1. INTRODUCTION

With their ultra-high spectral resolution (resolving power ${\gt}{10}$ million), infrared (IR) tunable laser spectrometers (TLSs) are widely used for in situ gas detection [1,2]. They offer high sensitivity (typically ${\le} {1}$ part-per-billion by volume (ppbv) for a variety of gases identified unambiguously through scanning over individual rotation-vibration lines. These compact, low-power instruments are particularly suited for measuring the abundances of low-molecular weight gases (e.g., CO, ${{\rm CO}_2}$, ${{\rm H}_2}{\rm O}$, ${{\rm N}_2}{\rm O}$, ${{\rm NO}_x}$, HCl, HF, ${{\rm O}_2}$, ${{\rm CH}_4}$, ${{\rm PH}_3}$, ${{\rm NH}_3}$, OCS, HCN, ${{\rm H}_2}{{\rm O}_2}$, ${{\rm O}_3}$, ${\rm HNO3}$, ${{\rm H}_2}{\rm CO}$, HOCl, and ${{\rm H}_2}{\rm S}$) and for the determination of stable isotope ratios in C, H, N, O, and S [3].

The Herriott cell [4] is the most common configuration for multipass gas detection. In a typical set-up, an IR semiconductor laser [diode, interband cascade (IC), or quantum cascade (QC)] is used as the source, with the beam bouncing multiple times (typically up to ${\sim}{100}$) between two spherical mirrors a fixed distance apart. The laser light is injected through a hole in the near mirror, exits either through the same hole or one in the far mirror, and then falls onto a single-element IR detector.

The sensitivity of TLSs depends on gas pressure, temperature, pathlength, and inherent molecular line strength in the chosen wavelength region. Classic Herriott cell configurations, even when properly aligned, are limited by the occurrence of optical interference fringes between a variety of reflecting surfaces (Herriott cell mirrors, the laser collimator, fore-optics elements, and the detector) that are at best equivalent in amplitude to a spectral line depth of ${\sim}{1} \times {{10}^{- 5}}$ over minutes of integration time [5]. Several approaches have been used to reduce the amplitude of the optical interference fringes, with limited success [6]. A particularly efficient method involves the use of a “fringe-wash” heater on the cell body to cyclically ramp its length by a few micrometers during the measurement time [7]. Tiny changes in mirror separation “move” the phase of fringes in a period shorter than the spectral recording time. In this way, the fringes are “washed out” over time, so that the optical fringe amplitude is greatly reduced through averaging. However, this method adds significant power requirement (up to tens of watts) to the spectrometer [7], which can be a limitation for many deployments (e.g., planetary probe missions). Piezo-electric transducers of much lower power that modulate the mirror separation have also been used successfully by several groups.

The two-channel TLS operating within the Sample Analysis at Mars (SAM) instrument suite on the Mars Science Laboratory (MSL) Curiosity rover [7] is configured as a traditional TLS (see Fig. 1). With laser sources at 2.78 and 3.27 µm first passing through a fore-optics chamber containing beamsplitters, steering mirrors, and reference gas cells with dedicated detectors, the laser light bounces multiple times between two gold-coated 4-cm-diameter aluminum spherical mirrors separated by 20 cm, entering and exiting through holes in the near and far mirrors. The light is collected by a single-element detector (Vigo PVI-4, 1 mm immersion lens) for each channel. To maximize the exiting beam intensity after multiple passes, traditional Herriott cell mirrors are highly reflective (${\gt}95\%$). The lasers scan at 1 Hz, and on-board averaging returns a spectrum every 30 s for analysis after a run lasting an hour or so that includes both empty and full cell spectra. The TLS-SAM spectrometer uses a 30 W fringe-spoiling heater to achieve high sensitivity to Martian gas detection.

 

Fig. 1. Top: schematic of the TLS-SAM Herriott cell showing the laser sources (L), beam-splitters (BS), steering mirrors (M), detectors (D), and wedge windows (W). Bottom left: one of the TLS mirrors showing holes in the aluminum spherical mirror for four-channel injection. Bottom right: visible laser demonstration of the orthogonal spot patterns for the two-channel TLS as operating on Mars. The cell detector records spectra only after the full multi-pass of either 43 or 81 passes.

Download Full Size | PDF

Operating successfully on Mars for over 8 years, the TLS on SAM on the Curiosity rover has produced first-of-its-kind, to the best of our knowledge, measurements on both the Martian atmosphere and on gases evolved from rock pyrolysis. This includes ${{\rm CO}_2}$ and ${{\rm H}_2}{\rm O}$ isotope ratios $^{13}{\rm C}/^{12}\rm C$, $^{18}{\rm O}/^{17}{\rm O}/^{16}{\rm O}$, and D/H, using the 2.78 µm channel [3] and ${{\rm CH}_4}$ and its $^{13}{\rm C}/^{12}{\rm C}$ isotope ratio using the 3.27 µm channel [8].

As with other traditional TLSs using Herriott cells, TLS-SAM has sensitivity limited by optical interference fringes coming from both the fore-optics chamber and the Herriott cell itself. Additionally, the dynamic range is limited by the fixed pathlength (43 passes for the 2.78 µm channel, 81 passes for the 3.27 µm channel). The recorded spectra also contain a contribution to the spectrum of gases in the fore-optics chamber through which the laser initially passes.

While Herriott cell fringes can be reduced by a variety of techniques, the most effective method to date, as used in the TLS-SAM instrument, is to cyclically heat the Herriott cell walls as described earlier. However, for TLS-SAM, this technique requires the use of a 30 W heater, because a motivation for developing this technique was to remove this large power requirement for future planetary mission applications.

We present here the first results of an innovative new approach using a detector array coupled to an IR-transmissive far mirror in place of the aluminum mirror. This novel approach enables imaging of the entire multipass spot pattern of the far mirror and collection of spectra for each pixel, resulting in improved sensitivity, increased dynamic range, laser power normalization, and contaminant subtraction.

2. EXPERIMENTAL DETAILS

Recognizing that subtracting the first pass spectrum from the final pass spectrum could potentially remove optical fringes and the contribution from contaminant gases in the fore-optics chamber, we first set up a simple laboratory demonstration using a small highly reflective but IR-transmissive mirror with two single-element detectors to produce detectable spectra only for the first and third passes. (The particular mirror used did not produce high enough throughput for additional passes.)

For the bulk of the experiments presented here, we used a duplicate of the flight TLS-SAM spectrometer coupled by a custom IR-transmissive far mirror to an IR camera from forward-looking IR (FLIR, model A6750sc) that was used for analysis of the Herriott cell patterns, individual spots, and spectra at the one-pixel resolution. The FLIR camera operated over the range of 3–5 µm with a thermoelectrically cooled (TEC) array of InSb detectors in a ${640} \times {512}$ pixel format operating at 125 Hz with 14 bit dynamic range. It was fitted with a 50 mm lens ($f\!/\!{2.5}$) and placed a few centimeters (cm) behind the IR-transmissive far mirror. The FLIR camera comes with software to record time series of the signal intensities on all pixels or a selected group of pixels. Many commercial cameras use a read out integrated circuit (ROIC) that can only read the entire detector array. For our application, this is wasteful, as a large percentage of the detector window has no useful information. It would also be unacceptably slow for laser scan operations, which, if using the entire ${640} \times {512}$ window, would take several minutes to complete a single-laser scan. The ROIC in the FLIR A6750sc camera, however, allows windowing, where only a selected portion of the pixels are read, and this results in both decreasing the telemetry data volume as well as speeding up the laser scan rate, both imperative properties for a spacecraft instrument. One advantage of the camera is that one can use it for multiple lasers, although we recognize that at about $\${60}\;\rm K$, these IR cameras are not inexpensive.

Our lab prototype system makes use of the entire ${640} \times {480}$ focal plane array (FPA). For each laser scan we capture about 500 frames (each frame is ${640} \times {480} \times {2}\;{\rm bytes} = {614}{,}{400}\;{\rm bytes}$) in a time series that shows the pixels changing in intensity with the scanning laser frequency. It is from those changing intensities that we can infer the absorption depth (i.e., abundance) of the gas of interest. Minimal CPU power is used in actually controlling the camera (the commercial software can be run from a modest laptop); although, there is considerable activity over a gigabit Ethernet connection to transfer all the data since the camera is a separate system. For the lab prototype, processing is done off line after data acquisition is complete. Because we are in a research mode and testing algorithms for fringe removal and ideal spot combinations, the lab computer is configured with a large amount of memory (1 TB) so that we can hold the entire sequence in memory to speed up processing and testing. This is not practical for flight of course, so the flight version would use FPA sub-windowing to choose a smaller subset of the pixels. This window size is adjustable in software and can be optimized for the particular laser scan as well as the current power budget, available bandwidth, and data taking scenarios. The sub-windowing results in faster scanning and much smaller data volumes. Our target platform, a Zynq system-on-chip (SoC), has multiple CPUs, programmable logic (field-programmable gate array, FPGA), and shared memory between them. This configuration allows high-speed, deterministic handling of the pixel frame grabbing, freeing up the CPUs to analyze the captured data in parallel. The data processing algorithms would, during the mission, be essentially summing and differencing a fixed number of well-known spot patterns chosen during the integration and test phase, and the Zynq SoC is more than capable of handling these computations.

For the new architecture described as Experiments 2 and 3 using an IR-transmissive output coupling far mirror, the mirror reflectivity was chosen to balance the reflection loss with the transmitted intensity. The IR-transmissive mirror used was purchased from LohnStar Optics in Escondido, CA. It has the same diameter (2 in.) and radius of curvature (0.44 m) as the TLS-SAM far mirror that it replaces. The mirror was made from ZnSe with a nominal thickness of 0.25 in. and was coated for a reflectivity of 98% and transmissivity of ${\sim}{1}\%$ at 3.27 µm.

The IR tunable laser was a type I interband multi-quantum-well laser. The laser structure was grown on a GaSb substrate by molecular beam epitaxy, based on a InGaAsSb/AlInGaAsSb three-quantum-well design [9] that was optimized for emission at 3.27 µm. Single-mode distributed feedback (DFB) lasers were fabricated with etched laterally coupled gratings [10], and 2-mm-long lasers were cleaved and mounted on submounts using AuSn eutectic solder. The laser used here was packaged in a custom transistor outline (TO-3) can by Achray Photonics with integrated collimating optics (Fig. 2). The collimating lens was a single aspheric element provided by LightPath, made of BD-2 (${{\rm Ge}_{28}}{{\rm Sb}_{12}}{{\rm Se}_{60}}$) with an anti-reflection coating suitable for operation at a 3.27 µm wavelength. The laser was focused midway into the multipass Herriott cell, resulting in beam spots with Gaussian widths of ${\sim}{1}\;{\rm mm}$ in diameter.

 

Fig. 2. TO-3 laser package (left) used in this study, compared to the traditional package used by TLS-SAM-MSL (right).

Download Full Size | PDF

The TLS Herriott cell was filled with low-pressure gas containing ${{\rm CH}_4}$ to produce readily observable line absorptions using the 3.27 µm tunable diode laser source scanning slowly (${\sim} 1\,\,{\rm Hz}$) over the $R(3)$ triplet of ${{\rm CH}_4}$ as targeted on TLS-SAM. A neutral density filter (mylar sheet) was placed in front of the laser source to reduce the signal amplitude and optimize the dynamic range seen by the highly sensitive camera with a noise-equivalent temperature difference of 18 mK at 3.3 µm.

3. RESULTS

We describe three experimental demonstrations using different Herriott cell configurations. The first experiment uses a small Herriott cell set-up for a 20-pass solution. The second experiment uses the TLS-SAM Herriott cell duplicate in the 43-pass configuration, as used on Mars for ${{\rm CO}_2}$ and ${{\rm H}_2}{\rm O}$ analysis. The third experiment uses the same TLS-SAM Herriott cell duplicate, but with the far mirror pulled out several millimeters to create a long-path solution.

For the second and third experiments reported below, mylar attenuators are put in front of the camera to allow the light intensity map to fall optimally within the FLIR camera’s dynamic range. While the 3.27 µm laser is scanned typically once per second over the ${0.5}\;{{\rm cm}^{- 1}}$ region, all spectra presented in these results are produced by averaging over 10 scans to improve SNR and therefore clarity of fringes.

There are numerous factors that go into the spectral noise in an actual gas measurement, such as the signal processing method [e.g., second harmonic or direct detection, detector noise, optical power on the detector(s), laser fluctuations, electronics noise, optical fringe amplitude, acoustic vibration, etc.]. In the studies presented here, we address only the contribution of interference fringe amplitude to the measurement sensitivity. Relative to a measured gas spectral line absorption depth, we consider a reduction in fringe amplitude to be associated with a corresponding improvement in gas detection ability.

A. First Experiment: First Pass Subtraction

In the initial demonstration, the FLIR camera was not used. A small-diameter (2 cm) Herriott cell of 12 cm length was fitted with an IR-transmitting mirror of very high reflectivity (${\sim} 99.9\%$) purchased for a cavity ringdown application. Two single-element, room-temperature detectors (Vigo) sample the first and third passes (Fig. 3). Because the IR-transmissive mirror was too reflective, only signals from the first two spots (first and third passes) produced enough signals for investigation. The first pass signal is subtracted from the third pass, resulting in the removal of the large-period fore-optics fringes, normalization of the laser power, and the contribution of any contaminant gases in the fore-optics. The extent to which Herriott cell fringes are removed is not optimum, since slight changes in the detector positioning could produce a range of results. To get a better understanding, an IR-transmissive mirror of lower reflectivity was used in conjunction with the FLIR camera imaging technique in the two additional experiments described below.

 

Fig. 3. First experiment demonstration using two single-element detectors sampling the first and third pass of a small 20-pass Herriott cell. The insert is a photo of the spot pattern as seen using a co-aligned visible HeNe laser. The spectra show features of the $R(3)$ ${{\rm CH}_4}$ triplet and the $R(4)$ $^{13}{{\rm CH}_4}$ quadruplet at lower intensity. In the upper trace, strong fringes (larger period) that are evident from a planar window inserted to mimic typical fore-optics fringes are removed with subtraction of pass 1. In the lower trace with the vertical axis expanded, fine period fringes associated with the Herriott cell mirrors are evident and reduced considerably by the subtraction. In both spectra, the difference spectrum is offset in the $y$ axis for presentation clarity. The wavenumber scale was established from the identification of the three main methane spectral lines listed in HITRAN 2016 [11].

Download Full Size | PDF

B. Second Experiment: Single Center Pixel Comparison

In a second experiment, the FLIR camera was used to capture the whole spot pattern, but we analyze data only for the center pixel for each spot (i.e., 22 pixels). The near mirror of the Herriott cell is the aluminum mirror with injection holes, as used in TLS-SAM. The far mirror and its associated Ge wedged window (no longer needed) of the duplicate TLS-SAM instrument were replaced with the IR-transmissive optic placed at the same distance from the near mirror as in the TLS-SAM instrument. As seen in Fig. 4, 22 spots are seen corresponding to the 43-pass configuration [10]. The injection angle is adjusted so that the 44th spot exits through the entrance hole on the near mirror.

 

Fig. 4. FLIR camera image of the output IR-transmissive mirror in the 43-pass configuration of the TLS-SAM instrument [10]. Note that because the beams are exiting at slightly different angles, the camera optical collection cannot capture all beams equally so that some upper spots are recorded only weakly.

Download Full Size | PDF

The spot shapes of Figs. 4 and 5 are not perfectly circular and are typical of Herriott cells in which the input beam quality is compromised by asymmetry in the laser emission profile and imperfect laser collimator optics. The central region of any one spot is seen by the camera over many tens of pixels. This high spatial (pixel) resolution allows us to record spectra from individual pixels for comparison. Looking at a moderate intensity spot, as shown in Fig. 5, we see a wide range of SNRs in the observed spectra. This gives insight into what happens using a traditional single-element detector that covers a large fraction of the central region. Further out from the spot center, as spot overlap increases, the relative fringe sizes increase.

 

Fig. 5. Intensity map of a single spot (37th pass in configuration from Fig. 4). The spectral graphs are laser scans over the ${{\rm CH}_4}$ triplet lines. The center placement of the spectral graphs identifies the region from which the single-pixel spectra originate. Looking at individual pixel spectra around the central portion, we see that both the signal size and relative fringe amplitudes vary dramatically across the spot image.

Download Full Size | PDF

In our second experiment, we compared the spectra from the center pixel for each of the spots and combined spectra to improve the signal-to-fringe ratio. Individual raw spectra are shown in Fig. 6, which are then normalized to the methane line absorptions to study relative signal-to-fringe results. The cleanest (best signal-to-fringe) results came from the 35th pass, which is notably cleaner than the 41st pass. By averaging individual spots/passes selected for good signal-to-fringe ratios, we are able to produce a spectrum that is four times cleaner than the 35th pass spectrum and ${\sim}{10}$ times cleaner than the 41st pass spectrum, compared to that recorded by a single-element detector. Clearly, there is the promise of increased improvement by further study of the tens of pixels that surround the central pixel, and this becomes the subject of our third experiment detailed below.

 

Fig. 6. Left column: spectra recorded from the center pixel of all spots seen on the transmissive mirror, normalized to the same ${{\rm CH}_4}$ line depth to compare fringe signal sizes away from the spectral lines. Right column: comparison of spectra for spots 35, 41, and a selected pixel average path. This selected average was made using spectra from 13 spots selected from the spot pattern to produce the best SNR. The average has a spectrum much cleaner than that of spot 41, with the selected pixel spectrum showing improvement of a factor of ${\sim}{10}$ over the spot 41 spectrum and a factor of ${\sim}{4}$ over that of the cleaner spot 35. With the laser off, the signal counts are around zero.

Download Full Size | PDF

We note that the spot imaging technique holds another advantage over single-element detection, whose output can suffer from misalignment producing power loss or increased fringe intensity at the spot edge. With imaging collection, changes in the pointing from misalignment can be mitigated by acquiring spectra from the spot center even if it has moved slightly.

We note that the full camera image is ${640} \times {512} = {327}{,}680$ pixels. In our second experiment using only center pixels, we have achieved significant improvement using spectra from only 22 pixels.

C. Third Experiment: Comparison with Traditional Detectors

In the third experiment, again using the FLIR camera, the IR-transmissive mirror is pulled back ${\sim}{3}\;{\rm mm}$ with a standoff ring to develop a multi-spot pattern of C3 symmetry (Fig. 7) [12]. Due to the camera imaging limitations for the beams exiting at different angles, the camera position was optimized for observation of the bottom arm of spots that are studied.

 

Fig. 7. Top: the complete spot pattern observed in the third experiment. Note that the camera does not faithfully capture the upper spots of the pattern due to the differing beam exit angles. Bottom: expansion and labelling of the bottom group of spots, the first 10 of which are analyzed in our third experiment study. Spot 9 is associated with 53 passes of the 20 cm Herriott cell achieving a pathlength of 10.6 m, while spot 10 is associated with 59 passes achieving a pathlength of 11.8 m.

Download Full Size | PDF

In order to obtain absolute absorption line depths and therefore gas abundances, the effective pathlength and zero laser light level must be known, the latter to set the scale of laser power and therefore spectral line depth. This can be done using a digital off-period at the beginning of each laser current scan (as is done on TLS-SAM) or in the laboratory by blocking the beam temporarily. Each pass produces spectra of different absolute size, so it is important to normalize all of the spectral scans (e.g.,  to the highest path scan with the deepest lines) before subtraction, addition, or manipulation. For example, referring to the caption of Fig. 7, if the best SNR for a spectral line was found by averaging spots 9 and 10 and then subtracting spot 1, the effective pathlength after normalization would be $({10.6} + {11.8}){\!/\!2}\;- {0.2} = {11}\;{\rm m}$. Absolute gas concentrations can then be calculated from knowing the average pathlength after manipulation in addition to gas temperature, pressure, and line parameters from the high-resolution transmission molecular absorption database (HITRAN) 2016. It is important to avoid non-linearities in the pixel detector response by using neutral density filters (e.g., mylar sheet) to reduce light levels into the middle of the camera dynamic range.

By way of example, Fig. 8 shows the intensity, fringe amplitude, and phase of two of the output spots, that of spots 8 and 9, corresponding to the 47th and 53rd passes of the Herriott cell (see Fig. 7). On the maps of Fig. 8, a superimposed circle represents the 1-mm-diameter field-of-view (FOV) of the single-element detector used in the traditional TLS-SAM Herriott cell that would average the individual pixel contributions seen in the images. The variation in fringe amplitude and phase is large, with sharp transitions from maximum and minimum fringe amplitudes and in phases from $\pi$ to ${-}\pi$.

 

Fig. 8. Sections of the FLIR camera image around spots 8 and 9, showing light intensity, fringe amplitude (peak-to-peak) and fringe period phase. Scales in $x$ and $y$ are the pixel number location. Intensity units are an inferred temperature from the IR camera, the fringe amplitude units are arbitrary, and the phase density units are degrees. The white circle overlaid represents the portion of the image that would be collected by a traditional single-element detector through a 1-mm-diameter hole in the mirror. Note that the phase maps cannot discriminate contributions from phase-folding, showing only the resulting phase between the limits of ${-}\pi$ and $\pi$.

Download Full Size | PDF

A typical laser beam transiting the Herriott cell interior will be associated with a spread in pathlength from beam center to edge. For a typical exit hole of 1 mm diameter, represented at the camera FPA by a circle of ${\sim}{20}$ pixels diameter (see Fig. 8), an exiting beam of this diameter will have a spread in pathlengths of $\pm {0.5}\;{\rm mm}$ over 20 cm or $\pm {1.3}\;{\unicode{x00B5}{\rm m}}$ in mirror-to-mirror distance. For the 3.27 µm laser, a 180° phase shift ($\lambda/2$) occurs for a change in distance of 1.65 µm. This estimated distance is close to that expected for the 1 mm diameter, so the fringe phase changes seen in Fig. 8 are not surprising, although the flips between the phase extremes are sharper than expected. The intensity, fringe amplitude, and phase fields seen in Fig. 8 are clearly complex and were seen to be different for every spot in the pattern studied.

In Fig. 9, for each spot, we show the spectra produced by averaging pixels over the white circular region (tens of pixels) of Fig. 8 that represents the traditional single-element detector FOV. As in Experiment 2, the spectra are first normalized to the methane line depths for assessment of signal-to-fringe ratios and prior to processing. Averaging of several spots of the lower C3 symmetry arm of Fig. 7 produced significant improvement over any single spot result. By averaging the spectra from the higher pathlength spots 6–10 (Fig. 9, bottom spectrum), we see an improvement in signal-to-fringe ratio of ${\sim}{10}$ over that of any single spot. By comparing the fringe phases seen in Fig. 9, we observed that spots 9 and 10 have fringes that are almost exactly 180° out of phase, so averaging the normalized spectra from these two spots alone produces excellent results. With a minor scaling factor for the spot 9 spectrum, this results in achieving an improvement in signal-to-fringe ratio (sensitivity) by a factor of 20 over that of a single spot (9 or 10) that would be recorded with traditional single-element detection.

 

Fig. 9. Top: raw spectra from the first 10 sequential spots in the lower arm of the C3 configuration shown in Fig. 7. Middle and bottom: spectra normalized to the ${{\rm CH}_4}$ line depth for spots 9 and 10 (showing phase difference) and the average of spots 9 and 10 that reduces the fringe amplitudes by a factor of ${\sim}{20}$. Added to the lower plot is the average of the spectra from five higher-pathlength spots (6–10) for comparison (trace is vertically offset down by 10 counts for clarity). Left to right is increasing laser current of the scan, which produces decreasing wave numbers. For the laser off, the signal counts are around zero.

Download Full Size | PDF

Regarding the fringe period seen in Fig. 9, the dominant fringe period is measured from the wavenumber scaled to be ${\sim}{0.007}\;{{\rm cm}^{- 1}}$, corresponding to a fringe generating length of ${\sim}{144}\;{\rm cm}$. Since the neighboring spots in the chosen cell patterns differ by six passes, we associate a ${\sim}{144}\;{\rm cm}$ fringe with scattering between neighboring spots in the circulation pattern, with a base mirror separation of ${\sim}{24}\;{\rm cm}$, consistent with our optical path determined by the combination of our mirror separation and added ZnSe element thickness.

4. PROPOSED ARCHITECTURES

In this paper, we report findings of three experiments using specific Herriott cell configurations and conditions. This study demonstrates the potential of improving the performance of TLSs by gaining insight into spot-by-spot and pixel-by-pixel discrimination. Implementation will depend on the available limits for mass (for a detector array and readouts or, alternatively, for possible additional single-element detectors and signal chain electronics) and for data volume and transmission rates. For the instrument architecture, we envision a set of detectors or a detector array set several cm behind the IR-transmissive far mirror, with appropriate collection optics.

By first using an IR camera to study the particular optical configuration selected, data collection at this high resolution could then be used to drive the final implementation design. We suggest here three basic configurations, all of which assume that an IR-transmissive mirror is used at the far end of the Herriott cell. Calculated data volumes and rates are given in Table 1. The suggested configurations are

Table 1. Data Volumes and Rates for Three Configurations

View Table

 

Fig. 10. Adapted from Ref. [12], showing two Herriott cell configuration solutions that make the case either for a small array architecture or one using several (e.g.,  six) individual detectors sampling the central regions of the neighboring spots within the yellow rectangles. By including the first spot, high-pass spots, and intermediate passes, sensitivity can be maximized while removing fore-optics fringes, contaminant gas spectra, and extending dynamic range. In the above figure, ${N}$ is the number of passes, ${M}$ is the number of orbits on a mirror, and $\theta$ is the angle between two successive reflections. The separation $d$ between the spherical mirrors is given by $d = ({1} - {\cos}(\theta)){/R}$, where ${R}$ is the radius of curvature of the mirrors. The configuration to the right represents the family and spot pattern used in the TLS-SAM instrument on Mars, but the mirror separation $d$ is 0.20 m, ${R}$ is 0.44 m, and $\theta$ is 57.07°.

Download Full Size | PDF

As given in Table 1, for a single-laser channel on the TLS-SAM-MSL instrument, the laser scans once per second over a 1024-point spectrum, each point needing 2 bytes, so that the instrument records 120 kB for each minute of data. Sampling a large array would require the capability to store and possibly downlink large data volumes and rates, so we calculate estimates for our three architectures for comparison in Table 1.

Scanning a laser every second and recording 1024-point spectra from a ${640} \times {512}$ pixel array would produce large data volume (${\sim} 40\,\,{\rm MB}$) in every 1 min of measurement, which is impractical for many applications, including planetary missions. Rather, we propose that the whole array is analyzed at the testing and calibration phase, during which algorithms for the best sensitivity are developed and, with the structure sufficiently stable over time, then implemented during subsequent mission data collection.

One potential disadvantage of this approach over the traditional use of aluminum mirrors attached to an aluminum cell is in regard to the coefficient of thermal expansion (CTE). With a ZnSe or other crystalline material optical output mirror, there is the possibility of CTE mismatch in extreme temperature change environments (e.g.,  on the lunar surface) that must be considered in instrument design.

5. CONCLUSIONS

An innovative approach to improving the performance of TLSs is demonstrated through three experiments using different Herriott cell configurations. This approach focuses on gaining insight into spot-by-spot and pixel-by-pixel discrimination seen by directly imaging all or part of the Herriott cell spot pattern on the far mirror and recording spectra for subsets of pixels.

This novel approach offers improved sensitivity, increased dynamic range, laser power normalization, resilience to misalignment, contaminant subtraction, and reduces the instrument power requirement by avoiding the need for fringe-wash heaters (30 W power consumption on the TLS-SAM instrument). With many tens of pixels representing each spot during the laser spectral scan, intensity and optical fringe amplitude and phase information are recorded. Selection and processing (e.g.,  co-addition, subtraction) of the pixel output spectra to minimize optical interference fringes thereby increases sensitivity. We demonstrate a factor of ${\sim}{20}$ sensitivity improvement over traditional single-element detection. Dynamic range increase of a factor of ${\sim}{100}$ is also demonstrated through spot selection representing different pathlengths. Additionally, subtracting the spectrum of the first pass from that of the higher pass normalizes the laser power and, for the fore-optics chamber, removes the contribution of contaminant gas and of fore-optics fringes. These initial results show that this imaging method is particularly advantageous for multi-laser spectrometers, and, once the image field is analyzed, pixel selection can be used to minimize data rate and volume collection requirements. We report these initial results in recognition of much needed further experimental development and analysis to discover more general algorithms applicable to a wide variety of instrument configurations.