Frequency modulation spectroscopy with a CO2 laser:results and implications for ultrasensitivepoint monitoring of the atmosphere
David E. Cooper and T. F. Gallagher
In a demonstration of frequency modulation spectroscopy with a CO2 laser, sidebands at 1 GHz were gener-ated using a CdTe electrooptic phase modulator driven either by 5-kW pulses from a microwave cavity oscill-lator or by 10 W of cw power from a solid-state microwave amplifier. Frequency modulation signals re-sulting from sideband absorption by Fabry-Perot resonances were measured using a room-temperature1-GHz bandwidth HgCdTe detector. Signal-to-noise ratios for the conditions of our experiments were-200:1 and limited by rf pickup noise in the detection electronics. Substantial improvements in SNR canbe made by providing better rf shielding for the detection electronics and by using a liquid-nitrogen-cooleddetector in conjunction with improved modulator designs.
1. Introduction
Derivative spectroscopy has become a standardtechnique for the sensitive monitoring of absorptionsdue to atmospheric trace gases. In conjunction withcommercially available tunable diode lasers,' whichcover the 3- to 3 0-,um spectral region, extremely sensi-tive point monitors have been constructed that candetect trace atmospheric constituents at the sub-ppblevel corresponding to absorptions in the 10-4-10-5range. This sensitivity limit is determined primarilyby diode laser noise and background-limited detectornoise.2 In considering further improvements to thissensitivity limit, one is led naturally to higher modula-tion frequencies (to eliminate laser noise) and to het-erodyne techniques (to obtain shot-noise limited de-tection).
The technique of frequency modulation (FM) spec-troscopy introduced by Bjorklund3 is essentially ahigh-frequency form of derivative spectroscopy.However, unlike conventional derivative spectroscopy,FM spectroscopy is in principle shot-noise-limited.This is due to the fact that it is essentially a heterodynetechnique, since the signals generated result from the
David Cooper is with SRI International, Electro-Optics SystemsLaboratory, 333 Ravenswood Avenue, Menlo Park, California 94025;T. F. Gallagher is with University of Virginia, Physics Department,Charlottesville, Virginia 22901.
beating of sidebands with the laser carrier. In addition,the modulation frequencies are high enough to place thesidebands outside the laser inewidth rendering thetechnique insensitive to laser amplitude noise. Indeedin the visible region absorptions as small as 10-6 can bedetected with 1 mW of laser power and a 1-sec integra-tion time.
Frequency modulation spectroscopy clearly has greatpotential in the IR spectral region for ultrasensitivereal-time optical monitoring of atmospheric trace gasesand hazardous materials. This is particularly true inthe 3-5- and 8-12-sAm atmospheric window regions,where numerous molecules have exceedingly strong andcharacteristic absorption signatures. In addition, thenarrow linewidths and optical powers available fromCO2 and semiconductor diode lasers make than excel-lent candidates for FM spectroscopy. Indeed FMspectroscopy with a CO2 laser has recently been con-sidered for long-path measurement of hydrazine fuelgases.4 For a double-ended long-path laser system itwas estimated that a minimum detectable absorptionof 3 X 10-4 could be achieved using sidebands separatedby 1 GHz from the laser carrier.
Several substantial problems are encountered in ex-tending FM spectroscopy to the IR spectral region.First, the IR modulators available are inherently muchless efficient than their visible wavelength counterparts.For example, the halfwave voltages for CdTe at 10.6 mand LiTaO 3 at 0.5 m are 44 and 2.8 kV, respectively. 5
Frequency modulation experiments are typically per-formed with modulation indices of the order of 0.1 re-quiring microwave drive power levels of typically a fewwatts for LiTaO3 and several kilowatts for CdTe. Since
one cannot drive a modulator at kilowatt power levelson a continuous basis without damaging the crystal, oneobvious approach to IR FM spectroscopy is to use apulsed microwave driver. Indeed, efficient CdTe CO2laser frequency shifters6 employ magnetron rf oscilla-tors to drive the modulator at power levels in the kilo-watt range. It is still possible, however, to drive CdTemodulators cw with -10 W of microwave power andobtain excellent FM SNRs. In fact we will show thatfor a fixed average microwave drive power, one alwaysobtains the highest sensitivity in a cw FM experiment.As a result, the extremely efficient GaAs thin-filmwaveguide modulators described by Cheo7 are an at-tractive possibility for IR FM spectroscopy.
A second problem in development of IR FM spec-trosocpy is the bandwidth limitation of available de-tectors. Since one is in general interested in monitoringpressure-broadened absorptions, sensitive FM detectionrequires modulation frequencies in excess of -10 GHz.This in turn requires detectors with 10-GHz band-widths, and whereas in the visible region GaAs Schottkybarrier photodiodes with 100-GHz bandwidths havebeen demonstrated,8 commercially available HgCdTephotodiodes are limited to 2-GHz bandwidths. 9 A wayaround this problem has been devised and demon-strated in the visible spectral region using two elec-trooptic modulators.' 0 This technique can be extendedeasily to the IR spectral region.
This paper reports the demonstration of FM spec-troscopy in the 9-11-,um CO2 laser wavelength regionusing sidebands displaced by 1 GHz from the lasercarrier. Two methods of driving the modulator togenerate the FM sidebands are used. In the firstmethod, sidebands with -1% of the laser carrier powerare generated using a CdTe electrooptic modulatordriven at 1 GHz by 5-kW pulses from a microwavecavity oscillator. Frequency modulation signals re-sulting from the absorption of sidebands by resonancesfrom a Fabry-Perot etalon are measured using a 1-GHzbandwidth room-temperature HgCdTe detector.Observed SNRs are better than 200:1 and are limitedby rf pickup noise.
In the second method, the CdTe modulator is drivencw by 10 W of microwave power at 1 GHz. Under theseconditions, only -0.005% of the laser carrier is shiftedinto each sideband. Nevertheless, SNRs obtained arecomparable with those observed when using the pulseddriver. This is primarily a consequence of the signifi-cantly reduced detection bandwidth in a cw FMsystem.
A shot-noise-limited CO2 laser FM system using aliquid-nitrogen-cooled 2-GHz bandwidth HgCdTe de-tector and 5 mW of laser power is realizable and shouldbe capable of detecting absorptions as small as 10-7.Our experimental apparatus for CO2 laser FM spec-troscopy and the observed FM signals are discussed inSec. II. Signal-to-noise calculations for our system anddifficulties that must be overcome in the approach toa shot-noise-limited system are presented and discussedin Sec. III along with a discussion of the implications of
FM spectroscopy for ultrasensitive monitoring of at-mospheric trace gases.
11. Experimental Results
In the following sections, we describe the experi-mental configurations and the equipment used to in-vestigate IR FM spectroscopy with a CO2 laser andCdTe electrooptic modulator. Frequency modulationsignals obtained using both pulsed and cw microwaveoscillator drivers are presented.
A. CdTe Modulator
The CdTe electrooptic modulator is similar in designto a LiTaO3 microstrip modulator." It consists of asingle 2- X 2- X 20-mm CdTe crystal (Cleveland Crystals,Inc.) placed between a thin copper strip and a brassbaseplate. The top and bottom faces of the crystal arethe (111) crystallographic planes and are metal-platedwith chromium gold to ensure a more uniform electricfield distribution within the crystal. The 2- X 2-mmendfaces of the crystal are polished and antireflection-coated for 10.6 ,um with PbF.
At microwave frequencies, the crystal dielectricconstant is EIEo - 11, and the impedance of the CdTe-loaded microstrip line is -58 Q, a good enough matchto the 50-Q transmission line that quarterwave match-ing sections are not necessary. A microwave circuitanalyzer verified that the modulator VSWR is •1.5 overthe 500-MHz-2-GHz frequency range, and it exhibits,reasonable microwave characteristics up to 4 GHz.
The modulator was tested at microwave drivefrequencies up to 3 GHz using various rf oscillatorplug-in units available with the pulsed cavity oscillator.The oscillator was adjusted to deliver 3-Asec pulses of5-kW peak power to the modulator. Figure 1 shows theoptical power spectrum of amplitude-modulated (AM)10.6-Mum CO2 laser light obtained by placing the mod-ulator between polarizers crossed at ±450 and using a2.5-GHz microwave drive frequency. The spectrum isrecorded using a Burleigh Instruments RC-140 IRscanning Fabry-Perot etalon with a 10-GHz free spec-tral range (FSR). Laser light transmission through theetalon is monitored with a liquid-nitrogen-cooledHgCdTe detector and gold-coated integrating sphere.The integrating sphere is necessary to avoid saturatingthe detector. The finite extinction ratio of the twopolarizers allows a significant amount of laser carrierlight to pass through in the absence of any microwavepower. Nevertheless, the two sidebands displaced 2.5GHz from the carrier are clearly distinguishable. Thecorresponding FM spectrum, recorded with the polar-izers removed, is shown in Fig. 2. To obtain this spec-trum, the two polarizers are placed in front of themodulator and nearly crossed, sufficiently attenuatingthe laser beam to avoid saturating the HgCdTe detector.As a result, the sidebands are deeper in the noise in theFM spectrum than in the AM spectrum. Nevertheless,-1% of the laser carrier power is shifted into the twosidebands corresponding to a modulation index of-0.15.
RELATIVE FREQUENCY (GHz)Fig. 1. Optical power spectrum of amplitude-modulated (AM) CO2
laser light with a 2.5-GHz microwave drive frequency.
B. Pulsed Modulator FM Experiments
The experimental configuration we have investigatedis illustrated in Fig. 3. The laser source is a Line-Litemodel 950 grating tunable CO2 laser operating singlemode, typically on the 10.6-Am 1P20 line with a 2-Wpower output and 100-MHz linewidth. The laser out-put is directed through a six-plate zinc-selenide (ZnSe)Brewster angle polarizer oriented at 0, a 500-Hz me-chanical chopper, and then gently focused through theCdTe electrooptic modulator by a 25-cm focal lengthZnSe lens. In these experiments, the modulator istypically driven at 1 GHz by 5-kW peak power, 3-Asecpulses from an Epsco PG5KB pulsed microwave cavityoscillator. The oscillator is triggered by pulses from aHewlett-Packard (HP) 214A pulse generator referencedto the 500-Hz mechanical chopper. During the mi-crowave pulse, 1% of the laser carrier power is shiftedinto the two sidebands, corresponding to a modulationindex of 0.15.
-2.5 0 2.5RELATIVE FREQUENCY (GHz)
Fig. 2. Optical power spectrum of frequency-modulated (FM) CO2laser light with a 2.5-GHz microwave drive frequency.
Fig. 4. Inphase (a) and quadrature (b) FMsignals resulting from the absorption of1-GHz sidebands by resonances of a Fabry-Perot etalon. The signals were obtained bydirecting -30 mW of 10.6-ptm optical poweronto a 1-GHz bandwidth room-temperatureHgCdTe detector. The sidebands, eachcontaining -1% of the laser carrier opticalpower, were generated by driving the CdTe
modulator with 5-kW microwave pulses.
The frequency-modulated laser light is next recolli-mated by a second 25-cm focal length ZnSe lens anddirected through a single ZnSe Brewster plate andFresnel rhomb onto the piezoelectrically scannableRC-140 IR Fabry-Perot etalon. The retroreflectionfrom the etalon is then focused onto a Boston Elec-tronics Corp. R005 room-temperature HgCdTe detec-tor, which has a D* -1 X 106 cm Hz"/2/W and a band-width of -1 GHz. The detector is biased by a 30-Vsupply through the dc port of a rf monitor tee. The1-GHz component of the detector photocurrent ismonitored through the rf port of the tee and amplified+40 dB by an Avantek AFT-2034 microwave amplifier.The amplified photocurrent signal is applied to the rfport of a Mini Circuits ZFM-150 mixer. The IF port ofthe mixer is driven by 5 mW of peak power from the
Epsco oscillator. This is obtained by splitting off -30dB of the modulator drive signal with a directionalcoupler and using a rf attenuator to obtain an additional30 dB of signal reduction. The IF signal from the mixeris sampled at 500 Hz by a Princeton Applied Research(PAR) 162 boxcar signal averager with a 3-jisec gatewidth triggered by pulses from the HP214A pulse gen-eator. Frequency modulation spectra are recorded bydriving an XY chart recorder with the boxcar output asthe frequency of the etalon is scanned.
Figure 4 shows the inphase and quadrature FM sig-nals resulting from sideband absorption and dispersionby resonances of the Fabry-Perot etalon. The etalonFSR is 10 GHz, and the resonances are observed tomimic a 50% gas absorption of the 10.6-gAm laser line.The optical power incident on the R005 detector is -30
Fig. 5. Inphase (a) and quadrature (b) FMsignals resulting from the absorption of1-GHz sidebands by resonances of a Fabry-Perot etalon. The signals were obtained bydirecting a30 mW of 1 0 .6 -,um optical poweronto a 1-GHz bandwidth room-temperatureHgCdTe detector. The sidebands, eachcontaining 0.005% of the laser carrier opti-cal power, were generated by driving theCdTe modulator cw with 10 W of microwave
power.
mW, and under these conditions the SNR is expectedto be detector-noise-limited. The observed SNRs are,however, limited by rf pickup in the detection elec-tronics resulting from inadequately shielded detectorand microwave amplifier power supplies. When stepsare taken to correct this problem, SNRs of better than1000:1 should be attainable with this system, corre-sponding to minimum detectable absorptions of 10-3in a 1-sec integration time.
C. Continuous-Wave Modulator FM Experiments
For the cw modulator experiments, the optical con-figuration is identical to that shown in Fig. 3. Theelectrical configuration, however, is somewhat different.The boxcar signal averager is replaced by a PAR5204 lockin amplifier referenced to the 500-Hz me-
chanical chopper. The pulsed cavity oscillator is re-placed by a HP8620 sweep oscillator and a solid-statemicrowave amplifier capable of 15-W output over the1-2-GHz range. In these experiments, the microwavedrive power delivered to the CdTe modulator is set at10 W. Frequency modulation signals are recorded bymonitoring the output of the lockin amplifier with a XYrecorder as the frequency of the Fabry-Perot etalon isscanned.
Figure 5 shows the inphase and quadrature FM sig-nals obtained under these cw modulator drive condi-tions. The etalon FSR is 10 GHz, and as before theresonances mimic a 50% gas absorption of the 10.6-jumlaser line. The optical power directed onto the R005detector is again -30 mW. The observed SNRs arecomparable with those obtained in the pulsed modu-
lator experiments. Apparently, even though themodulation index is quite small in this case (0.01), thereduced bandwidth resulting from the quasi-cw detec-tion system essentially compensates for this. Addi-tional details on the expected SNRs are given in Sec.III.
Ill. Discussion
The theory of line shapes and SNRs in FM spec-troscopy has been treated in the literature.12 For thecase of shot-noise-limited detection, the power SNRis
P (Ab)2 M2
SNR= (1)4Af
where j is the detector quantum efficiency, Po is thelaser carrier power at w, incident on the detector, A isthe sideband differential absorption, M is the modu-lation index, and Af is the detection system bandwidth.In the case of detector-noise-limited FM spectroscopyin the IR, with the detector noise characterized by thedetectivity D*, the SNR can be written
SNR= (D*p 0 )2 (A6)2M2 (2)
4AAf
where A is the detector area.In our pulsed modulator experiments, D* - 1 X 106
cm Hzl/ 2/W, A - 0.01 cm2, M - 0.15, Ab -~ 0.5, Po - 30mW, and Af - 3 X 105 Hz, corresponding to an -3-usecpulse. Inserting these values in Eq. (2) results in a SNRof -422 for one sample. In our experiment, the boxcarsignal averager is set up for exponential averaging, sincethe input signal is scanned. Under our experimentalconditions (3-Asec pulse duration, 500-Hz prf), the ef-fect of the boxcar averaging is to reduce the bandwidthby -10, giving an expected SNR -4 X 103. Our ob-served SNR of -200 is about an order of magnitudebelow this value as a result of rf interference prob-lems.
In the cw-driven modulator experiments, the detec-tion system bandwidth is determined by the lockin in-tegration time constant, which is set to 30 msec, corre-sponding to Af 33 Hz. The modulation index was notmeasured but is estimated as follows. With the CdTecrystal cut for frequency modulation in which the mi-crowave electric field is applied perpendicular to the(111) plane, it can be shown that the modulation indexis given by' 3
2- 1M = -,/3 d nor4lV. (3)
Using the values n3r4l = 1.2 X 10-10 mV-1 for CdTe, X= 10.6 ,um, Id = 10, and V - 22 V for 10 W of drivepower results in an estimated modulation index of-0.01. Inserting this in Eq. (2) with Af - 33 Hz givesa SNR of -1.7 X 104. Again the observed SNR of -200in this case is limited by rf pickup in the detectionelectronics. However, the expected higher SNR in thecw-drive case brings up an important feature of FMspectroscopy in general. The FM SNR is always pro-portional to M2/Af. For the case in which the modu-
lator is driven cw at average microwave power P., M2
P., and for a 1-sec integration time, SNR P.However, if a pulsed microwave driver is used, then M 2
P.r prf, where T is the microwave pulse width, andfor a 1-sec integration time, SNR -- P//prf. Hence,for fixed average microwave drive power, one alwaysobtains the highest sensitivity in a cw FM experi-ment.
Under shot-noise-limited conditions, which could inprinciple be realized with a liquid-nitrogen-cooled2-GHz bandwidth HgCdTe detector, Eq. (1) yields aminimum absorption detectable with unity SNR of
2 (h, 1/2
mnM nPo f
At 10.6 Am withq X- 0.7, Po - 5 mW, M - 0.01, and Af1 Hz, the minimum detectable absorption is Abmin -
5 X 10-7. This is several orders of magnitude moresensitive than conventional derivative spectroscopy andcould be further improved by designing a more efficientCdTe modulator to increase the modulation index.However, as discussed by Bjorklund et al.,' 2 the actualachievement of shot-noise-limited detection in FMspectroscopy requires careful attention to experimentaldetail. First, the laser optical power Po must be suffi-cient for shot noise to dominate over detector and am-plifier thermal noise. This condition is usually satisfiedin the IR provided the carrier power incident on thephotodetector exceeds the detector amplifier thermalnoise,14
2kThvPo >
7e 2 R(5)
For a liquid-nitrogen-cooled HgCdTe detector at 10 Amwith '- 0.7, T = 300 K (the noise temperature of atypical amplifier), and R = 50 Q, Eq. (5) yields P0 > 184,uW. This power level can be easily exceeded with theoutputs available from CO2 lasers and semiconductordiode lasers. It can also be reduced significantly byusing a liquid-nitrogen-cooled detector amplifier.
Several additional effects can degrade the sensitivityof a FM experiment and prohibit the approach toshot-noise-limited detection. These include rf oscil-lator frequency instability, mismatched signal propa-gation times between the rf oscillator and the signal andlocal oscillator ports of the RF mixer, residual amplitudemodulation resulting from etalon effects in the modu-lator crystal and various refractive optical elements orrefractive-index variations for the two sidebands, andinadequate rf shielding of the detection electronics.These potential problems have been discussed in somedetail in Ref. 12 along with methods to eliminate orsignificantly reduce their effect on system performance.Clearly, however, improved FM spectroscopy experi-ments need to be conducted in the IR with attention tothese potential noise sources to determine how closelyone may approach the shot-noise-limit with a practicalFM system.
The usefulness of FM spectroscopy measurementsof atmospheric trace gases using a CO2 laser is limitedby the requirement of a coincidence between one of thesidebands and the gas absorption line of interest. This
problem is mitigated somewhat by the fact that efficientGaAs waveguide modulators working in excess of -15GHz have been demonstrated allowing a continuoussideband tuning range of -30 GHz7 and hence nearlycomplete spectral coverage of the 9-12-,um range usingseveral isotopic CO2 lasers. The resulting high-fre-quency FM signals can be demodulated using one of theschemes discussed in Ref. 10 allowing the use of con-ventional bandwidth-limited HgCdTe detectors. Anadditional problem in using a fixed frequency laser todo FM spectroscopy is the fact that tuning a sidebandthrough an absorption feature requires that the rf beswept, which is difficult to do in practice at high SNRbecause of impedance-matching problems. This dif-ficulty may perhaps be overcome by performing mea-surements at low pressure in a multipass absorption cellwhere the absorption lines are narrowed to a few tensof MHz. In this case, sideband tuning and imped-ance-matching will be required only over a relativelynarrow frequency range. However, because of thesedifficulties in using a fixed frequency laser it is expectedthat practical IR FM spectrometers will require a con-tinuously tunable laser source such as a semiconductordiode laser or a waveguide-modulator-tuned CO2laser.
IV. Conclusions
We have demonstrated IR frequency modulationspectroscopy using a CO2 laser and a CdTe electroopticmodulator. Experiments were performed using theresonances of a Fabry-Perot etalon to simulate a gasabsorption. Signal-to-noise ratios for the conditionsof our experiments were -200 and limited by rf pickupnoise in the detection electronics. Further improve-ments in SNR can be made by providing better rfshielding for the detection electronics, use of a liquid-nitrogen-cooled 2-GHz bandwidth HgCdTe detector,and by using improved modulator designs to increasethe modulation index. Infrared FM spectrosocpy, inconjunction with either CO2 lasers or continuouslytunable semiconductor diode lasers, promises a sev-eral-order-of-magnitude increase in detection sensi-tivity over that currently achievable using derivativespectroscopy with diode lasers.
References
1. R. S. Eng, J. F. Butler, and K. J. Linden, "Tunable Diode LaserSpectroscopy: An Invited Review," Opt. Eng. 19, 945 (1980).
2. J. Reid, M. El-Sherbiny, B. K. Garside, and E. A. Ballik, "Sensi-tivity Limits of a Tunable Diode Lasr Spectrometer, with Ap-plication to the Detection of NO2 at the 100-ppt Level," Appl.Opt. 19, 3349 (1980).
3. G. C. Bjorklund, "Frequency-Modulation Spectroscopy: A NewMethod for Measuring Weak Absorptions and Dispersions," Opt.Lett. 5, 15 (1980).
4. L. T. Molina and W. B. Grant, "FTIR-Spectrometer-DeterminedAbsorption Coefficients of Seven Hydrazine Fuel Gases: Im-plications for Laser Remote Sensing," Appl. Opt. 23, 3893(1984).
5. F. S. Chen, "Modulators for Optical Communications," Proc.IEEE 58, 1440 (1970).
6. G. M. Carter, "Tunable High Efficiency Microwave Frequency-Shifting of Infrared Lasers," Appl. Phys. Lett. 32, 810 (1978).
7. P. K. Cheo, "Frequency Synthesized and Continuously TunableIR Laser Sources in 9- to 11-ym," IEEE J. Quantum Electron.QE-20, 700 (1984).
8. S. Y. Wang and D. M. Bloom, "100-GHz Bandwidth Planar GaAsSchottky Photodiode," Electron. Lett. 19, 554 (1983).
9. D. L. Spears, "Theory and Status of High Performance Hetero-dyne Detectors," Proc. Soc. Photo-Opt. Instrum. Eng. 300, 174(1981).
10. D. E. Cooper and T. F. Gallagher, "Double Frequency ModulationSpectroscopy: High Modulation Frequency with Low-Band-width Detectors," to be published in 15 Mar. issue of Appl.Opt.
11. N. H. Tran, R. Kachru, T. F. Gallagher, J. P. Watjen, and G. C.Bjorklund, "Generation of Microwaves by Mixing Two OpticalFrequencies in a Nonlinear Crystal: A Novel Approach toHigh-Bandwidth Optical Mixers," Opt. Lett. 10, 128 (1984).
12. G. C. Bjorklund, M. D. Levenson, W. Lenth, and C. Ortiz, "Fre-quency Modulation (FM) Spectroscopy, Theory of Lineshapesand Signal-to-Noise Analysis," Appl. Phys. B 32, 145 (1983).
13. S. Namba, "Electro-Optical Effect of Zincblende," J. Opt. Soc.Am. 51, 76 (1961).
14. R. H. Kingston, Detection of Optical and Infrared Radiation(Springer, New York, 1978).
The authors wish to acknowledge stimulating dis-cussion with and generous loans of equipment by J.Watjen. This research was supported by SRI Inter-national IR & D funds.
