USA

1. INTRODUCTION

Laser-induced breakdown spectroscopy (LIBS) is a laser based emission type diagnostictechnique for elemental analysis [1,2]. In LIBS, a laser beam is tightly focused on asample to ablate the material, thus creating a micro-plasma. The optical emission fromthe plasma contains the signature of the elements present in the sample material. LIBShas distinct advantages over other established analytical techniques. It requires very littlesample preparation. The analysis can be carried out at either contact or stand-off. LIBS,has the ability to provide depth-profiling on layered structures, and multiple elementalanalysis. It only consumes very small amounts of sample (nano-gram level) and analysiscan be completed very fast for a wide variety of materials and it can be applied directlyto a sample in situ. It provides a high degree of spatial resolution for measurement ofsample surfaces and gives qualitative as well as quantitative (∼ ppm range) information.It has been successfully applied to the analysis of artworks, biomaterials, cultural heritageobjects, environmental samples, explosives, industrial alloys, pharmaceuticals, and manymore. Not all LIBS applications require quantitative measurements. Qualitative LIBSmeasurements can be used for identification of various samples in terms of their elementalcomposition through the unique spectral signatures. With proper data base established,LIBS spectra of unknown samples can be compared with the chemical profiles of knownsubstances to find a possible match. It has been used to identify a volcanic rock to aparticular strain of bacteria [3–6].

Bio-samples (such as bone, teeth, hair) contain biological signatures from the livingphase. They store information on the habitat, nutrition, and other environmental con-ditions. The forensic analysis of trace elements in a biological sample, such as bone,hair, or nail can provide valuable crime scene evidence. The elemental analysis of thesesamples can also be used to identify health problems, such as identification of teethaffected by caries, detecting toxic metals in hair, nail or urine. The advantage of real-timeelemental analysis without sample preparation with LIBS is attractive for the analysisof biological samples. LIBS can provide spatial information, which is not possible by




326 F. Y. Yueh et al.

conventional elemental analysis methods without sample preparations such as ashingor acidic dilution. The application of LIBS to the problems of life science has beenexplored in recent years. In 1999, Samek et al. reported the application of laser-inducedbreakdown spectroscopy and laser-induced fluorescence (LIF) spectroscopy to the anal-ysis of boron, calcium, chromium, copper, iron, silicon and zinc, and toxic elements,such as aluminum, cadmium, lead and mercury, in the body [7]. Samples studied byLIBS may come from different parts of the body such as skin-tissue, finger-nails andteeth. Nakamura et al. [8] have carried out LIBS investigations of human nail, hair andtooth. They were able to record Ca signals from these samples with great sensitivity.Based on their initial study, they found that Ca signals from the hair of healthy femalesare significantly decreased with age in good agreement with medical reports. In thischapter, various life science applications of LIBS will be discussed.

2. BONE & TOOTH ANALYSIS

LIBS, in addition to identifying many of the basic elements in an unknown sample, canalso distinguish different substances by comparison of unknown spectra with those ofknown substances. Due to this capability, LIBS has been applied to identify and sortdifferent materials for forensic science applications (e.g. bone, hair, nail and others)and industry processes (e.g. to identify alloy [9], polymer [10], etc). In this section theapplication of LIBS for the analysis of bone and tooth will be addressed.

In forensic investigations, the bones found from a possible crime scene must first beidentified as human. The traditional methods used for bone identification are macroscopicand microscopic evaluations. The accuracy of the identification using these methodsgreatly depends on the experience of the specialist. Also microscopic evaluation requireselaborate sample preparation. It is a destructive and time consuming method. LIBS hasthe capability of real-time analysis and minimum sample consumption and it is a perfectalternative method for bone classification. Collins and Vass [11] have evaluated LIBSfor bone classification. They have collected LIBS spectra of bones from human andanimals (rabbit, pig, sheep, bear, cow). To achieve adequate sampling, each spectrum(200–800 nm) is taken with a Nd:YAG laser of 40 mJ in a representative region of thebone averaging 100 laser shots. Clear elemental differences between human and animalbone are found in their initial spectral comparisons.

The invention of the laser has played an important role in dentistry. When a laserbeam is applied to gums, teeth and cavities, the microscopic explosions caused by thelaser make it act like a drill or a knife. Laser applications to dental cure and carehave progressed rapidly. From 1960s, lasers have been developed and approved forsoft tissue procedures. In 1997 FDA approved the Er-YAG laser for treating toothdecay. Now lasers have been routinely used to fix dental problems (e.g. cavities, treatgum disease, canker sores, and other problems in the mouth) including teeth whitening.Niemz [12,13] has evaluated the LIBS physical parameters during the laser-inducedablation of teeth. He used a picosecond Nd-YLF laser system and surfaces of extractedhuman teeth as target material. The laser operates at a wavelength of 1�053 �m withpulse durations of 30 ps and pulse energies up to 1 mJ. The laser beam was expanded to adiameter of 4 mm and focused to spot sizes of about 30 �m. The laser generated plasmasparks were spectroscopically analyzed and found to have mean plasma temperatures



Life Science Applications of LIBS 327

of about 5 eV and mean electron densities of about 1018/cm3. This laser system wasalso used to remove healthy and carious enamel. Emission lines from neutral and singlyionized Calcium and the Sodium D-lines were observed in the LIBS spectra. To obtainreproducible data, the raw LIBS data were normalized with the signal from the diffusereflected second harmonic laser light. By comparing the spectra recorded from severalsound and artificial caries regions of different teeth, it was found that the intensityof all observed mineral atomic lines are weaker from caries regions as compared tosound/healthy regions. Therefore, it is easy to distinguish caries infected teeth from thehealthy ones using LIBS data. This is enabling a computer controlled caries removalautomated system in the near future [13].

Samek et al. [14] have used a standard Nd:YAG laser and a fiber delivery/collectionsystem to demonstrate the feasibility of using LIBS for identification of carious teeth.The Nd:YAG laser (1064nm, 20 Hz, 4–8 ns pulse width) of 10–30 mJ was focused ontothe launch end of fiber. The laser light exiting the fiber was directed onto the desiredsample area (1.5–2 mm above the sample surface) (see Figure 1a). The LIBS signalemitted from the sample surface was collected by the same fiber and directed to a secondfiber via a mirror with a 2-mm diameter hole in the center. The LIBS spectra wererecorded with a spectrometer equipped with an intensified photodiode array detector.The LIBS spectra of samples from both healthy and carious tissues are recorded andthus form a database of reference spectrum. To differentiate the healthy and carioustissues, the spectra of the sample are compared with the spectra in the database. Sameket. al. used a pattern recognition algorithm called ‘Mahalanobis Distance’ to determinethe identity of an “unknown” tooth sample in real time (see Figure 1b). ‘MahalanobisDistance’ is based on correlations between variables by which different patterns can beidentified and analyzed. They have demonstrated the possibility of distinguishing thetransition from healthy to caries-affected tooth material based on one spectrum obtainedfrom the sample. This work shows that LIBS can be an useful tool for dentist to quicklyidentify caries-affected areas in the process of laser drilling or cleaning.

Samek et al. [15] have presented a proof-of-concept demonstration to quantitativelymeasure the minerals and toxic elements in representative calcified tissue samples(e.g. teeth from infants, children and adults and shin and thigh bones) using LIBS.They have scanned the teeth and bone samples to get both one-dimensional and two-dimensional mapping of the elemental contents. They used calibration data which wereobtained from calcified tissue-equivalent material pressed in the form of pellets (majoritycompound of pellets is CaCO3 and about 100–10000 ppm Al, Sr and Pb compoundsadded) to quantify the trace elements. The LIBS analysis results of calcified tissue sam-ples based on the calibration method described above were compared with the resultsfrom atomic absorption spectroscopy (AAS) and were found to be in good agreement.

LIBS has been recently applied to the areas of anthropology and archaeology todetermine the elementary composition of unique objects that have cultural value. Bilmeset al. have used LIBS to identify the trace elements in Hominide teeth to analyze theireating habits [16]. Tawfik and El-Tayeb have studied 150 human enamel from Egyptiansfrom the Old Kingdom (2770-2200 BC) to recent age with LIBS [17]. They determinedthe elemental level (such as Ca, Pb, Al, Sr) in teeth of ancient and recent Egyptians toprovide information for studying the aetiology of various diseases during this period.They found higher Ca, Pb and Al and lower Sr levels in ancient Egyptians as comparedto the results found from recent Egyptian teeth. The high Pb and Al levels in ancient




EnamelDentin

PulpGingiva(gum)

(a)

(b)

Rootcanal

To thetarget site

Launchend

Spectrometerfibre

Focusinglens

Silvered mirror with2 mm diameter holethrough the centre

Polarisorand half-wave plate

Pulsed Nd:YAGlaser @ 1,064 nm

Nd:YAG mirror1064 nm

Reference spectra“carious tissue”

Reference spectra“healthy tissue”

Generation of discriminantmodels:

“healthy” and “carious”and reference library

Discriminant modelHealthy tissue

Discriminant modelCarious tissue

Spectrum fromunknown sample

Yes

No

Give alarm

HeNe laser foralignment (optional)(f = 250 mm)

(f = 75 mm)

2 × collectionlenses

Blood vesselBone

Fig. 1. (a) LIBS experimental setup for identification of carious teeth. (b) Principle of sam-ple identification/screening applications based on a discriminant analysis. Here a warning isgiven when healthy tooth material is targeted during laser drilling. (Reproduced with permissionfrom Ref. [14]).




Egyptians indicated that the environment might have been polluted by these metals atthat point of time. The study of the difference of elemental levels in teeth betweenpeople at different time periods can provide information on their environment and eatinghabits.

3. HAIR & NAIL ANALYSIS

Elements normally found in the body (e.g. Cu, Cr, Zn, etc.) can also be found in thehair but with inconsistent levels. Therefore, hair testing has a limited usefulness inmedical practice. However, hair analysis is useful for detecting those compounds thatare not normally found in the body. Since the structure of the hair remains unchangedafter collection, the minerals and trace elements remain fixed and the levels of theseelements will not change even a few years later. Thus, hair analysis is an excellent wayto determine mineral and trace element concentrations. It can be used to detect elementimbalance or toxicity in the human body. It is an ideal complement to serum and urineanalysis. Hair analysis has been shown to be quantitatively useful for the detection ofberyllium, lead, cadmium, nickel, arsenic, and methyl mercury. Hair analysis can also beused to find the presence of drugs (of abuse) or the presence of certain pharmacologicalagents. It is also noted, that hair analysis might be used to diagnose mineral deficiencies.Since hair can be contaminated by air, water, perspiration, shampoos, dyes and otherhair preparations, it is necessary to follow certain washing procedures to avoid false testresults. Hair analysis with traditional methods is simple, but test results greatly dependon using the correct sample preparation procedure. Nail analysis can be used to replacehair analysis when hair loss or other reasons prohibit hair analysis.

Haruna et al. [18] have demonstrated the use of LIBS to detect calcium in humanhair and nail. High sensitivity for Ca detection (0.1 percent in human hair and nail)can be achieved by the use of either a UV or a near-IR laser pulse. The result is veryencouraging and shows that Ca detection in human hair may lead to new diagnosis,including a monitor for daily intake of Ca and a screening method for osteoporosis. Intheir work, they were also able to detect sodium and carbon. Their experimental workuses a low-energy laser pulse to illuminate the tissue samples and needs no poisonoussensititizers like a fluorescent dye. These results also show that LIBS has the potentialto be developed as an optical biopsy tool in the future.

The research group at Instituto per i Processi Chimio Fisici (IPCF) in Pisa, Italy hasrecently applied LIBS to hair-tissue mineral analysis [19]. They have evaluated the hairfrom people who might have toxic metal poisoning problems. They focused a 150 mJlaser beam from a Nd:YAG laser (1064-nm, 10Hz, 8-ns pulse-width) on a single hair(held by an U-shaped sample holder). A broadband Echelle spectrometer equipped withan intensified CCD was used to record hair spectra. Twenty LIBS spectra were recordedalong the length of the hair. The spectra were averaged and analyzed to obtain theconcentration of the main minerals present in human hair. They used a calibration-freemethod to quantify the results. In this method, the concentration is calculated based on theintensity of the atomic lines, plasma parameters (i.e. plasma temperature and electronicdensity), and available spectroscopic constants. They have compared the results of theLIBS analysis with the results obtained from a commercial analytical laboratory andreasonable agreements between LIBS and the commercial analytical tool were obtained.




Branch et al. [20] have also applied LIBS to elemental analysis in hair and nail. Theyexplored LIBS ability to monitor the level of trace elements like magnesium, sodium andcalcium in human hair and nail. A frequency-doubled Nd:YAG laser was used to carryout the measurements. The gate width and delay time were adjusted for each elementto achieve the best possible signal-to-noise ratio. This work determined the optimalexperimental configuration to obtain maximum LIBS signal for each of the elementstested.

4. BLOOD ANALYSIS

In 1999, Samek et al. [7] developed a LIBS (in combination with LIFS) analysis systemfor mineral analysis in skin tissue, finger nails and teeth. The same system has beenused to screen blood samples [21]. To detect trace metals in blood, they transferred theblood sample to a standard filter paper and the LIBS measurement was conducted on thefilter paper. They have successfully detected trace amounts of Rb (level down to 0.3%)in blood by LIBS [21]. It shows that a semi-quantitative result can be used to trace theeffect of illegal drug doping in less than a few minutes.

Many clinical applications require the analysis of single-cells. This type of chemicalanalysis is very challenging. A sensitive analytical technique for a single biologicalcell significantly improves the early detection of some medical problems, and evenmonitor medicine uptake. The technique should provide a reasonably good samplingrate to achieve meaningful statistics for a clinical profile. It should also be compatiblewith water and the atmosphere. One challenging task in this type of analysis is toquantify the metabolic electrolyte (i.e. Na and K) in a single red blood cells (RBC). Thequantities of Na and K are generally much lower than the detection limit of conventionalatomic emission and absorption spectrometry. Cytometry and fluorescence can be usedfor this type of analysis but they rely on some specific fluorescent tags for detection.Recently, Ng and Cheung [22] have demonstrated the feasibility of quantifying sodiumand potassium in single human erythrocytes using a modified LIBS experimental setup (see Figure 2). High fluence lasers, however, produce a high temperature plasma(few eV) and can cause extensive ionization of most elements which is not ideal forthe quantitative analysis of most elements. To avoid the problem of generating a hotplasma plume with an IR or visible laser pulse, they decided to produce the plasmawith an ArF laser (193-nm, 15-ns pulse) at sub-breakdown fluences �∼4 J/cm2�. Theresulting plasma temperature is about 0.5 eV which is ideal for the excitation of neutralatoms of K and Na. Details of the experimental setup for this measurement is describedelsewhere [22]. They also designed a special jet sampling system to achieve streamlineblood flow of 8 �m diameter (see Fig. 2) to avoid the plasma emission quench problem.To ensure that the blood sample flowed on the outside of the sheath, a slightly bent quartzcapillary was used at a point 8-mm above the tip, and the capillary wall was etched to5 �m thickness at the slanted tip (see Fig. 2, inset). The blood sample (10 times dilutedRBC suspended in 8% glucose) is pressure-fed down the quartz capillary at a flow rateof 4 �L/min. Optical multichannel detection with a 0.5-m spectrometer equipped with a600 l/mm grating and ICCD detector and nanosecond gating (70 ns gate, 1 �s gate-width)was used in the spectrochemical analysis of the emissions from the plasma produced bylaser ablation of blood cells confined in the sheath during the flow. In this study, they




ArF laserpulse

Liquidjet

F

ICCD

S

Fig. 2. Schematics of the experimental setup for detection of Na and K in single human bloodcells. F: UV-blocking filter, S: spectrometer; ICCD: intensified charge coupled device; Inset:magnified nozzle of the flow cell. The tapered capillary (50-um i.d.) rests against the inside of theouter glass tubing (400-�m i.d.). Blood cells (8-�m diameter) are allowed to flow on the outsideof the sheath in single file for eventual sampling downstream (Reproduced with permission fromRef. [22]).

have used two schemes to capture a single blood cell. In the first scheme, they sampledsingle blood cells that happened to be in the ablation volume. In the second scheme,they used a synchronous sampling method in which a He-Ne laser focused above theablation area was used to “spot” the individual blood cells and send a delay signal tostart the ablation laser downstream. Using these sampling schemes, they were able torecord LIBS spectra of single ablated cells. The ratios of the analyte line intensityto the root-mean-square fluctuation of the continuum background were found to be about18 for sodium and 30 for potassium. Based on the signal-to noise ratio, they estimatedthe mass detection limits for K and Na on the order of 30 fg �10−15 gm�� and 2 fg in redblood cells, respectively. This pioneering LIBS work on single RBC, shows that LIBScan be used to analyze the elemental content of most aqueous samples and suspensionsof live cells.

5. URINE STONES ANALYSIS

Laser induced breakdown has been used to treat patients with urinary and kidneycalculi since 1987 [23]. It uses the shock resulting from the laser-induced breakdownto disintegrate the calculus into tiny fragments. LIBS has recently been used to analyzeand identify elemental constituents of urinary calculi. Fang et al. [24] have made LIBSmeasurements on seven different urinary stone samples. The key elements identified




from the urinary stone samples are Calcium, Potassium, Sodium, Magnesium, Lead andSamarium. The concentrations of these key elements in each sample were estimatedbased on pre-generated calibration data for each element. They found the concentrationsof these key elements to be quite different in different samples but it was noticedthat the concentrations of Ca and Pb have some correlation (i.e. high Pb concentrationsamples also contain high concentrations of Ca). This initial work on Urine stone analysisshows that LIBS has great potential for routine clinic applications in urological disorderdiagnosis.

6. TISSUE ANALYSIS

Although, the LIBS technique has been applied in biological investigations in thepast [17,25,26], literature about LIBS on bio-matrix materials is sparse. This may bedue to several reasons. First, the hardness of the biological tissue samples is less thanmetals or other solid materials and the laser ablation process destroys the sample surfacemuch more rapidly, leading to weaker focusing and thus creating poor reproducibility ofthe signal. Secondly, biological samples are more inhomogeneous in most cases whichagain gives rise to poor reproducibility of results. Finally, detection of molecular speciesis important in biology, which is normally beyond the capabilities of LIBS. However,application of LIBS to tissue analysis still attracts great interest, because LIBS canprovide rapid tissue analysis without any sample preparation.

To understand the physical mechanism involved in the ablation process in orderto obtain the best conditions for analysis of biological tissue, several research groupshave studied the ablation plume at various stages of its evolution [27,28]. In 2003,Souza et al. [29] used LIBS to investigate the relative elemental composition in chickmyocardium tissue. In their investigation, a Nd:YAG laser (1064-nm, 9 ns pulse-width)was used for tissue ablation. The LIBS signal was collected through a 600 �m fiber andsent to a 0.25m spectrometer equipped with ICCD detector. They were able to identifyNa, K, Ca, H and other elements in the chick myocardium tissue under the best detectionwindow (5 �s gate). Zheng et al. [30] have recently investigated the feasibility of usingLIBS to characterize animal tissues. LIBS spectra of samples including brain, kidney,liver, lung, muscle and spleen tissues from a dog were collected. All tissue sampleswere kept frozen under −20 �C in a refrigerator before being tested. A custom-designedsmall cooling unit was used to keep the tissue samples frozen at around −20 �C whilethe experiment was performed. The small cooling unit was being translated during themeasurement. To achieve good statistical data, they recorded 50 to 100 spectra fromeach type of tissue samples. Typical LIBS spectra of brain and kidney tissues are shownin Fig. 3. Gornushkin et al. have used a simple statistical correlation method for solidmaterial identification [31] and Zheng et al. have adopted the same technique for tissueidentification [30]. They used intensity ratios of the trace element analyte lines and theCa 393.367 nm line to establish the correlation data among the tested tissue samples.Figure 4 shows the correlation plot of Brain tissue with two different unknown tissues.The linear correlation coefficient is used to determine the goodness of the correlationbetween two samples. Using this simple technique about 80% of the unknown samplescan be correctly identified. It is believed that a better identification accuracy can beachieved with more advanced pattern recognition techniques.




50000

45000

40000

35000

30000

30000

25000

20000

15000

10000

5000

0

25000

20000

15000Mg

Mg

Mg

NaNa

Na

Fe

Fe

Ca

CaCa

Ca

CaCa

Brain

MgK

K

K

K

10000

5000

0270 370 470

Wavelength (nm)

Inte

nsity

Inte

nsity

570 670 770

Kidney

270 370 470

Wavelength (nm)570 670 770

Fig. 3. LIBS spectra of brain and kidney tissue.

Kumar et al. [26] have demonstrated the first LIBS experiments to explore thepossibility of using LIBS for cancer detection. They have analyzed malignant andnormal tissue from a canine haemangiosarcoma. Canine hemangiosarcoma, a model forhuman angiosarcoma, may be valuable to define and analyze these types of tumors andsuggest potential means of improving their classification in humans. To study caninehaemangiosarcoma with LIBS, haemangiosarcoma and normal liver samples �1 × 1 ×2 cm� were taken from the liver of a dog. Each sample was bisected and processed




R2 = 0.859

R2 = 0.9987

0

0.05

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.450

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

Intensity ratio from brain tissue

Inte

nsity

rat

io fr

om u

nkno

wn

tissu

e

Unknown #5Unknown #6

Fig. 4. Correlation plots of the intensity ratio from brain tissue and an unknown sample.

following a special procedures. Sections (8 �m thick) were cut using a “Cryocut” cryostat(8 �m; Reichert-Jung), placed onto plain glass slides, and either air dried, or air dried,fixed in acetone (10 min), and air dried. Some samples were manually cut (so called“thick-cut” samples) to about ∼2 mm thickness. A frequency-doubled Nd:YAG laser(10 Hz, pulse width 5 ns) was employed in all the tissue measurements and backwarddetection was used in the experimental configuration. The detection system included anUV-VIS-Echelle optical spectrograph equipped with a 1024 ×1024-element intensifiedcharge-coupled device (ICCD). The spectrograph covers the 200–780-nm spectral rangewith a spectral resolution ��/�� of 40,000. The detector was operated in a gated modeand was synchronized to the laser output. They used a detection window that providedthe best signal-to-noise ratio to record the LIBS spectra of normal and malignant tissues.Samples of ∼2 mm thick tissue were placed on a glass slide. The whole glass slide wasmounted on a rotating pad, and the rotation of the pad was adjusted so that the laserlight did not hit the same spot more than once. At least twenty spectra (each spectrumis ten laser shots averaged) for each sample were collected. Figures 5 and 6 show theLIBS spectra of malignant and normal tissue cells in two spectral regions. There is aclear difference between the spectra of normal and malignant tissue. The intensities ofvarious element-lines, which are related to the concentration of trace elements in normaland malignant tissue, are significantly different. The elements identified from the LIBSspectra of tissues were Ca, Al, Fe, Cu, Na, K, and Mg. The intensity of Fe lines frommalignant cells were found weaker compared with the intensity from the normal cells.They also noticed that the intensity of Ca lines and Al lines were weaker from themalignant tissue cells. However, Cu lines were found to be much stronger in the normaltissue. They have compared the spectra recorded with thin tissue and thick-cut samples.LIBS spectra from thin tissue showed a higher variation and a poorer signal-to-noise ratiocompared with that from the thick-cut sample and all the later data were obtained fromthe thick-cut samples. To reduce the effect of pulse-to-pulse laser variations, they usedthe K 766.491 nm line as the reference line in their analysis. Intensity ratios of the majorelements with the K reference line were analyzed. They found that the intensity ratios

ElsevierAMS

Jobcode:

LRN

Ch14-N51734

24-7-2007

8:07p.m.

Page:335

Trim:165 ×

240MM

TS:Integra

Font:

Times

Size:10/12pt

Margins:Top:13MM

Gutter:20MM

T.Area:125mmX198mm

4Color

COP:Recto

Floats:

Inline

Life

ScienceA

pplicationsof

LIB

S335

0

200

400

600

800

1000

1200

1400

1600

0.404 0.4045 0.405 0.4055 0.406 0.4065 0.407 0.4075 0.408

K

Fe

Fe

Fe

K

0

500

1000

1500

2000

2500

3000

0.317 0.319 0.321 0.323 0.325 0.327 0.329 0.331

Normal tissueMalignant tissue

NaCa

Cu

Cu

Wavelength (µm)

Normal tissueMalignant tissue

Inte

nsity

Fig. 5. LIBS spectra of the liver tissue of a dog in two different spectral regions.




1021.8

Normal

Malignant

816.9

612.0

407.1

202.2

C

C

Fe

Fe

FeFe

Fe

Fe

Fe

Mg

Mg

Fe

Al

Al

Cu

Ca

Ca

Na

Na

Zn FeFe

Fe

Fe

469.8

373.7

277.6

181.5

85.4

–10.7246.3 259.5 272.6 285.8 298.9

Wavelength (nm)

Inte

nsity

Inte

nsity

312.1 325.2 338.4 351.5 364.7 377.8

–2.6246.3 259.5 272.6 285.8 298.9 312.1 325.2 338.4 351.5 364.7 377.8

Fig. 6. LIBS Spectra of the liver tissue of a dog. Top: normal tissue; Bottom: malignant tissue.Top inset: the photo of the normal tissue sample; Bottom inset: the photo of the malignant tissuesample.

of Ca/K and Na/K are higher in the malignant tissue spectra whereas the concentrationof copper is low in malignant tissue in comparison with that in normal tissue. The Mg/Kand Al/K are comparable in the normal and malignant tissue spectra, and Cu/K is lowerin the malignant spectra. This indicates that the concentration of trace elements like Ca,Na and Mg might be higher in malignant cells in comparison with that in normal cells.Since they were not able to establish LIBS calibration data, the elemental compositionsare not quantified in this work. To compare LIBS and ICPES data, they simply comparedthe intensity ratio obtained from LIBS with the concentration ratio obtained from theICPES measurements. This comparison is valid for optically thin laser-induced plasmabecause the intensity ratio is linearly proportional to the concentration ratio. Figure 7




0

2

4

6

8

10

12

14

16

18

20

Al/K Ca/KC Cu/K Fe/K Mg/K Na/K

Nor

mal

tiss

ue/M

alig

nant

tiss

ueLIBSICP

Fig. 7. Comparison of the results of tissue analysis from LIBS and ICPES (Reproduced withpermission from Ref. [26]).

shows the intensity ratios obtained from LIBS data of malignant tissue and normal tissuewith the concentration ratios obtained from ICPES analysis for malignant tissue andnormal tissue. The LIBS data are found to be in reasonable agreement with the ICPESdata. The percentage difference between these two measurements was less than 12%for Al, Fe, Mg, and Na. The higher percent differences for Cu and Ca might be dueto the self-absorption effects. Since both Cu and Ca lines used in the LIBS analysisare resonant lines, the line intensity might not be proportional to concentration due toself-absorption of these lines. Although the result in this study is very preliminary, it stilldemonstrates LIBS’ potential for development as an in vivo diagnostic tool for cancerdetection. Extensive development in this area is needed to obtain quantitative results forpractical applications.

7. CONCLUSIONS

The application of LIBS to the analysis of biological and medical samples has beeninvestigated in recent years. Some preliminary LIBS studies on hard (bone, teeth, hair,nail) and soft tissue and even blood/urine samples have shown evidences that LIBS hasthe potential to be a useful tool for life science applications. However, like any tech-nique before it matures, it requires many researchers to solve the complex experimentalproblems to make the technique reliable. For example, one major problem for LIBSanalysis of biological samples is that the elemental composition of biological materialcan be quite different within a healthy individual of the same species. Therefore, it putsthe LIBS’ ability to discriminate between healthy and unhealthy individuals in doubt.The LIBS signature of a particular biological target, may change rapidly since the tissuefrom living animals can easily be affected by the environment. Also, the selections of




equipment are equally important for the development of this technique for life scienceapplications. A laser system that provides enough laser energy to ablate only the biologi-cal sample from the desired spot with minimum impact on the surrounding tissues is veryimportant to tissue analysis. Ultra-short pulse lasers seem to be good candidates to beused in tissue analysis because ultra-short laser pulse ablation removes material with therequired low energy fluence and causes minimal collateral damage. Kim et al. [27] havestudied the influence of pulse duration on ultra-short laser pulse ablation of biologicaltissues. This type of study will be helpful for biomedical applications because it willprovide the information of how the physical conditions change and the effects of shock-wave and heat on the tissue. A detection system that provides high sensitivity and alsoimaging capability will be very appropriate for the bio-samples. To perform quantitativeanalysis on biological samples, will require many experiments to determine the optimumarea on the sample for collecting data. The data processing techniques will also needfurther development to improve reproducibility and accuracy of the measurements.

It is notable that more LIBS research on bio-samples are conducted worldwide. Withthe technology advances and joint research efforts to solve the main barriers for LIBSbiological applications mentioned above, the in vivo analysis of living organism withLIBS is possible in near future.

REFERENCES

[1] L. J. Radziemski, D. A. Cremers (Ed.), “Spectrochemical analysis using plasma excita-tion,” in: Laser Induced Plasmas and Applications, Marcel Dekker, New York, NY, (1989)Chapter 7, p 295.

[2] F. Y. Yueh, J. P. Singh and H. Zhang, Encyclopedia of Analytical Chemistry, John Wiley &Sons, Ltd. 3 (2000). 2065.

[3] R. C. Wiens, S. Maurice, D. A. Cremers, and S. Chevrel, Lunar and Planetary ScienceXXXIV 2003.

[4] D. A. Cremers, L. J. Radziemski, Handbook of Laser-Induced Breakdown Spectroscopy,John Wiley & Sons, New York (2006).

[5] W. Lee, J. Wu, Y. Lee, and J. Sneddon, Appl. Spectrosc. Rev. 39 (2004) 27.[6] J. D. Hybl, G. A. Lithgow and S. G. Buckley, Appl Spectrosc. 57(10):1207–15 (2003).[7] O. Samek, M. Liska, J. Kaiser, and V. Krzyzanek, Proc. SPIE : Biomedical Sensors, Fibers,

and Optical Delivery Systems 3570 (1999) 263.[8] M. Nakamura, M. Ohml, and M. Haruna, CLEO 99 (1999). Paper TeE3.[9] I. V. Cravetchi, M. T. Taschuk, Y. Y. Tsui and R. Fedosejevs’, Analytical and Bioanalytical

Chemistry, 385 (2006) 287.[10] R. Sattmann, I. Monch, H. Krause, R. Noll, S. Couris, A. Hatziapostolou, A. Mavromanolakis,

C. Fotakis, E. Larrauri and R. Miguel, Appl. Spectroscopy 52 (1998) 456.[11] K. Collins and A. Vass, http://www.scied.science.doe.gov/scied/Abstracts2003/ORNLbio.htm[12] M. H. Niemz, Proceedings of SPIE: Laser Interaction with Hard and Soft Tissue II, 2323

(1995) 170.[13] M. H. Niemz, Proceedings of SPIE: Medical Applications of Lasers II, 2327 (1994) 56.[14] O. Samek, H. H. Telle, and D. C. Beddows,” BMC Oral Health. 1 (2001) 1.[15] O. Samek, D. C. S. Beddows, H. H. Telle, J. Kaiser, M. Liska, J. O. Caceres and

U. A. Gonzales, Spectrochim. Acta B56 (2001) 865.[16] G. M. Bilmes, C. Freisztav, D. Schinca and A. Orsetti, Proc. SPIE: Optical Methods for Arts

and Archaeology, 5857 (2005) 19.




[17] W. Tawfik and S. El-Tayeb, Science Echoes, 7 (2006) 28.[18] Haruna, Masamitsu; Ohmi, Masato; Nakamura, Mitsuo; Morimoto, Shigeto, Proc. SPIE

(Optical Biopsy III, R. R. Alfano; Ed) 3917 (2000) 87.[19] M. Corsi, G. Cristoforetti, M. Hidalgo, S. Legnaioli, V. Palleschi, A. Salvetti, E. Tognoni,

and C. Vallebona, Appl. Opt. 42 (2003) 6133.[20] J. W. Branch, A. Kumar, F. Y. Yueh and J. P. Singh, Pittcon 2005, Orlando, FL. (2005)

1190–1.[21] M. O. Al-Jeffery and H. H. Telle, Proc. SPIE : Optical Biopsy IV 4613 (2002) 152.[22] C. W. Ng and N. H. Cheung, Anal. Chem. 72 (2000) 247.[23] R. Hofmann, R. Hartung, H. Schmidt-Kloiber and E. Reichel, J. Urol. 141 (1989) 275.[24] X. Fang, S. R. Ahmad, M. Mayo and S. Iqbal, Lasers in Medical Science, 20 (2005) 132.[25] Q. Sun, M. Tran, B. Smith, and J. D. Winefordner, Contact Dermatitis 43 (2000) 259.[26] A. Kumar, F. Y. Yueh, J. P. Singh and S. Burgess, Appl. Opt. 43 (2004) 5399.[27] E-M. Kim, M. D. Feit, A. M. Rubenchik, E. J. Joslin, P. M. Celliers, J. Eichler, and L. B.

Da Silva, Appl. Surf. Science 127–129 (1998) 857.[28] J. T. Walsh Jr. and T. F. Deutsch, J. Appl. Phys B: (Lasers and Optics, Issue) 52 (1991) 217.[29] H. P. De Souza, E. Munin, L. P. Alves, M. L. Redigolo and M. T. Pacheco, XXVI ENFMC-

2003 Annals of Optics, Vol 5 (2003).[30] H. Zheng, F. Y. Yueh, S. Burgess and J. P Singh, LACESA (2006) Paper TuE7.[31] I. B. Gornushkin, B. W. Smith, H. Nasajpour, and J. D. Winefordner, Anal. Chem. 71

(1999) 515.



Chapter 15

Measurement of Carbon for Carbon Sequestrationand Site Monitoring

M. Z. Martin, S. D. Wullschleger, C. T. Garten Jr., and A. V. Palumbo

Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA

1. INTRODUCTION

A 2 to 6 �C increase in global temperature by 2050 has been predicted due to theproduction of greenhouse gases that is directly linked to human activities [1]. This hasencouraged an increase in the international efforts on ways to reduce anthropogenicemissions of greenhouse gases particularly carbon dioxide �CO2�, as evidence for thelink between atmospheric greenhouse gases and climate change has been established.Suggestion that soils and vegetation could be managed to increase their uptake andstorage of CO2, and thus become ‘land carbon sinks’ is an incentive for scientists toundertake the ability to measure and quantify the carbon in soils and vegetation toestablish base-line quantities present at this time.

The verification of the permanence of these carbon sinks has raised some concernregarding the accuracy of their long-term existence [2]. Out of the total percentageof carbon that is potentially sequestered in the terrestrial land mass, only 25% ofthat is sequestered above ground and almost 75% is hypothesized to be sequesteredunderground. Soil is composed of solids, liquids, and gases which is similar to a three-phase system [3]. The gross chemical composition of soil organic carbon (SOC) consistsof 65% humic substances that are amorphous, dark-colored, complex, polyelectrolyte-like materials that range in molecular weight from a few hundred to several thousandDaltons [4]. The very complex structure of humic and fulvic acid makes it difficultto obtain a spectral signature for all soils in general [5]. The humic acids of differentsoils have been observed to have polymeric structure, appearing as rings, chains, andclusters as seen in electron microscope observations. The humification processes of thesoils will decide the sizes of their macromolecules that range from 60–500 angstroms.The percentage of the humus that occurs in the light brown soils is much lower thanthe humus present in dark brown soils. The humus of forest soils is characterized bya high content of fulvic acids while the humus of peat and grassland soils is high inhumic acids. Similarly it is well known that the amount of carbon present in forest soilsis lower than the amount present in grassland soils [23].




342 M. Z. Martin et al.

Laser spectroscopic techniques offer real-time monitoring capabilities with high ana-lytical sensitivity and selectivity and have been used for their versatility in environmentalchemical analysis [6–8]. Solid-state NMR techniques [9] have had problems in beingadapted for its use in a field deployable instrument but have been shown to be very usefulin examining organic soils in the laboratory. Other laser based spectroscopy techniques,such as FT-IR, FT-Raman, and 1H-NMR spectroscopies have been applied to investi-gate molecular changes in soil organic matter (SOM) [10–12]. Dispersive Raman andsurface-enhanced Raman spectroscopy [13–15] have been used effectively to determinethe phenolic, alcoholic, and ketonic functional groups present in humic and fulvic acids.Again these techniques are valuable for in-laboratory research. However, another levelof research and development is needed to ensure ruggedness, stability, reliability, smallfootprint, and calibration algorithms that have been tested for a variety of matrices totake the technology to the field. The choice of particular laser technique will depend onthe specific problem at hand. For example, in the elemental characterization of airborneparticles, liquids, and solid surfaces if simultaneous in situ multi element determina-tion is desired then laser-induced breakdown spectroscopy (LIBS) is the technique ofchoice [16]. Taking the above considerations into account, we have applied this tech-nique in the determination of total carbon in various soils. In this article we have usedthe LIBS technique under controlled laboratory conditions in the determination of ele-mental carbon in various soils [17]. Our study builds upon and extends the preliminaryobservations of Cremers et al. [18], who have also demonstrated the unique capabilitiesof LIBS to detect soil carbon. The detection of total carbon from soil has been verysuccessful but the nitrogen data is very irreproducible and depends on the type of soilsespecially if the soils contain a considerable amount of titanium e.g., sandy soils. Wehave shown that we can reliably measure nitrogen if the amount of titanium presentis very low, in the range of ∼ few ppb. The best wavelength to monitor the nitrogencontent has been found to be N(I) at 746.83 nm which was determined by the work onsandy soils completed at Los Alamos National Laboratory [19].

2. LIBS MEASUREMENTS IN SOIL

Strong linear correlations were obtained for determination of carbon concentration resultsfrom – LIBS when correlated to a standard laboratory based technique (sample combus-tion). In our measurements we have used the LIBS technique on soils before and afteracid washing and the technique has been shown to be useful for the determination ofboth organic and inorganic soil carbon.

LIBS is a technique in which a focused laser pulse is directed onto a surface or sam-ple. The energy from the pulse heats, vaporizes, atomizes and ionizes a few nanogramsof material on the surface resulting in a small, hot and brilliant plasma, only millimetersin size. The atoms and ions in the plasma emit light, which are then detected with aspectrometer and detector. Elements in the plasma are subsequently identified by theiremitted unique spectral signatures. Recent advances in component instrumentation haveushered in a new generation of LIBS both in terms of capabilities and application areas.In particular, the advent of the broadband (multispectrometer) detector allows LIBSto be sensitive to molecular matter such as explosives, plastics, minerals, etc. In fact,the broadband spectral response from 200–980 nm means that LIBS is now capable of



Measurement of Carbon for Carbon Sequestration and Site Monitoring 343

PrismLaser

Collectionoptics

Echellespectrometer

Opticalfiber

Lens

Soil sample

ICCDdetector

Fig. 1. LIBS experimental setup for elemental detection from environmental samples.

detecting all chemical elements since all elements emit light somewhere in that spectralrange. The ability of LIBS to provide rapid multielemental microanalysis of bulksamples (solid, liquid, gas, and aerosol) in the parts-per-million (ppm) range with littleor no sample preparation has been widely demonstrated [16]. The experimental set upused to determine the concentration of carbon and other elements in soils is shown inFig. 1. In the experimental configuration we use a Spectra Physics model Indi-HG laserthat is a Q-switched Nd:YAG laser with output fundamental wavelength of 1064, whichwas frequency doubled to 532, and quadrupled to 266 nm. The laser was used at thequadrupled wavelength of 266 nm with a typical energy per pulse of 23 mJ. Optimumenergy/pulse was determined by ramping up the energy on the laser power supply untilbreakdown was achieved on the sample and increasing the energy to be 10% above thebreakdown threshold. This has been discussed thoroughly in another article [20]. A fiberbundle consisting of 19 fibers was used to collect the light emitted from the plasma atthe focal volume by a set of collection optics and focused into a low O-H silica fiberbundle. This fiber bundle (NA = 0.22, diameter = 4.66 mm) delivered the light to a0.5 m Acton Research model SpectraPro-500 spectrometer, (spectral bandwidth = 40 nmfor 1200 gr/mm grating and slitwidth of spectrometer = 25 �m) which was then incidenton an intensified charge coupled detector (ICCD) made by Andor Technologies. Thisdetector can be delayed with respect to plasma formation, and can be gated in order toprevent high background light intensity from the plasma in its early stages of formationfrom entering the detector which are some of the advantages of using an ICCD. Thusoptimization of the S/N (signal-to-noise ratio) of the acquired spectrum is achieved bygating and delaying of the detector. We have also added a new spectrometer-detectorsystem based on Echelle technology (Catalina Scientific Inc.). The detector has a1024 ×1024 pixel square chip and has a delay generator integrated on the detector head.This has enabled us to detect and analyze the soil samples in the laboratory over thefull wavelength region 200–800 nm with a spectral resolution of 0.04–0.06 nm over the




whole spectral region. The optimal time is unique for each element, providing a secondcharacteristic for identification in addition to the wavelength fingerprint. A computersoftware data acquisition program written by the Catalina Scientific Corp., KestrelSpec®

is used to acquire the spectra, identify the peaks, calculate the full width at half maximum(FWHM) of the peaks of interest, and also to calculate the area under the peak which canbe used in the semi-quantification of elements from a similar matrix. We have extendedthe analysis of the broadband spectra by including multivariate analysis of the data toenable the prediction of elemental content from any sample matrix. The latest additionto our experimental versatility is the inclusion of a translational stage that interfaceswith the spectrometer software. This enables us to collect very high spatial resolutionLIBS data (1 micron step) over a large length of any sample and save it automatically.

3. CARBON-NITROGEN ANALYSIS BY SAMPLE COMBUSTION

About 0.5 to 0.6 grams of soil is weighed in a ceramic sample boat in the combustionmethod. A LECO® CN-2000 elemental analyzer (LECO Corporation, St. Joseph, MI)heats the sample to a temperature of 1350 �C in the presence of oxygen after it is insertedinto the combustion chamber. Nitrogen and carbon that is present in soil organic matter(SOM) will be converted to N2� NOx, and CO2 after undergoing combustion. Watervapor is formed by the combination of hydrogen and oxygen. Infrared spectroscopy willbe used to analyze and detect CO2 and N2 will be detected by a thermal conductivitydetector. LIBS technique was calibrated to a LECO® CN-2000 elemental analyzer thatwas used to determine carbon and nitrogen in all soils that were sampled. The elementalanalyzer was calibrated with standards traceable to the National Institute of Standardsand Technology, Gaithersburg, MD.

4. ACID WASHING OF SOILS TO REMOVE INORGANIC CARBON

The soils were obtained from two different sites: (1) Oak Ridge National Labora-tory’s Natural and Accelerated Bioremediation Research (NABIR) Field Research Center(FRC) and (2) southwest Virginia mined lands. About two grams of homogenized soilwas mixed with 10 mL of deionized water in a vial and the solution was heated until itcame to a boil. Ten mL of 3 M HCl acid was carefully added to this solution while thesample was held near boiling temperatures for an hour and swirled frequently [21]. Afteradding 20 mL of deionized water, the sample was shaken for 30 minutes on a shaker.The supernatant was poured off after centrifugation at 2500 rpm for 10 minutes. Thevial and the solid mass were placed in an oven at 60 �C and dried overnight. A mortarand pestle was used to crush the dried soil sample in order to obtain a homogeneoussample for analysis. Pellets were formed from these homogeneous soil samples usingthe technique published in a previous article [17].

The results obtained by the combustion method (LECO® CN-2000) and the LIBStechnique was correlated for fifteen different soil samples, with total carbon concen-trations varying from 0.16% to 4.3%. A typical LIBS spectrum depicting the carbonpeak at 247.9 nm, is shown in Fig. 2. It has been observed that when soils containing a




246 248 250

500

1000

1500

2000

2500

3000C (I) 247.9 nm

Wavelength (nm)

LIB

S s

igna

l (co

unts

)

Fig. 2. LIBS spectrum for carbon in soil.

significant amount of iron were analyzed using the LIBS technique, a problem in resolv-ing the nearly overlapping iron (248.3 nm) and carbon (247.9 nm) peaks was observedfor our data. Thus, the resolution of the spectrometer is an important parameter to beremembered while configuring an instrument for analysis of soils containing a high Fecontent by the LIBS technique. So too are complications associated with soil watercontent and attenuation of carbon signal in moist soils (data not shown).

It can be observed in Fig. 3, that the LIBS signal and the carbon content concen-tration obtained from the combustion method were highly correlated with a coefficientof correlation of 0.978. Fifteen soils were sampled from Oak Ridge National Labora-tory’s Natural and Accelerated Bioremediation Research (NABIR) Field Research Center(FRC). The carbon percentages by weight of these soils were: 0.16, 0.2, 0.25, 0.48, 0.52,0.85, 0.91, 0.95, 1.18, 1.62, 2.19, 2.61, 3.57, 4.22, and 4.32%. The standard deviationfrom the slope of the regression curve calculated for the soil carbon content is in therange 10–15%. Ten laser shots were used for each soil sample analysis. As expected, the

Soil carbon (%)

LIB

S s

igna

l (co

unts

)

00 1 2 3 4 5

2000

4000

6000

8000

10000

12000

14000

16000% of C vs LIBS signal

Regression

r2 = 0.978

Fig. 3. LIBS signal versus soil carbon content measured with the LECO® CN-analyzer.




standard deviation was higher for the soils with low carbon concentration because theexcitation and ionization of carbon at low concentrations is harder when competing withthe iron, which in turn smears the carbon peak, making it difficult to calculate the areaunder the carbon peak which adds additional variability. Also, the small sample amountof soil present in the plasma is not necessarily a good representation of the whole soilsample. To deal with this problem, the LIBS measurements were made using a ten shotaveraging for each homogenized soil sample while moving each pellet around to cover alarger area while sampling the pellet. The spectral variability can be reduced if multiplesamples of each soil are measured and more laser shots are averaged.

In order to test whether we can reduce the variability of the LIBS soil carbon signal,two different soils were analyzed at least ten times using LIBS technique and comparedto data from the LECO® CN-2000 analyzer. One of the soils was supplied by theLECO® Corporation as a calibration standard. Details of this work have been publishedelsewhere [22]. Initially we did observe a larger deviation from the known values ofcarbon content for the LIBS technique. This can be attributed to the amount of soil thatis sampled by each laser shot. In the case of combustion method, the amount of sampletested is 0.5–0.6 grams. In the case of LIBS measurements, the laser was focused to aspot diameter of 10 �m, corresponding to only tens of nanograms of material tested inone shot. When the number of shots is increased for each LIBS measurement we haveshown that this has reduced the standard deviation in carbon measurements. Ten shotshave been used in the measurements shown above, totaling less than 6 seconds at a laserrepetition rate of 1.65 Hz. In order to reduce the variability in the LIBS measurementsif we accumulate 100 shots, the measurements would only take ∼60 seconds, whichis still considered a “near real time” measurement for this environmental application.Even though carbon concentrations in soils change very slowly, the mapping for carbonconcentration in all of the terrestrial areas would be a mammoth task to undertake usingtraditional carbon combustion techniques. Well-established field portable techniques thatcan accomplish quick verification of base-line amounts of sequestered quantities ofcarbon are needed.

One additional area that needs to be explored is the simultaneous measurementof nitrogen and carbon concentrations in soils using LIBS. Preliminary LIBS data ofnitrogen concentrations in soils indicated that this technique could be used but not verysuccessfully to measure total nitrogen in these soils. Fig. 4 shows the typical spectra forelemental nitrogen for two different kinds of soils.

Since air contains 78% of nitrogen, care was taken to make sure that these peakswere not due to atmospheric nitrogen (i.e., the plasma formed at the soil surface willengulf the surrounding atmosphere, and nitrogen from the air will also be excited in theplasma plume). This very high percentage of nitrogen in air is a great concern whenLIBS nitrogen measurements are undertaken under normal atmospheric conditions. Thelaser power was reduced (typically 35 mJ of laser energy/pulse is used but in this caseonly 23 mJ was used) to ensure that no plasma was formed at a distance above the soilsample. The nitrogen peaks shown in Fig. 4 were obtained only when the laser beamwas focused on the surface of the soil sample. To make sure that the nitrogen signal wasobtained from the sample of interest and not from the surrounding air, another sampleof similar consistency, which did not contain nitrogen, was placed at the focus of thelaser beam. After plasma formation at the sample’s surface, these nitrogen peaks did not




Wavelength (nm)742 744 746 748

LIB

S s

igna

l (co

unts

)

0

20

40

60

80

100

120 High nitrogen content soilLow nitrogen content soil

742.36 nm744.23 nm

746.83 nm

Fig. 4. LIBS spectra for two soils with high (0.13%) and low (0.03%) concentrations of nitrogen.

appear. Hence it was concluded that the nitrogen peaks observed in Fig. 4 were due tothe presence of soil nitrogen and not from atmospheric nitrogen.

Figure 5 shows the concentration of carbon and silicon for three different soil typesfrom mined lands in southwest Virginia. These mined soils are very rocky, ranging from40 to 80% coarse fragments at all depths sampled. The carbon contents obtained usingLIBS correlated well with concentrations obtained using the LECO® CN-2000 analyzer.Originally the soils were not acid washed, thus the concentrations of carbon in Fig. 5reflect total soil C and Si concentrations. LIBS spectra of 10 shots were obtained for 20pelletized samples of each soil type.

The standard deviation associated with measurements made on these samples rangedfrom 7 to 10%. Three of the same soil samples were washed with acid to dissolve theinorganic carbon, and LIBS data was obtained for 20 pellets of each soil type, againaveraged over ten shots each. The LIBS signal due to carbon present in the acid-washed

7.44 % 4.84 % 2.65 % 1.68 %0

500

1000

1500

2000

2500

3000

3500CarbonSilicon

Soils from different sites

LIB

S s

igna

l (co

unts

)

% Carbon

Fig. 5. LIBS signal as a function of carbon and silicon in soils before acid washing.




Percentage of carbon in soils before and afteracid washing

4.84 % 2.65 % 1.68 %

LIB

S s

igna

l

0

200

400

600

800

1000

1200

1400

1600

Fig. 6. LIBS signal of carbon for non-acid washed �� and acid washed �� soils.

and non-acid washed soils is plotted in Fig. 6. The amount of silicon present in themined land soils was similar but not equal for each soil type. The carbon concentrationsfor the acid-washed soils varied from 1.6 to 7.4%. Acid washing reduced the LIBScarbon signal by almost 60% (shown in Fig. 6). The reduction in the carbon content ofthese soils is correlated to the presence of inorganic carbon present in these samples.The standard deviation for the acid washed soils increased to 15% from a variation of7% for pre-acid-washed soils, which was attributed to the change in the soil matrix afteracid washing. Acid washed soils had a different packing density (i.e., they appeared tohave more porosity) and were more difficult to pelletize than untreated soil samples dueto surface modifications attributed to acid washing of these soils.

There are three ways in which we can improve the reproducibility and reliability ofsoil analysis: (1) increasing the number of shots and averaging of the spectra over moreshots (100 instead of 10), (2) applying the method of calculating the ratio of intensitiesof two elements present in all soils (e.g., silicon or aluminum), and (3) using the linearcorrelation technique for quantification [24,25].

Increasing the number of spectra for averaging has proven to be quite successful asdiscussed earlier. We have used the ratio of carbon to silicon to improve the reliabilityof the soil carbon and nitrogen data and to reduce the standard deviation in carbon andnitrogen measurements using LIBS. The standard deviation was reduced from 15% afteracid washing to ∼8% for the acid washed soils after normalizing the carbon signal tothe silicon signal. The ratio of carbon-to-silicon was used in the calibration for the acidwashed soils (Fig. 7). The small variation in the silicon contents of each soil type wasused very effectively to reduce the standard deviation in soil carbon analysis. The C/Siratio method has been used by other authors [26] in case of other LIBS applications. Thethird way to reduce variation of the carbon signal in soil matrices is done by applyingtechnique of linear correlation that uses the theory of covariance [27]. Covariance is ameasure of the tendency of two variables to vary together (to co-vary). In other cases ithas been established that shape and position of the calibration curve are not affected bydetector sensitivity or baseline level variations, as long as the whole spectra is affecteduniformly. The outcome of the linear correlation technique is that it successfully filtersout the effect of signal intensity fluctuations.




Carbon concentration (%)1.68 C % 2.65 C % 4.84 C % 7.4 C %

LIB

S s

igna

l (C

/Si)

0.00

0.05

0.10

0.15

0.20

0.25

0.30r

2 = 0.905

Fig. 7. LIBS signal for acid washed soils (ratio of C to Si) as a function of the carbon.

5. FIELD MEASUREMENTS

We are in the process of developing portable LIBS instrumentation for in-field mea-surements for soil carbon and other elemental measurements. Outdoor environmentsmay require heated, insulated, weatherproof, explosion-proof, or chemically resistantenclosures. Environments with excessive vibration, radiation, or strong electric/magneticfields may affect instrument performance and require isolation or shielding. The mostimportant aspect of field measurement is the ability to accurately accomplish in-fieldcalibrations. A very prudent way to accomplish in-field calibrations would be to usea known and constant amount of a ubiquitous component as a calibration standard.For example, if the measurement involves detection of nitrogen in soils, then to ensurethat the detector is holding its calibration (the spectral wavelength of emission) is tovery easily make a laser spark in air (since air contains 78% nitrogen) to establishthat the nitrogen emission lines are present at their specific wavelengths. At ORNL wehave a plan in which we would like to take our field portable instrument and measuresoil carbon concentrations in situ. We have successfully developed a field deployableLIBS system and are in the process of developing a high throughput LIBS multivariateanalysis protocol to be able to not only measure the elemental concentration in dif-ferent environmental sample but also to predict the concentration of these elements inunknown samples especially in field samples. The fiber-optic probe development forthis application is also underway at ORNL. We have also set up a translational stage tomove the sample and acquire very high spatial resolution (1 micron resolution) LIBSmeasurements along the length of the sample and save the data automatically beforemoving the stage.

6. CONCLUSIONS

In conclusion, carbon measurements by conventional combustion methods and LIBSwere determined to be in good agreement in fifteen different soils. LIBS was also used




successfully to determine the nitrogen content of various non sandy soils. Comparisonof the LIBS signals for carbon, nitrogen and silicon before and after acid washingsuggests that a strategy to determine the inorganic and organic carbon present in thesoils can be established for future experiments. The ratio of two elements (C/Si) wasused to improve the correlation between the LIBS signal and conventional soil carbonmeasurements. This work has shown that instrumentation and operation of a LIBSsystem is simpler than some of the more sensitive laser based techniques (for exampleLA-ICP-MS), and analysis times of the order of minutes make it amenable for real-time, in situ analysis and environmental monitoring of soil carbon and in some caseseven nitrogen. Furthermore, this technique requires little or no sample preparation,thus making it an attractive alternative to existing methods of soil carbon, nitrogenand other multielemental analysis. Finally, we are in the process of developing a highthroughput LIBS multivariate analysis protocol to be able to not only measure theelemental concentration in different environmental samples, but also to predict theconcentration of these elements in unknown samples especially in field samples. Theseadvancements will greatly facilitate the use of this technique for carbon managementactivities.

ACKNOWLEDGMENTS

We would like to acknowledge and thank Bonnie Lu for analyzing the fifteen soilsusing the LECO® CN-2000. We would also like to extend our thanks to Deanne Bricewho helped with the LIBS measurements and made sure that the soils were analyzedin a reproducible and repeatable manner. The research on soils was sponsored by theLaboratory Directed Research and Development (LDRD) program of Oak Ridge NationalLaboratory, managed by University of Tennessee-Battelle, LLC for the U. S. Departmentof Energy (DOE) under contract number DE-AC05-00OR22725. Additional support wasprovided by the DOE Office of Fossil Energy though the National Energy TechnologyLaboratory (NETL).

REFERENCES

[1] Intergovernmental Panel on Climate Change. 2001. Climate change 2001. The scientificbasis. J. T. Houghton et al., eds. Contribution of the working group 1 to the third AssessmentReport of the IPCC. Cambridge University Press, Cambridge, UK. 881 pp.

[2] D. Read, D. Beerling, M. Cannell, P. Cox, P. Curran, J. Grace, P. Ineson, Y. Malhi,D. Powlson, J. Shepherd, and I. Woodward, The role of land carbon sinks in mitigatingglobal climate change, 1–27, The Royal Society, London, (2001).

[3] M. Schnitzer and U. Khan, Humic Substances in the Environment, Marcel Dekker, New York,(1972).

[4] D. Hillel, Environmental Soil Physics, Academic, New York, (1998).[5] F. J. Stevenson, Humus Chemistry, John Wiley and Sons, New York, (1982).[6] S. J. Weeks, H. Haraguchi, and J. D. Winefordner, Analytical Chemistry, 50 (1978) 360–68.[7] S. Sjostrom and P. Mauchien, Spectrochim. Acta B 15 (1991) 153.[8] S. Rudnick and R. Chen, Talanta, 47 (1998) 907.




[9] C. M. Preston, S. –E. Shipitalo, R. L. Dudley, C. A. Fyfe, S. P. Mathur, and M. Levesque,Can. J. Soil. Sci., 67 (1987) 187.

[10] O. Francioso, S. Sanchez-Cortes, V. Tugnoli, C. Ciavatta, L. Sitti, and C. Gessa, Appl.Spectrosc., 50 (1996) 1165.

[11] O. Francioso, S. Sanchez-Cortes, V. Tugnoli, C. Ciavatta, and C. Gessa, Appl. Spectrosc. 52(1998) 270.

[12] O. Francioso, C. Ciavatta, S. Sanchez-Cortes, V. Tugnoli, L. Sitti, and C. Gessa, Soil Science,165 (2000) 495.

[13] Y. Yang and H. A. Chase, Spectrosc. Lett. 31 (1998) 821.[14] T. Wang, Y. Xiao, Y. Yang, and H. A. Chase, J. Environ. Sci. Health, A 34 (1999) 749.[15] E. J. Liang, Y. Yang, and W. Kiefer, Spectrosc. Lett. 32 (1999) 689.[16] M. Z. Martin, M. D. Cheng, and R. C. Martin, Aerosol Sci. Technol. 31 (1999) 409.[17] M. Martin, S. Wullschleger, and C, Garten Jr., Proceed. SPIE, 4576 (2002) 188.[18] D. A. Cremers, M. H. Ebinger, D. D. Breshears, P. J. Unkefer, S. A. Kammerdiener,

M. J. Ferris, K. M. Catlett, and J. R. Brown, J. Environ. Qual. 30 (2001) 2202.[19] R. D. Harris, D. A. Cremers, M. H. Ebinger, and B. K. Bluhm, Appl. Spectrosc. 58 (2004) 770.[20] M. Martin and M-D. Cheng, Appl. Spectrosc. 54 (2000) 1279.[21] S. E. Trumbore, and S. Zheng, Radiocarbon. 38 (1996) 219.[22] M. Z. Martin, S. D. Wullschleger, C. T. Garten Jr., and A. V. Palumbo, Appl. Opt. 42

(2003) 2072.[23] V. P. Evangelou, Undergraduate/Graduate level textbook on Environmental Soil & Water

Chemistry: Principles and Application, John Wiley & Sons, Inc., New York (1998).[24] M. Schnitzer and U. Khan, Humic Substances in the Environment 2–3, Marcel Dekker,

New York (1972).[25] M. Schnitzer, Humic Substances in Soil, Sediment and Water, 303–325, John Wiley and

Sons, New Jersey (1985).[26] I. B. Gornushkin, B. W. Smith, H. Nasajpour, and J. D. Winefordner, Anal. Chem., 71

(1999) 5157.[27] G. Galbacs, I. B. Gornushikin, B. W. Smith, and J. D. Winefordner, Spectrochim. Acta B,

56 (2001) 1159.



Chapter 16

Remote Analysis by LIBS: Application to SpaceExploration

D. A. Cremers

Applied Research Associates, Inc., 4300 San Mateo Blvd., Albuquerque, NM 87110, USA

1. INTRODUCTION

One of the more outstanding capabilities of LIBS is the ability to provide a stand-off orremote elemental point analysis of a material at a significant distance from the instrument.Other methods (e.g. inductively coupled plasma analysis, X-ray fluorescence) requirethat some physical implement (e.g. electrodes, probe) come in contact with the sample orthat the sample be retrieved and introduced into the instrument. This requirement limitsthe applicability of these methods, excludes them from some important applications, andoften increases analysis times. Remote analysis is possible with LIBS because the plasmais formed by focused laser light which can be directed over a considerable distance fromthe optical system. The only requirement is for line-of-sight viewing of the target. Thisunique capability, combined with the other advantages of LIBS, opens up exciting newapplications of the technology that cannot be addressed by any other methods.

The majority of LIBS measurements are carried out at in-situ or close range in whichthe distance between the sample and the optical system is 10 cm or less. In these cases,the requirements on laser, spectrograph, detector, and optical system performance arethe easiest to satisfy and less than optimum operating parameters can be tolerated. Onlya few millijoules of laser energy may be required, single lens focusing will generate ananalytically useful plasma, and sufficient spark light can be readily collected simply bypointing a fiber optic at the plasma. As the distance increases, performance specificationsbecome more critical.

The remote analysis discussed here pertains to open path, direct line-of-sight mea-surements. Remote LIBS can also be carried out using fiber optic delivery of the laserpulses which are then focused onto the sample using a short focal length lens. Theplasma light is collected using either the same fiber optic or a second fiber. For fiberoptic delivery, line-of-sight to the sample is not required but this method requires that aprobe, although of small size, be positioned adjacent to the sample, thereby resemblingin-situ or close-up LIBS analysis. Each remote method, line-of-sight and fiber opticanalysis has certain advantages. Our discussion here is devoted to the former methodwith fiber optic LIBS discussed in Chapter 5.




354 D. A. Cremers

In the literature, LIBS analyses at distances of a few meters or more are referred to asstand-off or remote measurements. As a guide, the LIDAR (light detection and ranging)technique is usually described as a remote technique with data retrieved from distancesup to several kilometers from the laser/detection system. Currently, LIBS cannot provideanalyses at such long ranges although recent work indicates that a distance of a kilometermay be realized in the near future using new methods to generate the excitation plasma.In view of established usage for LIBS, here we will refer to stand-off and remotemeasurements interchangeably with the actual range realized listed for each application.

Some new applications of stand-off LIBS that cannot be addressed by more conven-tional analysis methods include:

• analysis of physically inaccessible targets (e.g. geological features on cliff faces)• targets located in hazardous environments (e.g. contamination by toxic, radioactive

materials)• rapid scanning of distinct, widely separated targets from a single vantage point• rapid interrogation of large surfaces by scanning laser pulses along a surface• industrial process control where analysis must be done rapidly and from a distance

(e.g. molten metals and glasses)

Some of these applications have been demonstrated and will be discussed below.Two distinct methods of stand-off LIBS measurements have been demonstrated which

rely on different physical processes to form the analytical plasma. The first method issimply “conventional” LIBS in which the laser pulse is focused to a distant point toproduce power densities sufficiently high to induce ablation and optical breakdown ofthe sample resulting in a microplasma. This generally involves nanosecond (ns) pulseshaving energies of at least several tens of millijoules and large diameter optics to producea small spot size on the remotely located sample to form an analytically useful plasma.Conventional stand-off LIBS can be carried out using the ns lasers typically used forin-situ LIBS. Analysis ranges up to 80 meters have been reported.

The second method, more recently demonstrated, uses self-guided filaments inducedby femtosecond (fs) laser pulses that can propagate over long distances and produceLIBS excitation of a sample. This method, exciting in its ability to analyze samples atlong distances (180 m demonstrated), with kilometer ranges predicted, currently requiresthe use of large sophisticated laser systems.

Although remote analysis provides unique sampling capabilities for real-world appli-cations, extreme care must be taken in deploying such LIBS systems. The laser pulse canrepresent an ocular and skin hazard to the public and operator and serve as an ignition sourcefor explosive/flammable materials. These hazards will generally increase as the path lengthincreases. These concerns may preclude the use of stand-off LIBS in some cases. In anyevent, procedures should be developed to ensure the safe use of all remote LIBS systems.

2. CONVENTIONAL STAND-OFF LIBS

Discussions in this section are limited to remote analysis using ns laser pulses focusedon a sample to generate power densities sufficiently high to ablate the material andproduce a plasma. This “conventional” method of generating the analytic plasma has



Remote Analysis by LIBS: Application to Space Exploration 355

been used most extensively for LIBS. Remote LIBS measurements made using fs pulsesare discussed separately in section 3.

2.1. Apparatus

An apparatus for conventional stand-off LIBS measurements is diagrammed in Fig. 1.Depending on the required analysis distance, the laser pulses may be focused using asimple lens, a pair of lenses (to vary the focal point), or a more elaborate optical systemmay be required to first expand the laser pulse spatially and then focus the pulse at adistance onto the target. In many cases the focusing system is adjustable so targets atvarious ranges can be interrogated from a fixed position. Remote LIBS systems havebeen developed in which the plasma light is collected collinear with and off-axis to thepath of the laser pulses to the target. Collection along the same axis as the laser pulseseliminates parallax as the distance to the target (r) changes. In the collinear arrangement,the plasma light is diverted from the path of the laser pulses using, for example, a beamsplitter or mirror with a central clear aperture. The diverted light is then directed intothe detection system. The intensity of the collected light varies as r−2.

For practical LIBS systems, the diameter of the laser beam, even after expansion by asimple Galilean telescope will typically be less than 4 cm whereas focal lengths (f ) willbe greater than 1 m with 3 m or more typical. In this case the minimum spot size (d )attainable at the focus will be determined by the diffraction limit rather than sphericalaberration which becomes important for shorter focal length lenses and small beamdiameters. For example, a typical beam diameter directly from a laser is 7 mm. In thiscase, spherical aberration determines the minimum achievable spot size for f < 3�5 cmwhereas diffraction is the determining factor at longer focal lengths. In the case wherediffraction dominates,

ddiff = 2�44 �f/D (1)

where � = wavelength and D = diameter of the laser beam. Therefore, if working in thediffraction limited focusing regime, expanding the beam will produce smaller spot sizesand higher power densities on the distant sample.

Laser

SpectrographDetector

Beamexpander

Mirror

Fiber optic

Telescope

Target

Computer

Fig. 1. Diagram of a conventional LIBS apparatus for stand-off analysis.



356 D. A. Cremers

An important consideration in collecting the plasma light is chromatic aberration.The apparatus shown in Fig. 1 is free of this effect because the light is diverted andfocused by mirrors. Unless achromatic lenses are used, focusing by lenses will result inwavelength dependent focal positions requiring that the fiber optic be moved relativeto the collecting lens to optimize the recorded intensity for different spectral regions.Chromatic aberration becomes an important consideration, especially when using anechelle spectrograph where the spectrum collected on a single shot can encompass alarge spectral range (e.g. 200–800 nm).

2.2. Results

Remote LIBS using conventional focusing has been used mainly for the analysis of solidtargets. In a few cases liquid samples have been interrogated at distances >1 m. Becauseof the high power densities required to induce air breakdown, analysis of gases at largedistances has not been reported using conventional focusing methods and practical sizelasers.

One of the first reports of remote LIBS analysis was described in 1985 [1].A “spectrochemical lidar” instrument was developed based on a Cassagranian telescope,spectrograph, and photographic or photomultiplier detection of the collected light. TheCO2 laser generated 300 �s pulses of 500 mJ. Although the focused power densitieswere not sufficient to induce an air plasma, plasmas were formed on aerosol particleswithin the telescope focus at ranges of 50 to 150 m. The elements Ca, Al, and Na fromthe particles were detected along with oxygen and nitrogen emissions from air.

Another early report of stand-off analysis of a solid material was published in 1987 [2].Laser pulses were focused on metal samples at a distance and the light was collectedby a bare fiber optic bundle pointed at the plasma. No attempts were made to optimizethe experimental set-up to extend the range but using this simple arrangement, usefulsignals were obtained as far as 2.4 meters. The use of the method for rapid identificationof metals according to their main element component (Cu, Zn, Al, Ni, Sn, Mo, Ti, or Fe)was demonstrated with 100% success. The use of repetitive ablation to clean a surfacewas described and the method was evaluated for the analysis of steel at 0.55 m.

In 1991, LIBS was demonstrated for the remote analysis of geological samples in airat a distance of 24 m using a laboratory laser and detection system [3]. The equipmentwas positioned on a cart and moved outside the laboratory to interrogate a cliff bank.Spectra were readily collected allowing at least a qualitative evaluation of the targetmaterial composition. The goal of this work was to promote LIBS as an instrument onfuture planetary missions. Subsequent investigations on the development of LIBS forspace exploration are discussed in section 4.

As noted above, stand-off analysis is usually performed using solid samples. In 2000,one study, however, describes the analysis of liquids at distances of 3 to 5 m using bothoff-axis and on-axis methods of plasma light collection [4]. The on-axis system employeda novel method of separating the laser beam from the plasma light using FTIR (frustratedtotal internal reflection). Different sampling methods were studied to minimize surfacemovements with a laminar water jet used for the majority of experiments. A radioactivesolution of Tc was interrogated, however, at a distance of 3 m at only 1 Hz to minimizesplashing of the solution in a Petri dish. From the concentration of Tc in the sample




and the strength of the signal it was estimated that Tc levels down to 25 mg/l should beobservable.

A custom, mobile LIBS system was developed in 2002 for stand-off analysis (1 m) ofmajor elements in a mineral melt �1600 �C� in an industrial environment [5]. The plasmalight was collected at an angle to the path of the laser pulses instead of collinearly.A variation in the LIBS signal of Si was observed due to changes in position of themobile instrument in relation to the melt. This was believed to be due to collection ofthe light at an angle which resulted in monitoring different regions of the expandingplasma. Variations in signals could probably be minimized through a collinear geometry(Fig. 1). Operational parameters such as pulse irradiance and gate delay and width ofICCD detection were optimized in terms of the Signal/Noise ratio. This permitted allmajor elements in the melt (Ti, Fe, Mn, Mg, Ca, Si, Na, Al) to be identified withinthe brief interrogation time of 1 s corresponding to 10 laser pulses. To evaluate processmonitoring, LIBS measurements were made from a melt at analysis intervals separatedby 60 s along with manual retrieval of a sample from the melt for subsequent analysisby X-ray fluorescence (XRF). Runs extending over 80 and 130 minutes were carriedout and compared. Good correlation was observed between the LIBS and XRF data(elements Si, Fe, Al, Ca, Mn, Mg) with a slight shift in the pattern (element signalvs. time) attributed to a small mismatch between the sampling times.

Analysis of stainless steel in 2002 at a distance of 40 m was demonstrated by Palancoet al. using an open-path LIBS system [6]. Light was collected off-axis to the path of thelaser pulses. The long Rayleigh length of the beam at 40 m was found useful to minimizethe effect of surface irregularities on the analysis with an RSD of 14% determined forthe absolute signal precision as the sample position varied +/− 1 m from the positionof optimum laser focus. The ratio Cr/Fe was found to be highly uniform over the range+/−5 m from the optimal laser focus. In addition, spectra from six stainless steels werecollected and emissions from Ni, Mo, and Ti compared. Using these data in a three-dimensional pattern recognition algorithm showed that the six steels could be accuratelyclassified using only three elements.

Based on a collinear optical design in which the plasma light was collected along thesame path as the laser pulses, Palanco and Laserna in 2004, developed and describedan open-path LIBS analysis system [7]. Performance of the system was evaluated andthe feasibility of extending the analysis range of a LIBS system based on conventionalfocusing was discussed.

Using an Nd:YAG laser, spectrograph, and ICCD, samples including plant material,soil, rock, and cement collected from an industrial environment were analyzed in 2004at 12 m distance in the laboratory [8]. The experimental arrangement was similar tothat of Fig. 1. Depth resolved measurements and the effect of surface condition onthe analysis were evaluated. Factors affecting analysis results such as moisture content,surface uniformity and sample orientation were evaluated. Detection limits for Cr andFe were determined to be about 0.2 wt.% from calibration curves prepared using a setof slag standards. These curves were then used to determine the Cr levels in some ofthe collected environmental samples. The change in Cr signals (concentration) with shotnumber on the different samples was used to draw conclusions regarding aspects ofenvironmental pollution. Using the large depth of focus of the optical system at 12 m,three-dimensional maps of Cr, Fe, Ca, and Mg distributions in a rock sample were madeover the volume 20 mm × 18 mm × 0.54 mm depth.



358 D. A. Cremers

Using the same apparatus, except that the collected plasma light was directed into afiber optic and a HeNe laser, collinear with the Nd:YAG pulses, was used to aim thesystem, the same group monitored the corrosion of stainless steel in a high temperatureenvironment [9]. The experiments were conducted in the laboratory at a distance of 10 mwith the steel samples maintained in an oven. The oxidation of the surface was monitoredand qualitative differences observed between the intensity of element emissions from thecorroded, scaled steel surface and a clean steel surface. Specifically, after only 10 min.exposure at 1200 C, the scaled layer was found to be Fe-rich with reduced levels ofCr, Ni, and Mo compared to the starting material. Depth profiling of the corroded layershowed depletion of Cr at the surface with the Cr signals increasing as a result ofrepetitive ablation that interrogated the underlying bulk material.

An open-path transportable LIBS apparatus was used to analyze molten steel ona factory floor [10]. The sample was heated in a crucible in a small scale inductionfurnace of 1 kg capacity. The distance between the instrument and the crucible wasabout 7.5 m with two mirrors in between to direct and focus the laser pulses verticallydownward onto the molten sample surface. Measurements showed the ability to monitorin real-time the changes in the composition of the melt (e.g. Ni added). Calibrationcurves for Ni and Cr were prepared by adding these elements to molten stainless steel.From the curves, detection limits of 1190 and 540 ppm were determined for Ni and Cr,respectively.

3. STAND-OFF LIBS USING FEMTOSECOND PULSES

3.1. Femtosecond Laser Pulses and LIBS

LIBS measurements are typically carried out using pulses from a nanosecond laser.These pulses have characteristics of 5–10 ns duration, energies from a few millijoulesup to 500 mJ, with the fundamental wavelength of 1064 nm preferred for most appli-cations. The powers produced by these lasers are in the range of 0.3 to 50 MW withcorresponding focused power densities of 4 to 640 GW/cm2 (for a 0.1 mm diameter spotsize). Nanosecond lasers are preferred for LIBS because they are technologically well-developed, rugged and reliable, and very compact systems are available for incorporationinto instruments.

Picosecond (ps) and femtosecond (fs) lasers, however, generating pulses of durationson the order of 10−12 and tens of 10−15 s, respectively, have been investigated for LIBSapplications and some advantages have been found.

An important concept in discussions of fs pulses is chirp. A chirped pulse is one inwhich the different wavelengths or colors are not distributed uniformly over the temporalenvelope of the pulse. Alternatively, chirp may be viewed as an increase or decrease inthe frequency of a light pulse with time as monitored from a stationary position as thepulse passes by. Because of the broad spectral content of fs pulses, chirp can be used tocontrol certain pulse properties. Positive chirp occurs when the leading edge of the pulseis red-shifted in relation to the central wavelength and the trailing edge is blue-shifted.Negative chirp is the opposite situation. Because of the dependence of refractive indexon wavelength, different wavelengths have different velocities when passing through amedium. Using this effect, by passing a chirped pulse through a sequence of prisms




the chirp can be adjusted to be either positive or negative. The refractive index ofglass and most materials (e.g. air) increases as the wavelength decreases so that longerwavelengths have higher velocities. Therefore, a positively chirped pulse will becomemore positively chirped and show an increase in pulse width when passing through air.For example, a 70 fs pulse of 16 nm spectral content will be chirped into a 1 ps pulseafter traversing 1 km of air. The peak power of the pulse is reduced by 10x [11]. On theother hand, a negatively chirped pulse transmitted through the atmosphere will exhibitreduced chirp and a shorter pulse width (Fig. 2). By controlling the chirp in an outputpulse, the pulse width and hence the pulse power at a certain distance from the laser canbe controlled.

3.2. Remote Sensing using fs Pulse Produced Filamentation

The remote analysis method of LIDAR is a well developed method of monitoring atkilometer distances, gases and aerosols in the atmosphere. LIDAR methods include laser-induced fluorescence, absorption, elastic scattering, and Raman spectroscopy. Thesemethods are useful to determine the presence of aerosol particles and to identify mole-cular species. Typically, ns pulse lasers are used for LIDAR, precluding the formationof a LIBS plasma at remote distances comparable to the analysis ranges achievableusing the other spectroscopic methods with LIDAR. Using pulses of reasonable energy(i.e. <500 mJ), ranges of a few tens of meters are possible with LIBS.

On the other hand, because of the very high optical powers generated by fs pulse lasers,extended range LIBS is possible based on atmospheric filamentation [12]. Filamentationoccurs when a sufficiently powerful fs pulse propagates through air or other mediumtransparent at the laser wavelength. The process is based on the Kerr effect in whichthe refractive index of a medium is changed by an applied electric field. The inducedchange ��n� is given by

�n = n2I (2)

Where n2 is the non-linear index of refraction particular to the medium and I is theoptical intensity. When a Gaussian-shaped light pulse passes through the transparentmedium, the change in refractive index will be greater at the spatial center of the pulseand less at the edges, following the intensity distribution across the pulse profile. For airn2 = 3×10−19 cm2/W and the induced changes will form a positive lens acting to focusthe light pulse. The lens produced resembles a GRIN (gradient-index) lens manufactured

Δt Δt ′

laser r

rb

b

Fig. 2. A negatively chirped pulse [blue (b) proceeds red (r)] is temporally compressed by travelingthrough air ��t>�t′�.



360 D. A. Cremers

by radially varying the refractive index value during manufacture. The focusing resultsin ionization of the transparent medium and formation of a plasma that acts to de-focusthe beam. When the pulse power exceeds a certain critical power, a dynamic equilibriumexists between the two processes resulting in self-trapping of the propagating laser beamnear the optical axis of the beam. For air, the critical power is several GW. The self-trapping generates filaments of “white light” or a spectral continuum over distancesmuch longer than the Rayleigh length of a conventionally focused beam. Images offilaments produced in air are shown in Fig. 3. About 10–20% of the laser pulse energyis present in the filaments.

Filament lengths of 200 m are typical with filaments of 2 km observed. Filamentdiameters have been characterized at about 0.1 mm. Measurements show that the intensityinside the filaments is on the order of 4 × 1013 W/cm2 [13]. At powers considerablyabove the critical value (x10–100), multiple filaments are formed that propagate alongthe beam. The “white light” is produced by self-phase modulation within the fiberthat generates frequencies other than the laser frequency. A significant portion of thiscontinuum is radiated in the forward and backward direction of pulse propagation.

3600

(a) (b)(a)

(d)(c)

zenith

20 km

44 km

6 km

3 km

9 km

4 km

3 km

1.00(b–d)

0.07

0.03

0.00

240

120

0

Fig. 3. Images of filaments produced by the Teramobile fs laser beam propagating verticallywere taken with a charge-coupled device camera. (a) Fundamental wavelength, exhibiting signalsfrom more than 20 km and multiple-scattering halos on haze layers at 4- and 9-km altitudes.(b to d) White light (385 to 485 nm) emitted by the fs laser beam. These images have the samealtitude range, and their common color scale is normalized to allow direct comparison with thatof (a). Reprinted with permission from J. Kasparian, M. Rodriguez, G. Méjean, J. Yu, E. Salmon,H. Wille, R. Bourayou, S. Frey, Y.-B. André, A. Mysyrowicz, R. Sauerbrey, J.-P. Wolf, andL. Wöste, White-Light Filaments for Atmospheric Analysis, Science, 301 (2003) 61. Copyright2003 AAAS.




3.3. Teramobile

Femtosecond laser systems producing powers on the order of 1014 W are laboratoryinstruments owing to their complexity, size, and requirements for a controlled operatingenvironment. Certainly, with improvements in technology, these laboratory devices willbecome more readily adapted to field-deployment. The driver for this development willbe the unique capabilities that fs lasers offer in a number of areas. A first importantstep toward field deployment has been the development of the Teramobile, a jointcollaboration between French and German organizations [14]. The Teramobile is acomplete mobile laboratory housing the transportable fs system producing TW laserpulses and associated detection and analysis instrumentation for evaluating the operationof the system. The Teramobile laboratory is contained inside a standard freight containerthat can be transported to different field sites.

3.4. Remote LIBS using fs Pulses

Remote LIBS analysis of targets has been carried out using fs pulses focused onto atarget as in conventional LIBS (Fig. 1) and using fs-pulse generated filaments. In theconventional LIBS configuration, fs pulses (795 nm, 10 Hz, up to 350 mJ, 75 fs) weredirected from the laboratory onto the solid target at 25 m using a simple mirror telescopeto expand (x3) and then focus the pulses [15]. The fs laser could also be adapted toproduce ps and ns pulses for comparison of results. A second telescope (10 cm primarymirror) was used to collect the plasma light at a position adjacent to but not collinearwith the path of the laser pulses. Some of the main results of the study were: (1) fs andps pulses can be used for remote analysis with a detection limit of 100 ng computedfor Cu at 25 m using fs pulses; (2) fs pulses produced a cleaner LIBS spectrum thaneither the ps or ns pulses, with the fs spectrum free of emissions from the ambient gas;(3) emissions from the fs- and ps-produced plasmas decayed slowly (microsecond timescale) compared to the laser pulse widths; (4) adjusting the fs-pulse chirp to producethe minimum duration laser pulse at the target (i.e. maximum power density) does notproduce the strongest emission signal. The chirp characteristics must be adjusted foreach type of material to produce the optimum signals.

The use of filaments for LIBS has recently been demonstrated and the method namedR-FIBS (remote filament-induced breakdown spectroscopy) [16]. The output pulses(80 fs, 250 mJ, 10 Hz) of the Teramobile were collimated (3 cm diameter) and directed ata solid target located 20–90 m distant. To begin filament production 7–8 m in front of thetarget, the pulse leaving the laser system was negatively chirped with a correspondingpulse width of 800 fs. This arrangement produced multiple filaments on the target asshown in Fig. 4. The light from the filament-target interaction region was collected by atelescope and recorded by a spectrograph and ICCD. The LIBS spectrum was examinedin the 500–550 nm region and emissions from Cu(I) and Fe(I) from copper and steelsamples were recorded at 90 m. Over the 20–90 m range investigated, the LIBS signaldid not depend on distance, except for the usual r−2 losses with distance. This indicatesthat the robustness of target excitation by the filaments did not change with increaseddistance. Considerations of signal-to-noise (S/N) changes with distance showed that with



362 D. A. Cremers

Fig. 4. Multiple filamentation pattern observed on a solid target. Reprinted with permission fromK. Stelmaszczyk, Ph. Rohwetter, G. Méjean, J. Yu, E. Salmon, J. Kasparian, R. Ackermann,J.-P. Wolf, and Ludger Wöste, Long-distance remote laser-induced breakdown spectroscopy usingfilamentation in air, Appl. Phys. Lett., 85 (2004) 3977. Copyright 2004, American Institute ofPhysics.

this non-optimized system, a distance of 150 m could be realized for LIBS detection�S/N∼1�. Expected improvements in the detection system, the authors conclude, shouldpermit measurements approaching 1 kilometer.

In a subsequent study by the same group, the range of R-FIBS was extended to an Altarget located 180 m from the laser [17]. Comparisons were made between the R-FIBSspectra and the spectra from conventional LIBS produced by ps and ns lasers. The mainresult was that the R-FIBS spectra were free of emission lines due to oxygen and nitrogensimilar to the spectra obtained using fs pulses for conventional LIBS [15]. Based onthe results of this work, the feasibility of kilometer range R-FIBS was considered andexperimental requirements to attain this range were estimated.

4. STAND-OFF LIBS FOR SPACE EXPLORATION

Perhaps one of the more exotic and certainly exciting applications of LIBS is forspace exploration. Preliminary tests have benchmarked the capabilities of LIBS forthis application at close-up and stand-off distances and for atmospheric pressures andcompositions simulating Mars [3,18–22], Venus, and the Moon. The method promises togreatly increase the scientific return from new missions compared to the data volumes thatcharacterize current elemental analysis techniques. Data provided by LIBS is importantto understanding planetary geology, one main goal of space exploration. Planetarygeology is important because it can answer questions dealing with (1) the physical andchemical evolution of the solar system, (2) what the early solar system was like, and(3) it can be used to compare processes that occurred on other bodies with geologicprocesses on Earth. Also, a geologic analysis can tell us something of a planet’s historysuch as whether earlier conditions were favorable for life (e.g. indications of pastwater).




4.1. Spectroscopic Methods of Planetary Analysis

A number of different spectroscopy-based methods have been successfully deployedduring the history of space exploration for planetary geology. To understand the uniquecapabilities of LIBS it is valuable to briefly review these other methods (Table 1). Notethat some methods are passive and others are active detection methods such as LIBS.

4.2. Prior Elemental Analysis Methods used on Landers/Rovers

The earliest chemical analysis measurements of a non-terrestrial body in 1967 wereconducted on the Moon by Surveyor landers 5, 6, and 7 using an alpha-scattering

Table 1. Spectroscopic methods for space exploration [23]

Method (capability,passive or active method)

Operating principles Information content(in-situ or remote method)

X-ray and gamma-ray(elemental analysis,passive)

Cosmic ray particles and lowerenergy solar x-rays excite x-rayradiation that is elementspecific. These radiationsdetected at gamma-ray andx-ray energies.

Na, Mg, Al, Si, P, S thatcompose minerals; H andnatural radioactive elements-K,Th, U (remote)

Reflectance (mineralogy,passive)

Sunlight reflected from targetrocks/minerals that absorbat sunlight wavelengths.Reflectance spectra containfeatures characteristic ofminerals and subtle shifts in thespectra can be related to someaspects of elemental make-up.

Mineralogy and some elementalinformation (in-situ, remote)

Thermal emissionspectroscopy(mineralogy, passive)

Thermal radiation from rocksand soils show spectral featuresin both emission and absorptionthat can be used to deducemineralogy.

Mineralogy (remote)

X-ray fluorescence withsources (elementalanalysis, active)

Sample irradiated by x-rays.Subsequent x-rays emittedby the sample throughrepopulation of inner shellelectrons produce elementspecific x-ray spectrum.

Elemental analysis (in-situ)

APXS or alpha-protonx-ray spectrometry(elemental analysis,active)

Expose sample to a radioactivesource. Record energy spectraof the alpha particles, protonsand x-rays returned from thesample.

Elemental analysis (in-situ)



364 D. A. Cremers

instrument [24]. This was followed by a series of Luna missions beginning in 1970 withLuna 17 carrying the rover Lunokhod 1. Soil analysis was provided by a RIFMA X-rayfluorescence spectrometer. Previous methods of elemental analysis deployed on space-craft to the surface of Mars and Venus have utilized either X-ray fluorescence or APXS(alpha-proton X-ray spectrometer). These missions include Viking [25], pathfinder [26],MER (Mars Exploration Rovers) [27], and Venera [28].

All methods used to date have been in-situ techniques that require that the sensingunit be positioned adjacent to the sample or that a sample be retrieved and then deliveredto the detector as shown in Fig. 5. Using LIBS, only optical access to the sample isrequired. Although these prior methods have useful detection characteristics and havereturned excellent data, the requirements of in-situ analyses limit the number and kindsof samples that can be accessed during the limited lifetime of a mission. A very smallnumber of samples were analyzed on the Surveyor missions 5, 6, and 7 (2, 1, and3 samples, respectively) [24]. Over an operational period of 322 days, the Lunokhodrover provided 25 soil analyses and traveled 10,540 m. In one month of operation, theSojourner rover of the Pathfinder mission returned 10 chemical analyses of Martiansoils and rocks [29] from a 100 m2 area. Clearly, a stand-off method of analysis havinga short analysis time per sample will greatly increase the scientific return from futuremissions.

The consideration of laser-based analysis methods for space exploration extends backat least two decades. For example, in 1986 a German firm conducted a study of instru-mentation for the Max Planck Institute for a flyby asteroid mission [30]. A conceptfor a multi-instrument analysis package (FRAS or “Facility for Remote Analysis ofSmall Bodies”) was developed. The instrument suite would include a laser to remotelyinterrogate the target surface. Instruments would include: time-of-flight laser ioniza-tion mass spectrometry, secondary ion mass spectrometry, laser-induced fluorescence,and UV spectrometry along with remote Raman spectrometry and surface profile mea-surements. Although laser plasma spectroscopy was known at the time, its use wasnot described. Perhaps the use of LIBS was considered but not implemented for somereason.

Fig. 5. Previously deployed elemental analysis methods require either that the detector be posi-tioned on the sample (left) APXS on pathfinder (1997) or that a sample be retrieved by amechanical arm (center) for x-ray fluorescence analysis on Viking (1977). Photographs showingdemonstrations of the two methods. Photos courtesy NASA/JPL. (right) Laser spark formed onsoil showing remote analysis capability requiring only optical access to the target.




Fig. 6. Artist’s conception of a laser beam interrogating the moon Phobos as planned for thePhobos 1 and 2 missions.

The use of a laser for chemical analysis on an interplanetary craft was first realizedon two Mars-bound Soviet missions, Phobos 1 and 2 launched in 1989. Operation of thelaser instruments, named LIMA-D, involved firing laser pulses at Phobos from a 30 mdistance during a flyby as depicted in Fig. 6 [31]. An area of 1–2 mm in diameter was tobe evaporated to a depth of 0.002 mm thickness. From the gas cloud of ionized particlesreaching the spacecraft, a mass spectrometer was to determine the chemical composition.The mass measuring range was between the elements hydrogen and lead. Unfortunately,a combination of equipment failures and ground control problems prevented successfuluse of the LIMA-D instruments on both spacecraft.

4.3. Advantages

There are many advantages to deploying LIBS for missions to planetary surfaces, butthe major advantage is stand-off analysis. This eliminates the time required to retrievea sample or what is even more time-consuming, piloting a rover remotely, from theearth, to a sample of interest. Autonomous navigation systems have been consideredbut not implemented for surface exploration by rovers. To deploy LIBS, the sampleneed only be optically acquired, the instrument-to-sample distance determined, the laserfired, and the plasma light collected. The value in stand-off analysis for geologicalanalysis can be seen by considering the trajectory of the Pathfinder rover over the



366 D. A. Cremers

LIBS

14 mradius

Rover

Fig. 7. (Inner circle) Area traversed by Pathfinder rover. (Outer circle) Area accessible by sta-tionary stand-off LIBS instrument on the lander.

mission lifetime. Figure 7 shows that the rover traversed an area around the lander(inner circle) that extended a maximum of 7 m radius. Using a stationary LIBS system,mounted on a pan/tilt mechanism on a mast on the lander, it is easy to envision thatan area of 14 m radius could be accessed without using the rover thereby speeding dataacquisition.

It has been demonstrated previously that remote analysis can be carried out using avery compact laser. Using the laser shown in Fig. 8, such as might be developed for aflyable instrument, useful LIBS measurements were obtained at a distance of 19 m withthe sample maintained in a Mars-like atmosphere (7 torr of CO2) [18]. As discussedbelow, this laser was also incorporated into a LIBS system tested on a rover.

beam

Fig. 8. Compact laser (35–80 mJ) used for stand-off LIBS analysis at 19 m. Power supplynot shown.




(a) (b)

(c) (d)

Fig. 9. (a), (c), (d) Images (courtesy NASA/JPL/Cornell) taken by the Opportunity rover at thelanding site (Meridiani Planum). (b) Laser plasma formed on a cliff face at 24 m distance in airon Earth. The horizontal strip in (b) is the result of moving the laser beam to interrogate differentlocations.

The stand-off analysis and point detection capability of LIBS allows interrogation ofinteresting geological features that may not be accessible to either the in-situ detectoror sample retrieval arm. Examples are shown in Fig. 9. The MER rover Opportunityobtained these images of layered deposits immediately upon landing. Using LIBS instand-off analysis mode, these layers would be directly accessible by the laser plasmaformed at a distance as shown in Fig. 9b and the layers, about 1 cm in thickness, couldbe individually sampled to evaluate any compositional differences.

Other advantages of LIBS, in addition to stand-off analysis, that make the methodparticularly attractive for space applications include:

• Rapid elemental analysis (one measurement per pulse)• Small analysis area of ≤1 mm, even at distance (Fig. 10)• Detects elements in natural matrix without sample preparation• Ability to detect all elements (high and low z)• Low detection limits for many elements (element specific, 2–1000 ppm)• Compact, lightweight, and able to operate in severe environments• Eliminates ambiguous results from current instruments (e.g. IR)• Laser ablation removes dusts and weathering layers (Fig. 10)• Easily combined with other spectroscopic methods (e.g. Raman and LIF)

Of particular importance is the ability to remove dust layers and weathered layers fromrock surfaces prior to analysis to determine the actual underlying rock composition. Anexample of hematite ablated under Mars atmospheric conditions is shown in Fig. 10.The ablation hole produced in the sample is visible after 100 shots. This figure alsoshows soil particles ablated from a basalt rock using repetitive laser pulses.



368 D. A. Cremers

Fig. 10. Samples at 5.3 m in 7 torr CO2. (Left) Large hematite sample ablated using 100 pulsesof 40 mJ each. The hole diameter is ∼200 microns. (Right) Loose soil ablated from a bulk basaltrock. The under-lying rock (within circle) is fully exposed after 50 shots of 40 mJ each.

4.4. LIBS Characteristics for Stand-off Analysis

Under typical LIBS analysis conditions, the laser plasma can be formed on all types ofmaterials and useful measurements made. At standoff distances, due to reduced powerdensities on the target and reduced light collected from the remote plasma, LIBS anal-ysis become more dependent on experimental conditions and sample properties. It isknown that the LIBS measurements are strongly affected by the ambient atmosphericpressure in which the measurements are carried out. Pressure affects ablation characteris-tics [32,33], element intensities [34,35], and detection limits [18]. Because atmospheresof target bodies are much different than earth’s atmosphere, under which LIBS has beenmainly characterized, it is important that LIBS capabilities for each target be carefullyconsidered. Current targets of interest include Mars (7 torr CO2�, Venus (90 atm CO2�and the Moon (∼10−9 torr) and asteroids. Figure 11 shows the dependence of signalsfrom three elements in soils as a function of pressure over the range 0.00002 to 580 torrwhich includes conditions on Mars and an airless body such as the Moon. Because forpressures below about 0.001 torr there are no observed changes in element signals withfurther pressure decreases down to 0.00002 torr (lowest pressure monitored in Fig. 11),measurements made at pressures below 0.001 torr should simulate an airless body suchas the Moon very well in terms of LIBS excitation. From the data we see that signalsare actually enhanced under Mars conditions (7 torr) compared to atmospheric and verylow pressures. On the other hand, LIBS signals are significantly degraded at the lowerpressures limiting the range of stand-off measurements [36].

The behavior shown in Fig. 11 can be understood as the result of the competingprocesses of collisional excitation of species in the plasma and ablation of the target. Asthe pressure decreases, the number of collisions per unit time decreases whereas the massof material ablated increases due to reduced plasma shielding [18]. For pressures above10 torr, increased sample ablation more than compensates for a loss in signal due todecreased pressure. For pressures below 10 torr, the ablation rate levels off with furtherpressure decreases whereas the number of collisions continues to decrease accountingfor the significant loss of signal for pressures <1 torr. Below about 0.001 torr, collisions




0

500

1000

1500

2000

2500

0.00

001

0.00

01

0.00

1

0.01 0.

1 1 10 100

1000

Pressure (Torr)

Inte

nsity

/100

0

Si(I) 288.16 nmMg(I) 285.21 nmMg(II) 279.55 nm

Mars

Moon,asteroids

Fig. 11. Element signals as a function of pressure determined at the in-situ analysis distanceof 7 cm.

become infrequent and material is readily ejected from the surface into the surroundingfree space so that pressure changes no longer affect emission signals.

Images of the laser plasma formed on basalt are shown in Fig. 12. At atmosphericpressure the plasma is a small intense ball of light about 3–4 mm high. As the pressure isreduced, the emitting plasma volume increases and at 7 torr (Mars atmospheric pressure)it appears significantly larger. At lower pressures simulating an airless body, only asmall plasma is observed at the target surface and as shown in Fig. 12, emitting materialcan be seen leaving the surface within a cone-shaped region.

Calibration curves for Li obtained at 4 m and at three pressures are shown in Fig. 13.A set of synthetic silicate samples were used to construct the curves. The concentrationsof elements in these samples resemble a typical soil. Between different samples the majorelements (the matrix) remain constant whereas the concentrations of minor elements�<1000 ppm� vary between samples. The data show good correlation between the Liconcentration and Li signal. There is a significant decrease in the slope of the calibrationcurve at the lowest pressure (55 mtorr) although these data do not show a loss of

Fig. 12. Images of individual laser plasmas formed on basalt rock at (left) 585 torr, (center)7 torr, and (right) 0.00012 torr pressures recorded using an ICCD. The gain of the camera wasadjusted so the intensity of the images for 7 and 0.00012 torr are enhanced by a factor of sevenfor display here. The delay and width of the ICCD gate pulse were 2 �s and 80 ms, respectively.



370 D. A. Cremers

Li 670.8 nm

R2 = 0.9905

R2 = 0.9991

R2 = 0.9907

0

100000

200000

300000

400000

0 100 200 300 400 500 600

Concentration, ppm

Inte

nsity

585 Torr

7 Torr

55 mTorr

Fig. 13. Calibration curves for Li at pressures of �� 585 torr, �� 7 torr, and �� 0.055 torr.Stand-off distance was 4 m.

sensitivity for the higher Li concentrations as observed for the other curves (2nd order fit).The reduced sensitivity of the curve at 55 mtorr was observed for all other elementsstudied (Al, Sr, Ba, Mn) and indicates that stand-off analysis distances for targetshaving low ambient pressures will be more limited compared to bodies with atmospherepressures of a few torr or greater. On the other hand, one study has shown that somematrix effects may be reduced at lower pressures [37].

The ability to remove dusts and clean weathered layers from a rock surface tointerrogate the underlying bulk material is an important LIBS capability that can beaccomplished at stand-off distances. Figure 14 shows the time required to ablate throughbasalt and limestone at different pressures. Aluminum metal is included for comparison.The data for basalt is scattered compared to the other data probably because of thenon-uniformity of sample composition. The basalt surface showed significant differencesin appearance, probably indicating differences in composition from spot to spot. In allcases, below a pressure of 10 torr, the ablation times were constant indicating that plasmashielding of the surface was no longer significant.

10

100

1000

10000

0.01 0.1 1 10 100 1000

Pressure (Torr)

Tim

e (s

)

Fig. 14. Time required to completely ablate through �� Al (1.56 mm thick), �� basalt (2 mmthick), and �� limestone (2 mm thick) at different pressures for 10 Hz ablation. Target distancewas 4 m.




The data presented above show the feasibility of LIBS to provide standoff analysisat reduced pressures as well as the characteristics of LIBS as a function of pressure.These data pertain to Mars, the Moon, and asteroids. Another target of interest is Venuscharacterized by pressures on the order of 90 atm and temperatures of 725 �C. Theuse of LIBS at high temperatures has not been shown to be a problem with moltenglass and metals being analyzed. There is fragmentary data concerning LIBS analysisat high pressures on the order of 30 atm [38]. However, recent work has shown thatmeasurements providing useful LIBS spectra can be carried out at higher pressures [39].Figure 15 shows basalt spectra obtained at 1 m at 90 atm and at 0.77 atm for comparison.At room temperature, CO2 liquifies at pressures above about 58 atm and so could notbe used. Nitrogen gas was used instead. Lines of some major elements in the sampleexhibit strong self-absorption whereas other lines do not appear affected by the pressure.This indicates that analytical lines will have to be carefully selected. Because of thehostile environment on Venus, a LIBS system will almost certainly be confined to thelander where insulation from the high pressure and high temperature can be provided,thereby necessitating stand-off analysis. Although the spectra of Fig. 15 were obtainedat only 1 m distance, the strength of the signals show that stand-off analysis of manymeters should be feasible.

0

100000

200000

300000

400000

500000

600000

700000

385 390 395 400 405 410 415 420 425

Wavelength (nm)

Sig

nal (

coun

ts)

Sig

nal (

coun

ts)

0

20000

40000

60000

80000

100000

120000

140000

160000

385 390 395 400 405 410 415 420 425

Ca CaCa

Al

Si

Fe

Sr

TiBaMn

Fig. 15. Comparison of spectra obtained at (top) 0.77 atm pressure and (bottom) at the Venussurface pressure of 90 atm. Gas was nitrogen [39].



372 D. A. Cremers

4.5. Capabilities

Several studies have addressed the feasibility of LIBS for space exploration. The resultsof some of these studies are summarized in Table 2.

For space missions both qualitative and quantitative information is important. Anexample of the use of qualitative data is shown in Fig. 16. Spectra of basalt and dolomiterocks recorded from field samples and certified rock powders are shown. Comparisonof the spectra shows that the two basalt spectra are very similar as are the two dolomite

Table 2. Summary of some studies of LIBS for space exploration applications

Results Distance(m)

Ref

Demonstration of LIBS analysis of a cliff bank in air at 24 m using labequipment outdoors. Discussion of requirements for a flyable LIBSsystem.

24 3

Remote analysis of an Apollo 11 rock stimulant at stand-off distances.Demonstration that stand-off LIBS has sufficient sensitivity to monitorthe elements Si, Ti, Al, Fe, Mg, Mn, Ca, Na, K, P, Cr at concentrationsin the rock stimulant.

10.5 40

Detailed study of stand-off LIBS using moderate pulse energy (80 mJ)with samples in 7 torr CO2. Preliminary evaluation of analyticalcapabilities. Use of micro laser for stand-off LIBS.

<19 18

Demonstration of (1) a compact LIBS system operated on board aNASA rover and (2) qualitative analysis capabilities such as rockidentification.

2–3 19

Study of factors affecting plasma emission under Mars atmosphericpressure and composition conditions.

1 20

Under Mars atmospheric conditions, a comparison was made betweenanalysis results obtained by CF (calibration free)-LIBS and SEM-EDX.

0.15 21

Study of plasma emission characteristics and determination of optimalexperimental parameters for samples interrogated in air and in asimulated Mars atmosphere.

0.225 22

Evaluation of stand-off LIBS for analysis of water ice and ice/soilmixtures under Mars atmospheric conditions.

4 & 6.5 41

Study of the S and Cl detection at stand-off distances in a Marsatmosphere.

3–12 42

Demonstration of LIBS at 90 atm for application to a Venus mission. ∼1 39

Study of the use of the vacuum ultraviolet (VUV) for monitoringelements in geological samples in a Mars atmosphere. The residual7 torr CO2 gas will prohibit detection of VUV lines at stand-offdistances.

in-situ 43

Study of the effect of atmospheric pressure on the analysis of soil andclay samples and the effect of pressure on some matrix effects.

in-situ 37

Comparison of LIBS capabilities at atmospheric, Mars, and lowpressures (simulating the Moon) for in-situ and stand-off analysis.

in-situ5.3

36

ElsevierAMS

Jobcode:

LRN

Ch16-N51734

25-7-2007

9:02a.m.

Page:373

Trim:165 ×

240MM

TS:Integra

Font:

Times

Size:10/12pt

Margins:Top:13MM

Gutter:20MM

T.Area:125mmX198mm

4Color

COP:Recto

Floats:

Inline

Rem

oteA

nalysisby

LIB

S:A

pplicationto

SpaceE

xploration373

Certified dolomite sample Dolomite rock samplefrom the field

Certified basalt sample

100

200

300

400

500

600

224 244 264 284 304 324

Wavelength (nm)

224 244 264 284 304 324

Wavelength (nm)224 244 264 284 304 324

Wavelength (nm)

Basalt rock samplefrom the field

224 244 264 284 304 324

Wavelength (nm)

100

200

300

400

500

600

700

800

100

300

500

700

900

1100

100

300

500

700

900

1100

Fig. 16. Comparison of LIBS spectra of powdered certified rock standards (left column) and corresponding bull rocks (rightcolumn) showing how strong similarities/differences can be used to identify rock types.



374 D. A. Cremers

3000

2500

2000

2000

1500

1500

1000

1000

500500

0 00 50

Shot number Shot number100 150 0 50 100 150

Fig. 17. Variation in Mn (left) and Si (right) LIBS signals as weathered granite is repetitivelyablated. Pulse energy was 100 mJ.

spectra thus allowing identification of the rock types. Similar results were obtained usingother rocks and their corresponding powders.

The ability to ablate through a weathered rock layer and monitor changes in compo-sition is demonstrated in Fig. 17. The weathered layer may have a composition differentfrom that of the bulk rock thereby complicating measurements using passive meth-ods. For Fig. 17, a heavily weathered granite rock was repetitively sampled by 100 mJlaser pulses and signals from Mn and Si were monitored. The strong Mn signal in theweathered layer (related to biogenic activity on Earth) decreases significantly as theinterior of the rock is sampled whereas the Si signal remains fairly uniform. From anestimate of the depth penetrated on each laser pulse, an estimate of the layer thicknesscan be made. In basalt, the depth of penetration is about 0�5 �m/pulse for a 40 mJ laserpulse.

In addition to removing weathered layers, the ability to remove dusts and soil par-ticles is important. Figure 10 demonstrates this capability visually. Measurements ofthe number of pulses required to ablate away soil thicknesses of 1, 2, and 3 mm froman underlying surface in 7 torr CO2 (oriented horizontally, pulses of 80 mJ directedvertically downward, 19 m) were carried out. Only 4, 14, and 28 pulses were required,respectively, to remove the soil particles so the underlying surface could be interrogatedby the laser plasma.

A LIBS spectrum of loose soil in 7 torr CO2 analyzed at 5.3 m is shown in Fig. 18. Thedepression produced in the soil by 100 pulses at 5.3 m is also shown. Even though theaction of the laser pulse moved the soil on each shot, useful spectra were obtained.The spectrum was obtained using a very compact spectrograph/detector system (HR2000,Ocean Optics, Inc.) of a type likely to be incorporated in a flyable LIBS instrument.

The use of LIBS to identify water ice (e.g. via OH emission at 306.4 nm), analyzeice/soil mixtures, and interrogate ice cores has been demonstrated [41]. This capabilityis important for Mars exploration because the polar regions may represent an archiveof past geologic activity in the form of layered deposits of ice and dusts. The abilityto sample these layers in the form of extracted cores could provide data reaching backmillions of years. A photo of a laser spark interrogating a water ice sample is shown inFig. 19.

Atmospheric conditions such as on Mars may in some cases significantly changethe LIBS spectrum increasing the detectability of some elements. For example, for




70

80

90

100

110

380 400 420 440 460

Wavelength (nm)

Inte

nsity

Fig. 18. Photo of loose soil having been analyzed at 5.3 m using 100 laser shots and the resultingLIBS spectrum. The width of the depression in the soil is 4–5 mm.

Fig. 19. Laser plasma formed on water ice. The use of LIBS for analyzing ice and ice/soil mixturesin 7 torr CO2 at several meters distance has been demonstrated [41].

Cl and S, two elements important to Mars geology, the Cl lines at 479.42, 480.98,and 481.91 nm and the S lines at 543.28 and 545.38 nm are observed only marginallyat atmospheric pressure even from samples having high concentrations. Under Marsatmospheric conditions, however, these lines become much more prominent and areuseful analytical lines and lie in a spectral region more easily detectable with an ICCDthan the IR lines of these elements [42].

Some representative LIBS limits of detection for stand-off analysis of samples at threedifferent distances are presented in Table 3. In general, LIBS has sufficient sensitivityto monitor the majority of elements of interest to geologists at useful concentrations.On the other hand, some elements such as Cl and Br may be present at levels belowcurrent LIBS detection limits (e.g. Cl ∼1�2% and Br ∼20–1000 ppm at certain locationson Mars established by the MER rovers).



376 D. A. Cremers

Table 3. Stand-off LIBS limits of detection (LOD)

LOD values for elements in soils and soil simulants(100 mJ/pulse; 19 m; 7 torr CO2� [18]

element LOD (ppm) element LOD (ppm)

Ba 21 Ni 2234Cr 39 Pb 95Cu 43 Sn 84Hg 647 Sr 1�9Li 20 Ba 12

LOD values for elements in ice/10% soil mixtures(100 mJ/pulse; 4 & 6.5 m; 7 torr CO2� [41]

element LOD @ 4 m (ppm) element LOD @ 6.5 m (ppm)

Ba 12 Ba 66Li 6 Li 3Mn 15 Mn 101Sr 1 Sr 2Ti 111 Ti 520

As noted above, LIBS can be readily combined with other spectroscopic methods forremote analysis. One example is Raman spectroscopy which can assist in determiningthe mineralogy of a sample providing information complementary to a LIBS elementalanalysis. The use of Raman at remote distances has been demonstrated [44] and acombined LIBS/Raman instrument has been demonstrated in the laboratory.

4.6. Instrumentation

Instrumentation for spacecraft must meet stringent requirements related to the harshenvironment encountered on the journey as well as on the surface of the missiontarget. This includes radiation protection and shielding from extreme cold and heat,especially during operations on the planetary surface. In addition, size, mass, andpower requirements are important engineering parameters. A compact LIBS systemhas been developed using mainly off-the-shelf components and the unit has been usedin the field [19]. A photo of the unit installed on a NASA/Ames rover is shownin Fig. 20.

As a result of work by an international team which included studies of LIBS capa-bilities for Mars analysis [3,18–20,37,40–43] and engineering work on development ofa flyable laser, optical system, and spectrograph, a design for a LIBS instrument wassubmitted to NASA for consideration for inclusion on the 2009 Mars Science Laboratory(MSL) rover. Following an evaluation of submitted instruments, it was announced thata combined LIBS instrument and micro-imager (named ChemCam) was selected for theMSL mission. An artist’s conception of LIBS operating on the MSL rover is shownin Fig. 21. Current specifications indicate this rover will be the largest ever landed on




WideEye

HawkEyeLIBS Spectrometer

Pantilt

Radio EthernetAntennas

ElectronicsEnclosure Front

Hazcams

WideEye

HawkEyeLIBS Spectrometer

Pantilt

Radio EthernetAntennas

ElectronicsEnclosure Front

Hazcams

Fig. 20. LIBS instrument for stand-off analysis on a NASA/Ames rover [19]. Photo courtesy ofNASA/AMES.

Fig. 21. Artist’s conception of MSL rover with LIBS instrument interrogating a remotely-locatedrock. Rover artwork courtesy of NASA/JPL.

Mars (900 kg mass) with a mission lifetime projected to be >1 year including a 6 kmtraverse of the Martian surface. The LIBS instrument is specified to have a stand-offanalysis range of 2–12 m and can perform an analysis (75 shots) approximately every2 minutes [45].



378 D. A. Cremers

REFERENCES

[1] V.E. Zuev, A.A. Zemlyanov, Y.D. Kopytin, and A.V. Kuzikovskii, High Power Laser Radi-ation in Atmospheric Aerosols, D. Reidel, Boston (1985).

[2] D.A. Cremers, Appl. Spectrosc., 41 (1987) 572.[3] J.D. Blacic, D.R. Pettit, and D.A. Cremers, Laser-induced breakdown spectroscopy for remote

elemental analysis of planetary surfaces, Proceedings of the International Symposium onSpectral Sensing Research, HI (1992).

[4] O. Samek, D.C.S. Beddows, J. Kaiser, S.V. Kukhlevsky, M. Liska, H.H. Telle, and J. Young,Opt. Eng., 39 (2000) 2248.

[5] U. Panne, R.E. Neuhauser, C. Haisch, H. Fink, and R. Niessner, Appl. Spectrosc., 56(2002) 375.

[6] S. Palanco, J.M. Baena, and J.J. Laserna, Spectrochim. Acta, B57 (2002) 591.[7] S. Palanco and J. Laserna, Rev. Sci. Instrum., 75 (2004) 2068.[8] C. Lopez-Moreno, S. Palanco, and J.J. Laserna, J. Anal. Spectrom., 19 (2004) 1479.[9] P.L. Garcia, J.M. Vadillo, and J.J. Laserna, Appl. Spectrosc., 58 (2004) 1347.

[10] S. Palanco, S. Conesa, and J.J. Laserna, J. Anal. At. Spectrom., 19 (2004) 462.[11] H. Wille, M. Rodriguez, J. Kasparian, D. Mondelain, J. Yu, A. Mysyrowicz, R. Sauerbrey,

J.P. Wolf, and L. Wöste, Eur. Phys. J. AP, 20 (2002) 183.[12] J. Kasparian, M. Rodriguez, G. Méjean, J. Yu, E. Salmon, H. Wille, R. Bourayou,

S. Frey, Y.-B. André, A. Mysyrowicz, R. Sauerbrey, J.-P. Wolf, and L. Wöste, Science, 301(2003) 61.

[13] J. Kasparian, R. Sauerbrey, and S.L. Chin, Appl. Phys. B 71 (2000) 877.[14] http://pclasim47.univ-lyon1.fr/[15] Ph. Rohwetter, J. Yu, G. Mejean, K. Stelmaszczyk, E. Salmon, J. Kasparian, J.-P. Wolf and

L. Wöste, J. Anal. At. Spectrom., 19 (2004) 437.[16] K. Stelmaszczyk, Ph. Rohwetter, G. Méjean, J. Yu, E. Salmon, J. Kasparian, R. Ackermann,

J.-P. Wolf, and L. Wöste, Appl. Phys. Lett., 85 (2004) 3977.[17] Ph. Rohwetter, K. Stelmaszczyk, L. Wöste, R. Ackermann, G. Méjean, E. Salmon,

J. Kasparian, J. Yu, and J.-P. Wolf, Spectrochim. Acta B60 (2005) 1025.[18] A.K. Knight, N.L. Scherbarth, D.A. Cremers, and M.J. Ferris, Appl. Spectrosc., 54 (2000) 331.[19] R.C. Wiens, R.E. Arvidson, D.A. Cremers, M.J. Ferris, J.D. Blacic, and F.P. Seelos, IV,

J. Geophys. Res. [Planets], 107 (E11) (2002) 8004.[20] R. Brennetot, J.L. Lacour, E. Vors, A. Rivoallan, D. Vailhen, and S. Maurice, Appl. Spec-

trosc., 57 (2003) 744.[21] F. Colao, R. Fantoni, V. Lazic, A. Paolini, G.G. Ori, L. Marinangeli, and A. Baliva, Planetary

and Space Science, 52 (2004) 117.[22] F. Colao, R. Fantoni, V. Lazic, and A. Paolini, Appl. Phys., A79 (2004) 143.[23] J.F. Bell, B.A. Campbell, and M.S. Robinson, Remote Sensing for the Earth Sciences: Manual

of Remote Sensing, John Wiley and Sons, New York, 1999, pp. 509–564.[24] “Surveyor Program Results” NASA SP-184 (1969).[25] B.C. Clark, A.K. Baird, H.J. Rose Jr., P. Toulmin III, R.P. Christian, W.C. Kelliher,

A. J. Castro, C.D. Rowe, K. Keil, and G.R. Huss, J. Geophys. Res. 82 (1977) 4577.[26] T. Economou, Radiat. Phys. Chem. 61 (2001) 191.[27] R. Rieder, R. Gellert, J. Bruckner, G. Klingelhofer, G. Dreibus, A. Yen, and S.W. Squyres,

J. Geophys. Res. B108 (E12) (2003) 8066.[28] D.M. Hunten, L. Colin, T.M. Donahue, and V.I. Moroz (eds.), Venus, The University of

Arizona Press, Tucson (1983) pp. 45–68.[29] M.P. Golombek, R.A. Cook, T. Economou, W.M. Folkner, A.F.C. Haldemann,

P.H. Kallemeyn, J.M. Knudsen, R.M. Manning, H.J. Moore, T.J. Parker, R. Rieder,J.T. Schofield, P.H. Smith, and R.M. Vaughan, Science, 278 (1997) 1743.




[30] Instrument is described in more detail in: A. Vertes, R. Gijbels, and F. Adams, (eds.), LaserIonization Mass Analysis, John Wiley and Sons, New York (1993) pp. 529–532.

[31] R.Z. Sagdeev, G.G. Managadze, I.Yu. Shutyaev, K. Szego, and P.P. Timofeev, Adv. SpaceRes. 5 (1985) 111.

[32] J.M. Vadillo, J.M. Fernandez Romero, C. Rodriguez, and J.J. Laserna, Surf. Interface Anal.26 (1998) 995.

[33] J.M. Vadillo, J.M. Fernandez Romero, C. Rodriguez, and J.J. Laserna, Surf. Interface Anal.27 (1999) 1009.

[34] Y. Iida, Appl. Spectrosc. 43 (1989) 229.[35] M. Kuzuya, and O. Mikami, J. Anal. At. Spectrom. 7 (1992) 493.[36] R.D. Harris, D.A. Cremers, K. Benelli, and C. Khoo, unpublished data.[37] B. Sallé, D.A. Cremers, S. Maurice, and R.C. Wiens, Spectrochim. Acta, B60 (2005) 479.[38] M. Noda, Y. Deguchi, S. Iwasaki, and N. Yoshikawa, Spectrochim. Acta, B57 (2002) 701.[39] Z.A. Arp, D.A. Cremers, R.D. Harris, D.M. Oschwald, G.R. Parker, and D.M. Wayne,

Spectrochim. Acta, B59 (2004) 987.[40] D.A. Cremers, M.J. Ferris, C.Y. Han, J.D. Blacic, and D.R. Pettit, Proc. Soc. Photo Opt.

Instrum. Eng. (SPIE), 2385 (1995) 28.[41] Z.A. Arp, D.A. Cremers, R.C. Wiens, D.M. Wayne, B. Salle, and S. Maurice, Appl. Spec-

trosc., 58 (2004) 897.[42] B. Sallé, J.-L. Lacour, E. Vors, P. Fichet, S. Maurice, D.A. Cremers, and R.C. Wiens,

Spectrochim. Acta, B59 (2004) 1413.[43] L.J. Radziemski, D.A. Cremers, K. Benelli, and C. Khoo, Spectrochim. Acta, B60 (2004) 237.[44] S.K. Sharma, P.G. Lucey, M. Ghosh, H.W. Hubble, and K.A. Horton, Spectrochim. Acta,

A59 (2003) 2391.[45] http://marsprogram.jpl.nasa.gov/msl/mission/sc_instru_chemcam.html



Chapter 17

LIBS for Aerosol Analysis

D. W. Hahna and U. Panneb

aDepartment of Mechanical and Aerospace EngineeringUniversity of Florida, Gainesville, FL 32611-6300, USA

bDepartment of Chemistry, Humboldt-Universitaet zu Berlin, Richard-Willstaetter-Str. 1112489 Berlin, Germany

1. INTRODUCTION

LIBS is well suited for the analysis of aerosol particles due to the unique point samplingnature of the laser-induced plasma. The discrete plasma volume corresponds well withthe discrete nature of aerosol particles to enable a wide range of data analysis options,including spectral averaging, conditional spectral processing, and single-pulse analysis.In this chapter, a detailed introduction to aerosol science and aerosol analysis is presentedto frame the overall problem of LIBS-based aerosol sampling. A detailed treatment of thelaser-induced breakdown process is focused on the gas phase processes associated withplasma initiation and propagation. Quantitative aerosol analysis is presented in termsof the aerosol-sampling problem, followed by direct and indirect quantitative aerosolmeasurements. The chapter concludes with a detailed discussion of LIBS applicationsto aerosol analysis and future directions in this challenging and important area.

2. FUNDAMENTALS OF AEROSOL ANALYSIS

The analysis of aerosol samples presents a challenging analytical problem due to thewide variation in both aerosol particles and measurement needs. For example, ambient airaerosol particles are a complex mixture originating from both natural and anthropogenicsources, and are generated either via direct emission processes (primary particles) or viagas-to-particle conversion processes (secondary particles). Aerosol particles in the ambi-ent atmosphere originate from several sources: wind-raised dust, agricultural activitiesand open fields, sea spray, industrial activity, traffic, volcanoes, wild fires and com-bustion processes, as well as photochemical conversion of gases to particles. Chemicalcomponents include sulfates, ammonium, nitrates, chlorides, trace metals, carbonaceousmaterials, crustal elements, and water. Because laser-induced breakdown spectroscopy(LIBS) is essentially an elemental analysis technique, attention must be given to aerosol




382 D. W. Hahn and U. Panne

particles containing elements that stand in contrast to the background gaseous matrix.For example, nitrogen (e.g. as nitrates), oxygen (e.g. as metal oxides), carbon (e.g. assoot or organic) may all be present in particulates, but such elements also arise from thepurely gas phase, thereby making the quantification of the gas-phase and particulate-phase partitioning difficult. Such measurements are discussed in detail below; however,given these difficulties, much of the LIBS analysis of aerosols has focused on elementsnot found in the gas phase, such as trace metals.

Trace metals are ubiquitous in various raw materials, such as fossil fuels and metalores, as well as in industrial products; hence many trace metals evaporate entirely orpartially from raw materials during the high-temperature production of industrial goods,combustion of fuels, and incineration of municipal and industrial wastes, thereby enteringthe ambient air with exhaust gases. Most of the time, aerosols from such sources havevery distinct particle distributions with a low geometric standard deviation and a meanparticle size on the order of 0�1 �m, yielding relatively low quantities on a particle-by-particle basis. For example, a solid spherical particle of 100 nm diameter with a densityof 2 g cm−3 has a total mass of about 1 femtogram (fg).

Natural sources are related primarily to the geological presence of trace metals inthe crustal material and are transformed into aerosol particles during various naturalphysical, chemical, biological and meteorological processes. For example, soil-deriveddust accounts for over 50% of the total Cr, Mn, and V emissions, as well as for 20–30%of the Cu, Mo, Ni, Pb, Sb, and Zn released annually to the atmosphere. Furthermore,volcanic emissions (which are perhaps the most extensively studied source) appear asa significant source, which account for 40–50% of the total natural Cd and Hg and20–40% of the total natural As, Cr, Cu, Ni, Pb, and Sb emitted yearly. Sea-salt aerosolsseem to account for <10% of atmospheric trace metals from natural sources. In certainparts of the world forest fires are the major emission source and more than 10% ofatmospheric Cu, Pb and Zn from natural sources can originate from fires.

Combustion of fossil fuels to produce electricity and heat is the main source ofanthropogenic emissions of atmospheric Be, Co, Hg, Mo, Ni, Sb, Se, Sn, and V [22] andan important source of As, Cr, Cu, Mn, and Zn. In general, the amount of emissions froma conventional thermal power plant depends on the content of trace metals in the fuels,the physical and chemical properties of trace metals during combustion, technologicalconditions of a burner, and the type and efficiency of emission control equipment. Themost commonly inventoried heavy metals (i.e. priority metals for emission reductions)are As, Be, Cd, Cr, Hg, Pb, and Zn. This is due to their effects on environmental andhuman health, as well as their ubiquitous appearance in the environment.

Given the broad range of aerosol size and composition, as briefly summarized above,the knowledge of aerosol size distributions is essential because the particle size signifi-cantly affects ambient transport and deposition processes as well as the described uptakein the respiratory system. Moreover, elemental size distributions can give an indication ofthe source of the element [2,3]. The typical characteristics of size distributions vary fromelement to element and from sampling location to sampling location, but generally theydisplay a trimodal or bimodal distribution [4]. Not surprisingly, many tasks in aerosolanalysis stem ultimately from the potential health effects of aerosols. The applicationsrange from aerosol analysis in combustion to process analysis and control in variousindustrial production processes. For example, the major issue in occupational hygiene isthe question of exposure, which may be considered as the time averaged concentration



LIBS for Aerosol Analysis 383

of the agent under study at the relevant interface between the environment and the bio-logical system, i.e. the worker. Not only the concentration, but also other characteristicssuch as composition, morphology, and particles size distribution are presently cominginto the focus of exposure threshold definitions.

Given the wide range of aerosol sampling needs in conjunction with the great vari-ation of particles presented to any analytical instrument, it is not surprising that onemight envision a wide number of desirable characteristics in such a tool. The “ideal”aerosol analyzer was described by Friedlander, who classified in his seminal 1970–71work different aerosol measurement techniques on the type of information given by theinstrument [5,6]. Although today’s aerosol analysis by mass spectrometry approachesthe ideal of a perfect single particle counter, which should provide the complete size-resolved chemical composition of an aerosol in real time, most current methods are farfrom this ideal. The application of LIBS for aerosol analysis offers inherent advantagesthat overcome many of the limitations inherent in current aerosol analysis tools. TheLIBS technique has been developed in recent years as a novel means for the quantitative,direct measurement of particle size and composition of individual particulates, althoughmany fundamental issues remain regarding the interactions of the laser-induced plasmawith aerosol particles. As a starting point for LIBS-based aerosol analysis, the followingsections provides a detailed review of the laser-induced breakdown process, as focusedon the gas phase processes associated with plasma initiation and propagation.

3. LASER INDUCED BREAKDOWN OF GASES

With the advent of the first laser more than 40 years ago, nearly simultaneously the firstlaser-induced plasmas in gases were reported by Maker et al. [7] and Meyerand et al. [8].Since then, a considerable body of literature aimed at understanding the various aspectsof gas and aerosol-induced breakdown has been written. For detailed information thereader is referred to the reviews [9–11] and corresponding chapters in the monographs[12–15] and references therein.

In contrast to a breakdown on solids or liquids, the irradiance for breakdown,i.e. plasma formation, in gases is in excess of 109 W cm−2 due to the involved mech-anism of multiphoton ionization (MPI) and and/or cascade ionization [16]. While highenergy photons can ionize gases in single-photon interactions, it is not obvious in whichway a laser photon of low energy (1–2 eV), compared to the ionization potential ofcommon gases, can generate a breakdown. The formation of a laser plasma in a neu-tral gas follows three distinct but overlapping stages: (i) plasma ignition, (ii) plasmagrowth and interaction with the laser pulse (in case of nanosecond laser pulse), and(iii) plasma development accompanied by shock wave generation and propagation in thesurrounding gas.

Plasma ignition comprises the growth in the free electron and ion concentration witharrival of the first laser photons. The growth stage is characterized by a fast amplificationof free electrons and ions. The term breakdown is rather arbitrarily and loosely defined inthe literature, often the luminous plasma emission or the acoustical detection of the shockwave is taken as the only criterion. In the following, breakdown will imply an electrondensity (Ne) >1013 cm−3 or a degree of ionization about 10−3, which permits a significantabsorption and scattering of the incident laser radiation and leads to a fully developed




plasma very fast (Ne typically 1017–1019 cm−3). With the onset of breakdown a highlyionized plasma develops in which – in the case of nanosecond pulse – further absorptionand photoionization occurs. With the end of the laser pulse, the plasma dies graduallyaway as a result of radiation and conduction of thermal energy, diffusion, attachment,and recombination of ions and electrons, until local thermodynamic equilibrium withthe surrounding gas is restored.

For nanosecond and microsecond pulses at pressures above 100 mbar �104 Pa�, thebreakdown at long wavelengths �>1 �m� proceeds via an electron avalanche or cascadeby inverse bremsstrahlung (IB), i.e. free-free absorption. In this way, electrons absorbphotons in the presence of a neutral gas atom for momentum transfer and acquireultimately sufficient energy for collisional ionization of gas atoms and albeit generationof further electrons. At visible or UV wavelengths highly excited states of the gasmolecules or atoms can be readily photoionized over times much shorter than the typicalnanosecond pulse widths. This reduces the observed breakdown thresholds considerably,and reduces the losses in electron density due to diffusion.

Under low-pressure conditions, the breakdown is initialized via electrons from mul-tiphoton absorption and ionization (MPI), while at latter time the cascade process oftenovertakes the electron generation. MPI comes also back into play when considering thesource of the “first” electron for the cascade process. In contrast to MPI, the cascadeionization is not self sufficient, but requires at least one electron in the focal volume.Due to natural local radioactivity (e.g. cosmic rays or ultraviolet radiation) ions occurnaturally in the atmosphere at a concentration of 102–103 cm−3, although free electronsare immediately attached to O2 yielding O−

2 . The mean lifetime of O−2 (which can be

treated as a free electron due to electron tunneling) is in the order of 10−7 s, so that theprobability of encountering an electron in an interaction volume of about 10−6 m3 duringa laser pulse with nanosecond pulse duration is rather negligible. The breakdown thresh-old depends weakly on the gas pressure through p−1/m, where m is the necessary numberof electrons for MPI. The breakdown starts when a fraction of atoms (on the order of10−3) present in the interaction volume is ionized. In air, a MPI-initiated breakdown atthe fundamental wavelength of a Nd:YAG (1064 nm) is an 8- or 10-photon process forO2 and N2 [17–22]. At visible and UV wavelengths highly excited states can be readilyphotoionized over times much shorter than the laser pulse duration, with thresholdssignificantly lower �5–10 GW cm−2�. Stark shift and broadening of intermediate levelscan additionally bring the levels into resonance with the laser wavelength. Also, if asufficient Ne is generated by MPI early enough in the pulse and affects the diffusion ofelectrons out of the focal volume, the diffusion becomes affected by the space charge ofthe ions remaining in the focal region. This ambipolar nature reduces the overall diffu-sion, and the breakdown threshold will be lowered especially in experiments with smallfocal spots where diffusion losses can be important [23,24]. Finally, it must be remem-bered that laser modes can lead to local fields that significantly exceed the average valueover the focal volume. Hence, the effective irradiance of reported experimental valuesare probably somewhat larger than the required irradiance. At longer wavelengths theproblem of generating the first electron in the absence of gaseous or aerosol impuritiesbecomes more serious. Breakdown at 1064 nm [25–28] can then increase to an irradiancelevel of 1012 W cm−2, where the electric field induces tunneling of an electron throughits potential barrier.




While production of the first electron is crucial for the start of the breakdown, theirradiance threshold is governed by the cascade or avalanche growth of ionization, whichis fed by absorption of the laser light. Electrons in a laser field will gain energy to ionizeand increase in number through electron-neutral inverse bremsstrahlung (IB). Microwavebreakdown theory [29–33] has been successful to describe breakdown thresholds atwavelengths beyond 1 �m, where IB is the dominant mechanism; while at visible andnear-UV wavelengths data are not so well understood [24,34–36]. If electron-impactionization was the dominant mechanism leading to gas breakdown, the breakdownthreshold for a given gas at a given pressure would scale inversely with the IB absorptioncoefficient, i.e. approximately with �−2. Experimental observations [37–43] revealedthat the breakdown thresholds peaked in the middle of the visible spectrum due to acompetition between IB and MPI of ground and of excited states [44–49].

The cascade process will lead to an exponential growth of electron density. Assuminga fully developed plasma at Ne ≈ 1017 cm−3, about 40 generations are required to growfrom an assumed initial value of Ne�0 ≈ 1–10 cm−3 in the focal volume, i.e. 99% ofthe ionization is produced in the last 7 generations. When the electron concentrationexceeds 1013 cm−3, i.e. the onset of the breakdown, electron-electron collisions will tendto populate the tail of the electron distribution function and this has a dramatic effect onthe cascade rate. Quantities such as the growth and losses from the cascade and the timeto breakdown are determined by conditions at times when the electron concentration issmall. Losses can be through several inelastic processes such as vibrational and rotationalexcitation (for polyatomic gases), excitation of electronic levels of atoms and molecules,elastic collisions, attachment, recombination, and diffusion.

Although the theoretical modeling of the breakdown in gases has developed to quite asophisticated level [24,34], to elucidate the exact mechanism of the plasma formation ingases, still a strict control of the experimental parameters such as focusing, laser modes,contamination of the gas and the cell is needed, accompanied by a suitable plasmadiagnostic [50,51].

The study of interactions between laser light and aerosols began in the 1960s with theavailability of the first lasers. Haught et al. reported already in 1966 the first observationsof particle-induced breakdown in gases, i.e. a breakdown on an electrostatically levitatedsingle 20-�m particle of LiH [52]. It became immediately clear that the propagation oflasers through the atmosphere and hence the operative efficiency of lasers for applicationsfrom nuclear fusion to military weapons research were intimately connected to aerosols(see [53–55] and references therein).

Once irradiated with the laser beam, an aerosol particle starts to absorb, whichleads to heating and further melting, boiling, and gradual evaporation (sublimation) ofparticle material (Reference [56] gives an elegant visualization through a moleculardynamics simulation). The process may not be uniform because the distribution ofthe electromagnetic field inside the particle is not necessarily uniform, especially withinternally mixed particles. Hot spots in the nodes of the field can virtually explode theparticle before the thermal conductivity smoothes out the temperature distribution. Forparticles small in comparison to the laser wavelength, the absorptivity decreases withparticle diameter, dp. The particle heating depends strongly on the heat losses causedby contact with the carrier gas and evaporation. The breakdown is promoted throughaerosol particles by heating the surrounding gas and providing the breakdown zonewith additional electrons caused by a number of mechanisms: heat explosion, shock




waves in the surrounding carrier gas, breakdown in evaporated matter, electron emission(thermo-, photo- and tribo-electrons). The increase in the initial electron density can beup to 107 cm−3. Naturally, this decreases the threshold and reduces the time for a fullplasma development. The presence of aerosol particles in the focal volume will lowerthe threshold for gas breakdown by several orders of magnitude (typically 107 W cm−2

compared to breakdown in pure air at 1011 W cm−2). Aerosol breakdown depends uponthe pulse width, wavelength, focusing of the laser and particle size, wherein the thresholdusually scales with the particle diameter d−1

p to d−2p . Aerosol induced breakdown has

been described by several authors with different wavelengths between 1064 nm and248 nm [57–68]. For pulse durations on the order of some microseconds, a wavelengthdependence of �−1 was observed, while with the usually employed nanosecond pulsea �−2 scaling was found [65]. The observed increase of the threshold with a decreaseof the focal beam diameter is explained through inclusion of larger particles in theinteraction volume, which produce a higher initial increase in electron density [69]. Thebreakdown of droplets is modified through the curved liquid-gas interface [55,70–74].The droplet can be envisioned as a lens that concentrates the incident light wave toa localized region just within the droplet shadow face and focuses it to a localizedregion just outside the droplet shadow face. For large transparent droplets and moderateirradiance, the breakdown occurs just outside the shadow face at which the irradiance ishighest. Whether the breakdown is initiated in the gas outside the droplet or within thedroplet depends on parameters such as the breakdown threshold of the gas and liquid,as well as the droplet morphology [75–77].

4. ANALYSIS OF AEROSOL PARTICLES BY LIBS

The analysis of aerosol samples presents a unique application for LIBS, in that the possi-bility exists to bring together the point-to-point sampling nature of laser-induced plasmaswith the discrete nature of aerosol particles. The following sections elaborate on differentapproaches to optimize the spectroscopic information from aerosol-derived analytes.

A fundamental issue with regard to the analysis of aerosol particles with the LIBS tech-nique concerns the relationship between the discrete aerosol particles and the finite-sizedlaser-induced plasma volume. With inductively-coupled plasma (ICP) atomic emissionspectroscopy, the plasma source is continuous, and analytes in solution are fed into theplasma at a constant rate. The resulting analyte signal represents an average analyteconcentration, and this mode of operation is characteristic of many analytical methods.However, with the LIBS technique, the coupling of the repetitive, finite-sized plasmawith the spatial distribution of aerosol particles must be considered. For example, if thespatial distribution of solid-phase aerosol is of such a value that the laser-induced plasmasamples particles with a high probability, then average analyte signals are reflective ofthe particle-phase contribution as well as the gas-phase species contributions. In con-trast, if the aerosol particle concentration is small such that the probability of samplingan aerosol particle in a given laser-induced plasma is very small, then the contributionof particle-derived analyte to the total signal may become a negligible fraction. Withthis latter case, the average LIBS signal may only reflect the gaseous components, withthe result being a non-detect condition with respect to the aerosol particle constituentspecies. The LIBS-based analysis of aerosol particles may be naturally partitioned into




these two statistical sampling regimes, namely one condition where the aerosol particlesampling rate is limited, and another condition where the probability of aerosol particlesampling is sufficiently high. The latter condition is well suited to traditional ensem-ble averaging techniques, where many LIBS spectra are averaged together in an effortto eliminate random, single-pulse spectral noise. With the former condition, advancedspectral processing schemes are required to compensate for the large amount of spec-tral data that contain no information regarding the aerosol-derived analytes of interest.The methodologies and implementation of LIBS-based aerosol analysis schemes arepresented and discussed for each of these operating approaches.

4.1. Spectral Ensemble-Averaging

Most laser-induced plasma spectral data are collected using intensified charge-coupleddevice (ICCD) array detectors, although increasingly, non-intensified CCD detectors arebeing examined [78,79]. The primary advantage of ICCD systems is the ability to tem-porally gate the detector, which provides for optimization of the analyte atomic emissionsignals with respect to the plasma continuum emission signals due to the differing decayrates of these two emission processes. An example of the optimal temporal gating forseveral toxic metals species is discussed in several publications [80–82], showing thatconsiderable time differences (some tens of microseconds) can exist between optimaldetection windows. The use of ICCD detectors; however, can result in a significantamount of spectral noise generally originating from the intensifier. Furthermore, theabsolute intensity of the LIBS signal (both continuum emission and atomic emission)can vary significantly on a pulse-to-pulse basis. Hence spectral data generated with theLIBS technique is inherently noisy due to the combination of natural plasma fluctuationsand intensifier/detector noise. In some of the early LIBS research toward analysis ofgaseous/aerosol samples, Radziemski, et al. noted that self-normalization of the atomicemission signal by the laser-induced plasma continuum emission provided a more robustanalyte signal [83–85]. In several recent papers, Hahn and co-workers examined thepulse-to-pulse fluctuations in detail for a number of gas-phase analyte species, includ-ing carbon, nitrogen, and hydrogen emission lines [86,87]. For example, it was foundthat absolute values of the continuum emission intensity and atomic emission intensityof the 247.8-nm carbon line varied by a factor of 4 and 6, respectively, between theminimum and maximum values for a sequence of 100 consecutive laser pulses [86]. Thecorresponding relative standard deviations (RSD) of the continuum and atomic emissionsignals were 9.5 and 11%, respectively. Alternatively, the carbon atomic emission peakintensity normalized by the nearby continuum emission intensity (Peak-to-Base or P/Bratio) varied by no more than a factor of two for the same 100 spectra, with a corre-sponding RSD of 6%. Clearly the benefit of spectral normalization was demonstratedby the nearly 50% improvement in precision, as realized in the reduction of RSD. Thespectral noise characteristics of LIBS are readily quantified through the examination ofrepresentative single-pulse and ensemble-averaged spectra.

Carranza et al. performed a detailed analysis of the noise associated with intensifiedCCD detectors as part of a comparison of ICCD and non-intensified CCD detectors [79].In that study, both single-pulse and ensemble-averaged data were examined. Figure 1




0

500

1000

1500

2000

385 405 415390 395 400 410 420

Inte

nsity

(a.

u.)

Wavelength (nm)

Fig. 1. LIBS spectra corresponding to a calcium-rich aerosol system for a 200-shot ensembleaverage and for a single shot. The ICCD detector delay was fixed at 5 �s following plasmainitiation for all measurements. Both spectra are presented at the same intensity scale, and thesingle-shot spectrum has been shifted up by 600 counts.

is adapted from that study, and shows the resulting spectra from a single-pulse mea-surement with the ICCD and a 200-pulse ensemble-averaged spectrum correspondingto the same characteristics. For these two spectra, the root-mean-square (RMS) noiseof the continuum emission intensity was calculated, as measured over relatively smoothspectral regions on each side of the calcium peaks. The average RMS noise was 2.37and 14.0 for the ensemble-averaged and single-pulse spectra, respectively. This six-foldreduction in spectral noise directly translates into enhanced detection limits, as reportedby Carranza and co-workers. For example, over a range of temporal delays between theincident laser pulse and initiation of the detector gate, the signal-to-noise ratio of the397.4-nm calcium emission line was increased by an average of 6.5 when comparingthe ensemble-averaged data to the single-pulse data. Similarly, when comparing thesingle-pulse vs. ensemble-averaged data for the non-intensified CCD, the signal-to-noiseratio was also improved by a factor of 7.

Clearly, significant benefit (i.e. improvements in analytical figures of merit) is realizedthrough the use of spectral ensemble averaging with the LIBS technique. Historically,ensemble-averaging of hundreds or thousands of spectra is by far the most widely imple-mented approach for LIBS-based analysis of gaseous samples. With such an approach,the resulting average spectrum is characterized by relatively low signal noise, and pro-vides an analyte signal representative of the average constitutive species. The constitutivespecies contribute to the analyte signal, which may contain contributions from both gas-phase and particulate-phase compounds as discussed above. Successful implementationof ensemble-averaging for LIBS-based analysis of aerosol samples is performed simi-larly to the analysis of gaseous species; however, several issues must be given attention,as summarized here.




4.1.1. Conditions for Aerosol System Analysis via Ensemble-Averaging

(A) The aerosol particle concentration (particles/volume) must be sufficiently largesuch that a sufficient number of laser induced plasmas (i.e. laser pulses) actuallysample aerosol particles.

(B) The size distribution of the targeted aerosol particles should lie within a sizeregime such that the largest particles can be totally vaporized within the laser-induced plasma, thereby presenting a linear analyte mass response with regard tothe atomic emission signal.

(C) The analyte calibration source stream must be designed to ensure that the trueLIBS analyte response is realized, including possible differences in analyteresponse due to gas-phase and solid-phase analyte sources.

First consider condition (A), namely, the exact percentage of required laser pulsesto ensure adequate sampling of an aerosol source, as measured by a significant analytesignal in the ensemble-averaged spectrum with respect to the method detection limit. Theexact sampling rate is difficult to specify in general, but must consider both the overallfraction of laser-induced plasmas that sample a particle, and the relative emission strengthof the targeted analyte. For hundreds of laser pulses, the baseline noise generally reachesa limiting value beyond which additional signal averaging provides no added benefit.Since this is the region desirable for ensemble-averaging, the following comments willassume that the spectral noise has reached the limiting value. Under such conditions, theanalyte signal intensity is reduced by a factor equal to the inverse of the aerosol particlesampling rate. For example, if the aerosol particle sampling fraction were 0.10 (i.e. 10%hit rate), then the analyte signal stemming from the particulate-phase species would bedegraded by a factor of 1/0.1, or a factor of ten. In other words, for every analyte-containing spectrum recorded, ten additional non-analyte containing spectra are recorded,which when ensemble-averaged together, produces the ten-fold decrease in particulate-phase derived emission signal. To estimate the necessary sampling rate for this modeof operation, divide the limit of detection by the expected analyte response of a singleparticle. However, with many signal responses unknown for different aerosol particlesizes and composition, it is often difficult to make such a calculation. In practice, one cansimply examine the nature of the analyte signal produced via ensemble-averaging. If theanalyte signal produces a significant signal-to-noise ratio, then ensemble-averaging is asuccessful strategy that enjoys the simplicity of the overall LIBS technique. Conversely,if no strong analyte response is observed in the generated average spectrum, whileanalyte-containing aerosol particles are known to be present, then additional spectralprocessing schemes, as related below, are required.

The second criteria (B) regarding the quantitative analysis of aerosol samples concernsthe upper particle size limit for complete particle dissociation, and the resulting indepen-dence of analyte signal on particle size (assuming an equal analyte mass). Cremers andRadziemski [88] explored LIBS for the detection of beryllium-rich particles depositedon filters using a range of conditions, including (i) particle diameters of about 50 nm,(ii) an ensemble collection of particles ranging from 0.5 to 5 �m and (iii) for nominally15-�m sized particles. They used a cylindrical lens to focus a laser beam directly onthe surface of the filters, thereby producing a laser-induced plasma that engulfed thedeposited beryllium particles. Plasma emission was collected and analyzed using theberyllium atomic emission line, which is one of the stronger emitting elements on a




mass basis with LIBS. Cremers and Radziemski reported a different analyte response,as realized with different calibration curve slopes, for these three different particle sizeclasses, and concluded based on their experimental observations that incomplete particlevaporization occurred for particles with diameters greater than about 15 �m. In otherstudies [89,90], this group also explored direct LIBS-based analysis of beryllium aerosolsusing particles less than 10 �m in diameter, noting in their latter study that such a particlesize is consistent with complete particle vaporization. Over the ensuing decades, manyLIBS researchers have cited an upper size limit for complete particle vaporization on theorder of 10 �m. While this value is consistent with the original research of Radziemskiet al. [89], one must consider the context of this early work. Clearly, no detailed studiesof the limiting range for linear analyte response as a function of particle diameter wereperformed. However, the premise of a linear analyte response for quantitative LIBSanalysis of aerosol particles is predicated on the complete vaporization; hence morerecent investigations have specifically addressed this issue.

Carranza and Hahn [91] investigated the laser-induced plasma vaporization of indi-vidual silica microspheres in an aerosolized air stream by examining the analyte response(silicon peak-to-base ratio) for progressively larger, monodisperse aerosol streams. A lin-ear mass response (noting mass is proportional to the diameter cubed) was observed to anupper diameter limit of 2�1 �m for a laser pulse energy of about 300 mJ. The results aresummarized in Figure 2, which shows the silicon analyte response (Peak-to-Base ratio)as a function of particle mass per plasma volume for single micron-sized microspheres.The most significant result is the clearly linear relation between the silicon P/B ratio as afunction of silicon particle mass (plotted as diameter-cubed) for the smallest three silicaparticle diameters investigated (1.0, 1.5, and 2�1 �m), and the abrupt deviation from thislinear trend for the particle diameters larger than 2�1 �m.

1

10

100

1 10 100

Sili

con

emis

sion

P/B

(a.

u.)

Diameter cubed (µm3)

2.1-µm particlediameter

Fig. 2. Peak-to-base (P/B) ratio of the 288.16-nm silicon atomic emission line for ensemble-averaged spectra of individually detected monodisperse silica microspheres as a function of thecube of the silica particle diameter [91]. The dashed line is a linear fit of the first three data points.




For complete silica particle dissociation and vaporization, the resulting analyte signalshould scale as the analyte mass contained in the particle, hence as the diameter cubed.With this in mind, Carranza and Hahn interpreted the threshold in the Figure 2 data asthe upper size limit in which an aerosolized silica particle is completely vaporized inthe laser-induced plasma, for the experimental parameters reported in their study. Toexamine if such effects as self-absorption were the reason for the departure from lin-earity, additional measurements were made by nebulizing solutions of dissolved silicon.As reported in their paper, such a configuration results in a high number density ofsubmicron-sized silicon-rich aerosol particles. The response to the silicon-rich nanopar-ticles was linear to an equivalent silicon mass of an approximately 8-�m sized silicateparticle. The authors speculated that the different results for different size particles aredue perhaps to a rate-limiting step of particle vaporization and dissociation. Therefore,the larger particles may simply not have time to completely vaporize by the time themeasurements were recorded, some tens of �s following plasma initiation. In addition,more recent imaging measurements, as discussed below, show that the plasma-particleinteraction is limited to a spatial region about the particle, hence local plasma condi-tions may be affected by the presence of large (i.e. micron-sized) particles. A morerecent study examined the complete vaporization of carbon-rich particles (specificallyglucose particles and sodium hydrogenocarbonate particles) in a laser-induced plasma,and reported an upper size limit of 5 �m for complete vaporization [92]. Such a largersize with the carbon-rich particles most likely reflects the marked difference in meltingpoints and volatility when compared to the more refractory silicon particles. Clearlyadditional experimental work and plasma modeling are needed to further determine theexact processes that govern particle vaporization, as well as to determine particle sizelimits for different laser pulse energies, wavelengths, focusing optics, and particle types.It is noted that this reported limiting particle size of 2 to 5�m is significantly below thefrequently used 10-�m particle diameter limit, and should be taken into considerationfor LIBS-based analysis of aerosol samples. Nonetheless, the upper size limits discussedhere are consistent with the general needs of PM2.5 monitoring, where assumption ofcomplete vaporization for particles smaller than about 2�5 �m is reasonable given thetypical uncertainties associated with measurement of atmospheric aerosols.

To gain additional insight into the role of plasma-particle interactions, it is usefulto consider the relative mass fraction due to an aerosol particle within a laser-inducedplasma. Complete particle dissociation and subsequent heating of the particulate massto the plasma state requires an amount of energy equal to the sum of the latent heat(i.e. heat of melting and heat of vaporization), the sensible heat (i.e. specific heat ofgas-phase species), and the dissociation energies of the constituent species. Assumingcomparable specific heats for various dissociated and gaseous species, the dissociationenergy will scale as the particle mass fraction with respect to total energy deposited intothe plasma. Using reasonable values for both aerosol particles and plasma properties,about 0.1 percent of the plasma energy will be consumed to heat the dissociated mass of anominal 10-�m sized particle. However, heats of vaporization are typically 2 to 3 ordersof magnitude greater than specific heat; hence the energy required to initially vaporizethe particle is expected to be significant. Assuming a conservative value of the heat ofvaporization a factor of 100 greater than the specific heat, the initial vaporization of the10-�m diameter particle could consume a quantity of energy on the order of 10 percent ofthe energy required to heat the particle-free gaseous plasma. In contrast, a 2-�m particle




would only consume approximately 0.1 percent of this energy budget. A significantassumption of quantitative LIBS is that the presence of an aerosol particle does not exerta significant influence on the overall plasma evolution and plasma characteristics. Sucheffects are reviewed in more detail below, but clearly, the energy budget for particlevaporization of a ∼2 to 2.5-�m aerosol particle is consistent with this primary tenant ofLIBS-based aerosol analysis: the additive nature of the overall plasma spectral emissionand the aerosol-derived analyte atomic emission. An important caveat, however, concernsthe degree of homogeneity of the particle-derived analyte with respect to the overallplasma volume. If the atoms derived from the aerosol particle remain locally confined,as demonstrated and discussed in Section 7 (Future Directions), and then perhaps thedirect comparison of analyte mass to the total plasma is too conservative of an approach.

The final issue, to address with LIBS-based ensemble-averaging, remains the need forcalibration schemes, which must produce an accurate analyte response for a known massconcentration. While it is traditionally accepted that the resulting atomic emission from alaser-induced plasma is independent of the actual analyte source (i.e. atomic, molecular oraerosol particle), recent research suggests significant departures in emission response forgas-phase and particulate-phase analyte sources. Specifically, Hohreiter and Hahn [93]reported marked differences in the atomic emission signal from carbon when comparingcalibration streams of gas-phase and submicron-sized solid-phase carbon species. Theresulting calibration curve slopes varied by a factor of eight over a comparable range ofatomic carbon concentrations for five different analyte sources, while the plasma electrondensity and temperature remained essentially constant. As an example, Fig. 3 presents

0

100

200

300

400

500

600

700

800

240 245 250 255

Inte

nsity

(a.

u.)

Wavelength (nm)

CO2

30-nm PS

C

Fig. 3. LIBS spectra in the vicinity of the 247.9-nm carbon atomic emission line for two differentsources of carbon: CO2 and 30-nm polystyrene particles. Both spectra are the ensemble average of1000 laser shots, and the concentration of atomic carbon was equal for both measurements [93].Both spectra have the same scale and baseline intensity, and have been shifted vertically forclarity.




the background-corrected LIBS spectra for two carbon sources (particulate and gaseous)present under equal overall mass loadings. Such findings challenge the widely heldassumption that complete dissociation of constituent species within the highly energeticlaser-induced plasma results in independence of the analyte atomic emission signal onthe analyte source. A physical model of the plasma-analyte interaction was proposedthat provides a framework to account for the observed dependence on the physical stateof the analyte [93]. The framework put forth must account for the enhancement ofanalyte signal with particulates less than the ∼2 �m limit for complete vaporization;hence incomplete vaporization is not an issue. The authors suggested that the relativeinertial differences between molecular and particle species play a role as the laser-induced plasma and subsequent plasma wave force species outward from the plasmakernel. The heavier particulates resist this outward push, resulting in a “concentration”of particulate-phase analyte species within the hotter plasma central region, resulting inan enhanced particulate-phase analyte response.

Clearly, one must pay attention to the calibration scheme in the context of the actualtargeted analyte species. An ideal calibration approach would be the ensemble-averagingof thousands of laser pulses in an aerosol stream comprised of a high aerosol particlenumber density of particles that are matched to the particle size of interest. Theseconditions may be realized by nebulizing aqueous solutions or suspensions of the targetedanalyte. For example, the nebulization of aqueous solutions and subsequently dryingthe droplets in a gaseous flow stream can produce a high-number density of nominally100-nm aerosols, as documented in the past for a range of analytes [94]. Clearly, thehighly linear analyte response that is often characteristic of the LIBS technique iscertainly achievable, but care must be given to provide as accurate a match as possiblebetween the targeted analyte and the calibration source stream.

4.2. Statistical Aerosol Sampling with LIBS

As outlined in the above sections, the analysis of aerosol particles with laser-inducedbreakdown spectroscopy differs significantly from other common analytical techniquesdue to the discrete nature of the laser-induced plasma, which forms the sample volume.Consider once again the necessary conditions enumerated above for ensemble-averaging,namely the first condition. If the aerosol number density is not sufficient to provideneither a suitable sampling rate nor a suitable analyte signal in the ensemble-averagedspectrum, then additional steps must be taken. Perhaps the most unique aspect of LIBSfor analysis of aerosols is that the individual spectrum may contain additional informationabout the discrete analyte nature of the sample stream. For aerosol analysis, the LIBStechnique is well suited to utilize the discrete sampling nature of laser-induced plas-mas, thereby enabling optimal sampling strategies and single-particle analysis schemes.A number of LIBS implementation strategies for aerosol analysis are available and dis-cussed in this chapter. An appropriate starting point for further development of LIBSaerosol analysis is the consideration of aerosol sampling rates, which are presented herefollowing the treatment reported by Hahn and co-workers [95,96].




The overall aerosol particle mass concentration (i.e. particulate mass per unit volume),YA, may be given by the expression

YA = �

6N∫ �

r=0r3p�r�dr� (1)

where � is the bulk particle mass density (mass/volume), N is the aerosol number density(particles/volume of gas), and the integral represents the cube of the volume mean diam-eter of the aerosol particle size distribution, which is defined in terms of the normalizedaerosol particle size distribution function p(r). Typically, p(r) is parameterized using alog-normal size distribution, which best reflects a wide range of actual aerosols. Note,that the assumption of spherical particles is inherent in Eqn. (1), although the distribu-tion function may in fact represent the distribution of mass-based equivalent sphericaldiameters.

An additional parameter is needed to define the aerosol sampling problem, namelya measure of the effective sampling volume of the laser-induced plasma. The plasmavolume is a complex function of the laser beam geometry, focusing optics, irradiance,and gas stream conditions. Plasma data based on plasma imaging studies, transmissionmeasurements, and sampling considerations have been reported in the literature (see forexample references [89], [97–99]). As reported in Carranza and Hahn [99], the actualvolume of the laser-induced plasma can be defined in a number of manners. In their work,a physical plasma volume was mapped out using a secondary low-energy probe laser toprobe the plasma boundary based on transmissivity (i.e. plasma absorption). The plasmais nearly opaque to incident laser energy in the first few 10s of ns following plasmabreakdown [100]. Using this approach, an elliptical plasma shape was defined, enablingcalculation of the plasma volume. Note, however, that such an approach is dependent onthe definition of the plasma edges, which was defined by a transmission value of 90% inthe Carranza study, resulting in a plasma volume of 1�4 mm3. Alternatively, the plasmavolume was calculated in the same study using statistical methods, as detailed below.With that approach, the effective plasma sampling volume was reported as 1�2 mm3,and overall comments were presented discussing the nature of agreement between thesetwo values, with conclusions citing the differences in the plasma-particle interactionregion for the sampling volume verses the region of strong absorptivity for the physicalmeasurement. In general, the plasma volume may be considered on the order of 10−3

to 10−4 cm3, with variations depending on the laser pulse energy, the configuration ofthe focusing optics, as well as the sample matrix. Knowing the plasma volume andaerosol size distribution parameters, one can then calculate the average number of aerosolparticles expected within a single plasma volume, as well as the overall aerosol particlesampling rates. The product of the plasma volume and the aerosol number density yieldsthe average number of aerosol particles per laser-induced plasma, �,

= N · Vplasma� (2)

Equations (1) and (2) may be combined to define the average number of particlesper plasma volume in terms of the aerosol mass concentration, particle size distribution,and the effective plasma volume,

= 6YAVplasma

�∫ �

r=0 r3p�r�dr� (3)




Note that the particle size distribution and resulting integral may be replaced withthe cube of the volume mean diameter. For discrete aerosol particles, the probabilitydistribution of the expected number of aerosol particles sampled per plasma volume maybe expressed using the Poisson distribution,

Pn = n

n! exp�−�� (4)

where Pn defines the probability of finding n discrete aerosol particles within a givenplasma volume. The aerosol sampling rate �RA� may be defined as the percentage oflaser-induced plasmas (i.e. laser pulses) expected to sample one or more aerosol particles.The sampling rate is directly calculated as the sum of all probabilities for sampling afinite number of aerosol particles, namely

RA = 100 ·∑�n=1

Pn� (5)

noting that the sum of all probabilities must be unity, it is observed that one minus theprobability of sampling zero particles also yields the sampling rate,

RA = 100 · �1−P0� = 100 · �1− e−�� (6)

To demonstrate these LIBS-based sampling statistics, a set of data recorded in ambientair is analyzed. Laser-induced breakdown spectroscopy was used to monitor ambient airin a manner that was previously reported [101,102]. Spectral analysis of each LIBS spec-trum was performed based on the pronounced 393.4-nm calcium atomic emission peak.Calcium is a common element in ambient air particulates, originating from many miner-als or calcium hydroxide. For one session, LIBS data were collected in 1000-pulse lasersequences, for a total of 20,000 laser pulses. During this period, 37 spectra were identifiedas containing a pronounced calcium atomic emission peak; hence they were consideredto represent the sampling of a calcium-rich aerosol particle. Assuming a plasma samplingvolume of 1�2 mm3, the corresponding number of sampled particles and total volume ofair sampled by the plasma �1�2 mm3/pulse×20�000 pulses� yields the measured numberdensity of calcium-rich aerosol particles. This value is calculated as 1541 calcium-richparticles/l of air. Using Eq. (2) above, the Poisson parameter is readily calculated as� = 0�00185. This value of � may then be multiplied by the number of laser pulsesper sampling interval, namely 1000, to yield the expected value of 1.85 particle hits per1000 pulses. This value of 1.85 may then be used with Eqn. (4), to predict the samplingdistribution of calcium-based particle hits per 1000 pulses. This distribution is plottedin Fig. 4, along with the experimentally measured sampling distribution over the twenty1000-pulse data collection intervals. The ideal Poisson distribution and the experimentalsampling rates are in excellent agreement. Specifically, the probability of recording zeroparticle hits is equal to 15.7% and 10% for the predicted and experimental data, respec-tively, while the most probable sampling rate of 2 particles per sequence correspondsto 29.1% and 35% for the predicted and experimental data, respectively. Overall, theFigure 4 data demonstrate the statistical nature of LIBS-based aerosol sampling, andprovide corroboration of Poisson-based models to describe the sampling problem.

The aerosol particle sampling rate enables an examination of the LIBS-based aerosolanalysis problem in the context of discrete aerosol particles and a finite number of




0

5

10

15

20

25

30

35

40

0 1 2 3 4 5 6 7

Experimental

Predicted

Pro

babi

lity

(%)

Hits per 1000 shots

Fig. 4. Particle sampling probability as a function of the expected number of particle hits per1000 laser shots. Experimental data are presented for sampling of calcium-based aerosol par-ticles in ambient air for 20 consecutive 1000-shot trials, and the corresponding distribution ispredicted using Poisson statistics per Equation (4).

discrete plasma sample volumes, thereby elucidating key regimes suited to ensembleaveraging or perhaps better suited to more sophisticated data analysis approaches dueto aerosol sampling limitations. Low aerosol particle sampling rates for LIBS-basedanalysis brings with it a potential decrease in method sensitivity (i.e. atomic emissionintensity) for a given analyte species if ensemble averaging of all collected spectra isused. For example, for large numbers of laser pulses (e.g. ∼1000 pulses), the reductionin analyte signal scales directly as the sample rate, as ensemble-averaging reaches anupper limit in the reduction of spectral noise as discussed above. In the following section,LIBS-based sampling strategies and data analysis schemes are formulated in the contextof aerosol sampling rates with a goal of maximizing the analyte signal and correspondingdetection limits.

4.3. Conditional Analysis for Spectral Processing

The above data set nicely demonstrates the concept of a limiting particle sampling ratein the context of aerosol analysis. Specifically, for the above data, the 37 calcium-rich particle hits were recorded from a total of 20,000 laser pulses, which results ina particle sampling rate of 0.185%, or about one particle hit per 550 laser pulses.Clearly with a sampling rate much less than 1 percent, the large fraction of spectral datawith no analyte information (i.e. no calcium atomic emission lines present) will resultin a greatly diminished analyte response with an ensemble-averaging approach. Thispoint is demonstrated in Fig. 5, where the ensemble-average of the 20,000 laser pulses




0

500

1000

1500

2000

2500

3000

3500

4000

385 390 395 400 405 410

Inte

nsity

(a.

u.)

Wavelength (nm)

Ca II

Ca II

Fig. 5. Ensemble-averaged spectrum of 20,000 laser shots recorded in ambient air, along with theensemble-average of a subset of 37 spectra corresponding to calcium-rich particle hits, as basedon the presence of the 393.4 and 396.7-nm calcium atomic emission lines. The correspondingparticles sampling rate is 0.185% [102]. Both spectra have the same scale.

is presented, along with the ensemble average of the identified 37 pulses containingsignificant calcium emission (i.e. calcium-rich particle hits). In the ensemble-averagedspectrum, the 393.4 and 396.7-nm calcium atomic emission lines are undetectable, whilein contrast, these two lines are very pronounced in the spectrum corresponding to the37 individual calcium-based particle hits. By averaging together so many blank spectrawith respect to the calcium emission line, the effect of null data on the actual analytesignals is readily apparent. Data such as those reported in Fig. 5 suggest the use of asuitable conditional data analysis approach to identify the subset of analyte-rich spectraldata corresponding to the presence of particles within given plasma volumes as a meansto greatly enhance the overall LIBS signal response. Such an approach was reportedspecifically for the analysis of aerosol particles based on the conditional analysis ofplasma emission spectra on a pulse-to-pulse basis [96], while other researchers haveused conditional schemes to increase the analyte response by rejecting both irregularspectra as well as spectra with poor analyte response [103]. Prior to implementation ofa conditional data analysis routine for LIBS-based aerosol analysis, several fundamentalissues must be addressed.

4.3.1. Considerations for Aerosol Analysis via Conditional-Processing

(A) A threshold criteria must be established for processing individual LIBS spectrato determine the presence of an aerosol particle-derived analyte signal, which isused to identify and confirm the sampling of an aerosol particle in a given plasma.




(B) The total number of actual particles to be sampled must be established such thata valid statistical sample of the aerosol distribution is achieved.

(C) A means for processing the identified spectrum corresponding to aerosol particlesampling must be identified for use with traditional calibration approaches.

For conditional processing of LIBS spectra, the most fundamental issue is the deter-mination of a suitable means to detect the presence of a targeted analyte-rich aerosolparticle in a given laser-induced plasma; that is, identification of a particle hit for a givenlaser pulse. A suitable threshold criterion should be selected taking into consideration therelatively large spectral noise associated with typical single-pulse LIBS spectra, notablywith intensified CCD detectors, as demonstrated above in Fig. 1. In addition to noise,the recorded intensity of the plasma emission is generally characterized by considerablevariance on a pulse-to-pulse basis, as discussed above; hence it is desirable to normalizethe analyte signal. To obtain optimal performance for any thresholding technique, themost useful metrics are the ratio of the analyte atomic emission peak (either integratedpeak or peak intensity) to the intensity of the adjacent continuum emission (Peak-to-Base,P/B), or the ratio of the analyte atomic emission to the noise of the adjacent continuumemission (Signal-to-Noise Ratio, SNR). With a suitable metric selected, all that remainsis to select a threshold above which the metric is considered to correspond to an actualaerosol particle-derived emission signal, and below which the metric is considered tocorrespond to random spectral noise.

Hahn et al. [95,96] outlined a basic approach for the selection of thresholds froman overall strategy point of view, while in a more recent work, Carranza et al. [104]explored in detail the various trade-offs between relatively high and low threshold values.The treatment here will follow the approaches put forth in these studies. The moststraightforward method of selecting a threshold is to force the spectral processing to bein effect a binary process, which involves the selection of a threshold such that no spectraexceed the value as a result of spectral noise alone. With this scheme, the threshold valueis selected such that it exceeds the maximum value of the spectral metric (P/B or SNR)that is obtained with the absence of any aerosol-derived analyte species (i.e. no aerosolparticles present). Hence, the threshold value is selected to exclude any fluctuationsin the normal spectral noise. This is readily accomplished by increasing the thresholdvalue, with no aerosol particles present, until no spectrum exceeds the threshold formany thousands of laser pulses. The approach dictates that the extreme fluctuations ofthe spectral metric are insufficient to trigger an analyte hit. For monitoring of manysample streams, it is possible to pass the gaseous stream through a HEPA filter prior tothe LIBS sample volume, thereby eliminating the aerosol source for calculating of theappropriate threshold. However, if the aerosol source cannot be readily removed fromthe source stream for this procedure, the threshold value may be obtained as based on anadjacent, surrogate spectral region of comparable intensity that is free from any analyteatomic emission lines. Generally, the continuum emission signal is sufficiently smoothsuch that a suitable surrogate region is available within a few nm of the targeted analyteatomic emission line.

Carranza et al. provide a more detailed discussion of threshold selection in the contextof the signal-to-noise ratio of the resulting ensemble-average of spectra correspondingto selected particle hits [104]. Clearly one must consider that there exists a region wherethe resulting atomic emission peak from aerosol-derived analytes will fall below the




noise-free threshold related above, but above the method detection limit. This is becausethe extreme spectral noise, although infrequent at a specific spectral location, resultsin a necessarily large threshold value if all such noise events are to be rejected. Toavoid such a strict spectral filter, the threshold value may be relaxed to yield a relativelylow false hit rate, on the order of 0.05% to 0.1% of total pulses, but not exactly zero.This method enhances the sensitivity of the conditional analysis method and extendsdetection to smaller analyte-containing particles; however, one must carefully optimizethis procedure, as false analyte signals are being introduced into the spectral data set.Toward this end, Fig. 6 presents analysis of the analyte peak-to-base ratio (P/B) and theexpected contribution of noise to the analyte peak signal (i.e. false hits) as a functionof the statistical threshold based on the 288.1-nm silicon emission line resulting fromthe detection of 2-�m sized silica spheres, as adapted from reference [104]. The figuredemonstrates that an optimal value exists for a false hit rate of about 0.05–0.1%, asreflected in the threshold of the P/B and expected noise contribution curves. Clearly afew false hits (∼1 per 1000 pulses) are tolerable, in that contribution to the genuineparticle-hit data is negligible, while the extension of the detection limits into the extremeenvelope of statistical noise is beneficial.

While the above comments provide additional insight into the selection of appropriatethresholds for conditional spectral analysis, attention must always be paid to the issue offalse hits when only a single analyte emission line is utilized. The difficulties of spectralthresholding are significantly reduced if additional analyte atomic emission lines areavailable in a given spectral window. If two or more emission lines are present on the

5

10

15

20

0

5

10

15

20

0.01 0.1 1

SNR approach

P/B approach

P/B

rat

io

Exp

ecte

d no

ise

cont

ribut

ion

to p

eak

(%)

Hit threshold (%)

P/B ratio(288.1 nm)

Percentnoise

Fig. 6. Influence of the conditional processing threshold (as expected false hit rate) on the averageP/B intensity of the silica emission line for identified hits, along with the expected noise as aweighted percentage of the on-peak signal [104]. The conditional processing rates are performedusing the Silicon emission line as a Signal-to-Noise ratio (SNR approach) and as a Peak-to-Baseratio (P/B approach).




recorded spectra, the additional atomic emission lines should also be used for quantitativeanalysis. Specifically, because spectral noise is considered white noise, it is improbablethat the extreme noise fluctuations that trigger false hits will appear simultaneously onmultiple atomic emission lines. Therefore, a useful approach for conditional processingis to relax the threshold criteria to something on the order of 1–2% false rates, whichfurther lowers the overall detection limit with respect to identifying individual aerosolparticles. Then all candidate spectra (i.e. those which exceed the threshold based on asingle atomic emission line) are checked for atomic emission above the noise thresholdon additional emission lines.

This approach was first detailed by Hahn and Lunden [95], and then utilized in anumber of aerosol studies [101,105]. An example of such an approach is presented inFig. 7, which shows two single-pulse spectra recorded in a sample stream containing adilute concentration of Bacillus spores. Conditional-based sampling was performed usingthe 396.9 nm Ca II atomic emission line, as reported by Dixon and Hahn [105]. In Fig. 7,both spectra exceeded the sampling threshold based on the single 396.9-nm calciumemission line; however, only the upper spectrum would be confirmed to correspond to anactual Bacillus spore hit, based on the correctly-proportioned presence of the 393.4-nmCa II emission line. This additional emission line is absent in the lower spectrum, hencethis spectrum would generally be dismissed as a noise event, rather than an actualsampled spore. Together, the two spectra nicely illustrate the issues associated withconditional spectral processing of LIBS data, including the corresponding single-pulse

0

500

1000

1500

2000

2500

3000

385 395390 400 405 410

Inte

nsity

(a.

u.)

Wavelength (nm)

Ca II

Ca II

Fig. 7. Representative single-shot spectra corresponding to LIBS-based measurements in anaerosol stream seeded with B. atrophaeous spores. The upper spectrum corresponds to an indi-vidual B. atrophaeous spore, while the lower spectrum corresponds to spectral noise only, astriggered on the 396.9-nm calcium emission line using a conditional analysis routine. Both spectrahave the same intensity scale, with the upper spectrum shifted vertically for clarity.




spectral noise and statistical fluctuations, as well as the excellent analyte sensitivity thatmay be realized with proper detection algorithms.

The above treatment of conditional analysis processing schemes provides insight intosegregating LIBS spectral data into those spectra that contain significant analyte emis-sion signals corresponding to a targeted analyte (presumably derived from analyte-richparticles), and those spectra that contain no detectable analyte emission. The identifiedanalyte-containing spectra, which may be referred to as particle hits, may then be pro-cessed together or treated individually. The former will be treated here, and the latterwill be discussed in the following section. The spectra corresponding to identified par-ticle hits may then be ensemble-averaged, yielding an analyte emission-rich spectrumthat enjoys the benefits of pre-concentration with respect to the resulting analyte atomicemission lines in combination with the reduction in signal noise always realized withensemble-averaging. An example of this outcome is presented in Fig. 5, as discussedabove, in which the corresponding particle sampling rate (i.e. hit rate) was 0.185%, orabout 2 particles per 1000 pulses. The rejection of 99.8% of the spectral data, namelythe null data with respect to the targeted analyte calcium, results in a nearly 500-foldincrease in the analyte emission peak intensity as compared to ensemble-averaging ofall collected spectra. The RMS of the continuum emission intensity, a direct measureof signal noise, in the spectral regions adjacent to the calcium peaks increased by only15% when comparing the 37-pulse ensemble-average to the average of all 20,000 spec-tra. Taking all factors into account, the calcium emission signal, as measured by thesignal-to-noise ratio, was enhanced by a factor of 470 with the conditional analysis rou-tine, enabling excellent analyte sensitivity. In contrast, the corresponding 20,000-pulseensemble-averaged spectrum is characterized by essentially a non-detect with respect tothe two calcium emission lines.

Using conditional-analysis, the actual concentration (mol/volume) of a given analytespecies is determined using a linear combination of the analyte signal as based on thesubset of the conditionally-analyzed spectra and the corresponding aerosol sampling rate(i.e. hit rate). The ensemble-averaged spectrum corresponding to the selected analytehits should be used with a traditional LIBS calibration curve relating the LIBS emissionsignal (e.g. P/B) to a known analyte concentration range. As related above, care mustbe taken to carefully match the analyte source in the calibration stream to the actualanalyte source in the stream of interest, with specific attention given to the issue ofgaseous-phase and particulate-phase analyte. Once the equivalent concentration of thehits spectrum is known, Xhits, this value is then adjusted by the sampling rate to yieldthe actual analyte concentration, Xtrue, as reflected by the relationship:

Xtrue = �Xhits�∗ �total hits/total pulses�� (7)

where the latter term corresponds to the particle sampling rate. It is noted that as thesampling rate approaches 100%, the conditional-analysis scheme converges to a classicLIBS methodology, where the ensemble-average of all pulses is directly related to thecorresponding calibration curve. If two analyte emission peaks are used as discussedabove (one for triggering and one for analysis), the effect of any false analyte hits isminimal because a false hit reduces the intensity of the analysis emission peak, butsimultaneously increases the sampling rate, thereby producing the correct actual analyteconcentration based on the above relation.




A final issue concerns the statistical nature of the particle sampling rate, namely thenumber of hits divided by the total number of laser pulses. Because the sample rate canrange considerably over limited sample periods, care must be taken to accurately quantifythis parameter. The total number of particle hits necessary to adequately assess theoverall mass concentration of aerosol particles follows the same guidelines as discussedabove regarding ensemble-averaging, namely that on the order of 20 aerosol particlesare sufficient to characterize the average aerosol mass for rather broad distributions ofsubmicrometer to micron-sized aerosol distributions. For a given average aerosol particlesampling rate, the total number of laser pulses should then be selected to ensure atleast 20–30 aerosol particle hits are recorded. Hahn et al. explored this topic in detail,and noted that better accuracy is obtained if the conditional processing algorithm isterminated after a pre-selected total number of pulses rather than a pre-selected numberof particle hits [96].

4.4. Analysis of Individual Aerosol Particles

The above treatment of LIBS spectra via conditional processing schemes is all basedon the concept of discrete particles interacting with a specific plasma volume in away such that characteristic spectral features (i.e. atomic emission lines) make theiridentification possible. Once individual spectra are identified as corresponding to anaerosol particle sampling event, they may be grouped together via ensemble-averagingto enhance the sensitivity for detection of aerosol-derived species, as related above.Alternatively, each individual spectrum may be analyzed on its own, thereby revealinginformation about individual aerosol particles on a laser pulse-to-pulse basis. Suchan approach takes advantage of the single-point sampling nature of LIBS, which isunavailable with continuous analysis systems such as inductively-coupled plasmas (ICP).

The LIBS-based quantitative analysis of individual particles was first reported byHahn [106], where a calibration scheme was proposed making use of microsphereswith a predetermined concentration of analyte, in that case iron. As implemented, themethodology enabled determination of the absolute mass of the targeted analyte perparticle, as determined from a single-pulse LIBS spectrum under dilute aerosol samplingconditions. In a follow-up paper, Hahn and Lunden [95] presented a formal treatment ofthe statistical sampling problem combined with the concept of an analyte signal and theassociated method calibration curve response. Following their analysis, in combinationwith the concepts related above on the use of conditional analysis for determination oftotal analyte concentrations, the following relation may be developed

Xhits RA = f �

6r3 N� (8)

In this equation, the left side is an expression for the true analyte concentration aspresented above in Eqn. (7), where RA is the sampling frequency equal to the ratio ofthe number of particle hits to the total number of laser pulses, and Xhits is the equivalentanalyte mass concentration based on the spectrum of hits and the calibration curve.The right side of Eqn. (8) is an alternative expression for the calculated analyte massconcentration (see Eqn. 1) given by the bulk particle density �, the volume mean particlesize r, the analyte mass fraction of the aerosol f, and the aerosol number density N.




As related by Hahn and Lunden, when the aerosol sampling rate RA is small (e.g.∼1–10% or less), the sampling rate may be approximated based on Poisson statistics asthe product of the aerosol number density N and the effective plasma sampling volumeVplasma. Using the above relations, the following expression is derived

Xhit = �analyte mass�/

�plasma volume�� (9)

that presents a direct relation between the equivalent mass concentration based on acalibrated LIBS analyte response, the discrete plasma volume, and the absolute analytemass. If the plasma volume is known, the above equation may be used to calculatethe absolute analyte mass corresponding to a single aerosol particle as the simpleproduct of the characteristic plasma volume and the equivalent mass concentration(i.e. calibrated response). The latter quantity is readily calculated from the correspondingsingle-pulse spectrum and the corresponding LIBS analyte calibration curve. One mayreadily calculate an equivalent spherical diameter of the individual particle present inthe plasma volume based on the calculated analyte mass, via the relation

req =(

6 Chit Vplasma

�f

)1/3

� (10)

where � is the particle bulk density (mass/volume), and f is the mass fraction of theanalyte with respect to the overall bulk particle mass. For a pure, homogeneous particlef equals unity and the density is the actual density of the analyte.

Single-pulse LIBS-based spectral analysis can yield the absolute analyte mass and thecorresponding equivalent spherical diameter using the relations developed above basedon two parameters, the equivalent concentration (i.e. atomic emission response) and thecharacteristic plasma volume. A calibration approach was outlined by Hahn [106] usingmicrospheres of known mass composition and size. To gain further insight in the issuesof plasma volume, Carranza and Hahn investigated plasma size under laser-inducedplasma conditions frequently used for aerosol and gas-phase analysis [99]. In their study,three distinct plasma-volume measurements were made, including a physical volumebased on transmission measurements using a spatially resolved probe laser as describedabove, an emission-based diameter similar to the approach of Hahn [106] in which theemission response and a known mass are used to solve directly for plasma volume viaEqn. (9), and thirdly a statistical plasma volume was recorded based on aerosol samplingrates modeled with Poisson sampling statistics. The resulting plasma volumes variedfrom 1.2 mm3 to 2.4 mm3, with the former corresponding to the statistical samplingvolume and the latter corresponding to the emission-based volume. Additional discussionis offered below regarding the uncertainty in plasma volume and the propagation of theuncertainty in LIBS-based absolute mass measurements.

The second parameter necessary for single-pulse mass measurements is the equivalentconcentration, which as discussed above, is readily calculated directly from the spectraldata (e.g. analyte Peak/Base ratio) and the corresponding calibration curve. The issuefor concern arises from the nature of the calibration source stream. Recall the abovediscussion regarding the effects of analyte phase on the analyte atomic emission response,with particulate species yielding as much as an 8-fold increase in analyte signal ascompared to identical mole fractions of analyte in the gas phase [93]. Hence to achieve




the most accuracy in the calibration methodology, one should make use of a source streamthat closely matches the range of aerosol data expected in the unknown sample stream.

Figure 7 above provides an illustrative spectrum generated from conditional-basedprocessing of LIBS for the identification of single-pulse spectra corresponding to indi-vidual aerosol particles. Specific applications and detailed discussion of various imple-mentations and resulting performance are presented in Section 6. However, the absolutecalcium mass corresponding to the Fig. 7 data is about 3 fg based on Eqn. (9) analysis,noting that the calibration curves used were generated from a particulate-phase calciumcalibration source. For the nominally 1-�m sized spores used to generate the spectraldata, ∼3 fg calcium mass corresponds to a nominal calcium mass concentration of about0.5% in the original spores. Such data demonstrate the overall sensitivity of the LIBStechnique for single particle analysis, as observed by the relative strength of the atomicemission line corresponding to only a few fg of analyte.

Finally, a few additional comments are offered on the overall sensitivity and precisionof single-pulse LIBS-based aerosol analysis, including consideration of key parame-ters such as laser energy, laser beam stability, and the presence of aerosol particlesthemselves. Carranza and Hahn explored the role of laser pulse energy on the resultingprecision of gas-phase analyte signals, where precision was quantified by the relativestandard deviations of the pulse-to-pulse measurements of absolute analyte atomic emis-sion peaks, and normalized P/B ratios [86]. The relative standard deviation of the carbonatomic emission signal, the gas-phase analyte used in their study, was found to decreasewith increased laser pulse energy up to a certain level, and then plateau for additionalpulse energies. Interestingly, the threshold in the RSD vs. laser pulse energy plot corre-sponded exactly to the saturation value for laser-pulse energy absorption by the plasma.A significant conclusion was that single-pulse LIBS based measurements should bemade with sufficient laser pulse energy to achieve saturation with respect to absorbedlaser pulse energy, as well as made with suitable collection geometry (e.g. backscattermode) to minimize spatial variability.

A more recent study examined the role of laser beam stability and the presence ofaerosol particles on the initiation of laser-induced breakdown, and on the overall analytesensitivity as measured for three different gas-phase species (hydrogen, nitrogen, andcarbon) [87]. Improvements in the temporal stability of the breakdown ignition volumewere observed with laser cavity seeding. However, laser cavity seeding produced nosignificant improvement in analyte precision for the range of gas-phase atomic emissionlines. In contrast, greater pulse-to-pulse analyte precision, as measured by a nearly 60%reduction in relative standard deviation, was realized with the elimination of concomitantparticles from the analyte sample stream. Clearly, one must live with the presence ofparticles for LIBS-based analysis of aerosols; however, this study provides insight intothe precision and accuracy to be expected for the interactions of ambient aerosol particlesand the laser-induced breakdown process.

5. ALTERNATIVE METHODOLOGIES FOR AEROSOL ANALYSIS

Much of the above discussion is focused on the statistical nature of discrete particlesand how aerosol sampling couples to the finite laser-induced plasma volume. While suc-cessful strategies for direct aerosol analysis were enumerated, including single-particle




analysis, an alternative approach is the collection (i.e. concentration) of particles onan appropriate filter medium prior to LIBS analysis. Perhaps the greatest advantageof filter collection and subsequent LIBS analysis is the potential to enhance detectionlimits. With automated filter handling, including continuous spools of filter material, thismethodology can be implemented on-line and in near real-time. An additional benefitstems from the minimally destructive nature of LIBS, which enables the filter mediato be archived for future analysis, including conventional bench top chemistry such asdigestion and ICP-MS analysis. The primary limitation of filter collection is linked to thevery nature of particle concentration, namely the loss of information regarding individualparticles. By collection of particles on a filter substrate, individual laser-induced plasmasmay sample tens to hundreds of particles, with the result being a spectral signaturerepresentative of the entire particle ensemble. As such, this approach is naturally suitedto emission monitoring, notably for toxic metals, where regulations are typically set fortotal elemental concentrations, which is consistent with this LIBS-based approach offilter collection and analysis. It is not surprising, therefore, that many of the publishedstudies have been applied to environmental monitoring.

Beryllium is an excellent analyte for this methodology, as its high toxicity makesit of interest in emissions monitoring and industrial hygiene. Furthermore, berylliumenjoys excellent detection limits in atomic emission schemes, including LIBS, whileit is undetectable with alternative schemes such as X-ray fluorescence spectroscopy(XRF). Cremers and co-workers at Los Alamos National Laboratory successfully usedthe filter collection approach for analysis of beryllium and tantalum aerosols [88,107].Detection limits were below 1 ng cm−2 for beryllium, and increased to 40 ng cm−2

for tantalum. As discussed above, one issue with laser-induced plasma analysis is thecomplete vaporization of large diameter particles, namely those in the micron sizerange. In their studies, Cremers et al. did report some signal saturation at high particleloadings, which was attributed to particle size effects. Due to the large particle diameters(�m size range) investigated and the corresponding high filter loadings, a particle sizedependence of the signal was observed due to saturation. In more recent work, Panneet al. used the filter collection approach for emission monitoring and ambient aerosolanalysis, successfully extending the method to ultrafine aerosol particles [108–110].For the waste incineration facilities, analyte volume concentrations between 0.1 and5 �g m−3 were detected with this LIBS scheme [110], however, the authors noted thatproblems may exist with independent reference analysis at such low concentrations.Rather surprisingly, because the laser-induced plasma samples only a certain depth ofthe filter (i.e. the surface layer that is enriched with deposited aerosols), matrix effectsmay actually be much more pronounced with traditional reference analysis that digeststhe entire filter substrate. As with any filtration method, filter selection must take intoconsideration temperature resistance for emissions monitoring applications, samplingefficiency (notably size-dependency), and blank values for a range of analyte species.

6. APPLICATIONS OF LIBS-BASED AEROSOL ANALYSIS

A number of reviews have been published regarding the LIBS technique, includingapplications to aerosol analysis [111–116]. Belaev and co-workers are perhaps the firstrecognized group to suggest the use of laser-induced gas breakdown for chemical analysis




of aerosols [117]. However, the modern foundation of LIBS as an analytical techniquefor aerosol analysis was established in a series of papers by Radziemski and Cremersin the early 1980s [83–85,89,90]. These papers systematically explored the issues ofdetection, including temporal gating (i.e. time-resolved analysis) to optimize the analyteatomic emission signals with respect to continuum emission, plasma formation andgrowth, and the issues of local thermodynamic equilibrium. In their landmark 1983paper, they presented detection limits for Cd, Pb, and Zn-based aerosols on the order of200 �g m−3 for particles in the micron size range [89], while also exploring the issuesof analyte source and plasma temperature. In additional communications, Cremers andRadziemski reported detection limits for beryllium aerosols below 1 �g m−3. In all ofthis early work, ensemble-averaging schemes were used to reduce spectral noise, whileaerosol source streams were generally of sufficient number density such that samplinglimitations were of no concern.

Following the pioneering work of the Los Alamos group, a wide range of studies andapplications has been reported in the literature regarding LIBS-based aerosol analysis.What follows below is not a thorough review of the published body of related papers, butrather an overview of select papers that focus on either unique applications or elucidatefundamental issues associated with this topic.

As a starting place for aerosol analysis, one may first consider breakdown of thegas-phase matrix. A number of investigators have examined the wavelength dependenceof gas breakdown and the ensuing plasma characteristics [39,118–122]. For calibrationaccuracy, it is desirable to have as much independence of the analyte emission on theoverall gas composition. Yalcin et al. [97,123] investigated the temporally and spa-tially resolved plasma temperature and electron density for several gaseous species atambient pressure, including N2, He, and SF6. Consistent values were found for electrondensities and plasma temperatures with changing gas species, laser energy, particulatelevels, and humidity levels. This important work has been widely cited in support ofoverall plasma invariance with respect to changing ambient conditions. While the effectof background matrices is obviously of importance for aerosol analysis, the issue ofcalibration (i.e. emission response of analyte) is the key step for any practical applica-tion. Essien and co-workers produced a submicron-sized aerosol for LIBS calibration[90]. Generally, linear calibration curves were generated for cadmium, lead, and zinc,although various degrees of saturation were observed at higher concentrations, whichwas attributed to incomplete vaporization of particles, as discussed above. An importantfinding was their general agreement (within 10%) of lead atomic emission signals whennebulizing either lead acetate, lead chloride, or lead nitrate, thereby showing indepen-dence of the analyte source. However, cadmium revealed a 27% difference in analyteresponse when comparing nebulized solutions of cadmium nitrate and cadmium chlo-ride, although detailed size measurements of the calibration aerosol were not reported.In analogous experiments, the relative independence of analyte signals on molecularsource for purely gas-phase species was reported in several studies [124,125]. Specifi-cally, Dudragne et al. demonstrated that analyte signals for fluorine, chlorine, sulfur andcarbon scaled with the number of analyte atoms in the constituent molecules for a widerange of compounds, concluding that the parent molecules were fully dissociated in thelaser-induced plasma of their study [124]. Similarly, Tran et al. verified that SF6 and HFyielded identical fluorine atomic emission signals when the gas composition of differentmixtures was adjusted to the same atomic fluorine mole fraction [125].




The studies above lead to two general conclusions, namely that the laser-inducedplasma, as measured by temperature and electron density, is remarkably robust withrespect to changes in gas-phase composition, and that the resulting analyte emissionis generally independent of the analyte source. However, a very important point mustbe made with regard to the latter comment, namely, analyte independence has beenobserved when the various analyte sources are all in a similar state (i.e. all gas phase,or all particulate phase with a similar particle size distribution). As discussed above,the effects of analyte phase on the calibration response for a typical LIBS system wasinvestigated for a range of carbon species [93]. Significant differences in the atomicemission signals from carbon were observed when comparing calibrations of gas-phaseand submicron-sized solid-phase carbon species. Consistent with the above comments,the plasma electron density and temperature remained essentially constant. Such findingschallenge a widely held assumption that complete dissociation of constituent specieswithin a highly energetic laser-induced plasma results in independence of the analyteatomic emission signal on the analyte source. The authors proposed a physical model ofthe plasma-analyte interaction to account for the observed dependence on the physicalstate of the analyte by considering the rapidly expanding plasma as a shockwave-like phenomenon, in which the pressure and electron wave rapidly expand from theoriginal plasma kernel. As the plasma wave expands, molecular and particulate speciesare pushed toward the edge of the plasma volume, however, due to the many ordersof magnitude difference in the mass of gas-phase species (molecules) and solid-phasespecies (particulates), it was proposed that the efficiency at which a given species iscarried by the plasma wave will scale inversely with particle mass. The authors referredto this behavior as an effective analyte slip factor as the plasma wave expands. Animportant point from such studies is the need to produce calibration schemes that reflectas much as possible the physical state of the analyte species of interest. Overall, clearlymore research is necessary to fully understand the exact nature of the plasma-particleinteractions associated with LIBS-based aerosol analysis. Nonetheless, a number ofpractical applications have been reported.

The analysis of toxic metals, such as the US Resource and Conservation RecoveryAct (RCRA) metals As, Be, Cd, Cr, Pb, and Hg, has received considerable attentionover the years, most notably in the context of a LIBS-based real-time emissions andprocess monitor. In the early 1990s, a LIBS-based monitoring effort was initiated atSandia National Laboratories, with early trials reported for detection of a chromium-richaerosol [126,127]. In a contemporary effort to the Sandia work, Singh et al. from theDIAL (now ICET) at Mississippi State University initiated a LIBS program similarlygeared toward process analysis and emissions monitoring [128–130]. They successfullystudied different optical arrangements at a coal-fired flow facility and achieved detectionlimits for different heavy metals between 1 and 600 �g m−3. In fact, both the DIAL andSandia LIBS teams participated in a jointly sponsored DOE/EPA demonstration test ofcontinuous emission monitoring technologies conducted in September 1997 [129,131].The demonstrations were conducted using a pilot-scale rotary kiln incinerator that wasseeded with the toxic metals As, Be, Cd, Cr, Pb, Hg, and Sb at concentrations between15–100 �g m−3, which were consistent with the U.S. maximum achievable control tech-nology (MACT) rules. Although the two employed LIBS systems demonstrated a fastresponse time, their overall precision and accuracy (between 50–150%) did not meet therequired 20% relative accuracy mandated by US regulatory agencies.




However, on-line LIBS air monitoring schemes have continued to improve. Forexample, Neuhauser et al. reported the successful detection of Cr-rich aerosol usingLIBS for fast monitoring of particulate emissions in an electroplating facility [132].Their prototype LIBS system was tested in cooperation with an independent laboratorythat provided corroborating data as to the total chromium content. The LIBS systemprovided both the necessary time-resolution and detection limits �14 �g m−3� for emissionmonitoring. Much of the Sandia LIBS-based emissions monitoring was summarizedin a more recent paper, where conditional analysis as described above was found toprovide considerably better results in terms of detection limits and accuracy, includingexcellent agreement with independent extractive sampling [133]. LIBS remains a viabletechnology for real-time emissions and process monitoring, with several research groupsstill pursuing associated issues. However, to date, the widespread implementation of real-time monitoring technologies have not been widely mandated for toxic metal species,hence this potential LIBS application remains a work in progress.

Due to the detection limits in the order of �g m−3, lower with conditional analysis,not many studies were focused on atmospheric aerosols [134]. Carranza et al. measuredambient concentrations of Al, Ca, Mg, and Na as low as 5 ng m−3 using conditional signalprocessing, and were able to detect transient changes in magnesium and aluminum-basedaerosols due to the discharge of fireworks [101]. Nunez et al. studied the detectionof sulphuric acid, which is relevant for several atmospheric processes [135]. A directmeasurement of sulphur via the S I at 182.034 nm emission yielded a detection limitof 165 ppbV after a 15 min integration time. However, laser photo-fragmentation (LPF)after interaction of NaCl and H2SO4 gave improved detection limits of 46.5 ppbV inonly 10 s.

Lithgow et al. reported on a LIBS-based ambient air measurement campaign as partof the Pittsburgh Aerosol Supersite study [136]. They used single-pulse conditionalanalysis similar to the methodology developed by Hahn et al. [96,101], and reportedaerosol measurements that contained the elements Al, Ca, Cr, Cu, Mg, Mn and Na.A significant difference between their measurements and the Carranza et al. [101] studydiscussed above was the overall particle hit rates, with Lithgow et al. reporting hit ratesthat were generally one to two orders of magnitudes less. Differences can be attributedto actual differences in ambient air loadings of the targeted particles, or possibly toadditional combinations of reduced pulse energy (40 mJ vs. 375 mJ) resulting in a smallerplasma sampling volume, orthogonal plasma emission collection vs. back-collection, ordifferences in sampling transport.

In more recent measurements, Hohreiter et al. reported a direct comparison of LIBS-based aerosol particle sampling rates (calcium-based and sodium-based particles) withindependent light scattering-based particle sampling rates [102]. A typical comparisonis presented in Fig. 8, where the correlation between the LIBS-based sampling and thelight-scattering sampling (particles between 500 nm and 2�5 �m) is very strong. Theabove studies all support the use of LIBS for real-time measurement of ambient airparticulate matter, although outstanding issues include overall limits of detection forselect species, calibration and particle size effects, and maximization of particle samplingrates.

In a complementary scheme, laser photo-fragmentation (LPF) and subsequent opticaldetection of atomic emission can be in some cases a valuable alternative to LIBS,especially for alkali species, which are actively involved in the breakdown, corrosion,




0.0

0.2

0.4

0.6

0.8

1.0

1.2

0

500

1000

1500

2000

2500

3000LIBS

Light scattering

LIB

S s

ampl

e ra

te (

%)

Ligh

t sca

tterin

g pa

rtic

le c

ount

rat

e (C

ts/m

in)

6/30

/200

3

7/5/

2003

7/10

/200

3

7/15

/200

3

7/20

/200

3

7/25

/200

3

7/30

/200

3

8/4/

2003

8/9/

2003

Fig. 8. LIBS calcium-based particle sampling rate (calcium-containing spectra/total spectra) andthe light scattering-based total aerosol sampling rate for the cumulative size range between 500 nmand 2�5 �m [102]. The data are presented for thirty, 1-hour sampling intervals as measured forambient air sampled through a PM2.5 inlet.

and erosion of power plant materials [137–139]. Similar observations were reported fordetection of Hg by Tong et al. [140,141]. The metalloids as well as the halides areproblematic elements for current LIBS systems due to the low sensitivity of the detectionsystems and the low emission strength in the NIR. Consequently, a buffer gas (e.g. He)can be employed to increase the emissivity for the NIR lines. In this way, Tran et al.demonstrated the detection of gaseous and particulate fluorides [142]. For particulatefluorides in air, limits of detection were 9 mg m−3 and for detection in He 0�5 mg m−3.In addition, significant improvements were reported for filter-sampled aerosols, whichallowed for a 5 �g m−3 detection limit in He after a 10 min sampling time at 10 � min−1.

Another extension of the direct analysis approach is the use of the laser-inducedplasma only as an atomization reservoir, with a subsequent single element analysis byexcitation of atomic fluorescence with a second laser (LEAF: laser enhanced atomicfluorescence). Neuhauser et al. demonstrated for ultrafine lead aerosol particles producedwith a DMA a detection limit in the ng m−3 range [143]. However, the detection limitincreased from 55 ng m−3 for a particle diameter of 48 nm to 130 ng m−3 for a particlediameter of 300 nm. The increasing detection limit with increasing particle diameter wasprobably due to the incomplete atomization of larger particles in the colder periphery ofthe plasma. A suitable way for an improved LIBS analysis of some of the relevant heavymetal species (e.g. Sn, Hg, and As) is the formation of the volatile hydrides [128,144]respectively, and in the case of mercury the direct detection of the mainly gaseousspecies [145]. In all cases detection limits in the order of 50 �g m−3 were observed.




7. FUTURE DIRECTIONS

Possibilities for improvement of the direct analysis approach are either technology drivenor in the methodology. The advent of echelle spectrographs extends the possibilities tocope with spectral interferences and improve the approach to calibration. Especially,when calibration strategies without any a priori information on the sample are employed,the simultaneous detection of all major constituents is an essential prerequisite [146,147].The combination of a kHz-diode laser-pumped Nd:YAG and fast readout ICCD camerascould improve the sampling efficiency and hence the overall analytical performanceof the direct approach to LIBS aerosol analysis and conditional sampling by orders ofmagnitude. The approach described by Noll et al. for LIBS microanalysis with a kHzsampling rate, i.e. a multiplexed Rowland spectrometer with photomultiplier, might bea suitable alternative to improve the analytical figures of merit [148]. The extension ofspectrometer, especially, echelle technology to the VUV range will permit the analysisof non-metals such as S, P, Cl, or Br, which are of considerable relevance for bothprocess analysis and atmospheric aerosols [149,150]. The direct approach can alsobenefit from aerodynamic lens systems which not only allow the focusing of an aerosolstream for a short distance [151–158], and hence a considerable enrichment, but also asimultaneous sizing of ultrafine aerosols. Another interesting option to reduce possiblematrix effects is on-line condensation of a matrix on the aerosols. This could reducethe variability of the plasma ignition and subsequent elemental emissions [159,160].Necessary detection limits on the order of 1�g m−3 and lower make future applicationsto atmospheric sciences challenging. Process control is expected to be a significantLIBS application in the future. Due to the improved emission control measures for alltypes of industrial emission, the requirements for the detection limits will, however,be tightened in the future and emphasis will shift to ultrafine particles. In that case,conditional analysis will be a sine qua non condition to meet user requirements. TheEPA/DOE test of 1997 showed that aerosol LIBS has similar problems as other LIBSapplications in terms of accuracy and precision, so there is certainly room for considerableimprovement in the general LIBS methodology. The filter based approach offers someadvantages for long-term monitoring of atmospheric aerosols, especially from remoteor rural areas. An automatic filter band sampling could be easily combined with otheraerosol characterization methods such as optical absorption for black carbon, Ramanand/or fluorescence spectroscopy. The latter could be also combined with the directapproach utilizing the 4th harmonic of a Nd:YAG laser for both plasma ignition andfluorescence spectroscopy, e.g. for bioaerosols. In this context, echelle systems, whichpermit a combination of different techniques, i.e. Raman, LIBS or molecular fluorescencein a single instrument, could provide a decisive advantage.

As discussed in chapter 6, the use of double-pulse or dual-pulse LIBS has proveda successful configuration for enhancing the analyte response with analysis of solidor liquid phase systems. However, the use of double-pulse LIBS for aerosol analysiswas only recently examined. Windom et al. reported on the effects of an orthogonaldouble-pulse laser configuration on the atomic emission response for purified air seededwith calcium-rich particles [161], which revealed a marked improvement in calciumatomic emission peak-to-base ratio (∼2-fold increase) and signal-to-noise ratio (∼4-foldincrease) with the double-pulse configuration. Representative spectra as recorded witha single laser and a double-pulse system are shown in Fig. 9, which clearly show




0

500

1000

1500

2000

390 391 392 393 394 395 396 397 398

Inte

nsity

(a.

u.)

Wavelength (nm)

Double-pulse LIBS

Ca II

Ca II

N2+

Fig. 9. Spectra corresponding to the single-pulse LIBS configuration and to the double-pulseLIBS configurations with a 750-ns delay. Both measurements were performed in an aerosol flowof submicron-sized calcium-rich particles. The Ca II atomic emission lines at 393.4 and 396.9 nmare enhanced with double-pulse LIBS, while the purely gaseous-phase emission lines given byN+

2 are degraded with double-pulse [161]. Both spectra have the same intensity scale.

the increased particle-phase analyte response, as measured from the calcium emissionstemming from calcium-rich aerosol particles. In addition to increased analyte response,the double-pulse LIBS system yielded an enhanced single-particle sampling rate whencompared to conventional LIBS. Additional plasma transmission measurements withrespect to the plasma-creating laser pulse were recorded for both single and double-pulse methods over a range of temporal delays, where it was found that an optimaltemporal region existed which yielded moderate coupling of the second laser pulseinto the plasma created by the first laser. Too short of a laser-laser delay (<100 ns)resulted in near total absorption of the second laser pulse into the first plasma, but onlymodest improvements in analyte response. Too long of a laser-laser delay �>20 �s�resulted in significantly reduced coupling of the second laser pulse into the first plasmadue to rarefaction within the plasma volume, negating the benefits of double-pulseLIBS. The optimal region for enhanced analyte response was a pulse-to-pulse sepa-ration between about 750 ns and 5 �s. In consideration of the overall spectroscopicand transmission data, the plasma-analyte interactions realized with a double-pulsemethodology were explained in terms of the interaction with the expanding plasma,which differs between gaseous and particulate phase analytes, as reported in a recentstudy [93].

Perhaps the final frontier in fully understanding the governing processes in LIBS-based aerosol analysis concerns the overall interactions of the laser-induced plasma andthe individual aerosol particle. Such processes speak directly to the issues of complete




Fig. 10. Laser-induced plasma image recorded in the presence of a borosilicate glass particlecorresponding to a delay time of 2 �s with respect to plasma initiation [162]. Image was recordedusing a 396.2-nm narrow band (3-nm fwhm) interference filter. Scale bar = 600 �m.

particle vaporization, size-dependency of analyte response, and issues of plasma inho-mogeneity, method detection limits, and analyte signal precision and accuracy. Recentefforts toward this end include plasma imaging studies in the presence of aerosol parti-cles, as reported by Hohreiter and Hahn [162]. Figure 10 presents an image of a singlelaser-induced plasma that captured an individual borosilicate glass particle. The imagewas captured at a delay of 2 �s following plasma initiation using a narrow line filtercentered at the Ca II line (396.2-nm, 3-nm fwhm) and an ICCD detector. The imageclearly shows that at this delay time, the material that has vaporized and subsequentlydissociated to calcium atoms remains localized about the original aerosol particle. Infact, the high degree of spherical symmetry in the reported images suggests that theatoms diffuse away from the particle on a time scale of some tens of microseconds.Time-resolved measurements provided an estimate of the overall diffusion coefficientin the range of 0�04 m2/s, the first direct measurements of atomic diffusion in a laser-induced plasma subsequent to particle vaporization/dissociation. Such measurementsprovide critical data to complement more advanced plasma modeling efforts, for exam-ple, as reported by Gornushkin et al. in a series of recent studies [163–164]. Overall,as LIBS researchers start to discern a time scale for the dissociation of aerosol particlesand subsequent atomic diffusion within laser-induced plasmas, a more complete pictureof the analyte response begins to emerge. To further advance the analytical rigor oflaser-induced plasma spectroscopy, rate-limited particle dissociation and mass transportshould be further explored in the context of LIBS, as well as for plasmas from othersystems (e.g. inductively-coupled plasmas) and plasma interactions with other forms ofmatter.




REFERENCES

[1] J. O. Nriagu and J. M. Pacyna, Nature, 333 (1988) 134.[2] B. Veronesi and M. Oortgiesen, Neurotoxicology 22 (2001) 795.[3] A. G. Allen, E Nemitz, J. P Shi, R. M. Harrison and J. C. Greenwood, Atmos. Environ. 35

(2001) 4581.[4] A. E. Suarez and J. M. Ondov, Energy & Fuels 16 (2002) 562.[5] S. K. Friedlander, J. Aerosol Sci. 2 (1971) 331.[6] S. K. Friedlander, J. Aerosol Sci. 1 (1970) 295.[7] P. D. Maker, R. W. Terhune and C. M. Savage, Optical third harmonic generation in

various solids and liquids, P. Grivet and N Bloembergen, Eds. Columbia University Press:New York, (1963) p. 1559.

[8] R. G. Meyerand, Jr. and A. F. Haught, Phys. Rev. Letters 11 (1963) 401.[9] C. DeMichelis, IEEE J. Quantum Electron. 5 (1969) 188.

[10] C. G. Grey Morgan, Rep. Prog. Phys. 38 (1975) 621.[11] N. Bloembergen, J. Nonlin. Opt. Phys. Mat 6 (1997) 377.[12] T. P. Hughes, Plasmas and Laser Light; Wiley: New York (1975).[13] Y. B. Zel’dovich and Y. P. Raizer, Physics of Shock Waves and High Temperature Hydro-

dynamic Phenomena; Dover Publications: New York (2002).[14] G. Bekefi, Principles of Laser Plasmas; Wiley: New York (1976).[15] J. F. Ready, Effects of High-Power Laser Radiation; Academic Press: New York (1971).[16] J. R. Bettis, Appl. Opt. 31 (1992) 3448.[17] G. S. Voronov, Zhurnal Eksperimental’noi i Teoreticheskoi Fiziki 51 (1966) 1496.[18] G. S.Voronov, G. A. Delone and N. B. Delone, Zh. Eksperim. i Teor. Fiz.Pis’ma v

Redaktsiyu 3 (1966) 480.[19] P. Agostini, G. Barjot, G. Mainfray, C. Manus and J. Thebault, IEEE J. Quantum Electron.

6 (1970) 782.[20] P. Agostini, G. Barjot, J. F. Bonnal, G. Mainfray, C. Manus and J. Morellec, IEEE J.

Quantum Electron. 4 (1968) 667.[21] G. Baravian and G. Sultan, Physica B+C 128 (1985) 343.[22] A. L’Huillier, Journal De Physique 48 (1987) 415.[23] C. H. Chan, C. D. Moody and W. B. McKnight, J. Appl. Phys. 44 (1973) 1179.[24] G. M. Weyl and D. Rosen, Phys. Rev. A 31 (1985) 2300.[25] E. Yablonovitch, Appl. Phys. Lett. 23 (1973) 121.[26] E. Yablonovitch, Phys. Rev. Lett. 32 (1974) 1101.[27] E. Yablonovitch, Phys. Rev. A 10 (1974) 1888.[28] L. V. Keldysh, Zh. Eksperim. i Teor. Fiz. 47 (1964) 1945.[29] A. D. MacDonald, Microwave Breakdown in Gases; Wiley: New York (1966).[30] A. V. Phelps, Physics of Quantum Electronics; P. Kelley, B. Lax and P. Tannenwald, Eds.;

McGraw-Hill: New York (1966) p. 538.[31] P. F. Browne, Proc. Phys. Soc. 86 (1965) 1323.[32] J. K. Wright, Proc. Phys. Soc. 84 (1964) 41.[33] E. K. Damon and R. G. Tomlinson, Appl. Opt. 2 (1963) 546.[34] N. M. Kroll and K. M. Watson, Phys. Rev. A 5 (1972) 1883.[35] G. M. Weyl, D. I. Rosen, J. Wilson and W. Seka, Phys. Rev. A 26 (1982) 1164.[36] D. I. Rosen and G. M. Weyl, J. Phys. D 20 (1987) 1264.[37] H. T. Buscher, R. Tomlinson and K. Damon, Phys. Rev. Lett. 15 (1965) 847.[38] K. C. Byron and G. J. Pert, J. Phys. D 12 (1979) 401.[39] J. B. Simeonsson and A. W. Miziolek, Appl. Phys. B 59 (1994) 1.[40] R. Tambay and R. K. Thareja, J. Appl. Phys. 70 (1991) 2890.[41] J. P. Davis, A. L. Smith, C. Giranda and M. Squicciarini, Appl. Opt. 30 (1991) 4358.




[42] J. Stricker and J. G. Parker, J. Appl. Phys. 53 (1982) 851.[43] R. A. Armstrong, R. A. Lucht and W. T. Rawlins, Appl. Opt. 22 (1983) 1573.[44] R. J. Dewhurst, G. J. Pert and S. A. Ramsden, J. Phys. B 7 (1974) 2281.[45] R. J. Dewhurst, J. Phys. D 10 (1977) 283.[46] C. L. M. Ireland and C. G. Morgan, J. Phys. D 6 (1973) 720.[47] Y. Gamal and M. A. Harith, J. Phys. D 14 (1981) 2209.[48] M. Hanafi, M. M. Omar and Y. E. E. D Gamal, Rad. Phys. Chem. 57 (2000) 11.[49] L. M. Davis, L. Q. Li and D. R. Keefer, J. Phys D 26 (1993) 222.[50] C. L. M. Ireland, A. Yi, J. M. Aaron and C. G. Morgan, Appl. Phys. Lett. 24 (1974) 175.[51] H. Sobral, M. Villagran-Muniz, R. Navarro-Gonzalez and A. C. Raga, Appl. Phys. Lett. 77

(2000) 3158.[52] A. F. Haught and D. H. Polk, Phys. Fluids 9 (1966) 2047.[53] V. E. Zuev, A. A. Zemlyanov, Y. D. Kopytin and A. V. Kuzikovskii, High-Power Laser

Radiation in Atmospheric Aerosols; D. Reidel Publishing Company: Dordrecht (1984).[54] A. A. Lushnikov and A. E. Negin, J. Aerosol Sci. 24 (1993) 707.[55] R. L. Armstrong, Appl. Opt. 23 (1984) 148.[56] T. A. Schoolcraft, G. S. Constable, L. V. Zhigilei and B. J. Garrison, Anal. Chem. 72

(2000) 5143.[57] S. V. Zakharchenko, L. P. Semenov and A. M. Skripkin, Proc. Intl Conf. on Lasers

(1980) 864.[58] D. E. Lencioni and L. C. Pettingill, J. Appl. Phys. 48 (1977) 1848.[59] D. E. Lencioni, Appl. Phys. Lett. 25 (1974) 15.[60] D. E. Poulain, D. R. Alexander, J. P.Barton, S. A.Schaub and J. Zhang, J. Appl. Phys. 67

(1990) 2283.[61] I. Y. Borets-Pervak and V. S. Vorob’ev, Quantum Electron. 23 (1993) 224.[62] D. C. Smith and R. T. Brown, J. Appl. Phys. 46 (1975) 1146.[63] D. C. Smith, J. Appl. Phys. 48 (1977) 2217.[64] G. M. Weyl, J. Phys. D 12 (1979) 33.[65] D. C. Smith, Opt. Eng. 20 (1981) 962.[66] S. T. Amimoto, J. S. Whittier, F. G. Ronkowski, P. R. Valenzuela, G. N. Harper,

R. R. Hofland, Jr.; G. L. Trusty, T. H. Cosden and D. H. Leslie, AIAA J. 22 (1984) 1108.[67] J. E. Lowder and H. Kleiman, J. Appl. Phys. 44 (1973) 5504.[68] A. A. Boni and D. A. Meskan, Opt. Comm. 14 (1975) 115.[69] G. H. Canavan and P. E. Nielsen, Appl. Phys. Lett. 22 (1973) 409.[70] R. G. Pinnick, A. Bsiwas, J. D. Pendleton and R. L. Armstrong, Appl. Opt. 31 (1992)

311–317.[71] P. Chylek, M. A. Jarzembski, N. Y. Chou and R. G. Pinnick, Appl. Phys. Lett. 49

(1986) 1475.[72] A. Biswas, H. Latifi, P. Shah, L. J. Radziemski and R. L. Armstrong, Opt. Lett. 12

(1987) 313.[73] A. Biswas, H. Latifi, L. J. Radziemski and R. L. Armstrong, Appl. Opt. 27 (1988) 2386.[74] R. G. Pinnick, P. Chylek, M. Jarzembski, E. Creegan, V. Srivastava, G. Fernandez,

J. D. Pendleton and A. Biswas, Appl. Opt. 27 (1988) 987.[75] P. Chylek, M. A. Jarzembski, V. Srivastava, R. G. Pinnick, J. D. Pendleton and

J. P. Cruncleton, Appl. Opt. 26 (1987) 760.[76] W. F. Hsieh, J. H. Eickmans and R. K. Chang, J. Opt. Soc. Am. B 4 (1987) 1816.[77] R. K. Chang, J. H. Eickmans, W. F.; W. F. Hsieh, C. F. Wood, J. Z. Zhang and J. B. Zheng,

Appl. Opt. 27 (1988) 2377.[78] M. Sabsabi, R. Heon, V. Detalle, L. St-Onge and A. Hamel, OSA Trends in Optics and

Photonics (TOPS) (OSA Technical Digest) 81 (2002) 128.[79] J. E. Carranza, E. Gibb, B.W. Smith, D. W. Hahn and J. D. Winefordner, Appl. Opt. 42

(2003) 6016.




[80] E. Voigtman, Appl. Spectroscopy 45 (1991) 237.[81] U. Panne, C. Haisch, M. Clara and R. Niessner, Spectrochim. Acta B53 (1998) 1957.[82] B. T. Fisher, H. A. Johnsen, S. G. Buckley, and D. W. Hahn, Appl. Spectrosc. 55

(2001) 1312.[83] L. J. Radziemski, D. A. Cremers and T. R. Loree, Spectrochim. Acta B 38 (1983) 349.[84] T. R. Loree and L. J. Radziemski, Plasma Chem. Plasma Process. 1 (1981) 271.[85] L. J. Radziemski and T. R. Loree, Plasma Chem. Plasma Process. 1 (1981) 281.[86] J. E. Carranza and D. W. Hahn, Spectrochim. Acta B 57 (2002) 779.[87] V. Hohreiter, A. Ball and D. W. Hahn, J. Anal. At. Spectrom. 19 (2004) 1289.[88] D. A. Cremers and L. J. Radziemski, Appl. Spectrosc. 39 (1985) 57.[89] L. J. Radziemski, T. R. Loree, D. A. Cremers and N. M. Hoffman, Anal. Chem. 55

(1983) 1246.[90] M. Essien, L. J. Radziemski and J. Sneddon, J. Anal. At. Spectrom. 3 (1988) 985.[91] J. E. Carranza and D. W. Hahn, Anal. Chem. 74 (2002) 5450.[92] E Vors and L. Salmon, Anal. Bioanal. Chem. 385 (2006) 281.[93] V. Hohreiter and D.W. Hahn, Anal. Chem. 77 (2005) 1118.[94] D. W. Hahn, J. E. Carranza, G. R. Arsenault, H. A. Johnsen and K. R. Hencken, Rev. Sci.

Instrum. 72 (2001) 3706.[95] D. W. Hahn and M. M. Lunden, Aerosol Sci. Technol. 33 (2000) 30.[96] D. W. Hahn, W. L. Flower and K. R. Hencken, Appl. Spectrosc. 51 (1997) 1836.[97] S. Yalcin, D. R. Crosley, G. P. Smith and G. W. Faris, Appl. Phys. B 68 (1999) 121.[98] Y. L. Chen, J. W. L. Lewis and C. Parigger, J. Quant. Spectros. Radiat. 67 (2000) 91.[99] J. E. Carranza and D. W. Hahn, J. Anal. At. Spectrom. 17 (2002) 1534.

[100] V. Hohreiter, J. E. Carranza, and D. W. Hahn, Spectrochim. Acta B 59 (2004) 327.[101] J. E. Carranza, B. T. Fisher, G. D. Yoder and D. W. Hahn, Spectrochim. Acta B 56

(2001) 851.[102] B. Hettinger, V. Hohreiter, M. Swingle, and D. W. Hahn, Appl. Spectrosc. 60 (2006) 237.[103] I. Schechter, Anal. Sci. Technol. 8 (1995) 779.[104] J. E. Carranza, K. Iida and D. W. Hahn, Appl. Opt. 42 (2003) 6022.[105] P. B. Dixon and D. W. Hahn, Anal. Chem. 77 (2005) 631.[106] D.W. Hahn, Appl. Phys. Letters 72 (1998) 2960.[107] S. D. Arnold and D. A. Cremers, Amer. Ind. Hyg. Assn. J. 56 (1995) 1180.[108] R. E. Neuhauser, U. Panne and R. Niessner, Anal. Chim. Acta 392 (1999) 47.[109] R. E. Neuhauser, U. Panne and R. Niessner, Appl. Spectrosc. 54 (2000) 923.[110] U. Panne, R. E. Neuhauser, M. Theisen, H. Fink and R. Niessner, Spectrochim. Acta B 56

(2001) 839.[111] M. Z. Martin, M. D. Cheng and R. C. Martin, Aerosol Sci. Technol. 31 (1999) 409.[112] L. J. Radziemski, Microchem. J. 50 (1994) 218.[113] L. J. Radziemski and D. A. Cremers (editors), Laser-Induced Plasmas and Applications,

Marcel Dekker: New York (1989).[114] J. Sneddon, Trends Anal. Chem. 7 (1988) 222.[115] B. W. Smith, D. W. Hahn, E. Gibb, I. Gornushkin and J. D. Winefordner, Kona 19 (2001) 25.[116] D. A. Rusak, B. C. Castle, B. W. Smith and J. D. Winefordner, CRC Crit. Rev. Anal.

Chem. 27 (1997) 257.[117] E. B. Belyaev, A. P. Godlevskii and Y. D. Kopytin, Kvantovaya Elektronika 5 (1978) 2594.[118] Y. L. Chen, J. W. L. Lewis and C. Parigger, J. Quant. Spectrosc. Radiat. 66 (2000) 41.[119] T. X. Phuoc and F. P. White, Combust. Flame 119 (1999) 203.[120] T. X. Phuoc and C. M. White, Opt. Commun. 181 (2000) 353.[121] A. Sircar, R. K. Dwivedi and R. K. Thareja, Appl. Phys. B 63 (1996) 623.[122] R. J. Nordstrom, Appl. Spectrosc. 49 (1995) 1490.[123] S. Yalcin, D. R. Crosley, G. P. Smith and G. W. Faris, Hazard. Waste Hazard. Mater. 13

(1996) 51.




[124] L. Dudragne, Ph. Adam and J. Amouroux, Appl. Spectrosc. 52 (1998) 1321.[125] M. Tran, B. W. Smith, D. W. Hahn, and J. D. Winefordner, Appl. Spectrosc., 55 (2001) 1455.[126] L. W. Peng, W. L. Flower, K. R. Hencken, H. A. Johnsen, R. F. Renzi and N. B. French,

Process Cont. Qual. 7 (1995) 39.[127] W. L. Flower, L. W. Peng, M. P. Bonin, N. B. French, H. A. Johnsen, D. K. Ottesen,

R. F. Renzi and L. V. Westbrook, Fuel Process. Technol. 39 (1994) 277.[128] J. P. Singh, H. Zhang, F. Y. Yueh and K. P. Carney, Appl. Spectrosc. 50 (1996) 764.[129] H. Zhang, F. Y. Yueh and J. P. Singh, Appl. Opt. 38 (1999) 1459.[130] J. P. Singh, F. Y. Yueh, H. Zhang and R. L. Cook, Process Cont. Qual. 10 (1997) 247.[131] P. M. Lemieux, J. V. Ryan, N. B. French, W. J. Haas, S. Priebe and D. B. Burns, Waste

Manage. 18 (1998) 385.[132] R. E. Neuhauser, U. Panne, R. Niessner and P. Wilbring, Fresenius J. Anal. Chem. 364

(1999) 720.[133] S. G. Buckley, H. A. Johnsen, K. R. Hencken and D. W. Hahn, Waste Manage. 20

(2000) 455.[134] M. Casini, M. A. Harith, V. Palleschi, A. Salvetti, D. P. Singh and M. Vaselli, Laser Part.

Beams 9 (1991) 633.[135] M. H. Nunez, P. Cavalli, G. Petrucci and N. Omenetto, Appl. Spectrosc. 54 (2000) 1805.[136] G. A. Lithgow, A. L. Robinson and S. G. Buckley, Atmos. Environ. 38 (2004) 3319.[137] S. G. Buckley, C. S. McEnally, R. F. Sawyer, C. P. Koshland and D. Lucas, Combust. Sci.

Technol. 118 (1996) 169.[138] S. G. Buckley, R. F. Sawyer, C. P. Koshland and D. Lucas, Combust. Flame 128 (2002) 435.[139] K. T. Hartinger, P. B. Monkhouse and J. Wolfrum, Ber. Bunsenges. Phys. Chem.

97(1993)1731.[140] X. Tong, R. B. Barat and A. T. Poulos, Rev. Sci. Inst. 70 (1999) 4180.[141] X. Tong, R. B. Barat and A. T. Poulos, Environ. Sci. Technol. 33 (1999) 3260.[142] M. Tran, Q. Sun, B. W. Smith and J. D. Winefordner, Appl. Spectrosc. 55 (2001) 739.[143] R. E. Neuhauser, U. Panne, R. Niessner, G. A. Petrucci, P. Cavalli and N. Omenetto, Anal.

Chim. Acta 346 (1997) 37.[144] E. A. P. Cheng, R. D. Fraser and J. G. Eden, Appl. Spectrosc. 45 (1991) 949.[145] C. Lazzari, M. De Rosa, S. Rastelli, A. Ciucci, V. Palleschi and A. Salvetti, Laser Part.

Beam 12 (1994) 525.[146] A. Ciucci, M. Corsi, V. Palleschi, S. Rastelli, A. Salvetti and E. Tognoni, Appl. Spectrosc.

53 (1999) 960.[147] E. Tognoni, V. Palleschi, M. Corsi and G. Cristoforetti, Spectrochim. Acta B 57 (2002) 1115.[148] H. Bette, and R. Noll, J. Phys. D 37 (2004) 1281.[149] I. Radivojevic, C. Haisch, R. Niessner, S. Florek, H. Becker-Ross and U. Panne, Spec-

trochim. Acta B, 59 (2004) 335.[150] I. Radivojevic, C. Haisch, R. Niessner, S. Florek, H. Becker-Ross and U. Panne, Anal.

Chem. 76 (2004) 1648.[151] G. A. Petrucci, P. B. Farnsworth, P. Cavalli and N. Omenetto, Aerosol Sci.Technol.33

(2000) 105.[152] B. Dahneke, Nature 244 (1973) 54.[153] J. Schreiner, C. Voigt, K. Mauersberger, P. McMurry and P. Ziemann, Aerosol Sci. Technol.

29 (1998) 50.[154] P. Liu, P. J. Ziemann, D. B. Kittelson and P. H. McMurry, Aerosol Sci. Technol, 22

(1995) 293.[155] P. Liu, P. J. Ziemann, D. B. Kittelson and P. H. McMurry, Aerosol Sci. Technol. 22

(1995) 314.[156] M.-D. Cheng, L. Karr, J. Kornuc, D. Staat, T. Wainman, B.Harre and B. Sugiyama,

Microchem. J. 72 (2002) 209.




[157] M. D. Cheng, Fuel Process. Technol. 65–66 (2000) 219.[158] J.-W. Lee, M.-Y. Yi and S.-M. Lee, J. Aerosol Sci. 34 (2002) 211.[159] S. N. Jackson and K. K. Murray, Anal. Chem. 74 (2002) 4841.[160] D. B. Kane and M. V. Johnston, Anal. Chem. 73 (2001) 5365.[161] B. C. Windom, P. K. Diwakar and D. W. Hahn, Spectrochim. B 61 (2006) 788.[162] V. Hohreiter and D. W. Hahn, Anal. Chem. 78 (2006) 1509.[163] I. B.Gornushkin, A. Y. Kazakov, N. Omenetto, B. W. Smith and J. D. Winefordner,

Spectrochim. Acta B 59 (2004) 401.[164] A. Y. Kazakov, I. B.Gornushkin, A. Y. Kazakov, N. Omenetto, B. W. Smith and

J. D. Winefordner, Appl. Opt. 45 (2006) 2810.



Chapter 18

Scope of Future Development in LIBS

J. P. Singh and F. Y. Yueh

Institute for Clean Energy Technology, Mississippi State University,205 Research Boulevard, Starkville MS 39759, USA

1. INTRODUCTION

The development of LIBS has progressed rapidly during the last decade. Various groupshave worked on improving LIBS measurements of different samples by using advancedlasers, detection systems and data processing techniques. This has opened more applica-tion areas for LIBS in the future. In this charter, we would like to discuss future LIBSdevelopments both for current LIBS applications and for possible near-term and far-termLIBS applications.

Many existing LIBS applications have not reached the state of practical use due toinsufficient accuracy and precision. The current precision and accuracy is enough for aprocess monitor and control unit, or for use as an on-site/on-line screening tool in anindustrial environment. However, for LIBS to be accepted as an analytical technique forquality control, its precision and accuracy need to be further improved. LIBS research inthis area needs to be strengthened before the technology reaches the maturity needed forcommercialization. Beside the general problems of analytical figure of merit, the differentsample types also have different technical challenges. The directions of research workneeded for effective handling of different sample types, for developing theoretical modelsto understand various physical and chemical processes in LIBS and for developing therequisite technology, are briefly summarized in the following sections. Although it is notpossible to indicate the exact course of development in a rapidly expanding field of LIBSapplications, we have tried to enumerate some of the areas on the basis of current trends.

2. GAS PHASE LIBS

The early LIBS development in the gas phase concentrated on using LIBS as a multi-metal Continuous Emissions Monitor (CEM). LIBS technology has been evaluated andcompared with other techniques, such as an ICP developed by DOE-EPA in several fieldtests at an EPA site. [1] During these tests, LOD for some of the RCRA metals (suchas Be, Cr, Pb, Cd) satisfied the EPA requirements, but it was not so for Hg and As.




420 J. P. Singh and F. Y. Yueh

To meet EPA requirements for a CEM, LIBS researches have to focus on improvingcalibration methods, sensitivity and accuracy. Also a reliable on-line calibration methodis needed to provide real-time, on–line LIBS analysis for off-gas emission measurements(see chapter 9). LIBS sensitivity for some toxic metals such as Hg, As, Sb, and Ce alsoneeds to be further improved to reach EPA’s requirement. The sensitivity of the LIBSmeasurement can be improved either by increasing the signal or reducing the back-ground noise from the plasma. Researches which can significantly improve the LIBS’signal-to-noise ratio are needed. The pulse-to-pulse variation in laser energy and effectsof particle size can also cause large LIBS signal variation. To improve the measure-ment precision, work is needed to reduce pulse-to-pulse fluctuations of the laser-inducedspark. Data processing techniques to select data over a certain threshold and to applystatistical averaging methods to this data might be very helpful for further improvementof the LIBS analytical figure of merit. Some of the toxic metals have their most sensitiveanalytical lines at wavelengths shorter than 200 nm, however, sensitivity of the detectorsdecreases tremendously in this spectral region. Work to improve measurements in thisspectral region will greatly enhance the sensitivity of detection for such elements. LIBSof particles suspended in air and aerosols is expected to be a major area of future investi-gations in view of the harmful biological species associated with such particulate matter.

3. LIQUID PHASE LIBS

There are many types of samples in the liquid phase that have been studied by LIBS[2–6] (also see chapter 10). Most common liquid samples studied are the toxic metalcontaminated water. Others include solid metal or ceramic pieces in water, and slurrysamples which are a mixture of solids and water. The exploration of the deep sea usingLIBS has also been reported [7,8]. LIBS has also demonstrated its potential for thedirect and rapid analysis of pharmaceutical liquid formulations. Due to the relativelyshorter plasma lifetime, however, LIBS measurements in liquids have poor sensitivity.The splashing of the liquid due to the laser produced shock waves can change thelocation of the measurement and affect its precision, as well as accuracy. Recent studieshave shown that multi-pulse LIBS can improve LIBS sensitivity in liquid measurements[9–13]. Further work in this area needs to be carried out to obtain the best configurationfor multi-pulse LIBS. Signal enhancement with multi-pulse LIBS is needed for allliquid phase samples to improve the LOD. The other main research areas of interestare the studies of the S/N for the laser spark produced with different liquid surfacesand bulk liquids; and techniques to measure concentrations at various depths. Furtherresearch work in handling different types of liquid samples is also needed. ImprovedLIBS experimental geometry should be able to increase the LIBS signal and reduce thebackground noise from the plasma.

4. SOLID PHASE LIBS

Solid samples such as soil, glass, alloy, concrete, paint, etc. have been explored byLIBS [14–19] (see also chapter 11). The main issue for solid samples is to achievehigh degree of quantitative precision in measurements. Although, at present the LIBS



Scope of Future Development in LIBS 421

measurements on solid samples have better precision compared to liquid and gas phasesamples, further improvements in instrumentation are needed to compete with otheranalytical instruments in the market. Lately, the use of short-pulse lasers (ps and fs) forLIBS measurements have been demonstrated [20–23]. The femtosecond lasers can ablatethe sample more effectively and cause less damage to the target compared to that usingthe nanosecond LIBS. Fs-LIBS also produces less continuum background as comparedto ns-LIBS leading to higher sensitivity. Further developments in fs-LIBS systems mightbe able to greatly improve their analytical capabilities. Hybrid techniques, combiningLIBS with other discharge excitation sources provide alternative ways to improve LIBSanalytical capabilities. Matrix effects have been known for years to affect compositionmeasurements with LIBS and this area also needs further investigations.

5. LIBS OF MOLTEN SAMPLES

Applications of LIBS as an on-line process monitor for molten materials such as alloys,steel, and glass, have been demonstrated [24–27] (see also Chapter 11). The major focusin this field is improving the precision and accuracy in order to establish LIBS as an on-line compositional analysis tool for process control. Improvements are also needed in thedesign of the LIBS probe for long-term operation under high temperature environment.The design needs to be perfected so that the optics in the probe remains clean in thehigh temperature furnace and the laser produced spark is at the melt surface, not in themetal vapor above the surface. On-line calibration techniques also need to be evaluatedand tested for long-term operation.

6. THEORETICAL MODELS OF LASER INDUCED PLASMA

The laser induced plasma (LIP) is a product of very complicated process of laser-matterinteraction. When a short pulsed, high peak power laser beam is focused onto any tar-get, the processes of absorption of laser energy, vaporization; ejection of atoms, ions,molecular species, and fragments, take place in quick succession. Plasma initiation isfollowed by its expansion, production of shock waves; and many other processes. Manytheoretical models have been developed to describe these processes but they are usuallyapplicable under some idealized conditions [28–31]. A model to completely describethe processes of laser ablation and plasma formation does not exist at present and thedevelopment of such a model will be a great help in selecting the optimum experimentalparameters and configurations for quantitative LIBS measurement in various environ-ments. The power density of the focused laser beam and the thermo-optical propertiesof the material are two critical parameters that influence the laser ablation process andthe model, that could predict the ablation processes for different materials, will be veryhandy in determining the LIBS operating conditions for different applications. Two (ormore) pulse LIBS excitation have shown great promise for investigations of liquid andsolid samples [32,33] and theoretical work in this area is needed to understand themechanisms involved in multi-pulse-LIBS.




7. COMMERCIALIZATION OF LIBS

LIBS has great potential for field applications for both the military and industry becauseof its use as an in-situ and remote detection system. Recently, some commercial LIBSsystems have found their way in the market for some specific applications. The com-mercialization of a LIBS instrument for general composition analysis still needs a lotof research and development. The research efforts should not only focus on the ana-lytical figure of merit, but also on methods to minimize the system components fora compact, light weight, portable system. A cheap LIBS detection system, that couldsimultaneously measure all the elements of interest, is yet to be developed. A telescopetype optical design is also needed for standoff detection. The developments in the fieldof micro LIBS can produce future affordable and portable LIBS systems for industrialand military applications [34,35] (see also chapter 8).

8. FUTURE APPLICATIONS

LIBS has been used in the exploration of many areas related to scientific, indus-trial, medical, environmental security and social problems of current interest [36–45].Possible directions of research and development that need further attention are brieflysummarized below:

• Ambient Air Particle AnalysisFine particulate matter (PM) in air has posed serious risks to human health. LIBShas great potential in real-time analysis of ambient air particles.

• Continuous Emission MonitorMonitoring of the toxic metals emission from coal-fired power plants/incinerators/industrial processes by LIBS needs further refinements.

• Detection of Toxic Elements in WaterGround water contaminated by metals dissolved in water due to increased acid rain,livestock and pesticides have become a health issue. LIBS can monitor the toxicelements in effluents from water treatment plants or groundwater.

• Soil AnalysisLIBS has been successfully used, in detecting toxic metals in contaminated soil andin the measurement of total carbon content in soils.

• Slurry Sample analysisAnalysis of slurry samples in various industrial processes or waste disposal processescan be very efficiently performed using LIBS.

• Powder Sample CompositionLIBS applications for on-line analysis of powder samples (e.g. glass batch, fly ash,ash added to cement, etc.) in various industrial and waste management processeshave been demonstrated.

• Radioactive AnalysisThe remote sensing capability of LIBS is very attractive for detection and mea-surement of radioactive nuclei in different radioactive samples. It can be used tomonitor radioactive glasses used to store nuclear waste.




• Molten MaterialOn-line analysis of the alloy composition in the molten state containing Al, Fe, Mn,Cu etc. has been demonstrated. This provides a process monitor for these moltenmetal industries.

• Military ApplicationsMilitary applications include mine/unexploded ordnance detection, detection ofchemical warfare agents, depleted uranium (DU) in aerosols/particulates, etc.

• Space Exploration:The study of LIBS for future space exploration was initiated in 1992. Recent researchis focused on developing stand-off LIBS for application to Mars exploration.

• Phytoremediation AnalysisPlant leaves can be analyzed on site to study the metals absorbed by the biomass.By combining LIBS data with other optical spectroscopy analytical techniques, onecould obtain optimum control of the phytoremediation process.

• Biomedical ApplicationLIBS has been successfully used for analysis of trace elements in nail, hair, blood,urine and dental samples. It has been used to distinguish between normal andcancerous tissue based on the difference in elemental concentration.

• Pharmaceutical ApplicationsLIBS has been applied for the direct and rapid analysis of drugs and lubricants intablets and saline solutions for the pharmaceutical industry.

• Food SafetyLIBS can be applied to determine toxic metal level in processed as well as unpro-cessed food.

The major difficulties common to most LIBS applications are:

1. Problems in achieving very high precision and accuracy in quantitative analysis.2. It is too expensive for many applications due to the cost of the laser and the

sensitive detection systems3. Matrix effects cause difficulty in obtaining suitable standard spectral lines for LIBS

measurements.4. Different elements have different optimal detection windows. It is hard to simulta-

neously obtain an optimum detection widow for all the elements of interest.

To overcome the known problems and advance LIBS for future applications, we needto improve our knowledge of the technology through basic research. The fundamentalsof LIBS plasma, improvement in LIBS quantification, increase in the understandingof pulse to pulse fluctuations, and study of the laser-matter interactions, are some ofthe broad areas for future research. Instrumentation developments that can providegood broadband detector sensitivity and fast response, improved overall S/N and cheaprugged laser sources are needed for future LIBS systems. For specific applications(e.g. biomedical applications) the study of specific preparation procedures are needed toachieve usefulness of LIBS data.

Since the first use of laser induced plasma as an ablation source in 1962, LIBS hasmade significant progress toward a mature analytical tool. Various LIBS applications insolid, liquid, and gas samples have been explored by many research groups. The future




applications for LIBS should focus on the real strengths of LIBS, that is, on-line, real-time, non-intrusive, and simultaneous multi-element analysis. With the rapid technologyadvances, more and more new LIBS applications will appear in the future. We shouldalso see more commercial LIBS instruments for different applications in the market innear future.

REFERENCES

[1] http://apps.em.doe.gov/OST/pubs/itsrs/itsr18.pdf, US Department of Energy, Office of Envi-ronmental Management, September (2002) 1.

[2] D. A. Cremers, L. J. Radziemski and T. R. Loree. Appl. Spectrosc. 38 (1984) 721.[3] Ota Shame, Dave C. S. Beddows, Jozef Kaiser, Sergei V. Kukhlevsky, Miroslav Liška,

Helmut H. Telle, Jim Young, Optical Engineering, 39 (2000) 2248.[4] J. R. Wachter and D. A. Cremers. Appl. Spectrosc. 41 (1987) 1042.[5] L. Barrette and S. Turmel, Spectrochim. Acta B56 (2000) 715.[6] L. St-Onge, E. Kwong, M. Sabsabi and E. B. Vadas, J Pharm Biomed Anal. 29;36(2)

(2004) 277.[7] A. De Giacomo, M. Dell’Aglio, F. Colao and R. Fantoni, Spectrochim. Acta. B59

(2004) 1431.[8] Anna PM Michel, Marion J. Lawrence-Snyder, S. Michael Angel, and Alan D. Chave,

Oceanic Applications of Laser Induced BreakdownSpectroscopy: Laboratory Validation(http://www.oceans06mtsieeeboston.org/docs/paper.pdf).

[9] J. Uebbing, J. Brust, W. Sdorra, F. Leis and K. Niemax, Appl. Spectrosc. 45 (1991) 1419.[10] R. Sattmann, V. Sturn and R. Noll, J. Phys. D28 (1995) 2181.[11] D. N. Stratis, K. L. Eland and S. M. Angel, Appl. Spectrsc. 5 (2001) 1297.[12] L. St. Onge, M. Sabsabi and P. Cielo, Spectrochim Acta B53 (1998) 407.[13] L. St. Onge, V. Detalles and M. Sabsabi, Spectrochim Acta B57 (2000) 121.[14] M. Hemmerlin, R. Meilland, H. Falk, P. Wintjens and L. Paulard, Spectrochim. Acta B56

(2001) 661.[15] G. A. Theriault, S. Bodensteiner and S. H. Lieberman, Field Anal. Chem. Technol. 2

(1998) 117.[16] T. L. Thiem, R. H. Salter, J. A. Gardner, Y. I. Lee and J. Sneddon, Appl. Spectrosc. 48

(1994) 58.[17] S. Rosenwasser, G. Asimellis, B. Bromley, R. Hazlett, J. Martin, T. Pearee, and A. Zigler,

Spectrochim. Acta B56 (2001) 707.[18] J. A. Millard, R. D. Dalling and L. J. Radziemski, Appl. Spectrosc. 40 (1986) 491.[19] R. Wisbrun, I. Schechter, R. Niessner, H. Schroeder and K. L. Kompa, Anal. Chem. 66

(1994) 2964.[20] Ph. Rohwetter, J. Yu, G. Mejean, K. Stelmaszczyk, E. Salmon, J. Kasparian, J.-P. Wolf and

L. Woste, J. Anal. At. Spectrom, 19 (2004) 437.[21] B. Le Drogoff, F. Vidal, Y. Von Kaenel, M. Chaker, M. Sabsabi, T. W. Johnston, S. Laville

and J. Margot, J. Appl. Phys. 89 (2001) 8247.[22] B. Le Drogoff, J. Margot, M. Chaker, M.Sabsabi, O. Barthélemy, T. W. Johnston, S. Laville,

F. Vidal and Y. Von Kaenel, Spectrochim. Acta B56 (2001) 987.[23] K. L. Eland, D. N. Stratis, T. Lai, M. A. Berg, S. R. Goode, and S. M. Angel, Appl. Spectrosc.

55 (2001) 279.[24] L. Peter, V. Sturm and R. Noll, Appl Opt. 42 (2003) 6199.[25] C. Aragón, J. A. Aguilera and J. Campos, Appl. Spectrosc. 47 (1993) 606.[26] A. K. Rai, F. Y. Yueh and J. P. Singh, Appl Opt. 42 (2003) 2078.




[27] J.-I. Yun, R. Klenze and J.-I. Kim, Appl. Spectrosc. 56 (2002) 437.[28] R. E. Russo, X. Mao and S. S. Mao, Anal. Chem. 74 (2002) 70A.[29] V. Stefan (Ed.), Select Topics in Laser-Matter Interaction, The Stefan University Press, La

Jolla, CA, (2003).[30] S. I. Anisimov and V. A Khokhlov, Instabilities in Laser-matter interaction, CRC Press, Boca

Raton, Fla. (1995).[31] R. Krasniker, V. Bulatov and I. Schechter, Spectrochim. Acta B56 (2001) 609.[32] R. Sattman, V. Sturm and R. Noll, J. Phys. D: Appl. Phys. 28 (1995) 2181.[33] De Giacomo, M. Dell’Aglio and F. Colao, Applied Surface Science 247 (2005) 157.[34] D. Menut, P. Fichet, J. L. Lacour, A. Rivoallan and P. Mauchien, Appl. Opt. 42 (2003) 6063.[35] J. J. Zayhowski, J. Alloys Compd. 303 (2000) 393.[36] A. W. Miziolek and J. Spicer, Report of the workshop on laser induced break-

down spectroscopy-state of current research and future directions, August 5–6, 2003,Baltimore, MA.

[37] J. F. Archambault, A. Vintiloiu and E. Kwong, AAPS Pharm.Sci.Tech. 6 (2005) E253.[38] L. St-Onge, J. F. Archambault, E. Kwong, M. Sabsabi and E. B. Vadas, J. Pharm. Sci. 8

(2005) 272.[39] A. K. Knight, N. L. Scherbarth, D. A. Cremers and M. J. Ferris, Appl. Spectrosc. 54 (2000)

331.[40] B. Sallé, J.-L. Lacour, E. Vors, P. Fichet, S. Maurice, D. A. Cremers and R. C. Wiens,

Spectrochim. Acta B59 (2004) 1413.[41] S. Morel, N. Leone, P. Adam and J. Amouroux, Appl. Opt, 42 (2003) 6184.[42] A. C. Samuels, F. C. DeLucia Jr, K. L. McNesby and A. W. Miziolek, Appl. Opt. 42

(2003) 6205[43] M. Baudelet, L. Guyon, J. Yu, and J.-P. Wolf, T. Amodeo, E. Fréjafon and P. Laloi, J. Appl.

Phys. 99 (2006) 084701.[44] C. Bohling, D. Scheel, K. Hohmann, W. Schade, M. Reuter, and G. Holl, Appl. Opt. 45(6)

(2006) 3817.[45] D. R. Alexander, Femtosecond LIBS and Second Harmonic Generation Techniques for

Detection of Chemical and Biological Agents, 2002 Merrill Conference, in A Merrill Centerpublication on the Research Mission of Public Universities.

Elsevier AMS Job code: LRN SubjectIndex-N51734 24-7-2007 8:12p.m. Page:427 Trim:165×240MM TS: Integra


Subject Index

Ablation, 49–79Aerosol, 381–412AES, 113, 151Aetiology, 327Angiosarcoma, 333Anthrax, 5, 316Anthropogenic, 341, 381–2API, 294–8, 301–302APXS, 364Atmospheric pressure effects, 70, 71, 78, 106,

145, 161, 164, 199, 229, 237, 362,369, 375

Atomic orbital, 26–7Autofocusing, 141Avalanche ionization, 52–3

Bioaerosol, 19, 316, 321, 410Biogenic activity, 374Bioparticle, 320Bioterrorism, 200, 322Blast furnace, 141, 271Blast wave, 59, 74, 191Blooming, 120Breakdown of gas, 10, 383–6Bremsstrahlung, 36–7, 69, 85, 86, 88, 99,

105, 137, 154, 233, 235, 250, 385Broadband LIBS, 187, 200Brominated polymers, 141

Cascade of ionization, 10Cavitation bubbles, 142CCD, 16–17, 24, 118–19, 120, 124, 216, 321,

329, 387, 398CEM, 127, 203, 209–13Chemometric data analysis, 148Chirped pulse amplification, 158, 178Combustion wave, 84Continuum emission, 61, 87Continuum radiation, 14, 88Coronal equilibrium (CE), 43Coulomb explosion, 51, 52, 54, 67, 154

Covariance, 348Cumulative intensity ratio (CIR), 319Cytometry, 330Czerny-Turner spectrograph, 18, 118,

292, 305

Daltons, 341Dead zones, 120, 124Debye radius, 40Degenerate, 26, 27, 29, 30Detonation wave, 11, 85Dichroic mirror, 114, 121, 210, 257Directionality, 4Disintegrant, 294, 295, 301Doppler profile, 43–4Droplet-free thin films, 152Dual-pulse LIBS, 125, 138–40

Echelle spectrometer, 18, 118–20Einstein’s coefficients, 30Electric dipole, 30–1Electric quadrupole, 32Electron temperature, 35–6, 41–3, 87, 90, 96,

105, 108Elemental fractionation, 67–9, 73Energy level, 23, 89, 123, 275EPA, 209, 210, 419–20Excipients, 297, 302Explosives, 314, 315, 325Explosive boiling, 65, 155Extraterrestrial analysis, 137, 148

F-sum rule, 35Femtosecond pulse, 8, 60–1, 75–8Fiber-optic probe, 121, 225, 349Filamentation, 359Fingerprinting of species, 319, 323Flux, 4, 11, 13, 34, 137Forbidden transitions, 32FRAS, 364Free-bound transition, 87



428 Subject Index

Frusted total internal reflection (FTIR), 356FWHM, 37, 43, 63, 157, 258

Gas monitoring, 127, 141Gated integrator, 15Gaussian profile, 37, 43Ghost line, 120Ground state, 26–7, 36, 69, 84, 94, 166,

264, 306

Haemangiosarcoma, 333HMX, 314Humification, 341Hund’s rule, 27

ICCD, 16–17, 94, 113, 118–19, 122, 193,200, 228, 257, 259

Impact approximation, 39–40Inter-combination lines, 32Internal standard, 47, 120, 201, 296, 299Inverse Bremsstrahlung, 10, 69, 85–6, 105,

154, 250, 384–5Irradiance, 4, 10, 51, 68, 76, 84–5, 97, 199,

228, 288, 357, 383–5Isochorically, 155

j-j coupling, 29

Keldysh parameter, 53Kirchoff’s law, 45

L-S coupling, 29Laporte rule, 32Laser fluence, 54, 154, 155, 159, 160,

162, 203Laser micromachining, 152Laser photo-fragmentation (LPF), 408Lead colloid, 142LEAF, 113, 409LIBS, 4–5, 15–16LIDAR, 124, 354, 356, 359Limit of detection (LOD), 107, 126, 141, 225,

232, 240–1, 251Line-strength, 33, 35, 36, 42Line to background ratio (LBR), 280Line to noise ratio (LNR), 275, 280LIP, 3, 15, 114, 116, 124, 130, 137, 142–3,

147, 153, 421Lorentzian profile, 37, 40, 91

LTE, 42, 43, 63, 88–90, 96, 162, 164,204, 261

Lunar missions, 364

Magnetic dipole, 32Mahalanobis distance, 327Mars science laboratory (MSL), 376Metastable level, 36Meteorite sample, 181, 188Microanalysis, 50, 114, 153, 168, 174, 181,

184, 187–8, 194, 410Microchip lasers, 173–4, 175–7, 177Mode-locking, 8Mode, 5–8Multi-elemental analysis, 118, 119, 148Multichannel detector, 17Multiphoton absorption, 52, 384Multivariate analysis, 344, 349, 350

Nanoparticles, 323, 391Near-field pattern, 7Nitroaromatics, 315Nuclear waste, 225

Oceanographic analysis, 148Off-gas emission, 200, 206, 420Optical biopsy, 329Optical breakdown, 155, 354Optical fiber, 121, 123–4, 206, 225, 255–8Optical multichannel analyzer, 117, 199, 318Oscillator strength, 33–5, 43, 165Osteoporosis, 329

Parity, 32Paschen-Runge spectrometer, 118Pauli’s principle, 27Periodic table, 18, 28Persistent lines, 45–7PETN, 314Pharmaceuticals, 143, 325Phase explosion, 65, 154–5Photo-diode array (PDA), 15, 16, 117, 118Photoconductive detector, 8Photoemissive detector, 8Phytoremediation analysis, 423Picosecond pulse, 7, 73–4Picture restoration, 152Pixel, 15–17, 114, 118, 120, 188Planetary geology, 362–3Plasma ignition, 52–7, 69, 383Plasma shielding, 51, 52, 56, 70, 97, 368



Subject Index 429

Pre-ablative spark, 139, 144, 147Principal component analysis (PCA), 317Process analytical technologies (PAT), 301Process optimization, 294Profilometry, 168

Q-switching, 7–8, 177, 242Quantum number, 26–9, 32, 44Quasi-static approximation, 40–1, 44

R-FIBS, 361–2R-LIBS, 117Radiation wave, 99Radiative recombination, 36, 37, 106,

230, 239RCRA metals, 140, 202, 203, 210, 213,

407, 419RDX, 314, 315Resonance line, 36, 92, 94, 224RSD, 130RSTD, 288–92Rydberg constant, 26

S/B ratio, 123, 231, 258–9S/N ratio, 116, 140, 201, 228, 282Saha’s equation, 42Scanning microanalysis, 186, 188–93Schrödinger equation, 25–6Selection rules, 30–2, 36Self-absorption, 71, 73, 92, 94, 97, 103,

105, 130SESAM, 176

Shock wave, 84–5, 146, 155, 181, 191,223, 421

Soil organic carbon (SOC), 341Spatial profile, 6, 93Spectrochemical LIDAR, 356Spectrometer, 24, 94, 104, 118–19Spin-orbit coupling, 29, 31Spontaneous emission, 30, 35, 37, 42, 49Stark broadening, 24, 37, 38, 44, 45, 91Stimulated emission, 30, 34, 35, 49Stochastic fluctuations, 118Stoichiometry, 313–15Surface ablation, 143

Teramobile, 361Thermistor, 9Time-gating, 15TNT, 314Toxic-metals monitor, 200, 206, 209, 213,

325, 387, 405, 407Transition of atom, 26Transition probability, 29, 35, 36, 65, 87, 106Tunneling ionization, 52, 53

Ulta-sonic nebulizer (USN), 201–203

Waste vitrification stimulant, 144Water-jacketed probe, 141Wavefunction, 25, 26, 29, 31Weisskopf radius, 39

X-ray fluorescence (XRF), 124, 226, 353,357, 364, 405

laser-induced breakdown spectroscopy

Documents