applied physics reviews—focused review secondary … · low vacuum scanning electron microscopy...

APPLIED PHYSICS REVIEWS—FOCUSED REVIEW

Secondary electron contrast in low-vacuum/environmental scanningelectron microscopy of dielectrics

Bradley L. Thiela! and Milos Tothb!

Polymers and Colloids Group, Cavendish Laboratory, Department of Physics, Madingley Road, Universityof Cambridge, Cambridge, CB3 0HE, United Kingdom

sReceived 28 July 2003; accepted 4 October 2004; published online 16 February 2005d

Low vacuum scanning electron microscopysSEMd is a high-resolution technique, with the ability toobtain secondary electron images of uncoated, nonconductive specimens. This feat is achieved byallowing a small pressure of gas in the specimen chamber. Gas molecules are ionized by primaryelectrons, as well as by those emitted from the specimen. These ions then assist in dissipating chargefrom the sample. However, the interactions between the ions, the specimen, and the secondaryelectrons give rise to contrast mechanisms that are unique to these instruments. This papersummarizes the central issues with charging and discusses how electrostatically stable, reproducibleimaging conditions are achieved. Recent developments in understanding the physics of imageformation are reviewed, with an emphasis on how local variations in electronic structure, dynamiccharging processes, and interactions between ionized gas molecules and low-energy electrons at andnear the sample surface give rise to useful contrast mechanisms. Many of the substances that can beexamined in these instruments, including conductive polymers and liquids, possess charge carriershaving intermediate mobilities, as compared to metals and most solid insulators. This can give riseto dynamic contrast mechanisms, and allow for characterization techniques for mapping electronicinhomogeneities in electronic materials and other dielectrics. Finally, a number of noteworthyapplication areas published in the literature are reviewed, concentrating on cases where interestingcontrast has been reported, or where analysis in a conventional SEM would not be possible. In theformer case, a critical analysis of the results will be given in light of the imaging theory put forth.© 2005 American Institute of Physics. fDOI: 10.1063/1.1861149g

TABLE OF CONTENTS

I. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . 1II. ELECTRON MICROSCOPY IN A LOW

VACUUM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2A. The instrument. . . . . . . . . . . . . . . . . . . . . . . . . 2B. Electron-gas scattering in a low-vacuum

SEM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31. Gas cascade amplification. . . . . . . . . . . . . . 32. Scattering of the primary beam. . . . . . . . . . 4

III. SECONDARY ELECTRON EMISSION FROMDIELECTRICS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5A. Intrinsic emission in the absence of

charging. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5B. Physical processes of charging and effects

on electron emission. . . . . . . . . . . . . . . . . . . . . 7C. The role of charge traps. . . . . . . . . . . . . . . . . . 8D. Time dependent behavior. . . . . . . . . . . . . . . . . 8

IV. IMAGING IN LOW VACUUM: THE ROLE OFIONS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

A. Space charge. . . . . . . . . . . . . . . . . . . . . . . . . . . 9B. Electron-ion recombination. . . . . . . . . . . . . . . 10

1. Spatial filtering. . . . . . . . . . . . . . . . . . . . . . . 112. Energy filtering. . . . . . . . . . . . . . . . . . . . . . 11

C. Schottky effect. . . . . . . . . . . . . . . . . . . . . . . . . 11D. Charge balance. . . . . . . . . . . . . . . . . . . . . . . . . 11E. Other ion effects. . . . . . . . . . . . . . . . . . . . . . . . 12

V. APPLICATIONS. . . . . . . . . . . . . . . . . . . . . . . . . . . . 12A. Ferroelectric domains. . . . . . . . . . . . . . . . . . . . 13B. Electronic devices. . . . . . . . . . . . . . . . . . . . . . . 13C. “Charge contrast imaging”. . . . . . . . . . . . . . . . 13

1. High-vacuum contrast. . . . . . . . . . . . . . . . . 142. Electron flux density. . . . . . . . . . . . . . . . . . 143. Pressure and anode bias dependence. . . . . 14

D. Electronic polymers. . . . . . . . . . . . . . . . . . . . . 15E. Liquids. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16F. X-ray microanalysis. . . . . . . . . . . . . . . . . . . . . 16

VI. SUMMARY AND OUTLOOK. . . . . . . . . . . . . . . . 17

I. INTRODUCTION

A major driving force behind the development ofenvironmental/low-vacuum scanning electron microscopes

adPresent address: College of Nanoscale Sciences and Engineering,University at Albany, Albany, NY 12203; electronic mail:[email protected]

bdPresent address: FEI Company, 1 Corporation Way #2, Peabody, MA01960.

JOURNAL OF APPLIED PHYSICS97, 051101s2005d

0021-8979/2005/97~5!/051101/18/$22.50 © 2005 American Institute of Physics97, 051101-1

http://dx.doi.org/10.1063/1.1861149

http://dx.doi.org/10.1063/1.1861149

sSEMsd was to enable secondary electronsSEd imaging ofuncoated insulators. In actuality, these microscopes do notstop charging processes, but rather stabilize them and simul-taneously eliminate many of the undesirable artifacts associ-ated with charging. When imaging conditions are carefullycontrolled, dynamic charging processes can be exploited toreveal spatially resolved information on the electronic inho-mogeneities in dielectric substances. As this technique ma-tures, it is emerging that complex interactions between high-and low-energy electrons, the weakly ionized environmentalgas, and the specimen can give rise to contrast mechanismswhich are potentially useful and unique to these instruments.This paper reviews and summarizes recent progress towardsunderstanding the various phenomena involved in low-vacuum imaging and then uses this foundation to interpretseveral examples of “anomalous” contrast reported in theliterature.

There is an unfortunate tendency in the literature to dis-tinguish high-vacuum, low-vacuum, and environmental SEMsolely as pressure regimes. This often leads to confusion withtrademarked names and acronyms used by manufacturersse.g., operating an ESEM® in low-vacuum moded. For clar-ity, the definition of terms used throughout this review isgiven here. The term “low vacuum” is applied when the gasperforms an electronic role, such as charge stabilization orsignal amplification. The term “environmental SEM” is usedspecifically in situations where the gasprimarily performs athermodynamic role, such as preventing the evaporation ofliquids from the specimen or initiating chemical reactions.Naturally, the two functions are not mutually exclusive, norare their pressure ranges. Alternatively, in the language ofphysical chemistry, low vacuum is appropriate when theab-solutegas pressure matters, and “environmental” applies toconditions where thepartial pressure of gas is relevant.Within this review, the term low vacuum is also used whengeneric issues are being discussed such as electron-gas inter-actions or practical/engineering considerations.

The simplified description of low-vacuum imaging putforward in most literature is as follows: If a small amount ofgas is fed into a SEM specimen chamber, positive ions arecreated via collisions between electrons and gas molecules.When the ion current reaching the specimen exactly offsetsthe rate of negative charge accumulation, charge balanceconditions are said to exist. Expressions describing each ofthese contributions can be summed together to provide anequation for charge balance.1–6 The parameter space is con-siderable, with dependencies on gas pressure, gas chemistry,primary beam energy and current, working distance, and thesecondary and backscattered emission coefficients for thespecimen among other things. However, real situations areeven more complicated because of the complex charge dis-tributions in and above the specimen, and because in fact,under typical operating conditions, the ion generation ratecan exceed the rate of negative charge accumulation by up tothree orders of magnitude.

A number of misconceptions surround the notion ofcharge neutralization. Usually, prescriptions for working un-der charge-neutral conditions actually achieve a state wherethe specimen surface potential is close to zero. This is true

for both high- and low-vacuum conditions. It is more accu-rate to describe these operating conditions as a regime inwhich the effects of charging on electron imaging are mini-mized. In fact, significant amounts of charge of both signsmay be present simultaneously, as will be discussed in detail.

The present review summarizes recent efforts to explainexactly how stable imaging conditions are achieved. It thengoes on to provide a description of the SE contrast mecha-nisms and imaging considerations relevant to working in alow-vacuum environment. The discussion on contrastmechanisms is divided roughly into three parts:sid SE gen-eration andintrinsic SE emission from dielectricssi.e., in theabsence of chargingd, sii d information present in SE imagesdue to localized charging, andsiii d contrast mechanisms re-sulting from interactions between ionized gas molecules andlow-energy electrons at/near the specimen surface. A fewselect applications are then reviewed, where contrast effectsare particularly interesting. Finally, we note that for many ofthe materials that can be examined only in low-vacuum in-strumentsse.g., liquidsd the charge carriers have intermediatemobilities, as compared to those of metals and common solidinsulators. This property can give rise to electron-flux-density-dependent contrast, which may be useful for thecharacterization of electronic inhomogeneities. Although thistechnique has not yet been developed in earnest, a few strik-ing examples are reviewed that highlight one of the uniquecapabilities of low-vacuum microscopy.

This review concentrates on peer-reviewed literature.However, much of the academic discussion has taken placein scientific meetings. In particular, the proceedings of theMicroscopy Society of America’s annual meeting, Micros-copy and Microanalysis, contain reports on a wide variety ofapplications. For the most part, these proceedings are ex-tended abstracts that report anomalous contrast effects or areprogress updates on efforts to understand various phenom-ena. Both categories are encompassed by the refereed paperscited here. Most of the earlier work and references can befound in the numerous papers authored by Danilatos, citedthroughout this review. A particularly good compilation ofearly applications can be found in a dedicated issue of Mi-croscopy Research and Technique.7 That same volume alsocontains a very extensive bibliography of early literaturecompiled by Danilatos.8 Finally, two reviews of general ap-plications with an emphasis on polymeric, hydrated, and liq-uid materials have been published recently.9,10

II. ELECTRON MICROSCOPY IN A LOW VACUUM

A. The instrument

In general, because this technique is still unfamiliarcompared to conventional high-vacuum scanning electronmicroscopes, most applications papers are written with abrief description of the instrument and its operation. As such,the present discussion will not delve into great detail on thesubject. For more detailed information, the reader shouldconsult one of the original references. In particular, Danilatosand Postle11 and Danilatos12–15 discuss in great detail thefundamental aspects underlying the design of modern com-mercial instruments.

051101-2 B.L. Thiel and M. Toth J. Appl. Phys. 97, 051101 ~2005!

The two major considerations in designing a microscopeto work under low-vacuum conditionsstypically, pressure=10–1300 Pad aresid isolating, insofar as possible, the low-vacuum region surrounding the specimen from the high-vacuum region comprising the electron optical column andgun; andsii d providing a means of detecting signals emittedfrom a specimen in a low-vacuum environment. Creating alarge pressure gradient between the specimen chamber andthe electron gun is readily achieved through the use of dif-ferential pumping zones, separated by pressure-limiting ap-ertures as depicted in Fig. 1. The Danilatos references in thepreceding paragraph discuss this in detail. The second pointis discussed in Sec. II B.

B. Electron-gas scattering in a low-vacuum SEM

1. Gas cascade amplification

Conventional Everhart–Thornley-type secondary elec-tron detectors use an intense electric field that causes the gasto break down at the pressures employed in low-vacuumSEM. To overcome this problem, most low-vacuum detectorsinvolve accelerating the emitted secondary electrons with amoderate electric field using a geometry similar to thatshown in Fig. 2. When the kinetic energy of an electronexceeds the ionization threshold of the gas, an ionizing col-lision can occur, creating a positive ion and an additional freeelectron. Both electrons are then accelerated by the field andthe process repeats, giving rise to a gas ionization avalanchethat amplifies the original emission signal. Over the gap dis-tanced from the specimen to the anode, the original currentwill be amplified by a gain factorg=expsadd. Townsend’sfirst ionization coefficienta gives the number of electron-ionpairs produced per unit length traveled by a low-energy elec-tron moving through the gas under the influence of a mod-erate electric field. In the simplest case, this arrangement isanalogous to a Townsend gas capacitor. Hence,a is a func-tion of gas pressureP and anode biasVa, and is given nomi-nally by16

a = C1Pe−C2Pd/Va. s1d

The coefficientsC1 and C2 are coupled to the overlap inte-gral of the energy distribution of the cascading electrons andthe ionization and total inelastic scattering cross sections ofthe gas, respectively. Accordingly, tabulated values of thecoefficients for different gases are reported for steady-state

or “swarm” conditions, where the average energy gain perunit length of electrons moving through the field is exactlyequilibrated by the stopping power of the gas.16 Thepressure-field combinations used in low-vacuum SEMss10–1300 Pa; 50–500 V/mmd typically do not allowsteady-state conditions to develop quickly. Thus, most of thecascade takes place in a regime where tabulated values ofC1

and C2 are not valid. Monte Carlo simulations have beenused to characterize the amplification behavior under theseconditions and suggest corrections.6

The flux of gaseous ions incident onto the specimen is atthe heart of the imaging phenomena that will be discussedhere. Accordingly, it is necessary to consider all the sourcesof ion production and how they depend on microscope oper-ating conditions. In addition to the secondary electrons emit-ted from the sample surface, ionization cascades can be ini-tiated by ionization events of the primary beam as it travelsthrough the gas to the specimen, as well as by backscatteredelectrons. These contributions are described in detail in anumber of papers.1–6,12,17,18The number of cascade eventstriggered by primary electrons ionizing gas molecules isgiven by the ratio of the distance that the beam travelsthrough the gasl si.e., the gas path lengthd to the ionizationmean-free-path. For the geometry shown in Fig. 2,l =d, al-though this will not be true for other configurations. Becausethe initiation points of these cascades will be distributed overthe entire gas path length, the total amplified cascade currentmust be obtained by integrating overd. As such, for a pri-mary beam current ofI0 the cascade currentIP due to thissource is given by

IP = I0SPEPl

adsead − 1d, s2d

whereSPE is the inverse mean-free-path of the primary elec-trons. A similar expression for the backscattered electrons isobtained by scaling Eq.s2d by the backscattered emission

FIG. 1. Schematized vacuum system of an environmental SEM. Pressure-limiting apertures separate regions of the column that are differentiallypumpedsRP=roughing pump; DP=diffusion pump; IP=ion pumpd.

FIG. 2. Idealized cross section of the region surrounding the specimen forone possible configuration. The cascade amplification field is determined bythe potential difference between the anode and specimen surface, divided bythe gap distanced. The anode can be a deliberately biased detector, such asthe gaseous SE detectorsGSEDd. In other configurations it is simply thegrounded pole piece, in which case, the signal can be collected from the stubsVa=anode bias, typically in the range of +300– +600 V; wd=working distanced.


coefficient of the sample, and substituting a suitable valuefor SPE that reflects the angular and energy distributions ofthe backscattered electrons.

Finally, Auger and photoelectrons can be generated dur-ing neutralization and deexcitation of the ions. These low-energy electrons will be multiplied by the cascade process,resulting in still more ions. Fortunately, this process can of-ten be neglected. It is discussed in more detail in the sectionon Imaging in Low Vacuum: The Role of Ions.

Putting all of the above terms together, the total electroncascade currentsand by symmetry, the ion fluxd for the ge-ometry of Fig. 2 is given approximately by6

I = I0kFd +hSBSEP

a+

SPEP

aGead, s3d

where d and h, respectively, represent the secondary- andbackscattered emission coefficients of the specimen, andk isa factor taking into account the secondary ionization pro-cessessAuger and photoelectron emissiond. For a given gas,the SE and the sum of the primary and backscattered com-ponents of Eq.s3d can be obtained experimentally, and theirdependencies onP and the gas-amplification field strengthcan be plotted onto a single master curve.18 Such curves canthen be used to compare gas-amplification characteristics un-der different conditions and of different gases in a meaning-ful manner.

2. Scattering of the primary beam

An inevitable consequence of having gas in the pressurechamber is that some of the primary beam electrons will bescattered out of the focused probe by gas molecules. In thepressure regime of concern here, scattering results in the cen-tral probe being surrounded by a diffuse “skirt” of low cur-rent density that can extend over several hundred microme-ters. The gas path length is only a few millimeters, whereasthe mean-free-path for scattering of high-energy electrons bygas moleculesfshown in Fig. 3sadg can be several millime-ters under typical operating conditions.19 Accordingly, thereare two distinct populations of primary electrons: those thathave been scattered and those that have not. This is the“oligo-scattering” regime described by Danilatos andPostle.11 The scattered fraction of the beam is distributedover an area that is usually quite large compared to the im-aging area. Conversely, the spatial distribution of the unscat-tered electrons remains defined by the electron optics. Thenet result is a high-resolution signal originating from theunscattered portion of the beam riding on a nearly constantbackground signal arising from the skirt.

Contrary to widespread misconception, scattering in thegas results in only a minor deterioration of the resolution.Figure 4 shows a series of gold-on-carbon resolution testimages taken at high vacuum, 665 Pas5 torrd and 1330 Pas10 torrd. Resolution was measured using the cross correla-tion function method in the scanning microscope analysisand resolution testingsSMARTd macro plug-in for theScion

Image analysis software.20,21 With these measurements, reso-lution decreased from 1.54 nm with high vacuum to 2.35 nmat 665 Pa and 2.85 nm at 1330 Pas10 torrd of water vapor.

Assuming the contributions to resolution add in quadrature,the resolution limit imposed by the gas broadening is1.77 nm at 665 Pa and 2.4 nm at 1330 Pa. Even so, much ofthe resolution degradation can be attributed to the fact that athigh pressures, the image becomes dominated by signals de-

FIG. 3. sad Inverse mean-free-pathsi.e., scattering events per millimeterd asfunction of electron energy for various gases at 100 Pa. Helium–dashed line;water vapor–solid line; nitrogen–dotted line. The information is presented inthis manner because the inelastic mean-free-pathsIMFPd scales linearly withpressure. For example, doubling the pressure halves the mean-free-path.sbdMean number of scattering events experienced by a primary electron en-route to the specimen as a function of the pressure-distance product in watervapor, computed for various electron energies. Assuming Poisson-scatteringstatistics, an average number of scattering events equivalent to unity corre-spond to 1/e of the electrons reaching the specimen unscattered. As a rule ofthumb, therefore, the shaded region represents the parameter space in whichhigh-quality images can be obtained.

FIG. 4. A gold-on-carbon resolution standard imaged at high vacuum andwith 665 Pas5 torrd and 1330 Pas10 torrd of water vapor. A conventionalEverhart–Thornley SE detector was used for the high-vacuum image, whilea gaseous SE detector was used in low-vacuum mode. In all cases, theaccelerating voltage=20 kV, with wd=7.5 mm. Resolutions measured usinga cross correlation procedure were 1.54, 2.35, and 2.85 nm, respectivelysRef. 20d. Signal-to-noise ratios were 7.57, 3.9, and 1.83, respectively. Me-chanical vibrations are present in the images. However, these have a mini-mal effect on the measured changes in resolution with gas pressure. 665 Paof water vapor is the minimum pressure necessary to stabilize hydratedspecimens. Images courtesy of Daniel Phifer, FEI Company.


rived from backscattered electrons.5,6 sThe authors wouldlike to emphasize that resolution broadening due to gas is afunction of the gas pressure, gas type, gas path length, andbeam energy. The values reported here are to demonstratequantitatively the effects of changing pressure, and shouldnot be taken as absolute values for resolution at the specifiedpressures.d It is worth noting here that in fact, low-vacuumSEM is being considered as a tool for critical dimensionmetrology in the semiconductor industry, where image reso-lution sand reproducibilityd is paramount.22

In truth, the limiting factor in image quality is the de-creasing signal-to-noise ratio under conditions of strong scat-tering. The SMART macro also returns a value for this met-ric. The signal-to-noise ratio measured from these imagesdecreased from 7.57 at high vacuum to 3.9:1 at 665 Pa and1.83:1 at 1330 Pa.sN.B. The signal-to-noise values for high-and low-vacuum cases should not be compared directly, asdifferent detectors are used.d A statistical analysis of scatter-ing probabilities, such as that presented for water vapor inFig. 3sbd, can be used to guide the selection of imaging pa-rameters such as pressure, gas path length, and beam energy.

Moncreiff et al. undertake a thorough analysis of skirtformation and provide detailed equations describing thebeam loss and intensity distribution as a function of pressure,gas path length, primary beam energy, and gas type.23 Sub-sequent papers have not improved significantly on this de-scription, and have been concerned mostly with the experi-mental verification of the intensity profiles. Experimentalefforts to measure the intensity profile have met with littlesuccess. Most of the approaches involve measuring somesecond-order effect, and eventually become mired in uncer-tainties of interpretation. Furthermore, all efforts are hin-dered because of the huge dynamic range of the electronflux. Even in conditions where a significant fraction of thebeam is lost to the skirt, Monte Carlo simulations suggestthat the current density in the central probe can be over 107

times greater than that found in the skirt at a radius of100 mm. To date the best experimental measurements re-ported are those by the group at NIST, who exposed self-assembled monolayer films to the beam, and then recordedthe damage using scanning secondary-ion-mass spectrometrysSIMSd.24,25 Wight, of the NIST group, gives a good over-view of the skirt phenomenon and compares several experi-mental techniques.26 In summary, the skirt effect represents aminor inconvenience for imaging. Conversely, it is a topic ofconsiderable concern for the quantification of x-ray spectra.This will be addressed in a later section.

A note of caution regarding the use of Monte Carlosimulations is appropriate here. Frequently, Monte Carlosimulations are used to justify various claims about the per-formance characteristics of low-vacuum SEMs, particularlywith regard to skirt formation, or for interpreting experimen-tal results. However, many of these algorithms use the “con-tinuous slowing down approximation.”27 In this model, elec-tron trajectories are determined by elastic scattering only,and energy is lost continuously between scattering events ata rate determined by the stopping power.28 Although thesesimulations are extremely powerful tools for providing in-sight into many problems, their treatment of a gas as a low-

density solid is not appropriate, as the electrons clearly can-not lose energy between discrete collisions. Furthermore,discrete ionizing collisions between electrons and gas mol-ecules also produce an angular distribution of trajectories,distinct from elastic-scattering events.19 For electron energiesin the keV range, the ionizing cross section is about twice aslarge as the elastic cross section for most simple gases.19

Thus, the above-mentioned models will not yield correct pre-dictions of the fraction of electrons scattered by gas mol-ecules, nor their energy and spatial distributions. Any correc-tion algorithms based on these models are also likely toproduce unsatisfactory results.

III. SECONDARY ELECTRON EMISSION FROMDIELECTRICS

Secondary electron contrast mechanisms accessible inlow-vacuum imaging of dielectrics fall into two categories:processes that are intrinsic to the nature of the dielectric, butare ordinarily obscured by charging effects, and extrinsicprocesses that result from the presence of gaseous positivelycharged ions. Some discussion of the former is necessary inorder to consider how these processes are altered in a low-vacuum environment.

A. Intrinsic emission in the absence of charging

An intriguing aspect of low-vacuum microscopy is thatin the absence of a metal coating, the SE emission from adielectric must depend on the electronic structure of thesample. While this may seem like an obvious statement, acomprehensive, predictive theory for SE emission of dielec-trics has not been developed. However, regardless of the de-tails, there are three aspects to SE emission from any sub-stance, and it is the local variations in these that give rise tocontrast. The three processes aresid generation of excitedelectrons through dissipation of the primary electron energy,sii d transport of the excited electrons to the surface, andsiii descape from the surface. Each of these is now discussedbriefly and contrasted to the simpler case of metals.

In both metals and insulators, the energy deposition pro-file of an electron traveling through a material with atomicnumberZ, molecular weightW, and densityr can quite ac-curately be predicted with the Bethe model,27 particularlywith recent improvements.29 Kanaya and Okayama consid-ered the randomizing effect of elastic scattering in conjunc-tion with the Bethe model to estimate the maximum rangeRmax of an electron in condensed matter. Their expressionyields the depth of the interaction volume as a function ofmaterial parameters and beam energy according to30

Rmax=0.0276WE0

1.67

Z0.89r. s4d

To a first approximation, energy deposition models canbe used to calculate secondary electron generation profilesby incorporating them into Monte Carlo simulations of elec-tron transport through solids. Figure 5 shows examples ofenergy deposition profiles in sapphire, calculated with theCASINO program,31 for primary electron energies of 1, 5, 10,and 30 keV.32 Also indicated in the figure is the typical range


of depths from which SEs are emitted in the case of uncoatedinsulators. The data illustrate that at the beam energies em-ployed in low-vacuum SEMs, the vast majority of excitedelectrons are generated well below any reasonable escapedepth. Consequently, the physics of SE transport through thebulk plays an important role in SE emission and is discussedin detail below.

A useful figure of merit for describing electron transportin a given material is the inelastic mean-free-pathl, an av-erage quantity determined by the magnitudes of all the pos-sible mechanisms through which an excited electron can loseenergy. For metals, the continuum of empty states above theFermi level allows this problem to be treated with simplemodels. However, the presence of discrete levels and a for-bidden energy gap in a dielectric require a more rigorousapproach.

From dielectric theory,l can be obtained from theimaginary part of the complex dielectric-response function«sq,vd,33–35

l−1sE,vd =1

pEE dvE

−q

+q Sdq

qDImf− 1/«sq,vdg. s5d

In the above,E is the electron energy, andq and"v are themomentum and energy transferred in an inelastic scatteringevent. The mean-free-path of electrons, and consequently thenumber of electrons reaching the surface, are therefore deter-mined by the factors that contribute to«sq,vd. These factorsinclude the size of the forbidden energy gap and the elec-tronic excitation spectrum.36 The final step is to convert Eq.s5d into an expression for the energy-loss rate, following theapproach of Ashley, for example.33

In theory, Eq.s5d could be very useful in interpretingcontrast from multiphased dielectric systems if the relevant«sq,vd functions were known. Unfortunately, this is seldomthe case for most nonmetals. Data in the optical limit, asq→0, are more readily available from such techniques as pho-

toabsorption. Groups at the Oakridge and NIST laboratorieshave given detailed discussions of how Eq.s5d can be modi-fied to give a reasonable approximation for theq.0 regimefrom «s0,vd data obtained experimentally.33–35 Subsequentwork has expanded this to include low- and subexcitationenergies.37

An energetic electron can lose energy via a number ofmechanisms such as the excitation of inner- and outer-shellelectrons and plasmons. However, once the kinetic energy ofthe electron falls below the thresholdEt for electron-holesorexcitond excitation, Imf−1/«sq,vdg vanishes. Energy canthen only be lost at a relatively low rate via scattering fromphonons and defects.

Howie provides a convenient means for estimating thethreshold energyEt for electron-hole pair production.38 Con-sidering energy and momentum conservation, he shows thatEt=s1+gdEg, whereEg is the band-gap energy. The coeffi-cient g reflects the effective masses of the electrons andholes and is defined asg=me/ sme+mhd for parabolic bands.Thus, to a first approximation,Et=1.5Eg. This implies thatl,and hence the maximum SE escape depth, in materials withlarge forbidden energy gaps should be considerably greaterthan the few nanometers associated with metals. Indeed, it iswell-known that the maximum SE emission coefficients fromdielectrics are considerably larger than for metals.39,40 Inshort, the existence ofEt gives rise to the possibility of elec-tronic structure contrast based on the size of the forbiddenenergy gap in the material. All other considerations beingequal, a substance with a large energy gap should have ahigher emissivity than one where the gap is smaller. Thismechanism has been used to explain the contrast from cer-tain water-oil emulsions as discussed in the Applicationssection.41 Another implication is that contrast may arise fromdielectric inhomogeneitiesse.g., defect or impurity clustersdwithin a single domain of a material.

A number of groups have developed sophisticated MonteCarlo simulations for SEM conditions which treat inelastic-scattering events explicitly by calculating the individualmean-free-paths for all the known electron energy-lossmechanisms.42–45Such models can be used to calculate morerealistic SE generation profiles, as well as energy spectra andangular distributions, and simulate SE transport through thebulk. However, to the best of our knowledge, in the case ofinsulators, accurate calculations have only been performedfor SiO2.

44,45For a very comprehensive review of the behav-ior of low-energy electrons in dielectrics, the aptly titledbook “Excess Electrons in Dielectric Media” edited by Fer-randini and Jay-Gerin should be consulted.46

Finally, it is necessary for the electrons to overcome anypotential barrier at the surface in order to be emitted. Again,the case of a metal is straightforward. The kinetic energyassociated with the component of the electron velocity that isnormal to the surface must exceed the work function of thatsurface.39,47 A larger work function, therefore, gives loweremissivity. Similar considerations apply to dielectric materi-als, but with the added complication of band bending at thesurface.38 Most dielectric materials have work functionssmaller than those typical for metals, which is another con-tributing factor to their high emissivities. Adsorbed gases

FIG. 5. Primary electron energy deposition profiles calculated for Al2O3

using the Monte Carlo programCASINO sRef. 31d. The simulations wereexecuted using initial electron energiesE0 of 1, 5, 10, and 30 keV, as shownin the figure. The curves are approximately proportional to SE generationprofiles. The shaded area represents the size of the SE escape region typi-cally encountered in the case of dielectrics in a SEMsRmax= maximumprimary electron penetration ranged.


also affect the secondary electron emission characteristics byaltering the escape barrier. Hence, measurements of electronemission profiles are performed under high vacuum. Someconcerns have been voiced as to the extent to which ad-sorbed gases in a low-vacuum environment will affect the SEsignals. However, investigations into this problem have notbeen reported to date.

Having relatively large SE mean-free-paths and smallescape depths imply that the SE emission, and hence contrastin images of dielectrics, will be very sensitive to extrinsiceffects that alter these. There is already evidence that underhigh-vacuum conditions, subtle variations in work functionand band bending can give rise to contrast in dielectricsystems.38,48 Particularly intriguing are the reports of dopantcontrast in semiconductors.49,50 Even if dopant concentra-tions are on the order of 1017 atoms cm−3, appreciable con-trast can be obtained.

B. Physical processes of charging and effects onelectron emission

The mechanisms and consequences of charging inelectron-irradiated dielectrics have been a topic of consider-able interest since the development of electron-beam instru-ments. Although many papers have been written, some of thebest descriptions of charging and its effects on electron emis-sion are to be found in the series of papers written byCazaux.51–54 He has compiled much of the known experi-mental data and put forward very elegant descriptions of thespatial- and time-dependent behaviors of charging for a setof idealized geometries and experimental conditions. In par-ticular, equations derived for the electric fields resulting fromvarious charge distributions, and how these affect electronemission and specimen surface potential. Other particularlyrelevant discussions of dynamic charging in electron-irradiated insulators can be found in the papers by Glav-atskikhet al.,55 Melchinger and Hofmann,56 and Ganachaudand Mokrani.57

The electric-field strength at any point above or belowthe specimen surface can be found by considering the distri-bution of charges in the system and solving the Poissonequation with appropriate boundary conditions. Cazaux de-scribes the relevant calculations for a variety of electron-beam applications.51 In the case of an uncoated insulator ona grounded support in high vacuum, there are two chargedistributions to be considered. The negative charge im-planted by the primary electrons has its center of gravity at adistance below the surface on the order of the primary elec-tron range,Rmax, given by Eq.s4d. Shown in Fig. 5 isRmax

calculated for Al2O3, irradiated using a number of beam en-ergies. Although the penetration depth has a strong depen-dence on the primary beam energy and the stopping power ofthe material, this distance is usually in the micrometer rangefor E0.5 keV. Secondary electron emission, conversely,originates quite close to the surface, with the SE escapedepth being on the order of nanometerssshaded region inFig. 5d. The surface layer is therefore depleted of electronsand the residual holes constitute a distribution of positivecharge. As is shown in Fig. 6, an internal dipole field iscreated between the centers of positive and negative charge,

which causes the upward drift of any electrons in the con-duction band and the downward drift of holes in the valenceband.

The electrostatic field resulting from the implanted elec-trons retards the primary electrons. In the case of an un-coated insulator, this effect can be expressed most easily interms of the surface potentialVs that develops. Primary elec-trons with kinetic energyE0 begin slowing down before theyencounter the surface such that they impact with a landingenergy ofE0−eVs. It is clear from Eq.s4d and the curvesshown in Fig. 5 that reducing the landing energy will fore-shorten the interaction volume and reduce the penetrationdepth of the beam. Accordingly, more secondary electronswill be generated within the escape depth, resulting in theincreased SE emission traditionally associated with “nega-tive charging.”

Coating the specimen with a grounded metal pins thesurface potential to zero and eliminates the gross image dis-tortions. However, electrons are still implanted in the speci-men, meaning subsurface charging must still be occurring.There are two main considerations in which this situationdeviates from the uncoated case described above. First, thefield caused by the implanted charge terminates on the coat-ing, and does not extend out into the vacuum. In fact, belowthe surface, the potential gradient is steeper, as the field ter-minates at the surface, rather than at the much more distant

FIG. 6. sad Schematic of the charge distribution encountered in the case ofan electron-beam-irradiated insulator in a high-vacuum environment. It con-sists of an electron-depleted near-surface regions,10 nmd and an under-layer of implanted electronss,103 nmd. A free electronsrepresented by theparticle in the middled drifts towards the surface under the influence of thefield. sbd Simplified electronic structure diagram corresponding to the situ-ation in Fig. 6sad. Negative charge implanted in the bulksnot shownd createsan electric field represented by the slope of the energy bands. Charge trapsare indicated by dashed lines inside the forbidden energy gap. Trapped elec-trons can be excited into the conduction bands1d after which they are free todrift towards the surface, losing energy through phonon scatterings2d. Elec-trons can either be retrappeds3d or eventually reach the surfaces4d, wherethey can recombine with excess holes within the SE escape region, or gas-eous ions at the surfacesin the case of low-vacuum SEMd. Only electronswith energy greater thanEb can be emitted as secondary electronssEv= topof valence band;Ec= bottom of conduction band;Evac=vacuum level;Eg

=forbidden energy gapd.


pole piece. Certainly, the steady-state concentration of im-planted electrons must be reduced, as the greatly increasedinternal field will help detrap and sweep excess electrons tothe grounded metal surface. Second, nearly all secondaryelectrons are emitted from the metal layer, eliminating anymaterial-dependent SE contrast mechanisms.

C. The role of charge traps

When excess electrons are injected into a perfect dielec-tric material, they initially must reside in the conductionband as all lower energetic states are filled. As this band isessentially empty, these electrons are free to move under theinfluence of an external field, and hence, conduct. All realmaterials contain a variety of charge traps that can bind elec-trons and holes. Traps are states created inside the forbiddenenergy gap by structural defects such as vacancies, disloca-tions, or impurities.46 Traps can either be preexisting defectsor those created by the irradiation process itself. In terms ofthe electron-transport model of Eq.s5d, traps act as addi-tional scattering centers and add new oscillator states to thecomplex dielectric-response function. Thus, when electronsmove through a material, they lose energy until their energyis low enough for them to be captured by traps. As thetrapped electrons are no longer free to move, the substancewill accumulate charge. Accordingly, traps decrease conduc-tivity, and a higher density of traps will allow a region tostore more charge at any given instant.sA very high densityof low-energy traps can give rise to an impurity conductionband within the forbidden energy gap.58 Such a region wouldexhibit enhanced conductivity and reduced charge storage.59d

An electron can be excited out of a trap by thermal fluc-tuations or some other event that imparts sufficient energy tothe electron. Accordingly, the trap depth—the amount bywhich the potential energy of the trap falls below the bottomof the conduction band—is important as well. It follows thatelectrons can escape from energetically shallow traps morefrequently than deeper traps, leading to the concept of trap-ping lifetimes. When an electron is detrapped, it will thendrift under the influence of the field created by other trappedelectrons until it meets one of three fates: recombination witha hole, capture by another trap, or diffusion to the surfacewhere it can be removed via surface conduction or, in thecase of low-vacuum SEM, by recombining with a gaseousion. These processes are illustrated in Fig. 6. Real specimenswill contain both deep and shallow traps. In such cases, theelectrons in deep traps will contribute to an electric field thatassists in preferential detrapping of electrons from shallowtraps. This concept is developed further in the section onCharge Balance.

D. Time dependent behavior

Finally, the dynamics of charging in a SEM must beconsidered. Based on the foregoing discussions, trappedcharge increases SE emission, which translates to a decreasein the rate at which additional charge is trapped. When thebeam moves away from an area, the trapped charge conductsaway at some rate before the beam revisits. If the charge candecay at a sufficiently high rate, the imaging conditions are

stable. Otherwise, charge will accumulate between scans un-til a different equilibrium state is achieved. Scan-rate depen-dencesor more accurately, flux dependenced has been under-stood for some time. Van Veld and Shaffner60 and Shaffner61

provide a phenomenological analysis of these dynamic pro-cesses that can be very helpful in interpreting experimentalbehavior. The charge implanted per unit areas0 during asingle pass of the beam is given by

s0 = I0s1 − d − hdF/A, s6d

whereF is the period time per frame andA is the area beingimaged. Clearly, the amount of implanted charge can be af-fected equivalently by changing beam current, magnification,and scan rate. The implanted charge decays exponentiallyaccording to

sstd = s0e−t/t, s7d

where t is a characteristic time constant for charge decay,and is equivalent to the ratio of conductivityg over permit-tivity « of the charged region. Any local variation in either ofthese material-specific parameters potentially can give rise todifferential charging behavior and, subsequently, contrast.On the length scales of interest in electron images, the valuet0 for a bulk, marcoscopic material will be altered locally bythe presence of defects and impurities,si.e., trapsd and willeven change during irradiation via the phenomenon ofradiation-induced conductivitysRICd. Any of these varia-tions can be represented by a change in the electronic prop-erties g and «. Consequently, at each pointsx,yd on thesample surface,tsx,yd must reflect all conduction processes,and sums as

1

tsx,yd=

1

t0+

1

tRICsx,yd+

1

tasx,yd+

1

tbsx,yd+ ¯ , s8d

wheretasx,yd, tbsx,yd, etc., represent time constants associ-ated with the local defect/trap population. To provide a feelfor the time scales involved, for a good conductor such ascopper,t has a value,10−20 s, whereas for an insulator suchas quartz,t0 is on the order of 400 s.60 In comparison, typi-cal frame times in a scanning electron microscope are in therange 10−1–102 s. The papers by Van Veld and Shaffner andby Shaffner provide a formalism for describing the chargecondition of a specimen as it evolves over successive scanframes.

IV. IMAGING IN LOW VACUUM: THE ROLE OF IONS

The physics of secondary electron imaging in lowvacuum is dominated by the behavior of gaseous ions. Thisis true to such an extent that the SE-derived images obtainedin these instruments need only bear a superficial resemblanceto those obtained in high-vacuum conditions. Ions affect thedetected signal through several mechanisms: perturbing thecascade amplification field and the landing energy of the pri-mary beam by the accumulation of space charge at and abovethe sample surface; damping the SE signal by recombiningwith emitted electrons; enhancing SE emission by alteringthe surface energy band structure; influencing the charge bal-


ance of the specimen; and limiting detector response time bytheir drift velocity. The seminal points of each of these pro-cesses will be discussed.

A. Space charge

The term space charge has multiple implications stem-ming from its use in several disciplines. In the present con-text, space charge describes the distribution of positive ionsin the gap region at and above the specimen surface. Thisdefinition is an adaptation of the space-charge concept asgiven in standard texts treating gas-filled capacitors.16,62,63

sMuch of the relevant physics underlying the phenomena inlow-vacuum SEM can be found in these sources.d Althoughthe majority of ions are created near the anode,6 they driftthrough the field towards the sample surface, where somefraction may be pinned by the electric field generated byimplanted electrons, and mutual repulsion causes the remain-der to drift sideways. The concentration maximum is justabove the surface.

The ionic space charge, together with the spatial distri-bution of subsurface trapped electrons and holes, determinesthe electric-field structure between the grounded specimenstub and the anode. General features of the electrostatic po-tential for this arrangement are shown in Fig. 7. A key fea-ture is the dipole field that is created between the centers ofmass of the negative chargesimplanted electronsd and theionic space charge. The specimen surface lies inside this di-pole field, shifting the surface potential away from ground byan amountVs. Provided that there is sufficient ion current,dynamic charge balance conditionssdiscussed in the ChargeBalance sectiond tend to limit Vs to a few hundred voltsspositive or negatived, and shifts the landing energy of pri-mary electrons accordingly.64

In addition to affecting the landing energy, space chargehas a strong influence on cascade amplification. In the ab-sence of space charge, the amplification electric field be-tween a flat sample surface and the anode is approximatelylinear and given bysVa−Vsd /d. A positive surface potentialtherefore reduces the gas amplification efficiency. This pro-vides a feedback mechanism, whereby the surface potentialis self-limiting: as the space charge increases, further ionproduction is suppressed by the reduced amplification ac-cording to Eq.s1d. If the actual distribution of positive ionsin the gap is considered, the electric field will clearly benonuniform, and simple formulations for signal amplifica-tion, such as those in Eqs.s1d–s3d, will not suffice.

The above descriptions are only valid as steady-state ap-proximations. With a real sample the SE emission variesfrom point to point in the imaged area, and the ion produc-tion rate follows as the beam is rastered over the specimen.In the gas, electrons have a drift velocity that is approxi-mately three orders of magnitude greater than gaseous ionsunder typical low-vacuum conditions.65 Consequently, theamount of space charge present at any instant reflects therecent history of electron and ion production, which in turnaffects the cascade amplification of subsequent emissions.These effects have been documented by Toth andPhillips,66,67who have shown that abrupt changes in electron

emission from an insulating specimen will give rise to brightand dark streaking in images. If the beam moves abruptlyfrom a region of high emission to low, there will be a tran-sient excess of ions above the specimen. Amplification willbe suppressed until the excess dissipatessvia recombinationwith electrons and lateral driftd and a new, lower steady-stateconcentration is achieved. This process can take several mi-croseconds, so a dark streak appears, with its length being afunction of scan rate. Conversely, if the beam movesabruptly from a region of low emission to high, there will bean immediate spike in amplification efficiency, which willdiminish as the ion concentration slowly builds up. Underextreme conditions, they show it is actually possible to invertthe contrast temporarily.

FIG. 7. Axial potential function for an electron-irradiated dielectric in alow-vacuum environment sketched at three length scales:sad anode-gas-sample-stub,sbd sample surface-maximum primary beam penetration rangeRmax, andscd sample surface-maximum SE escape range. The potential func-tion is sketched for the case where electrons are implanted within the inter-action volume, but the surface potential is positive due to the presence of theionic space charge at and above the sample surface, and excess holes belowthe surfacesz=0 at the sample surface;d=distance between the sample andthe anode;Vs=surface potential;Va=anode biasd.


B. Electron-ion recombination

In the classic literature on gaseous electronics, “electron-ion recombination” usually refers to the recapture of a newlyliberated electron by its parent ion, immediately after an ion-ization event.16,62,63 Except under conditions of extremelylow electric field and high pressure, this process is consid-ered negligible as, normally, the oppositely charged particlesare separated quickly by the field. In the early low-vacuumSEM literature, recombination in this context was mentionedoccasionally for completeness in describing electron-gas in-teractions. Here we are concerned exclusively with recombi-nation processes involving gaseous positive ions at/near thesample surface, electrons at the surfaceswithin the firstmonolayerd, and secondary electrons emitted from the speci-men. Excellent reviews of recombination processes of elec-trons at surfaces have been compiled by Hagstrum68 andVarga and Winter.69 Hahn provides an overview of the phys-ics behind recombination involving ions and electrons in freespace.70

If a secondary electron is captured by a positive ion im-mediately after it crosses the sample surface, it cannot con-tribute to the amplification cascade and the final electronimaging signal. This SE signal scavenging effect, docu-mented systematically by Cravenet al.,71 and later quantifiedby Tothet al.,72 can only occur appreciably at and just abovethe sample surface, where the ion concentration exhibits amaximum and most emitted SEs possess relatively low ki-netic energy on the order of a few eV.sAs SEs move awayfrom the surface, their energy spectrum shifts to higher en-ergies due to the action of the electric field between the an-ode and the surface.d In terms of SE signal amplification, thegap between the surface and the anodesshown in Fig. 2d cantherefore be divided roughly into the three regions shown inFig. 8. Just above the surface, most SEs do not possessenough energy to ionize gas molecules, but can recombinewith ions. The electron flux therefore decreases with distance

z above the surface. However, asz increases, SEs are accel-erated by the electric field. At distancez0 the gas ionizationavalanche initiates and the electron flux starts to increase.The resulting electron population profile is shown in Fig. 8.The net effect of this scavenging process on the detected SEsignal intensity is described by72

ISEsz= dd < GI0e−CI0, s9d

whereG represents gas gain in the absence of scavenging,andC is a signal-damping coefficient. Tothet al. also showhow G andC can be measured as a function of microscopeoperating parameters. This is illustrated in Fig. 9, whichshows such measurements obtained as a function of pressureP. The gas gain curveGsPd has a maximum at several hun-dred pascals, consistent with the Townsend model of gasamplification discussed earlier. At low pressures, dampingincreases rapidly withP because the ion generation rate isproportional toGsPd. Hence, the ion flux incident into theSE-ion recombination volume increases withP. sN.B. In theexample shown in Fig. 9, “low pressure” means less thanabout 150 Pa, but in general this value depends on param-eters such as the anode bias, gas type, working distance, andd.d At higher pressures, the damping coefficient becomes in-dependent ofP because the flux of ions into the recombina-tion volume approaches some limiting value. This is becausemutual repulsion prevents very high ion concentrations fromever being achieved.

The fact thatC depends on the concentration of ionsaround the beam impact point and on the SE energy spec-trum has two important implications on SE image contrast.Firstly, lateral variations in ion concentrationswithin the re-combination volume shown in Fig. 8d at length scales on theorder of the imaged area can give rise to image contrast. Theresulting features in SE images are said to be the conse-quence of “spatial filtering in the SE-ion recombinationrate.” Similarly, contrast caused by the effects of lateralvariations in the SE energy spectrum on the recombinationprobability can give rise to an “energy filtering” effect.

FIG. 8. Schematic illustration of the SE signal intensityISE, plotted as afunction of distancez, above the sample surfacesz=0d. Adapted from Ref.

72. fI0=beam current;d=SE emission yield;V̄= mean SE-ion recombina-tion probability sdetermined by the lateral SE and ion distributions withinthe SE-ion recombination volumed; zV= maximum height at which SE-ionrecombination can occur;z0= minimum height at which SE gas amplifica-tion can occur.g

FIG. 9. Gas gainG, and recombination-induced SE signal damping coeffi-cient C, measured as a function of gas pressureP. Data from Ref. 72ssample=grounded conductor; gas=H2O; accelerating voltage=10 kV;Va

=451 V; wd=3.2 mmd.


1. Spatial filtering

At high gas pressuresgreater than about 200 Pad the mo-tion of ions is randomized in collisions with gas molecules.If the pressure is sufficiently low, and if the electric field isnot constant above the specimen surface, ions will flow pref-erentially to regions of low potential. The resulting nonuni-form ion flux results in a contrast mechanism based on thelocal degree of signal scavenging. Inhomogeneities in theelectric field at the surface can result from subsurfacetrapped charge in insulators, and asperities on sample sur-faces. Various papers by Tothet al. detail experiments de-signed to illustrate this effect.32,72,73In extreme cases, signalscavenging can take place to such an extent that the normallybright contrast from a negatively charged region of an insu-lator can be made to invert. Similarly, enhanced emissionfrom sharp edges and corners on metal surfaces can be sup-pressed through deliberate manipulation of the ion flux.sWenote that ion focusing as it is described here is quite distinctfrom the focusing due to the electrostatic pinch effect de-scribed by Danilatos.12d

2. Energy filtering

If the SE emission spectra from adjacent regions on aspecimen are differentsfor example, due to charging as de-scribed in the section on the Schottky effectd, image contrastcan be altered due to corresponding differences in scaveng-ing efficiency.32,74,75The capture cross section for an electronby an ion decreases rapidly with increasing electron energy.Thus, the lower the average energy of the SE spectrum, themore susceptible the emissions are to scavenging.

In principle, if two neighboring regions had the sametotal SE yield, but different SE energy spectra, contrastwould be observed in the presence of ions, despite its ab-sence in the images obtained using conventional, high-vacuum SE detectors. This effect has not been verified ex-perimentally, but can uniquely account for some of theexperimental observations described below. Contrast basedon energy filtering could become a very useful technique forcharacterizing defects once it has been studied systematicallyand developed.

C. Schottky effect

A combined effect of the anode bias and the strong di-pole field formed between the trapped negative charge andthe space charge is that the work function of the surface willbe lowered via the Schottky effect.76 The work function af-fects secondary electron emission because it determines thecritical angle for internal reflection of electrons moving to-wards the surface.52,77,78Equivalently, it is sometimes statedthat reducing the work function increases the maximum SEescape depth, and increases the emission yield.

Lowering the work function also alters the energy distri-bution of emitted secondaries.73 Figure 10 shows the influ-ence of an ion and the applied electric fieldsgenerated by theanode bias, ionic space charge, and implanted electronsd onthe surface electronic structure of an insulator. Also shown isthe effect on the SE emission spectrum,N8sE8d. Prior toemission, the energy distribution of free electrons immedi-

ately below the sample surfaceNsEd depends on the densityof states in the dielectric and on the energy dependence ofthe SE stopping powerssince SEs generated below the firstmonolayer must travel to the surface prior to emissiond. Asshown in Fig. 10, the applied electric field lowers the surfacebarrier and preferentially enhances the emission probabilityof relatively low-energy SEs, as is indicated by the shadedpart of theN8sE8d curve.

D. Charge balance

As it is trivial to obtain stable SE images in low-vacuuminstruments, steady-state charging conditions must be estab-lished rapidly. In this case, “steady-state” implies that Kirch-hoff’s law is satisfied; that is, the probe current is exactlyequal to the sum of the emissive current and the conductiveflow of electrons out of the sample. To summarize the earlierdiscussion, as charge accumulates in traps, the resulting elec-tric field has three consequences. First, if the surface poten-tial is negative, it reduces the landing energy of the primaryelectrons, which diminishes the size of the interaction vol-ume and brings it closer to the specimen surface. More sec-ondary electrons are generated within the escape depth, thusincreasing emission.78 Second, the dipole field further in-creases the secondary yield via the Schottky effect, as shownin Fig. 10. Finally, the Fowler–Nordheim effectsakin to theSchottky effectd assists in detrapping electrons viatunneling,79 allowing them to be transported towards the sur-face sFig. 6d. As more charge is trapped, the internal fieldintensifies making the processes of electron detrapping,transport and, ultimately, removal from the sample more ef-ficient. The charge state thus becomes self-regulating.

In the case of uncoated insulators imaged by high-vacuum SEM, stable imaging conditions are typically

FIG. 10. Effect of ions and external fields on surface electronic structureand secondary electron emission from a dielectric.sad A simplified energydiagram of a surface infinitely far from an ion, in the absence of appliedelectric fields. Only the electrons that have a sufficient component of theirvelocity normal to the surface can surmount the surface barrier.sbd Anenergy diagram for an ion near the surface, in the presence of the electricfield generated by the biased anode, specimen charging, and the ionic spacecharge. The surface-potential barrier is lowered, and electrons in the con-duction band that could not normally escape can be emitted or captured bythe ion. The additional emitted electrons are indicated by the shaded part ofN8sE8d. Adapted from Ref. 32. fEv=top of valence band ; Ec

=bottom of conduction band;Evac=vacuum level;NsEd SE energy distribu-tion just below the specimen surface;N8sE8d SE energy distribution justabove the surface.g


achieved by deliberately lowering the beam energy to in-crease SE yield and implant electrons close to the samplesurface. In low vacuum, the flux of ions landing on the sur-face provides an additional pathway for the removal of ex-cess electrons through recombination. As Fig. 10 shows,when an ion approaches the specimen surface, excess elec-trons that lack sufficient kinetic energy to escape the speci-men via SE emission can be easily transferred to ionssi.e.,recombinationd and transported away. Unlike the scavengingeffect described in the section on Electron-Ion Recombina-tion, this process does not result in a loss of SE signal sincethe electrons involved would not ordinarily be emitted.Through the use of the electrostatic “mirror effect,” Thieletal. demonstrate that the rate at which charge dissipates in astrongly insulating specimen depends on the positive ion fluxto the surface, supporting this model.76

Circumstantial evidence also suggests that electronscomprising the skirt may actually assist in this process. Skirtelectrons can be considered a low-intensity defocused floodthat continuously irradiates the area surrounding the probe.Each of these electrons deposits most of its energy, whileonly adding a single additional negative chargesless second-ary and backscattered emissionsd. Some of the energy depos-ited assists in the removal of trapped electronssvia radiation-induced conductivityd. The skirt current densitysand thecorresponding electron implantation rated, typically, is nothigh enough to cause the charging effects seen directly underthe probe.

Bringing these concepts together, we can construct a hy-pothetical situation for instructive purposes. Let region A ofa dielectric specimen contain a population of energeticallydeep traps, and region B an equivalent density of shallowtraps. If both regions are irradiated by a flux of electrons,charge accumulates in both regions. However, in region B,the excess electrons are detrapped, transported to the surface,and removed by ions more efficiently than in region A.Hence, if the irradiating flux is sufficiently small, Region Astores more charge than region B in the steady-state condi-tion, causing the SE yield from A to increase. Thus region Aexhibits greater SE emission and appears bright in an elec-tron image. If the incident electron flux is increased such thatelectrons are entering the specimen at a rate greater than theycan be removed from region B through conductivity and re-combination at the surface, the amount of stored charge inregion B will begin to approach that of region A. The SEemission from region B will subsequently be increased rela-tive to A ssince the maximum total yield from either regioncannot exceed unity in steady-stated, thus reducing the SEimage contrast.

Toth et al. describe an experiment with similar implica-tions using gallium nitride.74 GaN is ann-type semiconduc-tor as-grown, and so has a reasonable conductivity. Regionsof high charge trap density were created by bombardmentwith high-energy helium ions. As expected, the damaged re-gions showed considerably higher SE emission than the sur-rounding matrix. If, however, the material is heated to300 °C, the traps can depopulate thermally and the damagedregions once again become conductive. Charge-induced con-trast is then lost between the damaged and as-grown regions.

E. Other ion effects

A few other ion-related processes should be mentioned,although they have not been treated explicitly in the low-vacuum literature. The most important of these are secondaryemission processes, whereby positive ions cause emission ofadditional electrons. It should be emphasized that under mostlow-vacuum conditions, these emission processes are due en-tirely to the potential energy state of the ion, and not to itskinetic energy.16 Ballistic impact of ions on surfaces does notresult in secondary electron emission until the kinetic energyof the ion approaches the keV range.19 Similarly, the stop-ping power of any solid for ions is so great that ion implan-tation below the surface is not likely to occur for sub-keVenergies.80

Secondary electron emission from low-energy ion im-pact is due primarily to three processes, two of which in-volve relaxation of the metastable state that results from thecapture of an electron by an ion.16,19,68,70The relaxation pro-cesses involve the emission of a photon, which can eject aphotoelectron from the specimen surface, or the emission ofan Auger electron from the excited neutral gas molecule.When an ion is neutralized by capturing an electron from thespecimen, an Auger electron can be emitted from the sample.In all three cases, the emitted electron is amplified in the gascascade and contributes to the detected signal. In low-vacuum SEM, secondary emission is generally not consid-ered a problem for polyatomic gases.

Under the conditions encountered in low-vacuum mi-croscopy, the specimen surface can be covered by adsorbedgas molecules. The resulting changes in the surface elec-tronic structure often alter the SE emission probability. In-deed, this is the basis for much of the interesting work ap-plying photoelectron emission microscopysPEEMd to thestudy of catalysis.81 Consequent effects on SE images havenot yet been investigated systematically in the context oflow-vacuum SEM.

Finally, it should be kept in mind that the drift time ofions generated in the gas is generally appreciable with re-spect to the time scales associated with scanning the primarybeam and the time constants of most detector systems. Oneissue is that the induced signals on electrostatic detectors willbe influenced by the drift times of the ions.4 A second issue isthat through the processes described above, the ions will in-troduce scan-rate-dependent effects, as the ion flux at anypoint in the image will depend on the electron emission inthe preceding few microseconds.67 This is obviously a com-plex feedback process, and is very dependent on specimenand operating conditions.

V. APPLICATIONS

This section reviews some of the published work usinglow-vacuum and environmental scanning electron micros-copy on materials applications, and interprets their resultswith respect to the fundamentals outlined in Sec. IV. Al-though a number of publications report the use of these in-struments to examine the morphology of poorly conductingspecimens, far fewer are concerned with either interpreting


contrast or exploring applications and uses of the technique.It is the latter two categories of papers that will be consid-ered here.

A. Ferroelectric domains

One of the most interesting applications papers pub-lished is the work of Zhu and Cao on the imaging of ferro-electric domains in LiTaO3 using an environmental SEM.82

They are able to obtain contrast between positive and nega-tive domains on polished specimenssFig. 11d, but only underconditions of extremely high beam current and relatively lowpressure s,250 Pad. Selective etching was used subse-quently to identify the domains unambiguously. Signifi-cantly, they note that the contrast is reversed from that whichis obtained under high-vacuum conditions; that is, positivedomains are seen to be brighter than negative domains in thelow-vacuum SEM. A conclusive explanation of the contrastis not given, but the authors speculate that it depends on theinteraction between the electron beam and the surface layerof the crystal, and that certain combinations of electrons andions may be able to screen intrinsic surface charges. Xiaoetal. obtained similar contrast from domains in lead titanates,but did not offer an explanation.83,84

Under high-vacuum conditions, SE emission from thenegative character faces is enhanced, but suppressed on posi-tive faces, due to voltage contrast.85 Inversion of the contrastunder low-vacuum conditions is consistent with positive ionsbeing focused on negative domains and scavenging the sig-nal svia SE-ion recombinationd. As the SE spectrum from thenegative regions should have increased intensity in the low-energy range relative to the positive region, the SE signalfrom these regions will be especially susceptible to scaveng-ing ssee Fig. 10d. As pressure is increased, the focusing effectdiminishes, but the ion flux increasessdue to increased cas-cade amplification, as indicated by theGsPd profile shown inFig. 9d. Hence, scavenging increases on average such thatintrinsic emission differences in the low-energy end of thespectrum are suppressed. The need to use high beam currentsto observe domain contrast is also consistent with this inter-pretation, as this is a means of increasing the ion flux withoutincreasing pressure. Because large excesses of negativecharge cannot accumulate in the specimen under low-vacuum conditions, this contrast is stable, as they report. It is

likely that the same contrast effects would be observed in awide range of ferroelectric systems, with the degree of con-trast obtainable, being a function of the polarizability.

B. Electronic devices

Phillips et al. illustrate the potential for low-vacuumanalysis of electronic devices by comparing contrast from ap-n junction in EBICselectron-beam-induced currentd, speci-men current, and induction images.86 In the latter, the deple-tion layer is clearly visible when either thep or n side isgrounded, but the other left floating. Contrast inverts be-tween the two casessFig. 12d. They attribute the distinctivecontrast of the depletion region to the fact that the junctionpotential separates the electron-hole pairs created by thebeam and the injected current biases the floating part of thedevice. The corresponding change in sample surface poten-tial induces current flow in the detector, giving rise to imagecontrast. The instantaneous bias state of the devicesand,therefore, induced signal intensityd depends on whether thebeam is striking the open-circuit side, the closed-circuit side,or the junction itself.

C. “Charge contrast imaging”

One of the most intriguing topics in low-vacuum imag-ing is the group of dynamic contrast mechanisms collectivelydescribed as “charge contrast imaging”sCCId or “charge in-duced contrast”sCICd.75,87–92 sThe term CCI will be usedhere for brevity.d This topic is discussed in considerable de-tail here, as it is a phenomenon unique to low-vacuum in-struments, and the present authors’ interpretation involvesmost of the processes described in Secs. III and IV, appliedto a single system.

A simple definition of CCI is not forthcoming becausethe term has been applied to a wide variety of effects seen inseveral different systems. However, the essential aspect ofCCI is that certain insulating materials exhibit dramatic, butstable, SE contrast variations that are extremely sensitive toexact imaging conditions such as gas pressure and scan rate.In most cases, the contrast is impossible to obtain in high-vacuum instruments. The contrast cannot be explained byatomic number, density, or topographic variations, but doesrequire that the specimen be in some steady-state chargingcondition. A rigorous explanation for these effects has notbeen established, but we will attempt to reinterpret the re-ported behavior in terms of the ion-related mechanisms de-scribed in Secs. III and IV. Numerous examples of CCI canbe found in the proceedings of the meetings of MicroscopySociety of America, the Microbeam Analysis Society, the

FIG. 11. Alternating positivesbrightd and negative ferroelectric domains inLaTiO3. Image adapted from Ref. 82saccelerating voltage=10 kV;P=260 Pad.

FIG. 12. An image pair of thep-n junction in a cross-sectioned diode. In theleft-hand image, thep-doped side is floating while then-doped side isgrounded. Conditions are reversed for the right-hand image.


Australian Society for Electron Microscopy, and others overthe past ten years. Rather fewer examples exist in peer-reviewed literature, but these will be the focus here.

The most widely discussed system exhibiting CCI issynthetically produced gibbsite, or AlsOHd3. Polycrystalsgrown by seeding supersaturated liquor baths developgrowth zones, resulting in grains with a nominally “onion-like” structure.89 When polished cross sections of such par-ticles are imaged under high-vacuum conditions, no contrastis apparent in either secondary or backscattered modes.However, under low-vacuum conditionssroughly between 20and 200 Pad and moderately high cascade field strengths.104 V/md, detailed contrast emerges from the growthzones, providing that the scan rate of the beam is sufficientlyfast.87,88 Contrast of the growth zones changes dramaticallywith scan rate, beam current, and magnificationsall varia-tions on changing the electron flux densityd, working dis-tance, anode bias, and gas pressure. The behavior is demon-strated in a set of images shown in Fig. 13. As a rough guide,contrast is greatest at low electron flux densities, low pres-sure, and high anode bias. The most thorough documentationof the CCI effect in gibbsite and its dependency on pressure,working distance, detector bias, primary beam energy, andscan speed is contained in a Ph.D. thesis by Baroni.91

Additional clues as to the nature of this contrast mecha-nism come from comparisons of cathodoluminescencesCLdimages with CCI images.87 These show a reasonable degreeof correlation between light emission and SE contrast ingibbsite. As CL emissions frequently are due to defects suchas impurities, vacancies, and lattice defects,93 this suggeststhat the CCI contrast may be associated with certain types ofdefects.

The various experimental observations will now be col-lated, and the mechanisms described in Secs. III and IV in-voked to propose one possible scenario for the behavior ofgibbsite. In turn, explanations will be offered for the lack ofcontrast in high-vacuum mode, and the dependency of thecontrast on electron flux density, pressure, and anode bias.

1. High-vacuum contrast

Successful imaging of an insulator in high-vacuum modeis predicated on choosing the primary beam energy such thatthe total emission coefficient is unity. That is, to work at theso-called E2 crossover point.85,94As no contrast is evident inthe backscattered images of gibbsite, the backscattered emis-sion coefficient must be constant everywhere. In order for thetotal emission to be unity everywhere, the secondary emis-sion coefficient must also be invariant. Any local variationsin charging or intrinsic emissivity tend to self-regulate toachieve this condition. Thus, charge contrast is difficult toobtain in high vacuum. In low-vacuum conditions, positiveions provide an additional pathway for the dissipation ofcharge, removing the constraint that the total emission mustbe unity. Thus, contrast variations due to inhomogeneities inthe electronic structure are allowed to emerge. Additionally,at the E2 crossoverftypically less than 2 keVsRef. 94dg mostof the interaction volume is contained within the SE escapedepth and the influence of charge trapping in determiningcontrast is greatly reduced.

2. Electron flux density

One possible explanation for the origin of charge con-trast can be found in Eqs.s6d–s8d. The trapped charge in-creases the SE emission coefficient by lowering the escapebarrier and compressing the interaction volume. The amountof charge trapped in a given region is determined by thedifference between the rate at which it is depositedfEq. s6dgand dissipatedfEq. s7dg. As the trap population variesthroughout the material, the local time constants for dissipa-tion must also varyfEq. s8dg. Thus, the steady-state amountof trapped charge varies locally, giving rise to SE contrast.

At very low electron flux density, the charging differen-tial between adjacent regions with different trap populationsis small, and the contrast is poor. As the flux density is in-creased, the charging differential increases, as does the con-trast. This continues until the traps in a given region saturate,or the current of electrons leaving the samplesi.e., the sum ofthe emission and electron-ion recombination currentsdmatches the electron injection current. Further increases influx result in diminishing contrast as the traps in adjacentregions fill. It is clear from Eq.s6d that the rate of chargeinput, or electron flux density, can be controlled equivalentlyby adjusting the beam currentI0, the scan ratefrelated toF inEq. s6dg, or the magnificationsrelated toAd.

3. Pressure and anode bias dependence

Gas pressure and anode bias are treated together, as theyindividually and jointly affect a number of processes. Theircomplex interdependence is one of the contributing factors tothe confusion surrounding CCI. The primary effect of these

FIG. 13. A polished grain of gibbsite imaged with 50-Pa water vapor. Thebottom image was taken at a fast scan rateseight frames integrated to reducenoised. Increasing the electron flux by slowing the scan ratestop imagedcauses many of the SE contrast features to invert.


two parameters is in determining the ion production rate.However, the gas pressure and the anode bias also influencethe extent to which the ions are focused to different points onthe specimen surface, thus affecting local concentration. Anincrease in local ion concentration, in turn, will have severaleffects:sid It can decrease local charge storage by enhancingthe recombination pathway for charge dissipation, that is, bydecreasingt in Eqs.s7d ands8d, sii d it can increase emissionand alter the SE energy spectrum by lowering the surfacebarrier via the Schottky effect, andsiii d it can decrease theSE signal by scavenging the low-energy portion of the SEspectrum according to Eq.s9d.

These concepts can be brought together to describechanges in contrast as a function of gas pressure, under con-ditions of high anode bias.sExact values for the pressureranges will not be given here, as they are dependent on de-tector geometry. It is the sequence in which various pro-cesses dominate the contrast that is of concern.d At very lowpressures, there are insufficient ions present to achievecharge balance conditions, that is, the situation is akin to thehigh-vacuum case and any contrast is weak. As pressure isincreased, steady-state charging conditions can be attainedregardless of primary beam energy, and contrast differencesdue to defect distribution can appear. At the same time, ionfocusing takes place and signal scavenging occurs. In thisregime, contrast is the result of a competition between en-hanced emission due to charging and signal loss due to scav-enging. As pressure continues to increase, the focusing effectis diminished and the ion flux increases, as well as becomingmore uniform. As “charge contrast” information is carriedprimarily by the lowest-energy SEs, contrast is lost due toenergy filtering and the image once again becomes feature-less. Supporting evidence for this is given by the fact that ifan ion collection grid is inserted to control the ion popula-tion, charge contrast can continue to be observed up to themaximum pressure tolerated by the microscope.71

Griffin notes that surface details such as scratches andcontamination patches are particularly easy to image underCCI conditions.87 This was attributed to the action of posi-tive ions suppressing all but the near-surface SE emissions,and altering the ratio of SE1 to SE2 contributions to theimage. It is difficult to imagine that the ions could preferen-tially suppress SE1 emissions, as these nominally are emittedfrom an area with the diameter of the probe, i.e., around10 nm. It is not likely that ions could be focused to suchprecision, nor is there a substantial difference in the energyspectra of the two SE populations.

A more likely explanation for the enhanced surface sen-sitivity emerges from the following considerations:sidCharging preferentially enhances the low-energy tail of theSE emission spectrum as shown in Fig. 10, andsii d the sen-sitivity of the SE emission probability to variations in theheight of the surface barrier increases with decreasing SEenergy; that is, the enhanced emission due to charging is verysensitive to subtle changes in the state of the surface, such asthose caused by residue contaminants, scratches, etc.32 Anexample of such contrast from scratches on the surface ofAl2O3 is shown in Fig. 14. Unlike the majority of charge

contrast examples in the literature, the image also demon-strates that high-resolution images can be attained under“charge-contrast imaging conditions.”

Other systems are reported to exhibit charge contrast ef-fects, including sphalerite, calcite, mica, alumina, quartz, bi-otite, and cordierite.87,88At the time of this writing, it is notknown exactly what attributes a specimen must possess inorder to demonstrate charge-induced contrast. As of yet, sys-tematic studies have not been performed. From the analysispresented, though, it is likely that the resistivity and polariz-ability of the substance both must fall within certain ranges,and that these characteristics could either be intrinsic or theresult of charge trap distributions. Theoretically, CCI couldreveal previously inaccessible information on subsurface in-homogeneities at high spatial resolution. However, untilthese questions are resolved, the technique will only providequalitative information.

D. Electronic polymers

An application area that stands to benefit immenselyfrom the use of low-vacuum SEM is the development ofelectronic polymers. As with conventionalssilicon-baseddelectronics, the active region of a device will be at the inter-face between phases, and characterization of the microstruc-ture at high spatial resolution is vital. Being carbon-based,there is little absolute density variation between the differentpolymers typically used, making them good candidates foridentifying phases based on electronic structure contrast.One example of using low-vacuum SE imaging to studyphase separation between the polymers F8BTfpolys9,9-dioctylfluorene-cobenzothiadiazoledg and PFB fpolys9,9-dioctylfluorene-co-bis-N,N-s4-butylphenyld-bis-N,N-phenyl-1,4-phenylenedi-aminedg is given by Ramsdaleet al., wherethey show a hierarchy of structures developed in films spincoated from solution.95 They report that the F8BT phase isconsistently brighter than the PFB, and attribute the differ-ence to their electronic structures. However, this is unlikelyto be the source of contrast, as F8BT has both the smallerenergy gaps2.36 vs 2.8 eVd and the larger electron affinitys3.53 vs 2.29 eVd, both of which would point to a smallerescape depth and lower relative emission.96 As the F8BTphase contains sulfur, it is likely that in this example thecontrast is due to differences in stopping power. Neverthe-

FIG. 14. High-resolution image of sapphire showing surface detail onlyvisible when the sample was allowed to charge, obtained using an acceler-ating voltage of 30 kV and a gas pressure of 530 Pa.


less, the stable, high-resolution images obtained demonstratethe potential for low-vacuum imaging in this area.

E. Liquids

Molecular liquids are a particularly interesting class ofspecimens, as they present a situation in which the electron–electron interactions most closely resemble those of a dielec-tric, whereas the charge dissipation characteristics are thoseof a reasonable conductor. Many liquids are classed as beingdielectric fluids, but that designation refers to electric currentprofiles that develop in response to an applied electric field.In the context of a SEM specimen, however, the issue isradiation-induced conductivity. Distilled water, for example,is known to be a very poor conductor, but charged particles,such as solvated electrons and the ionic products of radioly-sis, have reasonably high mobilities.97,98 Previous work onthe SE emission characteristics of liquids is limited, as it hasseldom been an issue of much practical importance. How-ever, as liquids and liquid-containing structures can now becharacterized routinely in an environmental SEM, the issuesbear renewed consideration.99

We now consider the processes of SE generation andemission from liquids by analogy to the processes in thesolid state. Typical primary electron energies in a SEMgreatly exceed the binding energies in the specimen. Thestopping power and the energy deposition profile are deter-mined by valence ionizations and collective, plasmonlike os-cillations of valence electrons.100 Thus, for the generationprofile of excited electrons, it makes no difference whetherthe substance is solid or liquid.

As with solids, the factors that determine the inelasticmean-free-path of low-energy electrons are the molecular-orbital structure, charge traps, and collective oscillations.Most molecular liquids can be classified as insulators orwide-band-gap semiconductors on the basis that usually themolecular-orbital structure is such that there is a significantenergy difference between the highest-occupied and lowest-unoccupied molecular orbitals. Water in particular has beentreated in this way.101–103An effective energy gap of 8.9 eVhas been reported for water in the liquid state.103

In polar liquids, such as water, molecules can reorient inorder to screen electric fields generated by charged particles.These local perturbations in the otherwise random orienta-tion distribution of molecules,si.e., polaronsd act in a similarfashion as charge traps in solid-state materials. These traplikestates comprise the Urbach tail of the conduction band, andaccount for the reduction of the electronic excitation thresh-old of liquid water relative to that of the vapor state.103 Coeet al. report a detailed study on the electronic properties ofvarious arrangements of water molecule clusters.104 Theyalso discuss the role of ionic impurities as creating dopantlevels within the energy gap.

The above phenomena make the determination of thestopping power and the mean-free-path of low-energy elec-trons in liquid water a challenging topic. The most compre-hensive compendium on the subject can be found in the ref-erence edited by Ferrandini and Jay-Gerin.46 Extensive workhas also been done in radiology, where workers are con-

cerned with the track structure of ionizing radiation as itinteracts with biological tissue. This is obviously an ex-tremely broad topic in itself and so will not be covered indepth here, but a few papers that the present authors havefound particularly useful are given.97,105–107Ritchie discussesthe relevant issues of energy-loss processes of low-energyelectrons in water, with the intent of describing the inelasticmean-free-path.37 Howie has discussed the problem in thespecific context of SE emission, and also considers the roleof elastic scattering in determining the escape depth fromwater.38 Finally, many of the same considerations apply topolymers and molecular substances to a greater or lesser ex-tent. A separate review on the use of environmental SEM forthe study of polymers has been published elsewhere.9

Stokeset al. have exploited the SE emission character-istics of liquids to obtain interesting and useful contrast ef-fects in water-oil emulsions using the environmental SEM.41

By virtue of its larger forbidden energy gap, the water phasewas demonstrated to have a higher emissivity than the spe-cific oil phase used. This interpretation is consistent withwater having a higher threshold energy for electron–electronscatteringsdue to a larger forbidden energy gapd, which leadsto a larger escape depth for secondary electrons. However,the mobility of solvated electrons in the oil was such that thecharacteristic time constant for charge decay was comparableto the frame time. As with gibbsite, a flux-density-dependentcontrast could be induced by appropriate changes in scan rateor magnification. The contrast could be manipulated to theextent that a type of contrast inversion was observed. A de-tailed study of the contrast effects was given in a secondpaper.92

Another popular use for environmental SEMs involvingliquids is the observation of contact angles in wettingstudies.108–110This is a promising technique, as the liquid canbe kept in thermodynamic equilibrium while the substrate-liquid contact geometry is examined with very high spatialresolution. It is tempting to extract quantitative values for thecontact angle, but Stelmashenkoet al. provide a detailedanalysis of the problem, and show that knowledge of the SEescape depth is essential.111

F. X-ray microanalysis

Much of the literature discussing x-ray microanalysis ina low-vacuum environment focuses on complications im-posed by the skirt scattering and its effects on quantification.Two articles by Mansfield112 and Newbury113 have summa-rized the work in this area, so interested readers are directedto those references.

In the context of this review, it is more appropriate tofocus on the effects of specimen charging on x-ray mi-croanalysis. It is well-known that the surface potential of aspecimen will alter the landing energy of the primaryelectrons.78 As the atomic ionization cross sections arestrongly energy dependent, quantification algorithms forcharacteristic x-ray emissions require that the landing energybe known quite accurately. As was discussed earlier, an in-sulator can develop either a positive or a negative surfacepotential under low-vacuum analysis conditions. Although


the ion flux will prevent the surface from developing a largenegative potential, a positive shift of a few hundred volts ispossible. If the primary beam energy is only a few keV, thisshift can be significant.

Toth et al. very clearly demonstrate the effect in lowvacuum analysis, using the Duane–Hunt limit to measure thelanding energy.64 sThe Duane–Hunt limit is the highest en-ergy continuum x-ray that can be produced by electron irra-diation, and is equal to the electron-beam landing energy.dHowever, as with imaging considerations, the surface poten-tial can be controlled through the use of an ion collectiongrid.64,71 Carlton demonstrates that if such a grid is used,most quantification errors are minimized and elaborate pro-cedures to correct skirt effects are less important.114

VI. SUMMARY AND OUTLOOK

Introducing a low pressure of gas, or more specifically apopulation of positive gaseous ions, into a SEM results in arich variety of electron–specimen–gas interactions. These in-teractions can be exploited as contrast mechanisms to revealinformation about dielectric specimens that cannot be ob-tained by conventional SEM methods. Furthermore, dynamicresponses to changes in input/output currents can provideinsight into processes, instead of merely producing static im-ages of structure.

The work reviewed here has been largely concerned withsimply identifying, isolating, and demonstrating the variousphenomena that are unique to performing electron micros-copy in a low-vacuum environment. A few of the investiga-tions have tried to show the way in which phenomena suchas charging, charge dissipation, recombination, and surfacebarrier modification depend on microscope operating param-eters. The applications papers discussed, however, highlightthe fact that the information acquired during investigationsof real specimenssrather than idealizedd is the result of acomplex interplay between these mechanisms, with differentprocesses dominating the signals under different conditions.Further advances in imaging theory will come in response toissues raised during investigations in increasingly sophisti-cated application areas.

A notable parameter that has not been explored in thecontext of the investigations included here is the effect ofusing different gas species and gas mixtures. It is possiblethat the exact molecular-orbital structure of the gas mol-ecules will have an effect on recombination and scavengingefficiency; that is, on the rate at which excess charge can beremoved from the specimen and how the emitted SE signal ismodified before detection. Naturally, changing the gas com-position will also affect the cascade amplification and there-fore the rate at which the ions are produced.

Another low-vacuum analysis technique that is likely tosee significant activity in the future is cathodoluminescence.Optical photon emissions can yield a wealth of informationfrom dielectrics, but the technique has seen limited popular-ity in conventional SEMs. The conductive coatings typicallyused for high-vacuum electron-beam analysis block the lightemissions. With the need for a coating eliminated, the poten-

tial for CL can be realized. When correctly interpreted, CLmay be able to provide information on subsurface defectscomplementary to charge contrast imaging.

A final point worth mentioning is that it is generallyobserved that surface contamination of specimens duringelectron irradiation is reduced considerably, as compared tohigh-vacuum analysis. Complications resulting from the ac-cumulation of a hydrocarbon layer during imaging and mi-croanalysis are well-known, and are often a limiting factor inhigh-resolution analysis. The processes responsible for thelack of contamination have not been reported in detail, butwarrant further investigation.

ACKNOWLEDGMENTS

The authors would like to thank Dr. Daniel Morrison andDr. Beth Bromley for useful discussions and critical feed-back on the manuscript. Professor Wenwu Cao from Penn-sylvania State University is thanked for giving permission toreproduce Fig. 11 and for discussing their results. This workwas supported in part by the FEI Company, The Isaac New-ton Trust, and the Engineering and Physical Sciences Re-search Council, UK. Dr. Brendan Griffin, University ofWestern Australia, kindly provided the gibbsite specimen.

1D. A. Moncrieff, V. N. E. Robinson, and L. B. Harris, J. Phys. D11, 2315s1978d.

2A. N. Farley and J. S. Shah, J. Microsc. -Oxf. 158, 389s1990d.3G. D. Danilatos, Scanning Microsc.4, 799 s1990d.4R. Durkin and J. S. Shah, J. Microsc. -Oxf. 169, 33s1993d.5P. Meredith, A. M. Donald, and B. L. Thiel, Scanning18, 467 s1996d.6B. L. Thiel, I. C. Bache, A. L. Fletcher, P. Meredith, and A. M. Donald, J.Microsc. -Oxf. 187, 143s1997d.

7Microsc. Res. Tech. 25s1993d.8G. D. Danilatos, Microsc. Res. Tech.25, 529 s1993d.9F. S. Baker, J. P. Craven, and A. M. Donald, inTechniques for StructuralCharacterization of Polymers, edited by R. E. Pethrick and C. VineysWiley, London, 2003d.

10A. M. Donald, Nat. Mater.2, 511 s2003d.11G. D. Danilatos and R. Postle, Micron 14, 42s1983d.12G. D. Danilatos, Adv. Electron. Electron Phys.71, 109 s1988d.13G. D. Danilatos, Adv. Electron. Electron Phys.78, 1 s1990d.14G. D. Danilatos, J. Microsc.162, 391 s1991d.15G. D. Danilatos, Microsc. Res. Tech.25, 354 s1993d.16A. von Engel,Ionized GasessClarendon, Oxford, 1965d.17A. N. Farley and J. S. Shah, J. Microsc. -Oxf. 158, 379s1990d.18B. L. Thiel, Ultramicroscopy99, 35 s2004d.19H. S. W. Massey and E. H. S. Burhop,Electronic and Ionic Impact Phe-

nomenasClarendon, Oxford, 1969d.20D. C. Joy, Y. U. Ko, and J. J. Hwu, Proc. SPIE3998, 108 s2000d.21D. C. Joy, http://www.web.utk.edu/;srcutk, 200022M. T. Postek and A. E. Vladár, Proc. SPIE5038, 315 s2003d.23D. A. Moncrieff, P. R. Barker, and V. N. E. Robinson, J. Phys. D12, 481

s1979d.24S. Wight, G. Gillen, and T. Herne, Scanning19, 71 s1997d.25G. Gillen, S. Wight, and T. Herne, Scanning20, 404 s1998d.26S. Wight, Scanning23, 320 s2001d.27H. Bethe, Ann. Phys.5, 325 s1930d.28D. C. Joy, Monte Carlo Modeling for Electron Microscopy and Mi-

croanalysissOxford University Press, New York, 1995d.29D. C. Joy and S. Luo, Scanning11, 176 s1989d.30K. Kanaya and S. Okayama, J. Phys. D5, 43 s1972d.31P. Hovington, D. Drouin, and R. Gauvin, Scanning19, 1 s1997d.32M. Toth, M. R. Phillips, B. L. Thiel, and A. M. Donald, J. Appl. Phys.91,

4479 s2002d.33J. C. Ashley, J. Electron Spectrosc. Relat. Phenom.46, 199 s1988d.34C. J. Tung, J. C. Ashley, and R. H. Ritchie, Surf. Sci.81, 427 s1979d.35D. R. Penn, Phys. Rev. B35, 482 s1987d.36Z. H. Levine and S. G. Louie, Phys. Rev. B25, 6310s1982d.


37R. H. Ritchie, R. N. Hamm, J. E. Turner, and W. E. Bolch, inWorkshop onComputational Approaches to Molecular Radiation Biology, edited by N.M. Varma sPlenum, New York, 1994d.

38A. Howie, J. Microsc. -Oxf. 180, 192s1995d.39H. Seiler, J. Appl. Phys.54, R1 s1983d.40D. C. Joy, http://web.utk.edu/;scrutk/htm/interact.html41D. J. Stokes, B. L. Thiel, and A. M. Donald, Langmuir14, 4402s1998d.42Z. J. Ding and R. Shimizu, Scanning18, 92 s1996d.43Z. J. Ding, X. D. Tang, and R. Shimizu, J. Appl. Phys.89, 718 s2001d.44H. J. Fitting, E. Schreiber, J.-Ch. Kuhr, and A. von Czarnowski, J. Elec-

tron Spectrosc. Relat. Phenom.119, 35 s2001d.45E. Schreiber and H. J. Fitting, J. Electron Spectrosc. Relat. Phenom.124,

25 s2002d.46Excess Electrons in Dielectric Media, edited by C. Ferrandini and J. P.

Jay-GerinsCRC, Boca Raton, 1991d.47A. Modinos, Field, Thermionic, and Secondary Electron Emission Spec-

troscopysPlenum, New York, 1984d.48M. R. Castell, D. D. Perovic, and H. Lafontaine, Ultramicroscopy69, 279

s1997d.49D. D. Perovic, M. R. Castell, A. Howie, and C. A. Walsh, Ultramicroscopy

58, 104 s1995d.50C. Schönjahn and C. J. Humphreys, J. Appl. Phys.92, 7667s2002d.51J. Cazaux, J. Appl. Phys.59, 1418s1986d.52J. Cazaux, J. Appl. Phys.85, 1137s1999d.53J. Cazaux, J. Electron Spectrosc. Relat. Phenom.105, 155 s1999d.54J. Cazaux, J. Appl. Phys.89, 8265s2001d.55I. A. Glavatskikh, V. S. Kortov, and H. J. Fitting, J. Appl. Phys.89, 440

s2001d.56A. Melchinger and S. Hofmann, J. Appl. Phys.78, 6224s1995d.57J. P. Ganachaud and A. Mokrani, Surf. Sci.334, 329 s1995d.58A. M. Stoneham,Theory of Defects in SolidssClarendon, Oxford, 1975d.59S. O. Kucheyev, M. Toth, M. R. Phillips, J. S. Williams, C. Jagadish, and

G. Li, J. Appl. Phys.91, 3940s2002d.60R. D. Van Veld and T. J. Shaffner, inScanning Electron Microscopy,

edited by O. M. JoharisITT Research Institute, Chicago, 1971d, pp. 17–24.61T. J. Shaffner, inScanning Electron Microscopy, edited by O. M. Johari

sITT Research Institute, Chicago, 1973d, pp. 293–300.62L. B. Loeb, Basic Processes of Gaseous ElectronicssUniversity of Cali-

fornia Press, Berkeley, CA, 1961d.63J. M. Meek and J. D. Craggs,Electrical Breakdown of GasessWiley, New

York, 1978d.64M. Toth, M. R. Phillips, J. P. Craven, B. L. Thiel, and A. M. Donald, J.

Appl. Phys. 91, 4492s2002d.65E. A. Mason and E. W. McDaniel,Transport Properties of Ions in Gases

sWiley, New York, 1988d.66M. Toth and M. R. Phillips, Scanning22, 370 s2000d.67M. Toth and M. R. Phillips, Scanning22, 319 s2000d.68H. D. Hagstrum, inElectron and Ion Spectroscopies of Solids, edited by L.

Fiermans, J. Vennik, and W. DekeysersPlenum, New York, 1978d, pp.273–323.

69P. Varga and H. Winter,Electron Emission IIsSpringer, Berlin, 1992d.70Y. Hahn, Rep. Prog. Phys.60, 691 s1997d.71J. P. Craven, F. S. Baker, B. L. Thiel, and A. M. Donald, J. Microsc. -Oxf.

205, 96s2002d.72M. Toth, D. R. Daniels, B. L. Thiel, and A. M. Donald, J. Phys. D35,

1796 s2002d.73M. Toth, B. L. Thiel, and A. M. Donald, J. Microsc. -Oxf. 205, 86s2002d.74M. Toth, S. O. Kucheye, J. S. Williams, C. Jagadish, M. R. Phillips, and G.

Li, Appl. Phys. Lett. 77, 1342s2000d.75M. Toth, B. L. Thiel, and A. M. Donald, Ultramicroscopy94, 71 s2003d.76B. L. Thiel, M. Toth, and J. P. Craven, Microsc. Microanal.10, 711

s2004d.

77T. J. Shaffner and J. W. S. Hearle, inScanning Electron Microscopy,edited by O. M. JoharisITT Research Institute, Chicago, 1976d, pp. 61–70.

78L. Reimer,Scanning Electron MicroscopysSpringer, New York, 1985d.79R. M. Marsolais, E. A. Cartier, and P. Pfluger, inExcess Electrons in

Dielectric Media, edited by C. Ferradini and J. P. Jay-GerinsCRC, BocaRaton, 1991d, pp. 44–72.

80H. Rothard, M. Jung, B. Gervais, J. P. Grandin, A. Billebaud, and R.Wünsch, Nucl. Instrum. Methods Phys. Res. B107, 108 s1996d.

81E. Bauer, Rep. Prog. Phys.57, 895 s1994d.82S. Zhu and W. Cao, Phys. Rev. Lett.79, 2558s1997d.83J. Xiao, M. Shao, Y. Hang, and S. Yin, J. Phys.: Condens. Matter13,

11567s2001d.84J. Z. Xiao, M. J. Shao, and S. T. Yin, J. Cryst. Growth240, 521 s2002d.85J. I. Goldstein, D. E. Newbury, P. Echlin, D. C. Joy, A. D. Romig, Jr., C.

E. Lyman, C. Fiori, and E. Lifshin,Scanning Electron Microscopy andX-ray MicroanalysissPlenum, New York, 1992d.

86M. R. Phillips, M. Toth, and D. Drouin, Appl. Phys. Lett.75, 76 s1999d.87B. J. Griffin, Scanning22, 234 s2000d.88G. R. Watt, B. J. Griffin, and P. D. Kinny, Am. Mineral.85, 1784s2000d.89T. C. Baroni, B. J. Griffin, J. R. Browne, and F. J. Lincoln, Microsc.

Microanal. 6, 49 s2000d.90T. C. Baroni, B. J. Griffin, J. B. Cornell, G. I. D. Roach, and F. J. Lincoln,

Scanning24, 18 s2002d.91T. C. Baroni, Ph.D. thesis, University of Western Australia, 2002.92D. J. Stokes, B. L. Thiel, and A. M. Donald, Scanning22, 357 s2000d.93B. G. Yacobi and D. B. Holt,Cathodoluminescence Microscopy of Inor-

ganic SolidssPlenum, New York, 1990d.94D. C. Joy and C. S. Joy, Microsc. Microanal.4, 475 s1998d.95C. M. Ramsdale, I. C. Bache, J. D. MacKenzie, D. S. Thomas, A. C.

Arias, A. M. Donald, R. H. Friend, and N. C. Greenham, Physica EsAmsterdamd 14, 268 s2002d.

96R. J. Wilson, Ph.D. thesis, Cambridge University, 2002.97M. A. Hill and F. A. Smith, Radiat. Phys. Chem.43, 265 s1994d.98C. P. Royall, B. L. Thiel, and A. M. Donald, J. Microsc. -Oxf. 204, 185

s2001d.99B. L. Thiel and A. M. Donald, J. Mol. Liq.80, 207 s1999d.100J. M. Heller, Jr., R. N. Hamm, R. D. Birkhoff, and L. R. Painter, J. Chem.

Phys. 60, 3483s1974d.101F. Williams, S. P. Varma, and S. Hillenius, J. Chem. Phys.64, 1549

s1976d.102A. Bernas, C. Ferradini, and J. P. Jay-Gerin, Chem. Phys.222, 151

s1997d.103T. Goulet, A. Bernas, C. Ferradini, and J. P. Jay-Gerin, Chem. Phys. Lett.

170, 492 s1990d.104J. V. Coe, A. D. Earhart, M. H. Cohen, G. J. Hoffman, H. W. Sarkas, and

K. H. Bowen, J. Chem. Phys.107, 6023s1997d.105S. Uehara, H. Nikjoo, and D. T. Goodhead, Phys. Med. Biol.37, 1841

s1992d.106S. M. Pimblott, J. A. LaVerne, and A. Mozumder, J. Phys. Chem.100,

8595 s1996d.107H. Tomita, M. Kai, T. Kusama, and A. Ito, Radiat. Environ. Biophys.36,

105 s1997d.108P. Forsberg and P. Lepoutre, Scanning Microsc.8, 31 s1994d.109A. Liukkonen, Scanning19, 411 s1997d.110M. L. Steen, A. C. Jordan, and E. R. Fisher, J. Membr. Sci.204, 341

s2002d.111N. A. Stelmashenko, J. P. Craven, A. M. Donald, A. M. Terentjev, and B.

L. Thiel, J. Microsc. -Oxf. 204, 172s2001d.112J. F. Mansfield, Mikrochim. Acta132, 137 s2000d.113D. E. Newbury, J. Res. Natl. Inst. Stand. Technol.107, 567 s2002d.114R. A. Carlton, C. E. Lyman, and J. E. Roberts, Microsc. Microanal.10,

753 s2004d.


applied physics reviews—focused review secondary … · low vacuum scanning electron microscopy...

Documents