cw integrated cavity output spectroscopy

9 July 1999

Ž .Chemical Physics Letters 307 1999 343–349www.elsevier.nlrlocatercplett

cw Integrated cavity output spectroscopy

Anthony O’Keefe ), James J. Scherer, Joshua B. PaulLos Gatos Research, 67 East EÕelyn AÕenue, Suite 3, Mountain View, CA 94041, USA

Received 9 March 1999; in final form 10 May 1999

Abstract

A new approach is described in which continuous, narrow band laser sources are employed with the recently developedintegrated cavity output spectroscopy technique to obtain sensitive, quantitative absorption spectra in a simple experimentalconfiguration. Absorption data obtained with cw-ICOS are related to the classical Fabry–Perot intracavity absorption model,which describes why the intracavity absorption is enhanced. A method of continuously injecting cw laser light into the cavityis described, as is a simple means of interpreting the ICOS data to extract accurate absorption intensities. Absorption spectraof vibrational combination bands of CO and H O in the 1.3 mm region are presented. q 1999 Elsevier Science B.V. All2 2

rights reserved.

1. Introduction

Recently, the integrated cavity output spectro-Ž .scopic technique ICOS was introduced and demon-

w xstrated using pulsed dye laser sources 1 . In theICOS approach, light is coupled into a highly reflec-tive optical cavity, and the cavity output is measuredin order to extract intracavity losses such as molecu-lar or atomic absorption. The initial ICOS workdemonstrated that the technique could be employedto obtain absorption spectra with high sensitivity,comparable to that obtained with other sensitivetechniques such as cavity ringdown spectroscopy. InICOS, the extinction coefficient for intracavity sam-ples is obtained by measuring the total transmittedoutput of the cavity as the input light source isscanned in frequency. Per pass extinction coeffi-cients are derived through knowledge of the cavity

) Corresponding author. Fax: q1 650 965 7772; e-mail:[email protected]

mirror reflectivity at the injection laser wavelength.The initial, proof-of-principle ICOS studies wereperformed with a pulsed dye laser, where normaliza-tion of the integrated output was achieved by mea-suring the intensity of the input light source, whichneeded to be measured at approximately 1% accu-racy to infer a per pass fractional absorption of1=10y6. This work demonstrated that very weakabsorptions could easily be quantitatively measuredusing a pulsed dye laser with 10–15% shot-to-shotamplitude fluctuations.

In this Letter, a variation of the ICOS technique,cw-ICOS, is described that enables absorption spec-tra to be obtained using continuous sources such asdiode lasers. This cw version of the ICOS methodemploys two strategies to effectively eliminate theproblems traditionally associated with the frequencyselectivity or sharp resonances of the optical cavity.Under coherent excitation, cw pumped optical cavi-ties exhibit high-energy build up when the cavitymode coincides with the excitation wavelength, re-

0009-2614r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved.Ž .PII: S0009-2614 99 00547-3

( )A. O’Keefe et al.rChemical Physics Letters 307 1999 343–349344

sulting in near-unity ‘transmission’ when on-reso-nance and near-zero transmission when off-reso-nance. The resulting contrast of the cavity is afunction of the cavity stability as well as the degreeof mode matching achieved between the laser andthe cavity. Small fluctuations in the cavity can resultin large deviations in transmitted energy, makingtraditional absorption measurements impossible. Ahigh finesse optical cavity also acts as an efficientspectral filter to narrow band light passed through it,which can result in strong variations in transmittedenergy with source frequency.

In cw-ICOS, both of these potential problems areeliminated by effectively randomizing the cavitymode structure on a time scale significantly fasterthan the frequency-scanning rate of the injectionlight source. This is achieved by dithering both thefrequency space of the optical cavity as well as thatof the input light source. The modulation of thesource laser frequency eliminates transmission fringeswith variation in wavelength. In the present study,we demonstrate that when the depth and frequencyof the cavity and laser source modulation is suffi-cient, the ICOS method of integrating the total cavityoutput can be employed to obtain accurate, linearabsorption spectra. Here absorption spectra are ob-tained in the 1.3 mm region using a frequencyscanned, narrow bandwidth diode laser source. Thesestudies establish the feasibility of the concepts em-ployed in cw-ICOS, and demonstrate the utility,generality, and simplicity of the technique.

2. Theory

In ICOS, the absorption signal is obtained throughthe integration of the total signal transmitted througha ringdown-type, optical cavity absorption cell, inmuch the same fashion as in conventional absorptionmeasurements. Single pass cavity losses are calcu-lated from the measured cavity output, which is afunction of the mirror reflectivity as well as scatter-ing and absorption losses, which occur due to thepresence of samples located between the mirrors. Ina fashion similar to that employed in cavity ring-

w xdown measurements 2 , molecular or atomic absorp-tion is determined by measuring the baseline trans-mission of the cavity, and normalizing the transmit-

ted signal to this value. This removes the effects ofspectrally broad effects such as Rayleigh scatterfrom the measurement. The enhancement of the ICOSabsorption signal results from the effectively infinitesample path length, as the light retraces the samepath on each cycle. In a manner different from thatused in either conventional multipass or ringdown-based absorption measurements, the ICOS measure-ment represents an asymptotic transmission value,due to this essentially infinite sample path length.The measured absorption changes to the integratedoutput are thus very large, even in the case of weakŽ .per pass absorption. The ICOS signal for a pulsed

w xcavity injection has been shown 1 to be:

y1X2 yk lIs I T e = 2 log R , 1Ž . Ž .0

where I is the transmitted intensity, I is the incident0

light intensity, T is the mirror transmission, k is theintra-cavity absorption per unit length, l is the cavitysample length, and RX is an effective ‘reflectivity’.The effective reflectivity embodies the effect of in-tracavity absorption on the Q of the cavity, and isgiven by

RX sR eyk l , 2Ž .

where R is the ‘true’ mirror reflectivity. The changein the ICOS signal is very nearly linear when themagnitude of the absorption is small relative to the

Žtotal mirror transmission losses ignoring mirror scat-.ter .The predicted ICOS sensitivity and dynamic range

have been demonstrated for pulsed injection into aw xsample cavity 1 . Here, we examine what occurs

when coherent, continuous optical injection of thecavity is applied. In cases where the cavity is ac-tively stabilized, the cavity behaves like an idealclassical Fabry–Perot etalon. The transmission char-acteristics of such cavities are well established andthe effects of factors such as absorption are easily

w xcalculated 3,4 . The ‘enhancement’ of intra-cavityabsorption in Fabry–Perot type optical cavities is awell known phenomenon, and results in a strongerpeak absorption value at the center of the cavitymode. In this case, the steady state cavity transmis-sion for the center of the mode is given by

2IrI s 1y Ar 1yRqA =A 3� 4Ž . Ž .0

( )A. O’Keefe et al.rChemical Physics Letters 307 1999 343–349 345

where A is the fractional absorption per optical passŽ .or reflection , T is the fractional transmission perpass in the cavity, and A is the Airy function. Here,it is important to note that this value represents thetheoretical value for the absorption enhancement online center for a stabilized optical cavity. This ‘en-hancement’ does not suggest that the actual molecu-lar absorption strength is modified. Rather it is adirect result of the increased optical residence timeof light within a highly reflective cavity.

In the present work, we demonstrate a method ofreproducibly injecting optical energy into a cavitywith amplitude stability such that small changes in

Ž .measured transmission or integrated output can beusefully related to intra-cavity absorption. In thisapproach, continuous laser light is injected into alinear, two-mirror cavity, which is aligned, such thatthe light retraces the same optical path with eachpass, as is the case in cavity ringdown cavities.When the laser frequency is scanned over severalfree spectral ranges, the expected cavity transmissionspectrum is the classic Fabry–Perot modulated pat-tern. When the laser beam and cavity mirrors areproperly aligned, the observed transmission patternclearly shows sharp longitudinal cavity mode struc-ture. Typically, low order transverse modes are alsoexcited, though to a much lesser extent. There aretwo approaches to achieving a stable and repro-ducible transmission as the injection laser frequencyis scanned. One approach would be to employ stabi-lized optical cavities and laser sources, locked toeach other, and simultaneously scanned such that the

injection laser frequency always matches a cavitymode. Such an approach is very sensitive to vibra-tions, and moreover results in high intracavity energybuild up, which can lead to undesirable saturationeffects. In the present approach, we spoil the finesseof the cavity such that the period of time in whichthe laser and cavity are in coincidence is not suffi-cient for the theoretical energy build up to beachieved. This is achieved by rapidly modulating thecavity length while simultaneously modulating theinput laser frequency, resulting in a rapid randomiza-tion of the cavity and laser dynamic mode matches.When averaged over a number of these dynamicmode matches, the time average transmission is foundto be nearly constant as the laser is scanned ininjection frequency. The cavity length is modulatedusing a piezoelectric actuator as one of the cavitymirror mounts, in a fashion similar to that employed

w xin the cw-cavity ringdown studies 5,6 , while thelaser is frequency modulated, in this case using apiezoelectric actuated grating mechanism in an exter-nal cavity diode laser. The experiments describedbelow demonstrate that when the modulation fre-quency and depth are properly adjusted, this strategyenables cw-ICOS measurements to be performed toyield quantitative, cavity enhanced absorption spec-tra.

3. Experimental

The cw-ICOS experimental apparatus is shown inFig. 1, which also displays an oscilloscope that was

Fig. 1. Schematic representation of the cw-ICOS experimental configuration. A single mode output diode laser operating near 7550 cmy1

was coupled to a two mirror cavity comprised of mirrors with reflectivities of 0.9994, and the averaged transmitted signal detected using aInGaAs detector interfaced to a LabView computer analysis system.


used for diagnostic purposes only. Absorption mea-surements were made using a two mirror optical

Ž .cavity Rs99.94% at 1.3 mm employing a singleŽ .mode, tunable diode laser New Focus model 6224

as the injection light source. The system was tunedcontinuously between 7630 and 7410 cmy1. Absorp-tion measurements were made in air for severalcavity sizes, and in a closed-cell that could be evacu-ated and filled with controlled pressures of variousgases. The cavities ranged from 30–50 cm in length,resulting in cavity free spectral ranges of 500–300MHz, respectively. The output of the diode laser wasaligned to the axis of the cavity by first observingthe back reflections from the ICOS cavity mirrors,and then tuning the entrance mirror angle whilemonitoring the transmission signal on a scope. Nocare was required to stabilize the cavities or thetables upon which they were placed. The cavity wasmodulated using two methods, which produced iden-tical results. The first method was to modulate theposition of one of the cavity mirrors using a piezodriven mount. The second approach was to leave theICOS cavity mirrors fixed and slightly modulate theangle of injection using a piezo driven final turningmirror. In both cases we employed a commercial

Ž .piezo driven mirror mount Thor Labs, KC1-PZ andŽapplied a low amplitude modulation 1–3 V input to

.controller, model MDT690 at a sinusoidal frequencyof ;1 kHz. This modulation resulted in an ICOScavity coverage of 5–10 longitudinal modes. Usingthis modulation alone it was possible to scan thediode laser and obtain absorption spectra which dis-played the expected ICOS intensities, however a

Ž .noticeable 10% amplitude modulation persisted.This amplitude modulation was identified as result-ing from the ICOS cavity from the frequency spac-ing, which matched the predicted etalon spacing ofthe cavity. We believe that this fringing results fromthe periodic coincidence of the laser frequency withan ICOS cavity mode at the turning points of thepiezo modulation. Frequency coincidence at the turn-ing point, where the mirror momentarily stops per-mits greater transmission into the cavity. This effectis periodic as the laser frequency is scanned.

This weaker modulation was effectively removedthrough a small modulation in the diode laser fre-quency over a range that was significantly smallerthan the absorption line widths under study. The

laser frequency modulation was achieved by apply-Ž .ing a 0.1 V peak to peak , 150 Hz signal to the

frequency modulation input of the laser, resulting ina 1 GHz frequency modulation. The use of thisdouble modulation was important in realizing atransmission amplitude profile, which was flat withlaser frequency. The transmitted signal exiting therear of the ICOS cavity was directed to an InGaAs

Ždetector whose output was digitized directly no RC.filtering and read into a Labview analysis program

which stored the average signal as a function of laserfrequency.

4. Results and discussion

Spectra of rovibrational overtone-combinationbands of water vapor and carbon dioxide were mea-sured, and were used to characterize the performanceof the cw-ICOS technique. These measurements werenot normalized to variations in the diode laser out-put. The diode laser exhibited an output power mod-ulation of 1–5% when scanned over spectral regionsof greater than several wavenumbers. While thiseffect could be easily normalized out for wide spec-tral scans, no attempt was made to do this in thepresent study since most absorption strength mea-surements were made on selected lines in regionswhere the laser output was stable to at least 1%. Inthe spectral region between 1.30 and 1.35 mm wave-length, absorptions resulting from H O and CO2 2

were recorded and compared to known absorptionw xstrengths calculated using the HITRAN database 7 .

Absorption measurements for single lines were ob-tained in spectral regions free of laser mode hops,although in some of the wide spectral range datataken of CO some mode hopping is evident. The2

New Focus diode laser is designed to minimizemode hopping, however, there was a noted fre-quency error between observed and predicted absorp-tion line position near the points of mode hops. Thismay have been the result of the mechanical gratingresetting in the laser head at these points, and couldpossibly have been reduced by reduction in thefrequency scan rate of the system. The absorptiondata were normalized to the base line cavity trans-mission level and plotted as fractional absorption asa function of laser frequency.


Absorption due to nascent water vapor in thelaboratory air were used in some tests, however therapidly increasing strength of the absorption for probewavelengths above 1.35 mms effectively turned theICOS cell black. Weaker water absorption featuresnear 1.346 mms were used to demonstrate the sensi-tivity and ease of use of the cw-ICOS technique. Atypical 30 cm cavity cw-ICOS absorption spectrumdue to atmospheric water is shown in Fig. 2, whereseveral absorption features near 7526 cmy1 are seen.This spectrum was recorded using the pico-motorscan mode of the New Focus laser. In Fig. 2 theabsorption spectra predicted using the HITRANdatabase are shown for a 290 m pathlength of atmo-spheric broadened water vapor at a relative pressure

Ž .of 8.75 Torr 50% humidity at 228C . The predictedspectra compares well with the observed spectra inboth line width and absorption strength.

Examination of the cw-ICOS spectra indicates afractional absorption of 70% for the line at 7528cmy1, and a full width at half-max line width of 0.23cmy1, which agrees well with the HITRAN gener-ated spectra showing a width of 0.2 cmy1. It isimportant to note that the ultimate resolution that can

Žbe realized using the cw-ICOS technique as de-.scribed in this Letter is limited by the depth of

modulation of the diode laser frequency. In this case,we have utilized frequency modulation of ;1 GHzwhich should not add significantly to the line width.However, it is clear that in order to achieve spectralresolution much greater than this it will be necessary

Žto use much lower frequency modulation and thus.longer ICOS cavities to reduce the mode spacing .

The spectral resolution of several GHz offered bythis experimental approach represents a good matchto the collision broadened line widths seen at atmo-spheric pressure in the near and mid-infrared spectralregions. This suggests that this approach will be veryuseful in atmospheric analysis and other monitoringapplications.

The 290 m path length used in the HITRANmodel shown in Fig. 2 was selected to give a goodline shape comparison. The absolute ICOS signal canbe checked using the HITRAN deduced absorptionstrength and the known cavity length. The predictedline center absorption coefficient for the water ab-sorptions can be deduced from the HITRAN gener-ated spectra. Using the line near 7528 cmy1 a line

Fig. 2. Five GHz wide scan of three water absorption lines near 7529 cmy1 taken using the cw-ICOS approach. The Y-axis representsfractional transmission through the 30 cm long cavity. The dashed line represents the HITRAN predicted transmission spectrum for a 290 mlong path of air at 50% humidity at 208C, at a total pressure of 1 atm.


center absorption strength of K s4.1=10y5 pera

cm is deduced. This results in an ICOS per passabsorption of 1.23=10y3 for the 30 cm long cavity.

Ž .Plugging this absorption value into Eq. 1 wherey3 X wk = l s 1.12 = 10 , R s 0.9994 = exp y1.23 =

y3 x10 , and normalizing to the baseline value whereRX sR, yields IrI s0.30, which agrees well with0

the observed absorption peak strength. The weakerabsorption peak at 7527.5 cmy1 can be similarlyanalyzed. The HITRAN predicted per-cavity passabsorption factor is 3=10y4 resulting in a predictedICOS cavity transmission 0.60 which is in excellentagreement with the observed absorption signal seenin Fig. 2.

Additional measurements were made using CO2

as a target gas using a pressure controlled ICOS gascell. Several overtonercombination bands of CO2

were measured and compared to the absorption spec-tra predicted using the HITRAN database. The over-tone bands of CO offer a convenient test of the2

variation of the cw-ICOS signal with intra-cavityabsorption strength because the rotational progres-sions provide a smoothly changing absorption profile

to compare with both the impulsive ICOS and coher-ent Fabry–Perot cavity absorption models.

A section of the CO absorption overtone band2

near 7580 cmy1 taken with a 50 cm long cellcontaining a fraction of an atmosphere of CO is2

shown in Fig. 3, which also shows a overlaid trace ofthe HITRAN predicted absorption signal for a 400 mpath of 1 atm CO . This spectrum was recorded in2

the fast scanning mode of the New Focus diodelaser, which accounts for some of the noise in theamplitude spectra. The excellent match between theHITRAN prediction and the observed cw-ICOS sig-nal clearly demonstrates that the cw-ICOS techniqueprovides an accurate absorption profile of this over-tone band, which required only 20s to acquire.

The results presented above demonstrate that thecw-ICOS spectra are well described using the ‘im-pulsive’ ICOS model, as well as the traditional

w xFabry–Perot absorption models 3,4 . Recently, En-w xgeln et al. 8 also demonstrated a somewhat more

complex approach to making integrated cavity outputspectroscopic measurements, in which a cw laser isfrequency scanned over a small frequency range

Fig. 3. cw-ICOS results recorded using a 1 atmosphere, 50 cm long cell of CO near the overtone band at 7580 cmy1. A HITRAN2

simulated spectrum corresponding to a 400 m path, 1 atm sample is overlayed for comparison.


Ž y1 .;1 cm and the transmitted cavity mode spikesfed into a transient recorder which then signal aver-ages the net spectrum as the frequency hoppingmodes fill the entire scanned frequency range. Whilethese authors did not discuss their cavity enhanced

Ž .absorption CEA work in relation to the well-knownFabry–Perot absorption models, the underlying opti-cal process is the same, and the absorption ‘enhance-ment’ resulting from the long cavity residence timeis equivalent to that described here and in the earlier

w xICOS paper 1 .The advantages of the cw-ICOS approach de-

scribed here are in the system simplicity and ease ofapplication, and the obvious potential for use in alock-in detection mode of operation. This offers thepotential for an increase of several orders of magni-tude in sensitivity. We are currently working on thisaspect of the technique for applications in quantita-tive trace analysis.

5. Conclusion

In this Letter, the integrated cavity output absorp-Ž .tion ICEA method, recently introduced, has been

extended to include continuous injection operationand the resulting absorption intensities demonstratedto follow the ICOS model predictions. These resultsare equivalent to the classic Fabry–Perot intra-cavity

Ž .model predictions in the short build up limit . Thecw-ICOS approach provides a far easier approach tocoupling energy into a high finesse optical cavity,and provides a more generally useful method of

making such ultra-sensitive measurements. The useof double modulation, where the ICOS cavity ismodulated as well as the frequency of the injectionlaser, minimizes the transmission amplitude modula-tion with laser frequency, making this approach pos-sible. Perhaps even more exciting is that the ap-proach can clearly be used in conjunction with post-cavity spectral resolution of broad band width lightsources to provide optical multi-channel array detec-

w xtion of extremely weak absorption 9 . This approachwould be of great importance in the concurrentmonitoring of several absorbing species.

Acknowledgements

This research was funded by the Office of NavalResearch, under Contract aN00014-98-C-0347.

References

w x Ž .1 A. O’Keefe, Chem. Phys. Lett. 293 1998 331.w x2 J.J. Scherer, J.B. Paul, A. O’Keefe, R.J. Saykally, Chem. Rev.

Ž .97 1997 25.w x3 E. Hecht, A. Zajac, Optics, Chap. 9.8, 1979.w x4 A. Yariv, Optical Electronics, Chap. 4.1, 1985.w x5 D. Romanini, A.A. Kachanov, N. Sadeghi, F. Stoeckel, Chem.

Ž .Phys. Lett. 264 1997 316.w x6 D. Romanini, A.A. Kachanov, F. Stoeckel, Chem. Phys. Lett.

Ž .270 1997 538.w x7 HITRAN 1996 database, US Airforce.w x8 R. Engeln, G. Berden, R. Peeters, G. Meijer, Rev. Sci. In-

Ž .strum. 69 1999 3763.

cw integrated cavity output spectroscopy

Documents