Optical cavity enhanced real-time absorption spectroscopy of CO2 using laseramplitude modulationJiahao Dong, Timothy T.-Y. Lam, Malcolm B. Gray, Richard B. Warrington, Edward H. Roberts, Daniel A.Shaddock, David E. McClelland, and Jong H. Chow Citation: Applied Physics Letters 105, 053505 (2014); doi: 10.1063/1.4892535 View online: http://dx.doi.org/10.1063/1.4892535 View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/105/5?ver=pdfcov Published by the AIP Publishing Articles you may be interested in Observation of a new electronic state of CO perturbing W Π 1 ( v = 1 ) J. Chem. Phys. 141, 144311 (2014); 10.1063/1.4897326 Chirped laser dispersion spectroscopy using a directly modulated quantum cascade laser Appl. Phys. Lett. 103, 191107 (2013); 10.1063/1.4829036 Generation of high-order rotational Raman sidebands in the deep-ultraviolet region using molecular phasemodulation induced by an intensity-modulated optical pulse J. Appl. Phys. 108, 056104 (2010); 10.1063/1.3467525 Highly accurate determinations of C O 2 line strengths using intensity-stabilized diode laser absorptionspectrometry J. Chem. Phys. 127, 084311 (2007); 10.1063/1.2759930 Beam quality enhancement for a radio-frequency excited annular CO 2 laser Appl. Phys. Lett. 73, 2549 (1998); 10.1063/1.122529

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

130.56.106.27 On: Thu, 17 Dec 2015 03:21:26

Optical cavity enhanced real-time absorption spectroscopy of CO2 usinglaser amplitude modulation

Jiahao Dong,1,a) Timothy T.-Y. Lam,1 Malcolm B. Gray,2 Richard B. Warrington,2

Edward H. Roberts,3 Daniel A. Shaddock,1 David E. McClelland,1 and Jong H. Chow1

1Research School of Physics and Engineering, Australian National University, Acton, ACT 0200, Australia2National Measurement Institute, Lindfield, NSW 2070, Australia3Australian Scientific Instruments, Fyshwick, ACT 2609, Australia

(Received 20 February 2014; accepted 26 July 2014; published online 5 August 2014)

We present a spectrometer based on the cavity enhanced amplitude modulated laser absorption

spectroscopy (CEAMLAS) technique for measuring molecular gas absorption. This CEAMLAS

spectrometer accurately measured a CO2 absorption line at 1572.992 nm with effectively 100%

measurement duty cycle. It achieved an absorption sensitivity of 5.2� 10�9 Hz�1=2 using a linear

Fabry-Perot cavity with a modest finesse of �1000. We also used the spectrometer to perform pre-

liminary measurements of the 13C/12C isotopic ratio in CO2, yielding an isotopic signature d13C of

�8369& for our CO2 sample. VC 2014 Author(s). All article content, except where otherwisenoted, is licensed under a Creative Commons Attribution 3.0 Unported License.

[http://dx.doi.org/10.1063/1.4892535]

Measurement of trace gases and their isotopic ratios are

of interest in a diverse range of applications. These include

the monitoring of atmospheric greenhouse gases;1 geologi-

cal, ecological, and biochemical research;2,3 illicit drug test-

ing;4,5 and natural gas detection in the mining industry.6

These high sensitivity measurements are primarily conducted

in analytical laboratories using mass spectrometry.3–5

However, with advances in optical technologies, similar or

surpassing levels of precision can be achieved using laser

absorption spectroscopy (LAS) techniques.1,2,6 These techni-

ques can significantly reduce instrument cost while their

compactness enables in-situ measurements. One of the lead-

ing optical techniques is cavity ring-down spectroscopy

(CRDS), which was first demonstrated by O’Keefe et al. in

1988.7 This method has been used in many field-deployable

spectrometers to date. The state-of-the-art absorption sensi-

tivity of a CRDS spectrometer is approximately

2� 10�10 Hz�1=2.8,9 Another notable technique is noise-

immune cavity-enhanced optical heterodyne molecular spec-

troscopy (NICE-OHMS) developed by Ma et al., which

holds the record sensitivity of 5� 10�13 Hz�1=2.10 However,

due to the complexity and high cost, portable NICE-OHMS

spectrometers are still under development.11

The high sensitivity of these LAS spectrometers is a

result of using a high quality Fabry-Perot cavity, where the

finesse F is usually larger than 104. This increases the effec-

tive light-gas interaction length by the effective bounce num-

ber b ¼ 2� F=p. The sensitivity in NICE-OHMS systems is

further improved by using frequency modulation of the laser,

and results in a continuous and zero-background absorption

signal with technical noises significantly suppressed. In

2008, Chow et al. demonstrated a proof of concept experi-

ment involving cavity enhanced amplitude modulated laser

absorption spectroscopy (CEAMLAS).12 Like NICE-OHMS,

the CEAMLAS method also provides a real-time, zero-back-

ground absorption signal. At optimal operating condition,

the CEAMLAS signal was immune to laser intensity noise,

and a quantum shot-noise limited sensitivity was achieved in

a fiber cavity with finesse of 60,12 with less complicated

setup than NICE-OHMS. This result offers a new architec-

ture for building a highly sensitivity, field deployable optical

spectrometer using CEAMLAS. In this Letter, we report the

construction and spectroscopic result for a CEAMLAS spec-

trometer, achieving an initial absorption sensitivity of

5.2� 10�9 Hz�1=2 with a cavity finesse of F � 1000. This

initial result is already on par with or even surpassing the

sensitivity of many existing NICE-OHMS11 and CRDS

systems.13

CEAMLAS uses radio frequency (RF) amplitude modu-

lation (AM) of the laser to interrogate the resonant reflection

response of the cavity.12 The modulation frequency xm is

chosen to be much higher than the cavity linewidth. When

the laser carrier is on resonance, the reflected optical field Er

consists of the reflected carrier and the off-resonance AM

sidebands, given by

Er ¼ E0

r1 � r2e�aL

1� r1r2e�aLþ b

2eixmt þ b

2e�ixmt

� �: (1)

In Eq. (1), E0 is the amplitude of the input beam, with optical

power P0 ¼ jE0j2. The first term in the bracket is the reso-

nant reflection response of the cavity, which characterizes

the impedance matching condition. r1 and r2 are the ampli-

tude reflection coefficients of the input and output cavity

mirrors respectively; a is the absorption coefficient of the

intra-cavity gas medium; and L is the length of the cavity.

The other two terms in the bracket correspond to the

reflected AM sidebands with small modulation depth of b.

As shown in Fig. 1, the resonance between the carrier

and the cavity is maintained using the Pound-Drever-Hall

laser frequency stabilization technique.14 The reflected beam

is fully detected on a fast photodiode by optimizing the

a)Author to whom correspondence should be addressed. Electronic mail:

j.dong@anu.edu.au

0003-6951/2014/105(5)/053505/3 VC Author(s) 2014105, 053505-1

APPLIED PHYSICS LETTERS 105, 053505 (2014)

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

130.56.106.27 On: Thu, 17 Dec 2015 03:21:26

orientation of the quarter wave plate. The optical power

Pr ¼ jErj2 is then converted into an electronic signal, and is

subsequently mixed with the local oscillator at xm to isolate

the beat between the sidebands and the resonant carrier.

After passing the down-mixed signal through a low pass fil-

ter and optimizing the demodulation phase, the demodulated

resonant signal Vres can be extracted. In the approximation

that the empty cavity is nearly impedance matched ðr1 �r2 � 1Þ and the absorption is small ðaL� 1Þ, Vres becomes

Vres � P0kFp

r1 � r2ffiffiffiffiffiffiffiffir1r2p þ aL

� �; (2)

where k is a proportionality constant that depends on the

modulation depth, photodiode response, and transimpedance

gain. ðr1 � r2Þ=ffiffiffiffiffiffiffiffir1r2p

quantifies the impedance mismatch of

the cavity mirrors, and is represented as a DC offset in Vres.

As shown in Eq. (2), this is a zero background signal and

immune to laser intensity noise when the empty cavity is im-

pedance matched. In the presence of small impedance mis-

match and small intra-cavity absorption, fluctuations in P0

are still highly suppressed by the resonant reflection response

of the cavity. When the carrier is far off resonance, it can be

shown that the signal Vof f � P0k, thus the AM readout can

be normalized as

VAM ¼Vres

Vof f� F

pr1 � r2ffiffiffiffiffiffiffiffi

r1r2p þ aL

� �: (3)

Equation (3) shows a simple conversion between the

experimental AM readout and the absorption aL. The signal

processing system was digitally implemented using an

analog-to-digital converter (ADC) and a field programmable

gate array (FPGA). Digital signal processing enables rapid

design and implementation of the low pass filters, data acqui-

sition, and subsequent data processing. To obtain a CO2

spectral line, an NKT Photonics Koheras Adjustik E15 fiber

laser was used. It covers the wavelength from 1572.4 nm to

1573.4 nm and has less than 1 kHz linewidth as specified by

the manufacturer. The cavity had a finesse of approximately

1000, and a length of 20 cm, with a stainless steel spacer for

mechanical stability. The output cavity mirror was mounted

on a piezo-electric transducer (PZT) which enabled cavity

length tuning. The cavity was enclosed within a vacuum tank

which can be evacuated and filled with CO2 gas.

Figure 2 shows the demodulated AM readout as we

scanned the laser frequency across a cavity resonance. When

the laser was off resonance, the readout was normalized to 1.

The central trough occured when the laser carrier and the

cavity were resonant (0 frequency detuning). Its positive

value indicated that the cavity was slightly under

coupled.15–17 This impedance mismatch was treated as a

DC offset in a spectral measurement. The two satellite

features in Fig. 2 occured when each of the AM sidebands

became resonant with the cavity. They were associated

with our modulation frequency sidebands of 37.5 MHz. We

used the drive voltage to the laser PZT to calibrate the

frequency axis in Fig. 2, obtaining a calibration factor of

14.6 MHz/V. However, this frequency calibration was only

approximate, as it is known that there can be significant non-

linearity between PZT displacement and drive voltage. This

was confirmed by the unequal spacing of cavity resonance

modes when we scanned the laser PZT with a linear voltage

ramp.

When CO2 is injected into a cavity and the cavity reso-

nant mode is coincident with a molecular transition, the

intra-cavity loss increases. As can be observed in the inset of

Fig. 2, such intra-cavity loss resulted in the rise of the reso-

nant AM readout shown as the red dashed line. This was a

first order effect, with the resonant AM readout strongly

dependent on the strength of gas absorption. One expects the

cavity finesse to be slightly degraded if it was on a transition.

However, this was a secondary effect for the weak absorp-

tion measurement of interest here. Hence, we assumed that

finesse remained constant across the absorption spectrum

when we calibrated the absorption signal.

There are three strong CO2 rovibrational overtone tran-

sitions within the wavelength tuning range of our fibre laser.

We injected 0.32 6 0.03 millibar of room temperature pure

CO2 into the evacuated gas chamber, and found an absorp-

tion line by a quick frequency scan. We then PDH locked the

laser onto the cavity resonance near the line. This represents

one fixed point on the absorption spectrum. The cavity

length was then scanned by a sawtooth voltage at a period of

50 s to tune the cavity resonant mode. The laser wavelength

closely tracked the scanning cavity resonance because it was

FIG. 1. Simplified schematic of the CEAMLAS spectrometer. Abbreviations

are PM: phase modulator; AM: amplitude modulator; PBS: polarizing beam

splitter; k=4: quarter waveplate; PZT: piezoelectric transducer; LO: local os-

cillator; PD: photodiode; VAM: AM readout signal.

FIG. 2. AM readout signal VAM obtained when the laser is scanned through

a cavity resonance, with the AM sidebands at 37.5 MHz. Inset: Comparison

of the AM readout around the resonance, when it is off (blue solid line) and

on (red dashed line) a gaseous molecular transition respectively.

053505-2 Dong et al. Appl. Phys. Lett. 105, 053505 (2014)

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

130.56.106.27 On: Thu, 17 Dec 2015 03:21:26

locked, thereby mapping out the CO2 absorption spectrum

while the mirror was translated longitudinally.

Figure 3 shows the absorption spectrum obtained for the

R12e line in the 30012 00001 band of CO2,18 with the

vertical axis calibrated by Eq. (3) using F � 1000. The drive

voltage to the laser PZT became the feedback control signal,

because the laser was locked to the cavity resonance. As in

Fig. 2, we used this drive voltage for calibration of our hori-

zontal frequency axis. This CEAMLAS spectrum was meas-

ured in less than 4 min.

Figure 3 shows that the spectral line measured by the

CEAMLAS spectrometer had a high signal to noise ratio

(more than 1000). At 0.32 6 0.03 millibar of CO2 pressure,

Doppler broadening dominates over pressure broadening;

therefore, the absorption line can be modeled by a Gaussian

line shape.18 The measured profile was well fitted by this

model. The full width at half maximum of the absorption

line yield a gas temperature of 20 �C. This was in good

agreement with an independent measurement taken with a

thermometer. A gas pressure of 0.29 millibar was obtained

from the fitted profile, using the spectral line intensity of

1.663� 10�23 cm�1/(mol cm�2) for this R12e transition

line.18 This was also in close agreement with an independent

measurement from a pressure gauge.

The sample spacing in Fig. 3 was approximately 50

kHz, and can be further narrowed by using a slower scan

and/or faster data acquisition. This spacing over a large mea-

surement span of the order of 1 GHz implied a continuous,

real-time spectral measurement, enabling effectively 100%

measurement duty cycle. With an absolute frequency refer-

ence such as a frequency comb, such small frequency spac-

ing will enable high precision measurement of the

fundamental Voigt lineshapes of molecular transitions.

To characterize the sensitivity of this first generation

spectrometer, we measured the AM readout at a fixed spec-

tral point in vacuum. The typical result of a measurement is

illustrated in Fig. 4. The standard deviation for the noise was

DaL¼ 1.6� 10�7 over a time span of 1 s, which agrees with

the inset of Fig. 3. As in a CRDS system, the noise equiva-

lent absorption (NEA) sensitivity of this CEAMLAS spec-

trometer was calculated as

NEA ¼ raLffiffiffifp ; (4)

where raL is the standard deviation of the absorption and f is

the data acquisition rate.9 In Fig. 4, the sampling rate was

1 kHz, so the NEA of the initial CEAMLAS spectrometer

was 5.2� 10�9 Hz�1=2. This was achieved with a relatively

modest cavity finesse of 1000, whereas the best CRDS sensi-

tivity of 2� 10�10 Hz�1=2 was achieved with a cavity of

much higher finesse (F � 20000 used by Spece et al.8 and

Hodges et al.9).

Our investigations suggested that this initial sensitivity

of CEAMLAS spectrometer was limited by technical noise

sources, such as time varying beam pointing error. This was

confirmed by gently dithering the cavity alignment and

observing the resultant noise increase. These technical limi-

tations will be overcome with improved opto-mechanical

design in the next generation CEAMLAS systems.

To demonstrate the potential application of the

CEAMLAS spectrometer, we measured the 13C/12C isotopic

ratio of our gas sample. This was achieved by changing the

laser to a New Focus Vortex II external cavity tunable diode

laser (ECDL). This laser has tunable wavelength from

1596.6 nm to 1597.1 nm. A transition line R60e in the

30013 00001 band of 12CO2 and an R30e line in the

30013 00001 band of 13CO2 are present within this wave-

length region.18 These two absorption lines have similar

absorption strengths at natural abundance. The solid line in

Fig. 5 shows this pair of isotopic absorption lines measured

by the CEAMLAS spectrometer at 10 millibar of pure CO2

and room temperature.

FIG. 3. R12e line in the 30012 00001 band of CO2 measured by the

CEAMLAS spectrometer, fitted by a Gaussian lineshape. Inset: Noise of the

CEAMLAS measurement.

FIG. 4. Fractional absorption measured by the CEAMLAS spectrometer at a

fixed wavelength, away from any gas absorption lines. This indicates the

time-varying noise of the spectral measurement.

FIG. 5. R60e line in the 30013 00001 band of 12CO2 and R30e line in the

30013 00001 band of 13CO2 measured by the CEAMLAS spectrometer,

fitted by a Voigt lineshape.

053505-3 Dong et al. Appl. Phys. Lett. 105, 053505 (2014)

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

130.56.106.27 On: Thu, 17 Dec 2015 03:21:26

Overlaid in Fig. 5 as a dashed line is the calculated spec-

trum. In contrast to Fig. 3, a Voigt profile was assumed

instead of a Gaussian fit, because, at 10 millibar, the effect of

pressure broadening becomes observable alongside Doppler

broadening.18 Using the independently measured total pres-

sure and room temperature, we fitted the 13CO2 absorption

line and obtained a concentration of 1.01%. From the fit

residual, we estimated an uncertainty in 13CO2 concentration

of 0.01%. The fitted 13CO2 profile stayed well within one

standard deviation of the measured absorption data, as evi-

dent from Fig. 5.

In comparison, we note that there is observable differ-

ence in the widths between experimental and calculated

Voigt line shapes for the 12CO2 absorption line. We attrib-

uted the width mismatch to the calibration error in wave-

length. This was likely due to the laser PZT nonlinearity as

previously discussed. The 12CO2 absorption line was

obtained near the end of the laser PZT tuning range, where

the most significant nonlinearity was manifested. There are

various methods to calibrate for this nonlinearity if it was

desired, for example, by using a wavelength meter or abso-

lute referenced frequency comb, which is planned for the

next generation of CEAMLAS spectrometer. Based on the

gas mixing ratios used in the two fit, the isotopic ratio of13C/12C was approximately 0.0103 6 0.0001. The sample13C isotopic signature d13C was then calculated as

d13C ¼13C=12Csample

13C=12Cstandard

� 1

!� 1000&; (5)

where 13C/12Cstandard is referred to an established standard of

0.0112372.2 Thus, for the CO2 sample we tested, d13C was

�8369& which was within the naturally occurring d13C

value from 0& to �110&.2 The signal to noise ratio in this

isotopic ratio measurement was degraded compared to that

in Fig. 3. This was because of the weaker transition and the

higher frequency noise of the ECDL. Since the isotopic ratio

of the CO2 we used has not been calibrated, the accuracy of

this isotopic ratio measurement awaits further confirmation.

In conclusion, this Letter reports the demonstration of

molecular gas sensing with a CEAMLAS spectrometer,

accurately measuring a rovibrational transition for CO2 in

real-time, with high spectral resolution. The initial noise

equivalent absorption sensitivity is 5.2� 10�9 Hz�1=2,

achieved using a modest finesse cavity with F � 1000. In

addition, we have demonstrated that CEAMLAS can readily

be adapted to measure the 13C/12C isotopic ratio of CO2.

This research was funded by the Australian Research

Council under the Project ID: LP100200604 and

DE130101361.

1E. Crosson, Appl. Phys. B 92, 403 (2008).2E. Crosson, K. Ricci, B. Richman, F. Chilesse, T. Owano, R. Provencal,

M. Todd, J. Glasser, A. Kachanov, and B. Paldus, Anal. Chem. 74, 2003

(2002).3J.-P. Godin, L.-B. Fay, and G. Hopfgartner, Mass Spectrom. Rev. 26, 751

(2007).4M. Becchi, R. Aguilera, Y. Farizon, M.-M. Flament, H. Casabianca, and P.

James, Rapid Commun. Mass Spectrom. 8, 304 (1994).5C. Saudan, N. Baume, N. Robinson, L. Avois, P. Mangin, and M. Saugy,

Br. J. Sports Med. 40, i21 (2006).6B. Fawcett, A. Parkes, D. Shallcross, and A. Orr-Ewing, Phys. Chem.

Chem. Phys. 4, 5960 (2002).7A. O’Keefe and D. Deacon, Rev. Sci. Instrum. 59, 2544 (1988).8T. Spence, C. Harb, B. Paldus, R. Zare, B. Willke, and R. Byer, Rev. Sci.

Instrum. 71, 347 (2000).9D. Long, G. Truong, R. van Zee, D. Plusquellic, and J. Hodges, Appl.

Phys. B 114, 489 (2014).10L. S. Ma, J. Ye, P. Dub�e, and J. L. Hall, J. Opt. Soc. Am. B 16, 2255

(1999).11A. Foltynowicz, F. M. Schmidt, W. Ma, and O. Axner, Appl. Phys. B 92,

313 (2008).12J. H. Chow, I. M. C. Littler, D. S. Rabeling, D. E. McClelland, and M. B.

Gray, Opt. Express 16, 7726 (2008).13Y. He and B. J. Orr, Appl. Phys. B 85, 355 (2006).14R. W. P. Drever, J. L. Hall, F. V. Kowalski, J. Hough, G. M. Ford, A. J.

Munley, and H. Ward, Appl. Phys. B 31, 97 (1983).15D. S. Rabeling, J. H. Chow, M. B. Gray, and D. E. McClelland, Opt.

Express 18, 9314 (2010).16J. H. Chow, M. A. Taylor, T. T.-Y. Lam, J. Knittel, J. D. Sawtell-Rickson,

D. A. Shaddock, M. B. Gray, D. E. McClelland, and W. P. Bowen, Opt.

Express 20, 12622 (2012).17J. Knittel, J. H. Chow, M. B. Gray, M. A. Taylor, and W. P. Bowen, Opt.

Lett. 38, 1915 (2013).18L. Rothman, I. Gordon, Y. Babikov, A. Barbe, D. C. Benner, P. Bernath,

M. Birk, L. Bizzocchi, V. Boudon, L. Brown et al., J. Quant. Spectrosc.

Radiat. Transf. 130, 4 (2013).

053505-4 Dong et al. Appl. Phys. Lett. 105, 053505 (2014)

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:

130.56.106.27 On: Thu, 17 Dec 2015 03:21:26