Frequency metrology with quantum cascade lasers

Invited Paper

Frequency metrology with quantum cascade lasers Saverio Bartalini, Simone Borri, Iacopo Galli, Davide Mazzotti, Pablo Cancio Pastor, Giovanni Giusfredi and Paolo De Natale Istituto Nazionale di Ottica Applicata (INOA) - CNR, Largo E. Fermi 6, 50125 Firenze, Italy, and Laboratorio Europeo di Spettroscopie Non-lineari (LENS), via N. Carrara 1, 50019 Sesto Fiorentino (FI), Italy ABSTRACT We recently reported the first Doppler-limited absolute frequency measurement of CO2 transitions around 4.4 μm wavelength, by linking a DFB Quantum Cascade Laser (QCL) to an Optical Frequency Comb Synthesizer (OFCS). We further achieved sub-Doppler recording of these transitions, improving of about three orders of magnitude the measurement precision. We are exploring techniques able to significantly reduce the QCL jitter, in order to get metrological-grade QCLs for very demanding experiments in the frequency-domain. The latest experimental results in our group will be reported. Keywords: Quantum Cascade Lasers, Frequency Metrology, Infrared Spectroscopy

1. INTRODUCTION Since their advent, Quantum Cascade Lasers (QCLs) have almost uniformly covered the mid and far infrared, a region of the electromagnetic spectrum which has suffered the lack of coherent sources for a long time. This has prevented the development of mid-IR-based spectrometers in many fields of applied physics but, thanks to the recent availability of commercial QCLs, this spectral region is gaining more and more importance for a huge number of applications. The areas of interest of this spectral region range from astrophysics to climatology, from trace gas detection in environmental monitoring to homeland security. Many of these applications, especially in industry and space fields, require not only high-performance, but also compact, reliable and affordable devices, and could greatly profit from the advent of next generation QCLs. Up to now, indeed, these sources have demonstrated to be very suitable for high-sensitivity and high-resolution spectroscopy,1–3 thanks to their good stability and relatively high emission power. In our group, in the last years, we demonstrated the versatility of QCLs for high sensitivity spectroscopic applications by implementing several the main frequency modulation techniques for sub-ppb detection of trace gases (N2 O, CH4 and CO2 ).4 The next important challenge is to provide QCLs with the frequency stability and the spectral purity required for any metrological-level application. In fact, the typical spectral purity of single-mode free-running QCLs has always been observed in the range of few MHz5, 6 and is mainly ascribed to the driving current noise, together with the absence of an effective frequency selection mechanism. Even with an external cavity configuration, as demonstrated by the first results published,7, 8 no significant narrowing of the QCL emission spectrum has been observed. A drastic linewidth narrowing can be achieved only by an active stabilization of the laser frequency, for example by implementing a feedback loop on the QCL current:9–11 with these techniques narrowing up to the kHz, or even the Hz,12 level has been obtained. However, in absence of an absolute frequency reference, these techniques are not suitable for providing metrological-grade QCLs. This manuscript describes our activity in referencing a mid-IR QCL to an absolute frequency standard through an optical link with a near-IR optical frequency comb synthesizer (OFCS) and our efforts in developing simple Further author information: S. Bartalini: E-mail: [email protected], Telephone: +39 (0)55 4572500 Quantum Sensing and Nanophotonic Devices VI, edited by Manijeh Razeghi, Rengarajan Sudharsanan, Gail J. Brown, Proc. of SPIE Vol. 7222, 72220C · © 2009 SPIE · CCC code: 0277-786X/09/$18 · doi: 10.1117/12.813519

Proc. of SPIE Vol. 7222 72220C-1

and reliable techniques for stabilization and narrowing of the QCL emission frequency. The basic experimental setup is sketched in Sec. 2. Preliminary results for absolute frequency measurements of CO2 transitions around 4.3 and 4.4 μm wavelengths are presented in Sec. 3. They have been performed both in a Doppler-limited and sub-Doppler spectroscopic experiments. In the second case, an important improvement came from locking the QCL frequency to the first derivative of the Lamb dip obtained with a conventional wavelength modulation technique. The same optical link to the OFCS also allows a direct analysis of the emission spectrum of the laser in free-running and locked operation and the measurement of its long-term frequency stability, as well as the effectiveness of the locking loop (Sec. 4). The analysis of the dependence of the laser emission on the driver current noise (Sec. 5) is performed by comparing different current sources, including an ultra low-noise home made supply that provides the best results. In Sec. 6 the first results of polarization spectroscopy with a mid-IR QCL will be described. This technique provides a dispersion-like signal without any modulation of the laser frequency, thus allowing the locking of the laser to the transition center with an enhanced bandwidth and a simplified electronics.

2. BASIC EXPERIMENTAL SETUP We have used single-mode distributed-feedback (DFB) QCLs emitting around 4.3 and 4.4 μm and operating at cryogenic temperatures. The lasers have a quite low threshold current (about 100 mA at 86 K) and a frequency tunability with current of 400 MHz/mA. They are housed in a liquid-N2 -cooled cryostat and supplied by a lownoise current driver. Since the electrical isolation of the current driver is not sufficient to reject additional noise coming from the AC line, which induces a spurious modulation of the QCL frequency, we adopt an UPS battery supply for powering the current driver. In Fig. 1, simplified schematics of the experimental set-up are shown, describing the two optical layouts used for Doppler-limited (direct absorption spectroscopy) and sub-Doppler (saturated-absorption spectroscopy) configurations. Their details are reported in Refs. 13, 14. In order to set an optical link with the near-IR OFCS, an up-conversion of the QCL frequency is required. For this reason, in both cases a fraction of the laser beam is sent to a second stage of the experiment, devoted to the sum-frequency generation (SFG) process and to the optical link with the OFCS. This stage is exactly the same for both the layouts. The OFCS used in this experiment is a mode-locked femtosecond Ti:sapphire laser operating in the 500÷1100 nm range. The QCL frequency has been mixed into a non-linear periodicallypoled LiNbO3 crystal with a Nd:YAG laser (λY AG = 1.064 mm) referenced to the comb. The obtained SFG

a

W ZnSewindow

to PC

RM: removable minor

BS beanr-aplitter Detector

/ I_I r'W

Current Dr uen

Absorption Cell

Tennp.

Fber-AnnpIifed @1064 nor Nd:YAG ever

Ditirackon

C

I

Gimbng

@858 nor

vPPLN Crystal

Phase-Lock

Freque cy-Cornb Filter

Controller

, / RM

Infrared Camera,,

DFB-QCL

© 44 pm

L1 Monitor

Sum-Frequency Radiation @858 nm

PPLN Cryetol

Cryostat

Lqu d N

Filter

Froqeeney.Cornb

DFB.QCL @4.3 jm Sem-Freqeency Redletlon @ 854 nn.

Freqoency Coonter

V V

Spectmm Maimer

Spectrorn Photodiode

MIizer

Figure 1. Schematic of the experimental set-up (left: Doppler-limited single-pass absorption spectroscopy; right: standard pump-probe saturation spectroscopy with electronic feedback to the laser). In the figure the main blocks of the apparatus have been highlighted by dashed lines: the QCL housing, collimation and driving electronics (a), the saturation spectroscopy set-up (b), the SFG assembly (c), the near-IR lasers phase-locked to the comb (d) and the beat-note detection and measurement (e). The cell is 12 cm long and has AR-coated CaF2 windows. P1 and P2 are wire grid polarizers, λ/2 is a tunable half-waveplate and BS is a beam-splitter.

Proc. of SPIE Vol. 7222 72220C-2

radiation, at a wavelength of about 858 nm in the first experiment and about 854 nm in the second one, is superimposed to the beam of a diode laser (DL), working at a close wavelength and referenced to the OFCS, too. A fast photodiode detects the radio-frequency beat-note. The signal is analyzed by a spectrum analyzer and its frequency measured by a counter. For referencing the Nd:YAG and the diode laser to the OFCS a direct digital synthesis (DDS) technique is used.15, 16 This makes the beat note spectrum to exactly reproduce the line-shape of the QCL emission spectrum. By counting the beat note frequency (νbeat ) it is possible to obtain the absolute value of the QCL frequency by (1) νQCL = νDL − νY AG + νbeat , where νY AG e νDL are the Nd:YAG and DL frequencies respectively. These two frequencies are known with the OFCS precision and have few kHz linewidths, negligible in comparison with the QCL linewidth. The beat-note frequency is counted by a commercial frequency counter and is analyzed by a real-time FFT spectrum analyzer, in order to characterize the QCL emission linewidth and stability as shown in Sec. 4

3. ABSOLUTE MEASUREMENT OF CO2 TRANSITION FREQUENCIES In this section, a comparison between the absolute measurements of two molecular transition frequencies, performed in Doppler-limited and sub-Doppler regimes, are presented. In the first case, the free-running laser frequency is slowly tuned across the 13 CO2 (000 1 − 0000) P(30) Doppler-broadened molecular line by controlling the driving current with a step-shaped signal. The absorption signal and the beat-note frequency are simultaneously acquired at rate of 500 ms, thus reconstructing the molecular profile on an absolute frequency scale, as shown in Fig. 2 and as described in details in Ref. 13. Several spectra have been acquired for the same transition, even at slightly different gas pressures. Each data set has been fitted with a Voigt profile (Fig. 2) in order to determine the corresponding line-center frequency, whose average value results in νc = 67678.415 GHz, with a spread of ±2 MHz. The result is consistent with the value provided by the HITRAN database17 (ν = 67678.413 GHz, with an error between 3 and 30 MHz). The achieved precision is limited by two instability sources: temperature fluctuations, leading to wide low-frequency fluctuations, and driving current instabilities, which add frequency noise in the whole current driver band (about 100 kHz). By implementing the sub-Doppler technique, we have been able not only to obtain a narrower reference, the Lamb dip, but to use its first derivative to actively stabilize the laser frequency (left part of Fig. 3, see Ref. 14 0,750,700,65(3

0,600,55-

C

0 0

0,50 -

Center Frequency v=67678,415 GHz

0,450,40-

P = 250 mTorr

0,35-

13CO2 (00°1-00°0) P(30)

0,30 67678,0

67678,2

j ci

0

67678,4

67678,6

67678,8

Transition absolute frequency (GHz)

0,020,00_-

:

%S.

-

-0,02 67678,0

67678,2

67678,4

67678,6

67678,8

Transition absolute frequency (GHz)

Figure 2. Absolute frequency measurement on the 13 CO2 (000 1 − 000 0) P(30) Doppler-broadened transition with a gas pressure P = 250 mTorr in the cell. The non uniform distribution of the points is due to a slow temperature drift of the QCL during the acquisition. The red line shows the Voigt fit of the data.

Proc. of SPIE Vol. 7222 72220C-3

180 40

149,8

laser unlocked

P=20 miorT

laser locked

4 aerages

170-

I N

10

149,7 -

-

160-

a)

Dip: OausEiSflhlt HWHM = 5.34 MHz

_j

150-

a) u_

.

149,6 -

C

r

1495

-

149,4-s..p.

140-

t%,

Zoom lOOx

130-

a -1

-50

0

50

Frequency N (MHz)

100

150

40

80

120

Acquisition Time (s)

s

149,3 -

149,2

120 -150

:.

b 100

300

500

700

900

1100

1300

Acquisition time (s)

Figure 3. Left: saturation spectrum of the (011 1 − 011 0) P(30) 12 CO2 transition in direct-absorption and first-derivative detection, with a gas pressure in the cell of 20 mTorr (which brings to a relative absorption of about 10%). The pump intensity interacting with the gas sample was about a factor 2 greater than the saturation level. The first-derivative signal was obtained by a lock-in amplifier with 1 ms integration time constant and its best fit is also shown. Right: (a) comparison between the spreads of the beat-note frequency values as measured by the counter in free-running and locked conditions; (b) 20-minutes-long acquisition of the beat-note frequency, shown in a 100× zoomed scale. The right part of the figure shows the frequency distribution of the data: a Gaussian fit is superimposed and the resulting 2 kHz standard deviation of the mean can be assumed as the precision of this measurement.

for further details). The best fit of the experimental data has been obtained with a convolution between two Gaussian functions, accounting for both the Doppler profile and the Lamb dip. The Lamb-dip HWHM resulting from the fit is about 5 MHz, more than 10 times narrower than the Doppler width, and is mainly determined by the QCL frequency jitter, as described in next section. A small bandwidth (500 Hz) feedback loop on the QCL current is good enough for a sensible reduction of the slow frequency fluctuations, as the right part of Fig. 3 clearly shows. This allows to perform long frequency measurement sessions, consisting in consecutive 1-s gate-time countings of the beat-note frequency. Each data set yields the mean frequency value νbeat with the associated uncertainty (standard deviation of the mean). The 3.5 kHz average value of the these errors can be assumed as a reasonable estimate for the ultimate precision (5 parts in 1011 ) achievable in the absence of all the systematics, that actually limit the uncertainty to 75 kHz. The improvement with respect to the Doppler-limited case is evident and the overall uncertainty is at least 2 order of magnitude smaller than the HITRAN error.

4. FREQUENCY STABILITY AND QCL LINEWIDTH MEASUREMENTS The results reported above are only a small step toward a metrological-grade stabilization of the frequency of a QCL. A further investigation of the frequency noise features and its sources are required. In the following, the first results of the analysis of the beat-note spectrum over different time-scales, in free-running and locking conditions, are presented. Figure 4 shows a comprehensive frame of the laser frequency noise as acquired with two independent techniques: the direct acquisition of the FFT spectrum of the beat note, as explained above, and the analysis of the laser frequency noise. The latter has been possible by means of a frequency-to-amplitude noise conversion of the laser frequency fluctuations by using the side of a Germanium etalon as discriminator. The first method allows to understand the statistical distribution of the noise itself over time, while the second is useful for a more quantitative noise evaluation, of its typical bandwidth and of the locking-loop effects on it. For the free-running QCL, the frequency noise spectral density clearly shows a 1/f behavior (Fig. 4, top) over the whole detection bandwidth (100 kHz). The emission spectrum, on the right, is averaged over timescales larger than 1 ms and is almost flat, since no correction to the frequency jitter is applied.

Proc. of SPIE Vol. 7222 72220C-4

M*: 9375 M9 -73.993

-121593$ -392 fr[

135

140

145

150

155

160

165

FroqoerIcy (MHZ)

- 70Ff beat-note

1

I I

TOWI QCLfr qeny nOise - Freqoeny noise when locked

0,0 135

F,eqency (k}k)

140

145

150

155

150

155

Frequency (MHZ)

Figure 4. Comparison between the laser frequency noise and jitter in free-running (top line) and locked (bottom line) conditions. The graphs on the left show the noise power spectral density of the laser radiation, recorded as explained in the text. The comparison shows the locking bandwidth (500 Hz), as expected by considering the 1 ms lock-in time constant. The central part of the figure shows the evolution, over 10 ms, of the beat note as detected by our real-time spectrum analyzer. The black profiles on the right are the FFT spectra of 70 ms-long acquisitions of the beat-note in free-running and locking conditions, while the gray trace is the FFT of a 70 μs-long acquisition.

In the bottom line of Fig. 4 the effects of the locking loop on the current noise spectral density are shown: a suppression of the noise power spectral density up to 20 dB arises, but only within a small bandwidth. However, it produces a visible reduction of the beat-note frequency fluctuations, that now lie well within the analyzer spectral window. It also allows a comparison between a fast shot (70 μs long) of the emission spectrum and its long-averaged profile (over 70 ms) both taken with the laser locked: the fast linewidth appears very narrow but, even in locked condition, the unsuppressed fast jitter determines a gaussian broadening of the line shape over longer timescales. Anyway, the regular shape of this gaussian envelope and its long-term stability, ensured by the locking, allows to count the frequency leading to the results presented in the previous section. The 70 μs-long shot of the emission spectrum probably reveals the existence of a very narrow fast linewidth of the QCL that, once achieved, would make this kind of sources suitable for any kind of high precision metrological experiment. In order to succeed in this challenge, two improvements are needed: on one hand the construction of a lower noise current driver, in order to reduce as much as possible its contribution to the frequency noise; on the other hand the development of a frequency locking technique with a wider bandwidth, able to compensate for the residual frequency jitter, thus further narrowing the emission linewidth down to its intrinsic value.

5. CURRENT NOISE MEASUREMENTS Even if the dependence of the QCL frequency noise from the supply current noise is experimentally proven, it is not clear if current noise can explain all the jitter that broadens the laser emission (see previous section). We are developing an ultra low-noise home made driver, which will substitute the commercial driver in use at present. At the same time, we are developing a low-noise, high bandwidth current noise amplifier that will enable simultaneous measurement of the laser frequency noise and of the driver current noise, thus allowing to look for a correlation between these two quantities. The work is still in progress, but a preliminary comparison between the beat-note spectra corresponding to the QCL powered by the commercial and the home-made drivers

Proc. of SPIE Vol. 7222 72220C-5

Marker: 913.1 P4t -61.0% m

-lola -3 frame

frame

I frame

Figure 5. Evolution of the beat note with the QCL powered by a commercial low-noise driver (left) and by a homemade ultra low-noise driver (right). Different current noise levels determine different fluctuation amplitudes of the QCL frequency.

is encouraging. In Fig. 5, the comparison is shown: the slow jitter is present in both cases but, with the homemade driver, the width of the envelope over short timescales is narrower, demonstrating a lower noise contribution at high frequencies.

6. POLARIZATION SPECTROSCOPY Together with the development of current drivers with better noise performances and higher modulation bandwidths, the new approach of polarization spectroscopy will be an important milestone towards the solution of the problems cited above. In the following, advances in this direction will be presented. Polarization spectroscopy is among the most sensitive spectroscopic techniques.18 It is a pump-probe technique, like saturation spectroscopy, and it is based on detection of an optical anisotropy induced in a medium by the pump beam. It has been first demonstrated for atomic transitions in the presence of degenerate Zeeman sublevels, and it works for all the molecular ro-vibrational molecular transitions in the mid-infrared region. Its basic principle is to induce a birefringence in a medium, with a circularly polarized pump beam, and to interrogate this with a linearly polarized counterpropagating probe beam, watching for the rotation of its polarization. It can be regarded as a form of saturation spectroscopy, with the change of the complex refractive index being proportional to the pump intensity. The key concept is the production of a dispersive signal suitable for frequency locking without any modulation (dither) on the driving current. The absence of the dither offers a direct benefit from

Linearly-polarized probe" beam

P2

-

Temp.

T-?J2

P1

Circularly-polarized "pump" beam

Controller

T-?J4

Current Driver

Ce Ce

0)

Absorption Cell

-o

0 0 0

0 to the SFG apparatus

3

0 lnSb Detector

P1, P2, P3: wire-grid polarizers T-?J2: tunable half-wave plate T-?J4: tan, quarter-wave plate

Detector

Amplifier

PID for servo lock

-10

-5

0

5

IS

Frequency sweeping signal (Volts)

Figure 6. Left: schematic of the polarization spectroscopy setup. The analysis polarizer (P3), is nearly crossed with respect to P2. The electronic apparatus required for locking is omitted. Right: by tuning the orientation of the analysis polarizer P3 it is possible to continuously move from the standard saturated absorption signal to the polarization spectroscopy signal. The polarization signal is intrinsically weaker since it is obtained with nearly crossed polarizers.

Proc. of SPIE Vol. 7222 72220C-6

50

p=20 mTorr 4 averages

Polarization signal

800-

- 1000

> - 800

E

Lock-in first derivative signal

- 1000

Polarization signal

600400-

- 600 0 Co

N

-400-

700

U-

-

0

-600-

Direct absorption signal

-150

-100

-50

0

0

50

100

150

-800-

p=20 mTorr 4 averages -40

-30

- 600 -20

-10

10

20

30

40

Frequency Af(MHz)

Frequency Af(MHz)

Figure 7. Left: comparison between the direct saturated-absorption signal and the polarization spectroscopy signal. Both are obtained without any dither on the driver current. Right: comparison between the sub-Doppler feature acquired with first-derivative lock-in technique and with polarization spectroscopy. In the latter case the signal is sharper and its bandwidth larger, thanks to the absence of the dither.

the point of view of spectral purity and allows full use of the available modulation bandwidth for the locking loop. Moreover, it simplifies the electronics involved in the locking loop (the lock-in is no more required). Also, from the point of view of the loop gain, polarization spectroscopy represents an important advance, thanks to the enhancement of the signal-to-noise (S/N) ratio. The changes made in the experimental set-up, compared with the saturated absorption layout, have involved only the manipulation of the beams polarizations, as shown in the left part of Fig. 6. The tunable λ/2 waveplate, in combination with the first wire grid polarizer (P1), splits the QCL beam into a high power pump and a weak probe beams. The polarization of the former is set to circular by the tunable λ/4 waveplate. The polarization of the probe beam is linear since it is already determined by the first polarizer, but it is further cleaned by the P2 polarizer, just before the spectroscopy cell. After the cell, the probe beam encounters the analysis polarizer (P3) and its transmitted component is detected. In the first test performed, the polarizers P2 and P3 are nearly crossed: this means that the detected signal is very weak, compared with the typical amplitude of a saturated absorption signal (Fig. 6, right). For this reason, a larger amplification has been applied to the detected signal. The dispersive profile appears in place of the Lamb dip, or better overlapped to it, since the background sub-Doppler profile is still present, and grows faster when moving far from the nearly crossed position. On the other hand, for perfectly crossed polarizers, no dispersion signal is present. Due to the residual background, the signal is not symmetric and suffers from offset fluctuations. A test of frequency locking performed with this signal did not lead to significant improvements with respect to the first-derivative locking loop, mainly because of amplitude fluctuations of the error signal itself. However, the improvements made in terms of sensitivity and bandwidth of the signal are noticeable. Fig. 7 (left) shows a comparison between a simple saturated absorption signal and the polarization signal: a S/N enhancement of more than a factor 5 is achieved without any effective bandwidth reduction. On the other hand, Fig. 7 (right) points out the improvement, in terms of bandwidth, in comparison with the first-derivative dither-locking technique. It is evident how, in that case, a comparable signal-to-noise ratio is obtained at the expenses of a smaller signal bandwidth. The comparison between the slopes is a further evidence of this fact. In alternative to the single detection scheme, a double-detection differential acquisition can be implemented: the analysis wire-grid polarizer (P3) is rotated by 45◦ with respect to its original position, and both the transmitted and reflected components are detected. In this way, the polarization effect is equally split into the two channels with opposite sign, while the common background is split with the same sign. The differential acquisition allows to cut off the background contribution, together with the related amplitude fluctuations. At the same time, the polarization signal is enhanced and assumes a symmetric shape, ideal for the locking signal. Even

Proc. of SPIE Vol. 7222 72220C-7

400

200

-200

Figure 8. The simulation (dark red line) describes very well the experimental data acquired in single detection mode (green points) and allows to reconstruct the signal expected in the same experimental conditions by switching to the differential detection mode (blue line).

if we still have not implemented the upgrade to double-detection polarization spectroscopy, we can simulate the expected signal. First of all we have checked if the theoretical model presented in Ref. 19 is able to reproduce the signals acquired in single detection polarization spectroscopy. This has been successfully verified, and Fig. 8 shows the good agreement between an experimental profile and its fit performed with the model equation. At this point it is possible to conveniently modify the parameters in order to reproduce the differential acquisition signal expected in the same experimental conditions. The result is shown in the same Fig. 8 and is consistent with the expectations: the signal is enhanced, perfectly symmetric and with zero background. We intend to use this signal for our next frequency locking loop, and we expect significant improvements on all the critical aspects: bandwidth, gain, sensitivity and stability. It has to be pointed out that here the width of the sub-Doppler feature is limited by the laser emission linewidth: with a better current driver we plan to see even sharper features, while with the standard sub-Doppler first-derivative spectroscopy the presence of the modulation would have prevented a better resolution.

7. CONCLUSIONS In conclusion, we have implemented several schemes to get “metrological-grade” OFCS-referenced-QCL absolute frequency measurements on Doppler-broadened CO2 ro-vibrational profiles and, later, we moved to a sub-Doppler saturation spectroscopy apparatus and to frequency stabilization of the laser to the Lamb dip. This allowed to increase the precision of our absolute frequency measurements up to 5 parts in 1011 . This represents a convenient way to extend the OFCS spectral coverage to the mid-infrared: QCLs, in fact, can ensure good stability combined with high-power outputs, thus allowing high-sensitivity detection levels. A new approach based on polarization spectroscopy has been implemented in order to improve the frequency-locking loop by overcoming many of the limitations occurred with the previous first-derivative dither-locking technique. Preliminary results and perspectives have been reported.

REFERENCES [1] Gagliardi, G., Tamassia, F., Natale, P. D., Gmachl, C., Capasso, F., Sivco, D. L., Baillargeon, J. N., Hutchinson, A. L., and Cho, A. Y., “Sensitive detection of methane and nitrous oxide isotopomers using a continuous wave quantum cascade laser,” Eur. Phys. J. D 19, 327 (2002). [2] Bakhirkin, Y. A., Kosterev, A. A., Curl, R., Tittel, F., Yarekha, D., Hvozdara, L., Giovannini, M., and Faist, J., “Sub-ppbv nitric oxide concentration measurements using cw thermoelectrically cooled quantum cascade laser-based integrated cavity output spectroscopy,” Appl. Phys. B 82, 149–154 (2006).

Proc. of SPIE Vol. 7222 72220C-8

[3] Castrillo, A., Tommasi, E. D., Gianfrani, L., Sirigu, L., and Faist, J., “Doppler-free saturated-absorption spectroscopy of CO2 at 4.3 μm by means of a distributed feedback quantum cascade laser,” Opt. Lett. 31, 3040–3042 (2006). [4] Borri, S., Bartalini, S., Natale, P. D., Inguscio, M., Gmachl, C., Capasso, F., Sivco, D. L., and Cho, A. Y., “Frequency modulation spectroscopy by means of quantum-cascade lasers,” Appl. Phys. B 85, 223–229 (2006). [5] Ganser, H., Frech, B., Jentsch, A., M¨ urtz, M., Gmachl, C., Capasso, F., Sivco, D. L., Baillargeon, J. N., Hutchinson, A. L., Cho, A. Y., and Urbana, W., “Investigation of the spectral width of quantum cascade laser emission near 5.2 μm by a heterodyne experiment,” Opt. Comm. 197, 127–130 (2001). [6] Weidmann, D., Joly, L., Parpillon, V., Courtois, D., Bonetti, Y., Aellen, T., Beck, M., Faist, J., and Hofstetter, D., “Free-running 9.1 μm distributed-feedback quantum cascade laser linewidth measurement by heterodyning with a C18 O2 laser,” Opt. Lett. 28, 704–706 (2003). [7] Maulini, R., Yarekha, D. A., Bulliard, J.-M., Giovannini, M., and Faist, J., “Continuous-wave operation of a broadly tunable thermoelectrically cooled external cavity quantum-cascade laser,” Opt. Lett. 30, 2584–2586 (2005). [8] Mohan, A., Wittmann, A., Hugi, A., Blaser, S., Giovannini, M., and Faist, J., “Room-temperature continuous-wave operation of an external-cavity quantum cascade laser,” Opt. Lett. 32, 2792–2794 (2007). [9] Williams, R. M., Kelly, J. F., Hartman, J. S., Sharpe, S. W., Taubman, M. S., Hall, J. L., Capasso, F., Gmachl, C., Sivco, D. L., Baillargeon, J. N., and Cho, A. Y., “Kilohertz linewidth from frequency-stabilized mid-infrared quantum cascade lasers,” Opt. Lett. 24, 1844–1846 (1999). [10] Taubman, M. S., Myers, T. L., Cannon, B. D., Williams, R. M., Capasso, F., Gmachl, C., Sivco, D. L., and Cho, A. Y., “Frequency stabilization of quantum cascade lasers by use of optical cavities,” Opt. Lett. 27, 2164–2166 (2002). [11] Taubman, M. S., Myers, T. L., Cannon, B. D., and Williams, R. M., “Stabilization, injection and control of quantum cascade lasers, and their application to chemical sensing in the infrared,” Spectrochim. Acta 60, 34573468 (2004). [12] Bielsa, F., Douillet, A., Valenzuela, T., Karr, J., and Hilicos, L., “Narrow-line phase-locked quantum cascade laser in the 9.2 μm range,” Opt. Lett. 32, 1641–1643 (2007). [13] Bartalini, S., Cancio, P., Giusfredi, G., Mazzotti, D., Borri, S., Galli, I., Leveque, T., Gianfrani, L., and Natale, P. D., “Frequency-comb-referenced quantum-cascade laser at 4.4 μm,” Opt. Lett. 32(8), 988–990 (2007). [14] Borri, S., Bartalini, S., Galli, I., Cancio, P., Giusfredi, G., Mazzotti, D., Castrillo, A., Gianfrani, L., and Natale, P. D., “Lamb-dip-locked quantum cascade laser for comb-referenced ir absolute frequency measurements,” Opt. Express 16, 11637–11646 (2008). [15] Stenger, J., Schnatz, H., Tamm, C., and Telle, H. R., “Ultraprecise measurement of optical frequency ratios,” Phys. Rev. Lett. 88, 073601 (Feb 2002). [16] Telle, H., Lipphardt, B., and Stenger, J., “Kerr-lens mode-locked lasers as transfer oscillator,” Appl. Phys. B 74, 1 (Jan 2002). [17] The HITRAN database is available at http://cfa-www.harvard.edu/HITRAN. [18] Wieman, C. and H¨ ansch, T. W., “Doppler-free laser polarization spectroscopy,” Phys. Rev. Lett. 36, 1170– 1173 (1976). [19] Demtr¨ oder, W., [Laser Spectroscopy], Springer, 2nd ed. (1973).

Proc. of SPIE Vol. 7222 72220C-9