Difference frequency generation spectrometer for simultaneous multispecies detection

Difference frequency generation spectrometer for simultaneous multispecies detection Petter Weibring, 1,* Dirk Richter, 1 J. G. Walega, 1 Lars Rippe, 1,2 and Alan Fried 1 1 National Center for Atmospheric Research, Earth Observing Laboratory, 3450 Mitchell Lane, Boulder, Colorado 80301, USA 2 National Center for Atmospheric Research, Advanced Study Program, 3450 Mitchell Lane, Boulder, Colorado 80301, USA *weibring@ucar.edu Abstract: A difference-frequency generation based spectrometer system for simultaneous ultra-sensitive measurements of formaldehyde (CH 2 O) and Methane (CH 4 ) is presented. A new multiplexing approach using collinear quasi-phase-matching in a single grating period of periodically poled lithium niobate (PPLN) is discussed and demonstrated for two pairs of pump and signal lasers to generate mid-infrared frequencies at 2831.64 cm 1 and 2916.32 cm 1, respectively. The corresponding absorption signals are discriminated by modulating the DFB diode lasers at modulation frequencies of 40 khz and 50 khz, respectively, and using a computer based modulation and de-modulation scheme. In addition, simultaneous measurements of CH 2 O, CH 4 and H 2 O are demonstrated utilizing both collinear and non-collinear quasi-phase-matching. 2010 Optical Society of America OCIS codes: (300.6260) Spectroscopy, diode lasers; (060.2390) Fiber optics, infrared; (190.0190) Nonlinear optics; (190.2620) Harmonic generation and mixing; (010.1120) Air pollution monitoring. References and links 1. A. Fried, and D. Richter, Infrared Absorption Spectroscopy, in Analytical Techniques for Atmospheric Measurement, Dwayne Heard, Editor (Blackwell Publishing, May, 2006). 2. R. W. Boyd, Nonlinear Optics, (Third Edition, Academic Press, 2008). 3. D. H. Jundt, Temperature-dependent Sellmeier equation for the index of refraction, n( e), in congruent lithium niobate, Opt. Lett. 22(20), 1553 1555 (1997). 4. D. Richter, D. G. Lancaster, and F. K. Tittel, Development of an automated diode-laser-based multicomponent gas sensor, Appl. Opt. 39(24), 4444 4450 (2000). 5. Z. Cao, L. Han, W. Liang, L. Deng, H. Wang, C. Xu, W. Zhang, Z. Gong, and X. Gao, Ultrabroadband tunable continuous-wave difference-frequency generation in periodically poled lithium niobate waveguides, Opt. Lett. 32(13), 1953 1955 (2007). 6. L. H. Deng, X. M. Gao, Z. S. Cao, W. D. Chen, Y. Q. Yuan, W. J. Zhang, and Z. B. Gong, Widely phasematched tunable difference-frequency generation in periodically poled LiNbO3 crystal, Opt. Commun. 281(6), 1686 1692 (2008). 7. Z. Cao, L. Han, W. Liang, L. Deng, H. Wang, C. Xu, W. Chen, W. Zhang, Z. Gong, and X. Gao, Broadband difference frequency generation around 4.2 µm at overlapped phase-match conditions, Opt. Commun. 281(14), 3878 3881 (2008). 8. J. J. Scherer, J. B. Paul, and H.-J. Jost, Quantitative trace gas sensing with mid-infrared difference frequency generation lasers, Proceedings FLAIR 2009, 52 (2009). 9. P. Malara, P. Maddaloni, G. Mincuzzi, S. De Nicola, and P. De Natale, Non-collinear quasi phase matching and annular profiles in difference frequency generation with focused Gaussian beams, Opt. Express 16(11), 8056 8066 (2008). 10. D. Richter and P. Weibring, Ultra-high precision mid-ir spectrometer I: Design and analysis of an optical fiber pumped difference-frequency generation source, Appl. Phys. B., doi: 10/1007/s00340 005 2072 2, (2005). 11. P. Weibring, D. Richter, A. Fried, J. G. Walega, and C. Dyroff, Ultra-high-precision mid-ir spectrometer II: system description and spectroscopic performance, Appl. Phys. B 85(2-3), 207 218 (2006). 12. P. Weibring, D. Richter, J. G. Walega, and A. Fried, First demonstration of a high performance difference frequency spectrometer on airborne platforms, Opt. Express 15(21), 13476 13495 (2007). 13. D. Richter, US Patent application 11276874, Precision Polarization Optimized Optical Beam Processor, filed March 17, 2006 with US Patent and Trademark Office. (C) 2010 OSA 20 December 2010 / Vol. 18, No. 26 / OPTICS EXPRESS 27670

14. C. Roller, A. Fried, J. G. Walega, P. Weibring, and F. K. Tittel, Advances in Hardware, System Diagnostics Software, and Acquisition Procedures for High Performance Airborne Tunable Diode Laser Measurements of formaldehyde, Appl. Phys. B 82(2), 247 264 (2006), doi:10.1007/s00340-005-1998-8. 15. B. P. Wert, A. Fried, B. Henry, and S. Cartier, Evaluation of inlets used for the airborne measurement of formaldehyde, J. Geophys. Res. 107(D13), 4163 (2002), doi:10.1029/2001jd001072. 16. B. P. Wert, A. Fried, S. Rauenbuehler, J. Walega, and B. Henry, Design and performance of a tunable diode laser absorption spectrometer for airborne formaldehyde measurements, J. Geophys. Res. 108(D12), 4350 (2003). 17. P. Werle, R. Mucke, and F. Slemr, The limits of signal averaging in atmospheric trace-gas monitoring by tunable diode-laser absorption spectroscopy (TDLAS), Appl. Phys., B Photophys. Laser Chem. 57(2), 131 139 (1993). 1. Introduction Sensitive and selective tunable mid-infrared absorption spectroscopy depends upon high quality laser sources such as lead-salt diode lasers, Quantum Cascade Lasers (QCL), Vertical Cavity Surface Emitting Lasers (VCSEL) as well as Difference Frequency Generation (DFG) based sources. For atmospheric conditions, many molecular species are present at mixing ratios of 1-1000 parts-per-trillion by volume (pptv, where 1 pptv is equivalent to 1 x 10 12 ), and in turn require a suitable measurement approach that permits extremely low limits of detection [1]. In addition, airborne platforms with limited payload capacities are becoming more frequently employed, especially for airborne studies reaching the upper troposphere/lower stratosphere (UT/LS). Compact and lightweight spectrometers that can provide multi-species detection, preferably by means of autonomous operation, are therefore highly desirable. DFG is a flexible and robust technology that has demonstrated very high quality airborne performance. This paper will discuss how a single molecule DFG-system can be extended to include multiple DFG operating wavelengths while retaining a single and fixed phase-matching condition. The usable bandwidth of the DFG process [2] is determined by the wavelength dispersion [3] and transparency of the non-linear crystal employed (e.g. PPLN is transparent from ~1 to 4.6 µm) over which photorefractive damage is negligible. The conversion efficiency is dependent on the signal and pump beam overlap, focusing conditions, crystal temperature and length, and these are further discussed in Section 2. To expand the relatively narrow inherent phase-matching bandwidth and allow the mixing of a range of tunable signal and pump lasers, one can adjust the phase-matching condition by changing the crystal temperature or effective grating period (fixed, fan-out) [4], or broaden it by waveguide (WG) multimode coupling [5]. Another reported approach is to utilize the degeneracy point, at which the widest tuning range is achieved for a fixed grating period, while keeping either the pump or signal wavelength fixed. In the 3-5 µm wavelength region, a tuning range of ~100 cm 1 can be achieved utilizing this approach for bulk PPLN [6,7]. DFG systems based on this approach have been commercially available for a couple of years [8]. In contrast to these previously reported approaches we discuss here an alternative approach using the inherent dispersion of pump and signal wavelengths employing a common single grating period and temperature. This approach can generate multiple mid-ir wavelengths in the 3-5 µm spectral region, which can have a substantial wavelength separation depending on the input wavelengths (see Sect. 2). In this paper we demonstrate the concept for two generated frequencies at 2831.64 cm 1 (CH 2 O) and 2916.32 cm 1 (CH 4 ). The feasibility and instrument performance for simultaneous measurements of both gases was assessed during long term laboratory tests (~30 days) and during the OASIS (Ocean Atmosphere Sea Ice Snowpack, ~27 days) ground based field campaign. In the latter, the instrument was deployed in a different configuration from that described herein, using only one pump laser in conjunction with two signal lasers. This will be the subject of another paper. The instrument was operated autonomously in the field, recording data 97% of the time. These aspects are important steps in moving towards an autonomous multispecies detection system on airborne platforms to study the UT/LS. (C) 2010 OSA 20 December 2010 / Vol. 18, No. 26 / OPTICS EXPRESS 27671

In the remainder of this paper we describe the theory for multiple wavelength generation in a single grating period crystal and details on multi-species detection capabilities, followed by a discussion of the instrument s performance. 2. Theory Efficient transfer of energy from the pump (p) and signal (s) waves to the idler (i) wave requires that both the energy and the momentum conservation are satisfied according to: Here are the wave vectors for the three different laser wavelengths. One of the highest conversion efficiencies can be achieved through quasi-phase-matching [2], in which beams propagate through a crystal where the nonlinear coefficient is periodically inverted by sign or magnitude (e.g. crystal orientation) every second coherence length (2L coh ). This allows compensation of a nonzero wave vector mismatch Δk, given by: Here is the grating period of the domain engineered crystal (e.g. LiNbO 3 ) and maximum output power is reached when perfect phase matching is fulfilled, i.e. momentum is conserved Δk = 0. Mixing a collinear pump and a signal beam in a non-linear crystal generates an idler beam with a power I i given by [2]: (1) (2) (3) where K i is proportional to: the idler frequency, the effective non-linear susceptibility, the index of refraction of the three waves, the intensity of the pump and signal beams and the crystal length L. Let us assume collinear plane wave interaction between the pump and signal wavelengths i.e. collinear quasi-phase-matching (CQPM), in which the vectors in Eq. (2), become scalars. For a given combination of pump and signal beams, an idler beam will be efficiently generated only if the grating period Λ is chosen in such a way that Δk is sufficiently small. Here, we utilize an approach that takes advantage of the fact that Δk in Eq. (2) for a single grating period can be maintained small enough over a wide idler wavelength range if both the pump and signal beams are tuned simultaneously to compensate for the phase mismatch while generating the desired idler wavelength. Thus, multiple combinations of the pump and signal wavelengths can generate the same idler wavelength while their imposed index of refraction are different. Equation (2) can be solved numerically to find the allowed pump and signal wavelength combinations for a given idler wavelength and crystal grating period. Alternatively this approach can be used to solve for different idler wavelengths from different combinations of pump and signal wavelengths. A CQPM PPLN simulation employing grating periods in the Λ = 28.9-30.0 µm and temperatures in the T = 19-250 C ranges reveals the potential of this approach. The widest tuning range ~799 cm 1 (2572-3371 cm 1 ) is achieved by Λ = 28.9 µm at T = 245 C, which puts challenging requirements on the PPLN oven temperature and its surrounding environment. On the other hand a tuning range of ~753 cm 1 can be achieved in the 2618-3371 cm 1 (2.97-3.82 µm) wavelength region for a grating period of Λ = 30.0 µm at T = 19.9 C. To achieve such large coverage and ability to target specific absorption lines, one either requires a pair of widely tunable pump and signal lasers (1030-1117 nm and 1530-1578 nm), or a selected set of conventional tunable DFB diode and fiber laser sources. (C) 2010 OSA 20 December 2010 / Vol. 18, No. 26 / OPTICS EXPRESS 27672

Figure 1 shows an idealized case i.e. Δk~0 for a grating period of Λ = 30.1 µm at T = 37.0 C, as a means to explain the more complex plots to follow. Here we show the possible idler frequencies in cm 1 that can be achieved with CQPM in a single grating period for different combinations of pump and signal wavelengths by solving Eq. (2) numerically for Δk = 0.01 cm 1. The idler frequency is shown on the x-axis and the corresponding pump and signal wavelengths are determined by drawing a vertical line until it intersects the signal trace (right axis, green) and the pump trace (left axis, blue) and the wavelengths are read on the corresponding y-axes. As an example, the idler frequency A 2831.64 cm 1 is generated by the indicated pump 1082.87 nm and signal 1561.76 nm wavelengths designated by the A on the vertical axes. Similarly, the idler frequency B 2916.75 cm 1 is generated by the indicated pump 1071.29 nm and signal 1558.19 nm wavelengths. Fig. 1. Collinear quasi-phase-matching (CQPM) in a PPLN crystal (Λ = 30.1 µm, T = 37.0 C), The idler frequency is shown on the x-axis and the corresponding pump and signal wavelengths are determined by drawing a vertical line until it intersects the signal trace (right axis, green) and the pump trace (left axis, blue) and the wavelengths are read on the corresponding y-axes, see text for details. In reality there is a phase-matching bandwidth, which we define as the full width (Δk = 1.3 cm 1 ) corresponding to 1/e maximum of the idler power. This is shown in Fig. 2 as wider pump and signal traces compared to Fig. 1. The bandwidth allows one to obtain the same idler frequency for multiple unique pump and signal wavelength pairs. For instance, around point B the 2916.75 cm 1 idler wavelength can be generated by a range of pump wavelengths (vertical cross-section) between 1070.85 and 1071.75 nm if combined with the matching signal wavelength in the 1557.24-1559.13 nm range. Alternatively, idler frequency scanning can be achieved. For instance, from B 1 to B 2 one calculates an idler tuning frequency range of 2915.9 to 2923.2 cm 1 (gray area), by employing a fixed pump wavelength of 1070.96 nm with variable signal wavelengths of 1557.27 to 1559.04 nm. As can be seen in this the pump profile bandwidth (horizontal cross-section) dictates the idler tuning range in this case. (C) 2010 OSA 20 December 2010 / Vol. 18, No. 26 / OPTICS EXPRESS 27673

Fig. 2. Collinear quasi-phase-matching (CQPM) in a PPLN crystal (Λ = 30.1 µm, T = 37.0 C) showing possible frequency combinations above the 1/e idler power level threshold (Δk = 1.3 cm 1). The idler frequency is shown on the x-axis and the corresponding pump and signal wavelengths are determined by drawing a vertical line until it intersects the signal trace (right axis, green) and the pump trace (left axis, blue) and the wavelengths are read on the corresponding y-axes, see text for details. The grey areas indicate the idler tuning range for fixed pump wavelengths 1082.97 and 1070.96 nm while tuning the signal DFB lasers around 1557.44 and 1561.96 nm. See text for details. In the wavelength range shown in Fig. 2, we demonstrate the wavelength multiplexing approach experimentally by mixing two DFB fiber lasers at 1070.96 nm and 1082.97 nm (pumps) with an external cavity diode laser (ECDL) tuned from 1557 nm to 1564 nm (signal). Figure 3 shows both experimental and simulated idler powers for the above mixing scheme around points A and B using Eq. (3) and the scalar representation of Eq. (2). As expected, this figure reveals a strong disagreement between the CQPM theory and the experimental results. This is due to the geometric focusing conditions in the PPLN crystal, which allows noncollinear quasi-phase-matching (NCQPM) to occur in different regions of the beam crosssection along the length of the crystal [9]. Instead of utilizing a strict NCQPM vector treatment according to Eq. (2), we adapt a scalar NCQPM approach for focused Gaussian beams proposed by [9]. Using this approach we only present results from the Malara formulations that are applicable to our geometry, which is determined mainly by the focusing condition of the pump beam [10]. For further details, we refer to the full-length paper by [9], in which the total idler power is given by: ( ) d L I i λ,t K e (w k ϕ ) p i 4 d L 2 L sinc 2 Δkeff dϕ 2 ϕ 2 (4) where K is proportional to the same factors as the constant in Eq. (3), d is the PPLN channel cross section, L is the crystal length, and φ is the angle between kλ and ki. keff is the scalar wave vector mismatch given by [9]: Δkeff = Δk + ki 2 ki 2 1 k ϕ s (5) (C) 2010 OSA 20 December 2010 / Vol. 18, No. 26 / OPTICS EXPRESS 27674

where Δk is the scalar of Eq. (2). Δk eff 0 as the NCQPM condition is satisfied in the idler off axis direction φ. By implementing Eq. (4) and (5) for our geometry, one achieves good agreement between theory and measurement (see Fig. 3). As can be seen, only by employing NCQPM theory can we faithfully model the observed longer frequency low power conversion tails that extend out to several tens of cm 1 from the peak centers. Fig. 3. Theoretical and measured idler power for the PPLN crystal (L = 50 mm, d = 1 mm, φ max = 1.14, T = 37.0 C and Λ = 30.1 µm). Points A, B are generated by mixing 1561.96 nm and 1557.44 nm with a 1082.97 nm and 1070.96 nm, respectively. The measured trace is recorded by mixing an ECDL laser tuned from 1557 nm to 1564 nm with a 1082.97 nm and a 1070.96 nm laser, respectively. Note that point C is generated by mixing 1561.96 nm with 1070.96 nm, which is further discussed in Sect. 2. Note that the y-scale is in arbitrary units and that the traces are normalized to each other. The asymmetrical shapes are due to the focusing condition [9] and references therein, where increasing angles between the signal beam/crystal grating (k s /k Λ ) and the pump beam (k p ) generate increasing idler frequencies with increasing annular far-field pattern behavior. A tighter focusing condition yields an even stronger tail behavior and a more pronounced annular beam profile. The opposite side of the phase-matching curve does not exhibit this behavior, as all the beams and the crystal grating (k p, k s, k i and k Λ ) become and remain collinear as the idler frequency decreases resulting in the ordinary CQPM Sinc 2 () behavior, as shown in Eq. (3). As can be seen in Fig. 3, the NCQPM peak position is displaced compared to the CQPM, resulting in peak powers with slightly annular beam profiles for focused Gaussian beams [9]. Another potential factor that may influence the bandwidth and phasematching condition can arise from small temperature gradients in the beam cross-section caused by thermal fluctuations in the crystal oven. In multispecies measurements using pre and post cell detector balancing [11], one has to be cautious, as one may be tempted to utilize the tails of the phase matching to achieve maximum wavelength coverage. However, this could result in idler beam profiles that may become too annular, making it even more difficult to effectively cancel out the large DFG amplitude modulation that we observe employing detector balancing. The ECDL was tuned from 1520 nm to 1570 nm, verifying that no significant phase matching occurred at frequencies other than those predicted by the theory. Note, the dynamic range of this measurement is only ~20 db, which is far less than the 40-50 db side mode suppression required in atmospheric trace gas measurements. Even in an ideal CQPM case, the Sinc 2 () function Eq. (3), predicts as little as 25 db side mode suppression as far as 60 cm 1 away from the phase-matching peak. As will be shown, the trace gases themselves are excellent for performing high dynamic range estimates of the side mode suppression for a particular trace gas measurement. A peculiar result of the long tails just illustrated in Fig. 3 is that phase matching at point C ~2935 cm 1 (1070.96/1561.96 nm) in Fig. 4. is now possible, where the 1070.96 nm (C) 2010 OSA 20 December 2010 / Vol. 18, No. 26 / OPTICS EXPRESS 27675

wavelength serves as a simultaneous pump for both the wavelength ranges B and C while the 1561.96 nm wavelength serves as a simultaneous signal for both the wavelength ranges A and C. This is illustrated in Fig. 4, which shows the NCQPM for idler powers above the 1/e and ~1/30 threshold of the peak power, respectively. Generation of such additional idler frequencies in DFG can be both beneficial and detrimental in spectroscopic studies. As will be discussed, this is beneficial in allowing one to probe additional idler scan regions without additional input lasers, thus potentially extending the number of atmospheric species detected. However, this same attribute can also be detrimental if one is not aware of the additional idler frequencies and such frequencies C are simultaneously scanned in time as the main frequencies (A and B in this case) and happen to be absorbed by abundant atmospheric gases like H2O. This situation is analogous to multimode tunable diode laser measurements. Fig. 4. Non-collinear quasi-phase-matching (NCQPM) in a PPLN crystal (Λ = 30.1 µm, T = 37.0 C), above the 1/e idler power level threshold (Δk = 1.8 cm 1, darker shading) and above the 1/30 idler power level threshold (Δk = 5.8 cm 1, lighter shading). The idler frequency is shown on the x-axis and the corresponding pump and signal wavelengths are visualized by drawing a vertical line until it intersects the signal trace (right axis, green) and the pump trace (left axis, blue) and the wavelengths are read on the corresponding y-axes, see text for details. The gray areas indicate the idler tuning range for fixed pump wavelengths 1082.97 and 1070.96 nm while tuning the signal DFB lasers around 1557.44 and 1561.96 nm. See text for details. Thus, when designing the multiple wavelength-mixing scheme one must carefully assess the selected wavelength pairs for spectral artifacts that may occur for coincidental phase matching in other wavelength regions. As shown in Fig. 4, it is possible to choose a pump wavelength of 1082.97 nm and a signal wavelength of 1561.96 nm as mixing wavelengths to generate 2831.64 cm 1 (the frequency employed for CH2O detection), but at the same time the geometric conditions and the wavelength acceptance bandwidth of the crystal may allow the same signal wavelength to be mixed with the 1070.96 nm pump generating a simultaneous wavelength sweep at ~2935 cm 1 (C), thus recording a convoluted spectrum from the two spectral regions. In this case H2O, which absorbs at 2935.19 cm 1, would be a spectral interference in detecting CH2O at 2831.64 cm 1. However, as indicated previously, one can take advantage of this special case as shown in Fig. 4, where a slight change of the wavelength pairs displaces the interfering H2O absorption feature, allowing CH2O and H2O to be measured simultaneously by the same signal laser. (C) 2010 OSA 20 December 2010 / Vol. 18, No. 26 / OPTICS EXPRESS 27676

Fig. 5. Second harmonic signals of CH 4, CH 2O and H 2O using frequencies generated by simultaneous collinear and non-collinear quasi-phase-matching (CQPM and NCQPM) in the same grating period of a PPLN crystal (L = 50 mm, T = 37.0 C and Λ = 30.1 µm). For the current focusing condition [11], CH 2O and CH 4 are collinearly phase matched by mixing 1082.97/1561.96 nm and 1070.96/1557.46 nm, respectively. H 2O is non-collinearly phase matched by mixing 1070.96/1561.96 nm. The geometric conditions allow the same signal laser, 1561.96 nm, to be mixed with both the 1082.97 nm and 1070.96 nm pump lasers, generating simultaneous wavelength sweeps at ~2831.64 cm 1 and ~2935.3 cm 1, allowing both CH 2O and H 2O to be measured on the same wavelength scan. The solid and dashed lines demonstrate the effect of turning the1070.96 nm pump laser on and off, respectively. Figure 5 demonstrates this situation and how the CH 4 and H 2 O features can be turned on/off by enabling/disabling the 1070.96 nm pump laser. In addition to detecting new species, NCQPM also allows one to tune over multiple absorption lines of the same species providing much better immunity to cross interferences from weak neighboring overlapping absorptions of other molecular species, and depending upon the baseline structure in the two frequency regions may also improve detection sensitivity. In addition, careful selection of the additional idler scan region may enable one to measure a single species over several orders of magnitude and/or isotopic pairs with widely differing ambient mixing ratios. One can exploit absorption cross-section differences and/or pump lasers with different powers that mix with one signal laser. In conclusion, the above tuning ranges (2827-2835 cm 1 and 2915-2924 cm 1 ) overlap with absorption lines of CH 2 O, NH 3, CH 4, OCS, HCl, HCOOH, NO 2, which are presented in Table 1 along with estimated detection limits (LOD) using our airborne platform [12]. Table 1. Estimated system performance limits of detection (LOD) a Species Wave number (cm-1) S int (cm 2 /cm*mol) LOD (pptv) 60s averaging NH 3 2831.48 1.1x10 22 10700 CH 2O 2831.64 5.33x10 20 22 CH 4 2916.32 1.25x10 20 94 HCOOH 2917.90 4.0x10 20 29 NO 2 2918.75 6.5x10 21 180 HCL 2923.74 1.32x10 19 9 OCS 2923.77 1.35x10 20 51 a Equivalent formaldehyde concentration in pptv (1σ), along with the integrated linestrengths for species in the 2827-2835 cm 1 and 2915-2924 cm 1 wavelength range, which is covered by the suggested wavelength-mixing scheme (C) 2010 OSA 20 December 2010 / Vol. 18, No. 26 / OPTICS EXPRESS 27677

3. System description The layout of the spectrometer is based on our previously published airborne instrument [12]. The design is basically comprised of a laser module, fiber optically connected by a single mode fiber to an integrated module consisting of DFG unit, detectors, and a multi-pass absorption cell. Additional wavelengths coverage was simply achieved by fusion splicing an DFB diode laser and a DFB fiber laser into the fiber optic train (see, Fig. 6), keeping all discrete optical components in their original alignment. The discrimination of the absorption signals from the respective gases (CH 2 O and CH 4 ) is accomplished by modulating the DFB diode lasers at modulation frequencies of 40 and 50 khz, respectively, and using a computer based modulation and de-modulation scheme. Both DFB diode lasers were also scanned with a 25 Hz triangular waveform and stabilized by actively locking the lasers center frequency to the CH 2 O and CH 4 absorption line center obtained from the reference cell. Fig. 6. Optical layout of the mid-ir spectrometer. To the left, the laser module consisting of two Ytterbium doped Distributed Feed Back (DFB) fiber lasers (Pump 1 and Pump 2), two Ytterbium (Yb) fiber amplifiers, a 1558 nm and 1562 nm (DFB-DL) laser, Polarization controllers and Wavelength division multiplexers (WDM). To the right, the detection module consisting of a Multi-pass cell and a combined detection unit and difference frequency generation stage consisting of Ball lens, Focusing Lens, Periodically Poled Lithium Niobate (PPLN), Collimation lens, Germanium filter (Ge), Detector focusing lenses, Sample detector, Noise detector, Reference Gas Cell, Reference Detector and Multi Pass Cell. The laser module is based on four fiber coupled laser sources and three fiber amplifiers. To demonstrate the feasibility we utilized commercial telecom fiber amplifiers that were designed for single wavelength amplification. However, these could be replaced by more efficient fiber amplifier designs that enable multiple wavelength power amplification. The signal lasers, 1558/1562 nm and 1562 nm DFB diode lasers are controlled by the computer for wavelength scanning, modulation and stabilization purposes. The signal laser outputs are spliced to a commercial Er/Yb fiber amplifier, which increases the optical power to ~200 mw for each wavelength. The pump source consists of two DFB fiber lasers (1071 nm and 1083nm), which are spliced to two separate commercial Yb fiber amplifiers (peak gain at 1083 nm) to increase each of the 1071 nm and 1083 nm powers to ~400 mw. The spectrometer simultaneously operates at 2831.64 cm 1 (CH 2 O) and 2916.32 cm 1 or 2831.92 cm 1 (CH 4 ). The mixing wavelengths are combined using a wavelength division multiplexer (WDM), and spliced to a single mode fiber, which connects the pump and signal beams to the DFG module, as shown in Fig. 6. The pump and signal beams are launched into the PPLN crystal by a two-stage lens design [10]. The unconverted signal and pump radiation are removed by a Germanium filter, while the remaining idler beam is imaged by a collimation lens into a Multi-pass absorption cell (path-length 100-m, astigmatic Herriot cell). A custom coated beam splitter (BS) [13], is used to split off and direct a portion of the input beam onto the Noise detector. This allows one to (C) 2010 OSA 20 December 2010 / Vol. 18, No. 26 / OPTICS EXPRESS 27678

closely match the beam intensities incident on the Noise detector and the Signal detector, which is an important requirement for removal of optical noise from the laser source [11]. The BS also transmits a small fraction of the beam through a sealed CH 2 O/CH 4 gas cell to the reference detector for computer controlled active wavelength locking using the target CH 2 O and CH 4 absorption features at 2831.64 cm 1 and 2831.92 cm-1 or 2916.32 cm 1. The idler beam passes through an AR-coated Sapphire window which pressure seals the multi-pass cell. Three 4 stage-peltier cooled photovoltaic HgCdTe detectors (D*~3.4x10 10 @1kHz, d = 1mm, 80 C) serve as signal, noise and reference detectors, which are connected to low noise transimpedance amplifiers. The transimpedance amplifier outputs are directly connected to a high precision, differentially coupled analog to digital conversion computer card and processed using software based lock-in algorithms and Singular Value Decomposition (SVD) [12,14]. 4. Results and discussion Spectrometer characterization and performance assessment was carried out both in the laboratory environment (~30 days) and during a field campaign in Barrow Alaska (~27 days). This allowed characterization of the instrument both under ideal controlled laboratory and harsh field conditions. We only discuss the former in this paper. 4.1 Experimental set-up and procedures For calibration and validation purposes, a computer controlled inlet/calibration system was connected to the mid-ir spectrometer. The inlet system has essentially three operational modes. Ambient, Zero air, and Calibration. In all three modes, air is drawn through the Multipass cell at flow rates of ~9 slm (standard liter per minute), resulting in a cell residence time of ~1-2 seconds. A pressure controller maintains the cell pressure around 50 Torr. In the Ambient mode, ambient air is continuously drawn through the system and used for actual ambient trace gas measurements. In the Zero air mode, air free of CH 2 O is produced by a heated Pd/Al 2 O 3 catalyst permitting acquisition of a true instrument background [15]. In the CH 2 O calibration mode, CH 2 O calibration standards from one of two permeation sources are mixed with Zero air and added to the cell inlet producing CH 2 O mixing ratios of ~4 ppbv or ~12 ppbv, with an estimated accuracy of better than 12% [16]. In the CH 4 calibration mode, flow from a CH 4 standard (1510 ppbv CH 4 in air) was introduced into the inlet at flows in excess of the cell flow. For CH 4 zeroing, we employed a second Pd/Al 2 O 3 catalyst that was operated at temperatures around 450 C. Before every performance test, the system was calibrated by injecting a known amount of CH 2 O and CH 4 in the Multi-pass cell using the inlet/calibration system. 4.2 Allan variance In order to obtain comprehensive estimates of the entire system s performance, multiple Allan variance measurements were carried out, while passing Zero air into the Multi-pass cell (see Fig. 7). Allan variance measurements yield two important performance characteristics: the entire system s sensitivity as a function of time and the system stability time [17]. The laboratory performance typically yielded a CH 2 O and CH 4 sensitivity of 10-15 pptv and 40-45 pptv respectively for 250s averaging, which corresponds to a detectable absorbance of ~5*10 7 for both species. (C) 2010 OSA 20 December 2010 / Vol. 18, No. 26 / OPTICS EXPRESS 27679

Fig. 7. Allan variance results for CH 2O (blue trace, 2831.64 cm 1 ) and for CH 4 (green trace, 2916.32 cm 1 ) during laboratory conditions. 4.3 Time series measurements and crosstalk Although the Allan variances with Zero Air are very useful, it has limits to completely describe the instrument performance. For example, the Allan plot does not capture any gas switching noise upon changes in the sampling mode and associated noise due to the background subtraction. Hence it is important to carry out additional time series measurements with constant mixing ratios added into the system inlet, which are shown in Fig. 8 for CH 2 O (0-12 ppbv) and CH 4 (0-1.8 ppmv). The sequence starts with CH 2 O and CH 4 calibrations (Event 1 and 2) and is followed by injecting CH2O (11.87 ppbv) to estimate short-term accuracy (Event 3). This is followed by Zero air replicate precision measurements during ~1.5 hours. The deduced CH 2 O and CH 4 replicate precisions of ~26 pptv and ~129 pptv (1σ) for multiple 60s measurements during this period compares well with the results obtained with the Allan variance measurements in Fig. 7. Event 4 records measurements of the ambient laboratory air over 45 minutes, and shows an elevated increasing CH 2 O concentration in the 8.55-9.16 ppbv range and a decaying CH 4 concentration of 1.85-1.81 ppmv. The time series from events 3, 5 and 6 shows negligible cross talk between the two channels when injecting high gas sampling concentrations of either species. A CH 2 O average concentration of 2 pptv was retrieved while injecting 1510 ppbv of CH 4. Also a CH 4 concentration of 75 pptv was retrieved while injecting 11.87 ppbv of CH 2 O. For both species these results are within the uncertainty of the measurements. The system long term accuracy, also shown by events 5 and 6, was tested by injecting CH 2 O and CH 4 calibration standards, which yielded less than 20 pptv discrepancy for 11.87 ppbv CH 2 O and less than 14 ppbv discrepancy for 1510 ppbv CH 4. This corresponds to 0.17% and 0.93% for CH 2 O and CH 4 respectively. (C) 2010 OSA 20 December 2010 / Vol. 18, No. 26 / OPTICS EXPRESS 27680

Fig. 8. The time evolution of the CH2O and the CH4 (2916.32 cm 1) concentration over 4 hours showing the system sensitivity to cross talk and long term calibration performance during laboratory conditions. The * indicates which gas is added to the inlet during an event and the numbers beneath are the instrument response. Events 1 and 2 are instrument calibrations. Events 3, 5 and 6 are crosstalk/accuracy tests, carried out by adding a known gas concentration to the inlet and measuring the instrument response. Event 4 is an ambient measurement of laboratory air and the ~1.5 h duration between events 3 and 4 is the instrument response when zero air was added to the inlet. 5 Summary The present study introduces a new concept and application of a mid-ir spectrometer for simultaneous multi-species measurements based on a DFG source that uses a single quasiphase-matching condition in a bulk PPLN crystal. The approach relies upon efficient DFG conversion over a broad wavelength range in which the dispersion of the pump and signal wavelengths are selected to simultaneously match a single grating period and generate a range of desired mid-ir wavelengths. Such a frequency agile mid-ir laser source permits the study of a number of atmospherically important trace gases with a single source optical architecture utilizing existing commercially available diode and fiber laser sources. The spectroscopic performance of this multi-wavelength/species design is on par with previously published single wavelength designs by our group. The practical use of such a spectrometer in a different configuration has been demonstrated during a 30 day field campaign in which the instrument was operated autonomously in very harsh environmental conditions in the arctic. Acknowledgement The National Center for Atmospheric Research is sponsored by the National Science Foundation and operated by the University Corporation for Atmospheric Research. The authors acknowledge valuable discussions with Dr. Scott Spuler regarding geometric optics simulations. Support from the Advanced Study Program at NCAR is gratefully acknowledged. (C) 2010 OSA 20 December 2010 / Vol. 18, No. 26 / OPTICS EXPRESS 27681