Open Access
Add to BibSonomy
Abstract
We have designed and built a wavelength-tunable optical source for standoff detection of gaseous chemicals by differential absorption spectrometry in the long-wave infrared. It is based on a nanosecond 2 µm single-frequency optical parametric oscillator, whose idler wave is amplified in large aperture Rb:PPKTP crystals. The signal and idler waves are mixed in ${\rm ZnGeP}_2$ crystals to produce single-frequency tunable radiation in the 7.5–10.5 µm range. The source was integrated into a direct detection lidar to measure sarin and sulfur mustard inside a closed chamber, in an integrated path configuration with a noncooperative target.
© 2020 Optical Society of America
1. INTRODUCTION
Standoff detection of chemical species—particularly in the gaseous phase—is of interest for atmospheric research, Earth observation, industrial process monitoring, defense, and security.
Infrared absorption spectroscopy is a powerful tool to identify chemical species because most polyatomic molecules display characteristic absorption features linked to their vibrational and rotational modes, and thus, to their chemical structure. Quantification is also possible provided that the absorption cross sections have been measured beforehand. In the laboratory, efficient implementations of infrared spectroscopy include Fourier transform infrared spectrometers (either based on blackbody or frequency combs), and tunable laser spectroscopy [using laser diodes, quantum cascade lasers (QCLs), or optical parametric oscillators (OPOs)]. The sensitivity of such techniques is increased using multipass geometries or optical cavities, while absolute accuracy and reproducibility are guaranteed by careful thermomechanical isolation and stabilization and calibration with gas reference cells [1]. In the field, standoff detection brings new challenges besides identification and sensitive quantification of the targeted species, such as a long detection range (for early warning), fast measurements (to be insensitive to atmospheric fluctuations or to movements of either the target or the instrument carrier), and 3D localization of gas plume sources. The differential absorption lidar method (DIAL), based on a wavelength-tunable laser emitter, is a candidate to fulfill these requirements [2,3]. In this paper, we report on the experimental results obtained with a novel lidar emitter architecture for detection of chemical warfare agents (CWAs) and toxic industrial chemicals (TICs) in the long-wave infrared (LWIR). Assessment of the ultimate performance of this lidar is the object of a companion paper [4].
A. Lidar Requirements for CWA Detection in the LWIR
In this work, we focus on DIAL with direct detection, which uses attenuation and time-of-flight measurement of light pulses through the atmosphere (Fig. 1). In its simplest form, a pulsed DIAL emits two laser pulses consecutively at two different wavelengths. One wavelength in coincidence with an absorption band for the targeted species is called ${\lambda _{{\rm ON}}}$, and a second wavelength on a minimum of absorption is called ${\lambda _{{\rm OFF}}}$. The laser pulses, after passing through the target gas cloud, are backscattered by the aerosols (dust or water droplets), collected by a telescope, and focused on a photodetector. The gas concentration $C(z)$ at distance $z$ is then retrieved from the ratio of the ON and OFF backscattered signals at time $t$ such that $z = ct/2$, where $c$ is the speed of light. Spatial resolution along the line of sight is usually limited by the detector bandwidth, so that $\delta z = c/2{\varDelta}\!f$, if we approximate the detection bandwidth $\varDelta\!f$ to the inverse of the detector response time. If light is backscattered from a solid target, spatial resolution is lost and the ratio of ON and OFF signals gives the integrated optical depth along the line of sight (IPDA configuration for integrated path differential absorption lidar). Obviously, additional ON wavelengths are required for identification of more than one molecule, or if atmospheric species present infrared spectra overlapping with the one from the target. Algorithms exist to determine the best choice of wavelengths for a given detection scenario [4–6].
Fig. 2. Absorption spectra of main CWAs [7] and atmospheric transmission over 2 km calculated using the 2012 HITRAN database [8].
For long-range detection ($\gt{100}\;{\rm m}$), the choice of wavelengths is bounded by atmospheric transmission windows, roughly located between 3 and 5 µm and between 7.5 and 13.5 µm. From now on, we will call the former range midwave infrared (MWIR) and the latter LWIR. For detection of CWAs, and many TICs, the LWIR is the most interesting spectral range because it contains the wavenumbers of fundamental vibration bands of characteristic chemical bonds; hence, these bands are very strong and often nonoverlapping between species, as shown in Fig. 2. For instance, it contains information on the P-O-C, P-F, and ${\rm P} {\text -} {\rm CH}_3$ bonds of organophosphorous compounds, on the C-S, ${\rm Cl} {\text -} {\rm CH}_2$, and ${\rm N} {\text -} {\rm CH}_2$ bonds often found in hazardous compounds but not in other molecules, as well as characteristic bands arising from the coupling of several bonds, such as the one located at 8.27 µm for sulfur mustards. In contrast, the MWIR only contains the C-H vibration band, common to all organic molecules.
Even for a set of chemicals restricted to the first few major CWAs, the characteristic vibration bands are distributed all over the LWIR. For a laser-based detection system, it demands a tunability range corresponding to 40% of the central wavelength value. Failing to do so would require targeting weaker bands, with an increased risk of interference from atmospheric species and of ambiguity arising from stronger neighboring bands from other species. For a single species, the required wavelength tunability can still be very large because absorption bands of heavy molecules in the LWIR are wide. An example is shown in Fig. 3 with sarin, for which the spectral distance between the peak maximum (ON) and the first minimum (OFF) is ${50}\;{{\rm cm}^{- 1}}$. Besides, the possible laser emission lines should match the absorption extrema, while avoiding the lines of interfering species, such as greenhouse gases, including water vapor, which is not always possible.
Fig. 3. Main absorption band of sarin in the LWIR, and a couple of laser lines yielding the highest absorption contrast.
B. State of the Art of Laser Sources for LWIR Lidars
The aforementioned spectral characteristics are quite challenging for a laser emitter to provide, even more so when a high peak power is also required for long-range detection. Until recently, only the ${\rm CO}_2$ laser has been able to meet those requirements [2–18]. Among other achievements, up to $\sim{1}\;{\rm J}$ has been demonstrated (an energy far beyond operational needs), with a Fourier transform limited spectrum [14]. However, because this laser is based on molecular transitions, its tunability is limited to discrete lines in the ${924 - 987}\;{{\rm cm}^{- 1}}$ (10.1–10.8 µm) and the ${1025 - 1090}\;{{\rm cm}^{- 1}}$ (9.2–9.8 µm) ranges, with a typical separation of ${1 - 2}\;{{\rm cm}^{- 1}}$ between adjacent lines. As explained previously, this prevents from choosing the ON/OFF doublet, the one maximizing the differential absorption while being free from absorption interference from atmospheric species. Worse still, the wavelengths emitted by the ${\rm CO}_2$ laser do not overlap with the main absorption band of sulfur mustard at 8.27 µm, but only with the second strongest available band that is at least 6 times weaker.
QCLs, based on intersubband transitions in semiconductor quantum wells, can be designed to address any part of the LWIR. In addition, schemes exist that produce a narrow linewidth ($\lt{{\rm cm}^{- 1}}$) while keeping a wide tunability ($\gt{100}\;{{\rm cm}^{- 1}}$) in the pulsed regime (e.g., distributed feedback and external cavity gratings). QCLs as such (without optical amplification) can be considered for IPDA over a few meters. However, for direct detection DIAL over a kilometer range, the required peak power would be several kilowatts, which is far out of the reach of any narrowband QCL, especially in the LWIR, where the electrical-optical efficiency is rather low. Nevertheless, a QCL could be very useful as a broadly tunable source in a larger architecture, either as a master oscillator [19] or as an injection seeder [20].
In contrast with the above-mentioned technologies, optical parametric oscillators (OPOs) and amplifiers (OPAs) are able to fulfil all the requirements for standoff detection of chemicals in the MWIR and the LWIR. Indeed, since they are based on a nonresonant nonlinear interaction, they are intrinsically broadly tunable and efficient at high peak powers [21]. Besides, since they use free-space optical elements, power scalability is readily achieved by increasing the optics diameter and the pump laser power. Several MWIR lidars based on optical parametric sources with long-range capabilities have indeed been reported [22–24].
So far, what has limited the use of parametric sources for standoff detection of chemicals in the MWIR and the LWIR is the difficulty in combining a narrow linewidth ($\lt{\rm a}\,{\rm few}\,{{\rm cm}^{- 1}}$) with a high peak power ($\gt{10}\;{\rm kW}$). Indeed, the output spectrum of an OPO is mainly driven by the acceptance bandwidth of the nonlinear crystal inside the cavity, which is directly related to the dispersion of the refractive indices of the crystal. For most of the nonlinear crystal currently available for efficient LWIR generation (such as ${\rm AgGaS}_2$, ${\rm AgGaSe}_2$, ${\rm ZnGeP}_2$, CdSe, GaAs, ${\rm BaGa}_4{\rm Se}_7$, and ${\rm BaGa}_2{\rm GeSe}_6$), the acceptance bandwidth exceeds a few ${{\rm cm}^{- 1}}$ [25]. In addition, these crystals have to be pumped at 2 µm (or above) to be efficient and yield radiation over the whole LWIR. Until now, ${\rm ZnGeP}_2$ has been the most efficient crystal for LWIR generation because it presents a high nonlinear coefficient and very good thermomechanical properties. However, its transparency beyond $\sim{9}\;{\unicode{x00B5}{\rm m}}$ is not high enough for use in a nanosecond LWIR OPO. Since ${\rm AgGaSe}_2$ is fully transparent over the LWIR and presents a narrower acceptance bandwidth, it has been used in some LWIR lidars. A few hundreds of microjoules have been produced across the whole LWIR, but nonetheless the linewidth exceeded ${2.5}\;{{\rm cm}^{- 1}}$ [16,26,27]. Moreover, ${\rm AgGaSe}_2$ has a low surface laser damage threshold, which compromises the emitter reliability. CdSe is another interesting crystal because it has a higher damage threshold than ${\rm AgGaSe}_2$ and can be grown in large dimensions. However, its birefringence is quite small, preventing the generation of radiation below $\sim{9}\;{\unicode{x00B5}{\rm m}}$ when pumped around 2 µm. Recently, ${\rm BaGa}_4{\rm Se}_7$ and ${\rm BaGa}_2{\rm GeSe}_6$ have emerged as promising crystals for LWIR generation [28,29]. The latter yields almost the same efficiency as ${\rm ZnGeP}_2$ with a similar damage threshold, while being transparent up to 16 µm. However, ${\rm BaGa}_2{\rm GeSe}_6$ was not mature enough at the start of this work.
Regardless of the type of crystal used, achieving a linewidth in the LWIR below a few ${{\rm cm}^{- 1}}$ requires additional spectral filtering elements. Insertion of a spectrally selective filter inside the OPO cavity, such as a diffraction grating or an etalon, has been attempted, but this dramatically increases the oscillation threshold and sometimes reduces the wavelength tuning range, while the linewidth stays above a few ${{\rm cm}^{- 1}}$ [30,31]. The recent development of volume Bragg gratings (VBGs) for spectral filtering inside and outside cavities was a breakthrough for high peak power lasers emitting in the near-infrared domain [32,33]. However, a narrow linewidth and a large wavelength tunability are conflicting goals when designing a VBG. In addition, the photorefractive glass used for VBGs absorbs radiation at wavelengths beyond $\sim{2.4}\;{\unicode{x00B5}{\rm m}}$, which limits its use in the mid-infrared. Injection seeding with an external single-frequency laser is a proven method for single-frequency operation of OPOs. However, the tunability becomes restricted by the seeding laser. For our application, since the tunability amounts to $\sim{40}\%$ of the wavelength, this would require a broad gain grating-tuned source, such as an external cavity QCL [19] or a continuous-wave OPO [34], which are both complex sources.
For spectral filtering in wavelength-tunable OPOs, in the mid-infrared in particular, we have been promoting a solution based on Vernier frequency filtering in nested signal and idler cavities [35]. This architecture provides pulsed single-mode operation without injection seeding or intracavity filters, as well as a low oscillation threshold thanks to its doubly resonant and double-pass pumping architecture. It has been extensively used by our team for high peak power 2 µm parametric sources for lidar applications [36,37], and for LWIR sources as well [38]. Using this concept for the core oscillator of the lidar emitter and combining it with parametric amplifiers provide the spectral purity, wavelength tunability, and energy output required for long-range detection [39–41]. Power scalability of this nested-cavity OPOs in the mid-infrared is made possible by 5-mm-aperture periodically poled nonlinear Rb:KTiOPO4 (PPRKTP) crystals [42] and commercially available high-energy lasers.
In the present work, we report on what we believe to be the first parametric source emitting in the LWIR, with an appropriate linewidth and tunability for detection of multiple chemicals. Moreover, this millijoule-level source was implemented into an IPDA lidar setup, and tested outdoors on actual CWAs. It was designed to provide arbitrary wavelength tunability and to allow further energy scaling for future DIAL applications.
2. LIDAR DESCRIPTION
The lidar consists of a laser emitter and an optical receiver, mounted on separate optical breadboards assembled in a structure made of aluminum rails, and the whole system is mounted on pneumatic wheels (Fig. 4). For this project, the monostatic lidar axis was steered manually.
The subsystems of the emitter are driven by electronics controlled by LabVIEW software, while the signal from the receiver detector is digitized by an acquisition card and MATLAB software. The whole lidar can be operated remotely.
The system was primarily designed to detect pure CWAs, such as sulfur mustard, sarin, and VX, as well as other gases absorbing light in the LWIR such as ${\rm SF}_6$ and ethylene. Because the lidar’s emitter is based on a widely tunable parametric source, it can be tuned to address various species on demand. The intended use case for this lidar was that of an early warning system: the system performs continuous scanning, and a concentration anomaly is detected at a long distance, giving enough time for people to protect themselves preventively, while the hazard is being confirmed by analytical detection systems (using either optical, physical, or chemical methods), that are slower but more accurate at close range.
A. Emitter
The emitter was designed with two goals in mind: arbitrary wavelength tunability and energy scaling. We chose to use a master-oscillator power-amplifier configuration to separate the functions of multiwavelength generation and energy increase. This architecture provides a better beam quality and a narrower linewidth than a single oscillator at high energies ($\gt{\rm mJ}$).
The core of the optical source (see Fig. 5) is a nested-cavity OPO (NesCOPO). The principle of this oscillator is described in full detail elsewhere [35]. It provides single-frequency radiation, without any restriction on the wavelength tunability other than the crystal tuning capability in a compact device with a very low oscillation threshold. For this particular work, the NesCOPO was based on a 16-period periodically poled ${\rm LiNbO}_3$ (PPLN) crystal (provided by HC Photonics). The NesCOPO is pumped by a single-frequency Q-switched Nd:YAG laser, using hybrid diode/flashlamp pumping (provided by Innolas). This laser is injection-seeded by a single-mode fiber laser from NP Photonics. The laser provides 7-ns-long, linearly polarized, Gaussian-shaped pulses with an energy of 280 mJ at a repetition rate of 50 Hz. The beam propagation factor is ${{ M }^{2}}\sim 2$, and the far-field intensity pattern is ${{\rm TEM}_{00}}$. The pulse spectrum is Fourier transform-limited to a FWHM width below 70 MHz, with a central stability of $\pm 55\;{\rm MHz}$ rms over 1 min, and the side-mode suppression ratio is over 30 dB.
The NesCOPO generates single-frequency radiation at 1.86–1.94 µm (signal) and 2.48–2.35 µm (idler). Wavelength tuning is achieved by three methods: (i) coarse tuning is made by piezoelectric-driven translation of the crystal, each grating period addressing a different wavelength region ($\sim{1}\;{{\rm cm}^{- 1}}$ wide); (ii) finer tuning is made using piezoelectric-driven displacement of the cavity mirrors (by wavenumber steps of ${0.13}\;{{\rm cm}^{- 1}}$), (iii) if required, the wavelength grid provided by the 16 grating periods can be globally shifted using temperature control of the PPLN crystal (${-}{1.7}\;{{\rm cm}^{- 1}}/{\rm K}$ for the signal wave).
Each grating of the PPLN crystal has a different quasi-phase-matching period, so that the downconverted LWIR wavelength matches one or several ON/OFF bands of the target chemical species, depending on which grating is in use. The wavelengths have been chosen so as to maximize the ON/OFF absorption contrast, while avoiding regions where atmospheric absorption is too high (mainly due to water vapor lines). With the current setup, the switching time between adjacent grating periods is $\sim{0.5}\;{\rm s}$. The measured wavelength grid is shown in Fig. 6, along with the normalized output LWIR energy.
Fig. 6. Emitted wavelengths compared to the absorption spectra of sarin, sulfur mustard, and the atmosphere.
The idler wave emitted by the NesCOPO is then sent in a parametric amplifier, based on 7-mm-long, high-aperture 5-mm-thick PPRKTP crystals with a single quasi-phase-matching period of 38.52 µm. PPRKTP has been selected for its high nonlinearity, high laser damage threshold, and wide acceptance bandwidth [42]. The crystals were specifically developed for this application. From 270 mJ of pump energy at 1.064 µm, we obtained 48 mJ of single-frequency radiation (${\rm signal} + {\rm idler}$). Up to 70 mJ has been obtained in the laboratory [39], but we kept a safety margin to avoid laser damage due to beam misalignment or pollution of the optics during field trials. The conversion efficiency could be further improved with longer PPRKTP crystals and a smoother pump beam profile.
The amplified signal and idler beams are then mixed in two ${\rm ZnGeP}_2$ crystals (provided by Eksma), of lengths 7 and 4 mm, respectively, to produce a difference-frequency wave tunable between 7.3 and 10.5 µm. These crystal lengths are not optimum, as the downconversion is far from being saturated. Nevertheless, we obtained up to 1.1 mJ of single-frequency LWIR radiation. During the field trials, the energy was reduced to $\sim{0.3}\;{\rm mJ}$ for safety reasons. Incidentally, this experiment also shows that, despite its well-known absorption issues at short and long wavelengths, ${\rm ZnGeP}_2$ can be pumped below 2 µm and at the same time emit radiation at wavelengths beyond 10 µm in a practical device. This is actually possible thanks to low 2 µm absorption in electron-irradiated ZGP crystals, and small thermal effects at moderate repetition rates (50 Hz in our case).
The LWIR wavenumber is calculated from the signal and the pump wavenumbers, using the energy conservation relation,
where ${{\sigma}_{{\rm LWIR}}}$, ${{\sigma}_{s}}$, and ${{\sigma}_{p}}$ are, respectively, the LWIR, signal, and pump wavenumbers.The 1.064 µm pump wavenumber is measured once a day with a lambdameter (WS-6 IR-II from HighFinesse, 50 MHz resolution), and stays stable within ${0.005}\;{{\rm cm}^{- 1}}$ during the day. The signal wavenumber is measured at each laser shot with the same lambdameter, enabling the removal of shots occurring at undesired wavenumbers during data postprocessing. For a given single longitudinal mode, its signal wavenumber is stable within ${0.004}\;{{\rm cm}^{- 1}}$ rms over 3 min even though the NesCOPO is not servo-locked. This means that, in the worst case, the LWIR wavenumber is stable within $\pm 0.013\;{{\rm cm}^{- 1}}$ rms. Obviously, this stability is more than enough for detection of CWAs, even when considering residual interference from atmospheric species. Actually, this stability would also be suitable for detection of species with narrower absorption features, such as TICs and greenhouse gases.
The LWIR beam is then filtered from the other beams using dielectric filters, sent to an expanding telescope, and aligned along the receiver axis by a pair of gold mirrors. The resulting half-divergence is 1 mrad. An almost negligible part of the LWIR energy is extracted and sent through a 152.4-mm diameter integrating sphere onto a HgCdTe detector, without focusing, to measure the relative energy on a shot-to-shot basis (reference), and normalize the lidar return signal accordingly.
B. Receiver
The receiver consists of a Newtonian telescope focusing the backscattered light on a HgCdTe detector (Fig. 7). The telescope was designed to get the largest field of view while using as few custom optical elements as possible. The aperture is 250 mm (limited by the primary mirror), and the half-field of view is 1.6 mrad (limited by the detector diameter of 1 mm), designed to be larger than the emitter divergence. Chromatic aberrations are low thanks to the use of catadioptric optics and because of the low dispersion of optical materials in the LWIR. The primary mirror is a 250 mm diameter, 1 m focal length (F/4), protected gold-coated on-axis parabolic mirror. The beam is collimated and taken out of the telescope aperture by an off-axis, bare gold-coated parabolic mirror. The beam was finally focused by an aspheric, antireflection-coated ZnSe lens (with less than 0.5% transmission variation in the 7.5–10.5 µm range) on the detector.
The telescope was prealigned in the laboratory using a collimated laser source. The detector position was finely adjusted in the field according to the target distance. Aiming at the target was made using a visible laser pointer, and overlap of the emitter and receiver field of views was optimized by maximizing the lidar signal.
The signal detector is a liquid-nitrogen-cooled photovoltaic HgCdTe detector with a cutoff wavelength of 11.5 µm, an active area diameter of 1 mm, and an antireflection-coated ZnSe immersion lens (provided by Infrared Associates). It is followed by a linear transimpedance amplifier, and the effective bandwidth has been measured to be 5 MHz for the average gain setting. Including amplification, the detectivity is $8.5\;{10^{10}}\;{{\rm cm}\,{\rm Hz}^{- 1/2}}/{\rm W}$ for a 5 MHz bandwidth. The signal is acquired with a 20 MHz bandwidth (software-limited), 8-bit, two-channel PCI Express acquisition card (Agilent U1084A). The card is triggered by the signal from the reference detector. Since each lidar waveform typically contains 10,000 or fewer points, at the maximum repetition rate of 50 Hz the data flow stays quite low ($\lt{200}\;{\rm K}\,{\rm bytes} \cdot {\rm s}^{- 1}$), so that every emitted shot can be recorded for further data processing.
3. FIELD TESTS
A. Environment
The lidar was installed at the Chemical, Biological, Radiological and Nuclear (CBRN) testing facilities of the Brno Military Research Institute (VVU), near Vyškov (Czech Republic), during August of 2017. It was installed under a tent with a small air conditioning unit, providing basic protection from the sun, wind, rain, and high temperature variations. Meteorological data were recorded during the tests. However, for the chosen wavelength grid, atmospheric absorption was less than 1%, and in any case, humidity was quite stable so that time-dependent correction of the data from water vapor absorption was not necessary.
B. Optical Configuration
A closed chamber is placed in the beam path, at a distance of 42 m from the lidar. Behind the test chamber, a ${1} \times 1\;{{\rm m}^2}$ target is placed in the beam path, at a distance of 150 m from the lidar (IPDA configuration).
The target is covered with black cotton fabric, previously tested to provide quasi-Lambertian backscattering with an albedo $\lt{0.1}$. However, cotton fabric displays nonnegligible spectral changes in reflectance ($\gt{5}\%$) across the 7.5–11 µm region. The target distance was chosen long enough so that the echo does not overlap in time with the echo from the cell, and short enough so that the target stays in the line of sight despite the upward slope of the terrain.
In a second series of tests, the cell is removed, and an ${\rm SF}_6$ cloud is released at a distance of 120 m from the lidar, with the target still at 150 m.
C. Sample Preparation
The test chamber for the preparation of defined concentrations of chemicals is made from stainless steel. The cell windows are made of polyethylene film, inclined at the Brewster angle to minimize reflection for a vertical polarization. The chamber diameter is 1 m, the length along the laser beam is 3.9 m, and the working volume is ${9.5}\;{{\rm m}^3}$.
The homogenization of the prepared test mixture is ensured by a vapor generator integrated into a bypass tube that sucks and opens tangentially into the chamber. The outer walls are heated to 50°C to avoid condensation of the chemicals.
Control of the vaporized concentration is done by combining different analytical methods, depending on the type of chemicals used. A gas chromatography/flame ionizer (GC/FPD) analyzer (MINICAMS from OI Analytical) and an infrared Fourier-transform (FT-IR) spectrometer (Nicolet from Thermo-Fischer) were placed inside a shelter, close to the chamber. The samples from the chamber are collected on sorbent sampling tubes (Tenax) and consequently analyzed in the lab by a gas chromatography/mass-spectrometer (GC/MS) system.
Another experiment was to generate a cloud of ${\rm SF}_6$ in an open space. A network of perforated pipes was created on a base measuring ${4.5} \times 25\;{{\rm m}^2}$ and connected to a bottle of pure ${\rm SF}_6$. To check the location and dispersion of the cloud, a thermal camera (FLIR Systems) with an optical filter centered on ${\rm SF}_6$ main absorption band was used. The wind strength and direction were also recorded.
D. Typical Test Sequence
As explained earlier, the system was designed to be operated as an early warning system. Thus, in this first demonstration, it constantly monitors a fixed line of sight. The background is periodically acquired, using one or several wavelengths, and a concentration anomaly due to the passing of a hazardous gas cloud is sought. Future improvement of the system would include scanning of this line of sight to cover a 360° surveillance zone.
For each chemical, three concentration levels were tested, in increasing order. The total test duration was about 2 h, including sample preparation and system calibration. The typical sequence was as follows: (1) lidar background measurement with an empty cell, (2) injection of the lowest concentration, (3) 5 min wait for homogenization inside the cell, (4) series of lidar measurements at the first wavelength, at the second wavelength, etc., (5) injection to increase the concentration, and back to (3), (6) emptying the cell, (7) background measurement (if necessary), (8) acquisition of the electric baseline of the lidar channel.
E. Data Postprocessing
During the field tests, typically 2000 raw lidar temporal profiles are saved at each wavelength. Then, the traces undergo the following postprocessing steps: (1) removal of traces associated with wavenumbers farther than $\pm 0.05\;{{\rm cm}^{- 1}}$ from the target value; (2) removal of shots saturating either the signal or the reference detector; (3) offset subtraction, based on the average value of the detector voltage during the first 1 µs pre-trigger signal; (4) normalization of the signal to its reference using peak values of the averaged signals; (5) removal of traces associated with very low or very high emitted powers compared to the average value, sometimes correlated with spatial mode changes; (6) trigger jitter correction; (7) averaging of remaining traces (between 600 and 1800 typically); (8) transmission is taken as the peak value of the normalized signal in the time window where the target echo is expected. Most of the selection is done during Step 5 because of high shot-to-shot fluctuations of the output energy in operation conditions ($\pm 50\%$). The raw transmission at a given wavelength is calculated as the peak ratio between the signal and reference signals, after averaging and offset subtraction. It can be formally written as
Gas transmission is obtained by the ratio of the transmissions at the ON and OFF wavelengths. In IPDA mode, the cell and the target bring a nonnegligible contribution to the spectral response that can be quite different at the ON and OFF wavelengths ($\gt{5}\%$ in all practical cases). This unpredictable bias must be corrected. This is why a blank measurement (without gas) is performed before each series of measurements. Thus the differential transmission of the gas writes
Finally, the concentration-length product of the gas inside the cell is obtained from
However, it has been shown that, for DIAL to be efficient, the correlation between the ON and OFF signals without absorption from the target gas must be close to 1 [43–45]. In our case, as will be discussed later, there was a correlation issue between signals recorded at different wavelengths due to a misalignment of the emitter oscillator when switching between PPLN periods. Thus, mixing information from different wavelengths can bias the measurements by 10%–20%. Another issue is the dependence of the target albedo with wavelength. We have measured that, for cotton fabric, this can amount to 5%–20% between the ON and OFF wavelengths. We have confirmed this after the field tests by calculating the concentration with and without the information from the OFF wavelengths: the results are more consistent without taking into account the OFF wavelengths. In our experiments, the concentration product was more accurately retrieved from the very simple equation,
A discussion on how the DIAL principle could be retrieved in range-resolved measurements without a hard target, and how it would benefit from OFF wavelengths, is given in the last section of this paper.
4. FIELD TESTS RESULTS IN INTEGRATED PATH MODE
A. Closed Chamber Tests on Sarin and Sulfur Mustards
The lidar is set in integrated path mode, with the Lambertian target placed 150 m away behind the cell.
Liquid sarin was injected into a test concentration generator, where it was immediately converted to vapors and then sent into the test chamber, following the sequence depicted in Fig. 8. As it was the first time the lidar was tested on an actual agent, the injected concentration was chosen so as to cause a high optical absorption.
Fig. 8. Relative concentration of sarin through time in the test chamber as measured by inline GC/FPD. The emission times of the ON and OFF lidar lines are shown.
In the case of sarin, the blank measurement was recorded after the cell had been cleaned (at 16:30). For sulfur mustards, we took blank measurements before the first filling.
Figure 9 shows postprocessed lidar signals in the presence of sarin gas. In this example, light absorption from sarin is clearly visible in the backscattered peak at 190 m. The echo at ${z} = 40\;{\rm m}$ is due to the test chamber windows. Around this echo, ripples due to electromagnetic interference from the laser $ Q $-switching cell are visible. Their intensity at the time of the target echo is negligible. These ripples are also visible in the reference signal because the detector is located inside the laser bench. We have checked that the error in intensity evaluation caused by these ripples is negligible compared to other sources of error.
Fig. 9. Example of postprocessed lidar and reference signals at the ON and OFF wavelength in the presence of sarin (case where the concentration was the highest).
The experiment was repeated twice by reinjecting the same quantity of sarin each time. The increase in concentration measured by the lidar is consistent with the one measured by the GC/FPD system and the FT-IR (Fig. 10). The results obtained at two different wavelengths indicate some bias, which will be discussed later.
Fig. 10. Relative concentration of sarin inside the test chamber, measured with different methods: our lidar, FTIR, GC/MS, GC/FPD. The errors bars indicate the average statistical error of each method (not including bias and possible accuracy errors). According to reference spectra [7], the absorption cross sections of sulfur mustards at 1023 and ${1025}\;{{\rm cm}^{- 1}}$ are equal.
Fig. 11. Relative concentration of sulfur mustard inside the test chamber, measured with different methods: our lidar, GC/MS, GC/FPD. The errors bars indicate the average statistical error of each method (not including bias and possible accuracy errors).
We then performed a series of tests on sulfur mustard, following the same protocol as for sarin. This time, the initial concentration of the gas was chosen so that the absorption is around 10%. Sulfur mustard was detected by the lidar, but the correlation with the concentration measured by the GC/FPD was not as good as for sarin (Fig. 11). Comparison with the FT-IR could not be made because it was not sensitive enough for the tested concentrations. In situ measurement of sulfur mustard is known to be tricky due to the very low saturation vapor of this chemical, making it more a liquid aerosol than a gas. We minimized this issue by heating the cell walls and the tubes leading to the analytical equipment, but we do not have any means to quantify how much this possible liquid state can affect measurements. Another known phenomenon is the diffusion of sulfur mustard inside the polyethylene windows. We are not sure whether adsorbed sulfur mustard inside a polymer film is still optically active at the probed wavelength. Note that these issues would not be present in a real-world scenario.
B. Open Air Tests on ${\rm SF}_6$
The lidar setup was still in IPDA mode with the target at 150 m. Because the size of the cloud is unknown, the lidar measurement gives the concentration-length product, not the concentration. A blank measurement at the ON wavelength was made before the cloud release. The result is shown in Fig. 12, where ${\rm SF}_6$ monitoring was successfully carried out. At 12:18, there was a strong gust of wind, keeping the cloud close to the ground, off the laser beam path. At 12:20 and 12:23, the lidar’s emitter wavelength was changed by temperature tuning of the crystal, resulting in an interruption of the measurements. This first demonstration shows the capability of the system for continuous, nonintrusive identification of a hazardous gas plume in open air. It also highlights the fact that such plumes are turbulent due to normal wind conditions, which makes the exact quantification difficult because the plume tends to appear and disappear within a fixed line of sight.
C. Discussion on Detection Noise
Analyzing the experimental noise type and magnitude is an important step in assessing whether the system can be optimized further.
Noise can originate from different sources: electronic (detector, amplifier, and digitization), optical background shot noise, speckle (especially in IPDA with a hard target), atmospheric turbulence (especially in DIAL at low repetition rates), and laser shot noise (significant only at very high-return laser energies). Noise can affect both the signal and reference channels, even though noise in the latter can often be neglected.
During the field trials, the limiting noise was the digitization noise on the lidar return signal. The acquisition card was set to 200 mV full-scale to avoid saturation in all cases, resulting in a digitization quantum of 0.8 mV. The typical echo from the target was about 100 mV and could be as low as 15 mV in some cases, depending on the emitter energy, cell transmission, target albedo, and beam divergence. Thus, the resulting single-shot signal-to-noise ratio (SNR) was between 18 and 125. This could be easily improved by implementing automatic gain in the software, increasing the number of bits, or by using a logarithmic amplifier to level the gain difference between low- and high-return signals.
Another nonnegligible source of noise in IPDA is laser speckle due to the target roughness. In most practical cases, the illumination pattern is made of speckle grains with a size comparable or slightly smaller than the photodetector surface area. Because this pattern randomly varies with the wavelength and the beam position, it creates noise in the signal. The signal standard deviation due to purely random speckle is given by [43]
where $\theta = 0.9\;{\rm mrad}$ is the laser divergence, and ${D} = 250\;{\rm mm}$ the pupil diameter, equal to the primary mirror diameter in our case. At $\lambda = 9\;{\unicode{x00B5}{\rm m}}$, we find $\sigma = 2.5\%$, equivalent to an SNR of 40. There is little leverage on this noise source, besides increasing the pupil size.A third common noise is thermal noise, coming from the optical background at ambient temperature and from thermally generated carriers inside the detector. This noise defines the detector detectivity $D^*$, whose value is provided by the manufacturer. Knowing $D^*$, the SNR associated with the lidar channel for a given range can be calculated ab initio using the elastic scattering lidar equation [9],
D. Discussion on Biases
There is a significant bias in our concentration measurements that will require further field tests for complete clarification. What matters is not so much the systematic overestimation of the concentration, but rather the bias instability observed during the tests. More specifically, results were not always consistent at different ON wavelengths, as can be seen in Fig. 10, in which the retrieved concentrations at two close wavelengths are not equal. One reason for this may be changes in the spatial profile and/or direction of the laser beam when switching between wavelengths. We have indeed observed that the spatial profile of the NesCOPO was not always single-mode when switching between PPLN periods. This behavior is unusual and was not observed with similar NesCOPOs in the laboratory. The facets of the crystal have been analyzed with an optical interferometer, and the flatness was found to be better than $\lambda /6$, so the problem is likely to be linked to the translation stage or the way it was attached to the overall housing.
Another explanation for the bias is that, since the blank measurement is taken several tens of minutes before or after the series of measurements on the gas, its validity may be affected by changes in the scenery (target albedo, speckle pattern, and cell windows transmission) or by changes in the system (drift of the beam direction). This issue could be mitigated by reducing the integration time and the delay between the blank and the measurements (as the SNR is already very high). Another way around this issue would be to use information from the OFF wavelengths during data inversion. To do this, not only should beam wandering be addressed, but the ON/OFF switching rate (currently limited to 2 Hz) should catch up with the laser repetition rate while this rate should be increased. An ON/OFF switching rate greater than 150 Hz could be sufficient to consider the atmosphere as frozen between the ON and OFF shots [43], even though a rate of 1 kHz would be safer to hold this assumption true in all cases [44]. Removing the spectral signature of the background or of the instrument itself could be attempted by performing additional measurements at several OFF wavelengths, with a polynomial fit of the spectral baseline, at the expense of the total acquisition time.
Note that complete elimination of biases is not always necessary for a detection system. Indeed, what matters is not the accuracy but rather the capability of detecting a substance beyond a given threshold with the lowest false alarm rate. This is usually presented as a receiver operating characteristic (ROC) curve for the system [46,47]. We have calculated such curves for our lidar and will present them in a separate paper [4], even though dedicated test campaigns will be necessary to confirm the theoretical predictions with enough statistical data.
5. PROSPECTS FOR A RANGE-RESOLVED LIDAR
Range-resolved DIAL requires much more energy than IPDA because light backscattering by aerosols is several orders of magnitude weaker than diffuse reflectance from objects. Noise and bias sources are also quite different. For instance, speckle noise from the atmosphere is much weaker than from a target. However, atmospheric turbulence partially decorrelates the ON and OFF pulses if they are emitted at a rate below $\sim{1}\;{\rm kHz}$, while an IPDA correlation can be maintained if the target is static and the beam does not wander.
For the sake of demonstration, we will consider the ideal case where the SNR is limited by thermal noise. It writes, for a single shot on the signal channel,
For the typical energy emitted during the field trials, $E = 0.3\;{\rm mJ}$, we find ${\rm SNR} = 0.6$ at $z = 150\;{\rm m}$. Having a single-shot SNR below 1 is very common in practical DIAL applications. Averaging over 1000 shots would have increased the SNR to 20, assuming stochastic noise.
An SNR of 6 (after averaging) on the measured concentration is usually considered as acceptable to define the limit of sensitivity of an early warning system. Indeed, when used as a detection threshold, and assuming Gaussian distributions for the measurements value (signal with agent and noisy background), it leads to a true-positive probability of 99.865% and a false-positive probability of 0.135% [46,47]. If the alert system is interrogated every 5 min, this yields a false alarm every 2 days. For low absorption by the target gas ($\lt{10}\%$), this also translates to a required SNR of 6 on the lidar channel. Reaching this SNR at a distance of 1.5 km would require an output energy of 10 mJ per pulse.
We must stress that specific bias sources can appear in DIAL. First, the OFF wavelengths should be emitted within a time of 1/150 Hz or less after or before the ON wavelength, so that the atmosphere can be considered as frozen (identical speckle pattern and backscattering coefficient). For CWA detection, this may not be sufficient, because Mie backscattering greatly varies between the ON and OFF wavelengths, which are $\sim{50}\;{{\rm cm}^{- 1}}$ apart. This is especially true for particles with a diameter close to the wavelength ($\sim{10}\;{\unicode{x00B5}{\rm m}}$, very close to the typical diameter of Sahara dust particles, for instance). Furthermore, the complex refractive index of some aerosol materials can undergo strong variations in the mid-infrared, near resonances, affecting in turn the backscattering efficiency. For instance, silica has its main infrared resonance near 9.1 µm. Even though these chromatic effects are smaller with a real-world mixture of aerosols of different natures and sizes, they are still more detrimental in our case than for DIAL systems dedicated to lighter molecules, for which the ON and OFF wavelengths are much closer to each other. Debiasing these effects would require some knowledge on the aerosol content, either from meteorological data or from an auxiliary aerosol lidar. This is a challenge for LWIR lidars in general, which was out of the scope of the work covered by this paper.
6. CONCLUSION AND OUTLOOKS
We have reported on a novel nanosecond optical source, entirely based on nonlinear optical oscillators and amplifiers, with an output linewidth narrow enough for detection of CWAs by absorption spectrometry. The main absorption band of sulfur mustard could be addressed, a useful addition over what ${\rm CO}_2$ lasers can offer. We have implemented the source in a direct-detection lidar that was field-tested on actual agents. Successful detection of sarin, sulfur mustard, and ${\rm SF}_6$ was performed in an integrated-path configuration. The spectral properties of the emission also make this lidar suitable for detection of lighter molecular gases with narrower spectral bands, such as greenhouse gases and TICs.
Improvement of the lidar to enable unbiased, range-resolved measurements will require additional work on the optical setup. This will first require reducing beam wandering issues that affect the correlation between the ON and OFF signals, biasing the concentration measurement. Further optimization of the telescope against chromatic and field aberrations might also help in this respect. Validation of the changes should be done by performing quantitative measurements in a cell on a gas with a high vapor pressure, such as acetone. Then, the output energy has to be increased. Increasing the length of the PPRKTP crystals inside the 2 µm amplifier would increase the 2 µm energy and also result in a better balance between the signal and idler energies [39]. Then, using longer ${\rm ZnGeP}_2$ crystals, and by implementing a two-stage downconversion scheme to reduce saturation [48], the LWIR converter efficiency could be maximized. This way we expect the output energy to reach at least 5 mJ. Further increase of the energy would require additional work on the 1 µm pump laser itself (energy increase and beam shape optimization). Now that the optical design has been validated, the volume and weight of the demonstrator could easily be halved by optimizing the mechanical design. Improved thermomechanical stability of the lidar can be dramatically improved by using carbon-fiber breadboards, isostatic optical mounts, thermal isolation and stabilization, optical path shortening, and active stabilization of the pump beam.
Thanks to the generic design of the emitter, which uses commercially available components to downconvert from 1 to 2 µm and then to the LWIR, the wavelength grid could be reconfigured and extended to 14 µm by adding CdSe or ${\rm BaGa}_2{\rm GeSe}_6$ crystals. This way, other species could be addressed on demand (Lewisite, phosgene, parathion, nitric acid, ${\rm H}_2{\rm S}$, etc.).
We believe that the energy could be increased up to 10 mJ, this upper limit being set by the maximum aperture of nonlinear crystals. This energy would enable range-resolved measurements on aerosols over a kilometric range. On the other hand, this energy requirement would be conveniently reduced if the detectivity of the LWIR detector could be improved, either by electron avalanche [49] or by frequency upconversion [50].
Funding
European Defense Agency (EDA) (A-1152-RT-GP, JIP CBRN call 1, MICLID).
Disclosures
The authors declare no conflicts of interest.
REFERENCES AND NOTES
1. J. Hodgkinson and R. P. Tatam, “Optical gas sensing: a review,” Meas. Sci. Technol. 24, 012004 (2013). [CrossRef]
2. J. Li, Z. Yu, Z. Du, Y. Ji, and C. Liu, “Standoff chemical detection using laser absorption spectroscopy: a review,” Remote Sens. 12, 2771 (2020). [CrossRef]
3. C. B. Carlisle, J. E. van derLaan, L. W. Carr, P. Adam, and J.-P. Chiaroni, “CO2 laser-based differential absorption lidar system for range-resolved and long-range detection of chemical vapor plumes,” Appl. Opt. 34, 6187–6200 (1995). [CrossRef]
4. N. Cézard and J.-M. Melkonian, “Long-wave infrared multi-wavelength IPDA lidar for standoff detection of chemical warfare agents: theoretical study,” Appl. Opt. 59, 11167–11179 (2020). [CrossRef]
5. M. E. Webber, M. Pushkarsky, and C. K. N. Patel, “Optical detection of chemical warfare agents and toxic industrial chemicals: simulation,” J. Appl. Phys. 97, 113101 (2005). [CrossRef]
6. R. E. Warren, R. G. Vanderbeek, and F. M. D’Amico, “Sequential detection and concentration estimation of chemical vapors using range-resolved lidar with frequency-agile lasers,” Proc. SPIE 4036, 66–75 (2000). [CrossRef]
7. S. W. Sharpe, T. J. Johnson, P. M. Chu, J. Kleimeyer, and B. Rowland, “Quantitative infrared spectra of vapor phase chemical agents,” Proc. SPIE 5085, 19–27 (2003). [CrossRef]
8. L. Rothman, I. Gordon, Y. Babikov, A. Barbe, D. C. Benner, P. Bernath, M. Birk, L. Bizzocchi, V. Boudon, L. Brown, A. Campargue, K. Chance, E. Cohen, L. Coudert, V. Devi, B. Drouin, A. Fayt, J.-M. Flaud, R. Gamache, J. Harrison, J.-M. Hartmann, C. Hill, J. Hodges, D. Jacquemart, A. Jolly, J. Lamouroux, R. Le Roy, G. Li, D. Long, O. Lyulin, C. Mackie, S. Massie, S. Mikhailenko, H. Müller, O. Naumenko, A. Nikitin, J. Orphal, V. Perevalov, A. Perrin, E. Polovtseva, C. Richard, M. Smith, E. Starikova, K. Sung, S. Tashkun, J. Tennyson, G. Toon, V. Tyuterev, and G. Wagner, “The HITRAN2012 molecular spectroscopic database,” J. Quant. Spectrosc. Radiat. Transfer 130, 4–50 (2013). [CrossRef]
9. P. H. Flamant, “Atmospheric and meteorological Lidar: from pioneers to space applications,” C. R. Phys. 6, 864–875 (2005). [CrossRef]
10. N. Menyuk, D. K. Killinger, and W. E. DeFeo, “Laser remote sensing of hydrazine, MMH, and UDMH using a differential-absorption CO2 lidar,” Appl. Opt. 21, 2275–2286 (1982). [CrossRef]
11. W. B. Grant, J. S. Margolis, A. M. Brothers, and D. M. Tratt, “CO2 DIAL measurements of water vapor,” Appl. Opt. 26, 3033–3042 (1987). [CrossRef]
12. P. Adam, J.-L. Duvent, and S. W. Gotoff, “Detection and reconnaissance of pollutant clouds by CO2 lidar (MIRELA),” Proc. SPIE 3127, 212–223 (1997). [CrossRef]
13. J. R. Quagliano, P. O. Stoutland, R. R. Petrin, R. K. Sander, R. J. Romero, M. C. Whitehead, C. R. Quick, J. J. Tiee, and L. J. Jolin, “Quantitative chemical identification of four gases in remote infrared (9-11 µm) differential absorption lidar experiments,” Appl. Opt. 36, 1915–1927 (1997). [CrossRef]
14. D. B. Cohn, E. J. Griffin, L. F. Klaras, M. E. Ehritz, C. R. Swim, and J. A. Fox, “WILDCAT chemical sensor development,” Proc. SPIE 4378, 34–42 (2001). [CrossRef]
15. H. Kariminezhad, P. Parvin, F. Borna, and A. Bavali, “SF6 leak detection of high-voltage installations using TEA-CO2 laser-based DIAL,” Opt. Lasers Eng. 48, 491–499 (2010). [CrossRef]
16. V. Vaicikauskas, V. Kabelka, Z. Kuprionis, and M. Kaucikas, “An OPO-based LIDAR system for differential absorption measurements of hazardous gases in mid-infrared spectral region,” Proc. SPIE 5988, 59880A (2005). [CrossRef]
17. See also the proceedings of the 3th, 4th, and 5th International Symposium on Protection against Chemical and Biological Warfare Agents, held in Stockholm, Sweden, in 1989, 1992, and 1995.
18. P. Gaudio, “Laser based standoff techniques: a review on old and new perspective for chemical detection and identification,” in Cyber and Chemical, Biological, Radiological, Nuclear, Explosives Challenges: Threats and Counter Efforts (Springer, 2017), pp. 155–177.
19. H. D. Tholl, F. Münzhuber, J. Kunz, M. Raab, M. Rattunde, S. Hugger, F. Gutty, A. Grisard, C. Larat, D. Papillon, M. Schwarz, E. Lallier, M. Kastek, T. Piatkowski, F. Brygo, C. Awanzino, F. Wilsenack, and A. Lorenzen, “Active multispectral reflection fingerprinting of persistent chemical agents,” Proc. SPIE 10434, 104340C (2017). [CrossRef]
20. K. M. Nowak, T. Suganuma, Y. Kurosawa, T. Ohta, Y. Kawasuji, H. Nakarai, T. Saitou, J. Fujimoto, H. Mizoguchi, A. Sumitani, and A. Endo, “QCL seeded, ns-pulse, multi-line, CO2 laser oscillator for laser-produced-plasma extreme-UV source,” Proc. SPIE 10254, 102540W (2017). [CrossRef]
21. I. T. Sorokina, “Broadband solid-state lasers,” in Mid-Infrared Coherent Sources and Applications (Springer, 2008), pp. 225–260.
22. S. Yakovlev, S. Sadovnikov, O. Kharchenko, and N. Kravtsova, “Remote sensing of atmospheric methane with IR OPO lidar system,” Atmosphere 11, 70 (2020). [CrossRef]
23. O. Romanovskii, S. Sadovnikov, O. Kharchenko, and S. Yakovlev, “Development of near/mid IR differential absorption OPO lidar system for sensing of atmospheric gases,” Opt. Laser Technol. 116, 43–47 (2019). [CrossRef]
24. S. Veerabuthiran, A. Razdan, M. Jindal, and G. Prasad, “Open field testing of mid IR DIAL for remote detection of thiodiglycol vapor plumes in the topographic target configuration,” Sens. Actuators B Chem. 298, 126833 (2019). [CrossRef]
25. V. Petrov, “Frequency down-conversion of solid-state laser sources to the mid-infrared spectral range using non-oxide nonlinear crystals,” Prog. Quantum Electron. 42, 1–106 (2015). [CrossRef]
26. S. Chandra, M. Wager, B. Clayton, A. Geiser, T. H. Allik, J. L. Ahi, C. Miller, P. Budni, P. Ketteridge, K. Lanier, E. Chicklis, J. A. Hutchinson, and W. W. Hovis, “2-µm pumped 8-12 µm OPO source for remote chemical sensing,” Proc. SPIE 4036, 200−208 (2000). [CrossRef]
27. D. B. Cohn, J. M. Fukumoto, J. A. Fox, and C. R. Swim, “Compact DIAL sensor: SHREWD,” Proc. SPIE 4378, 43–49 (2001). [CrossRef]
28. Y. He, Y. Guo, D. Xu, Y. Wang, X. Zhu, J. Yao, C. Yan, L. Tang, J. Li, K. Zhong, C. Liu, X. Fan, Y. Wu, and J. Yao, “High energy and tunable mid-infrared source based on BaGa4Se7 crystal by single-pass difference-frequency generation,” Opt. Express 27, 9241–9249 (2019). [CrossRef]
29. K. Kato, V. V. Badikov, L. Wang, V. L. Panyutin, K. V. Mitin, K. Miyata, and V. Petrov, “Effective nonlinearity of the new quaternary chalcogenide crystal BaGa2 GeSe6,” Opt. Lett. 45, 2136–2139 (2020). [CrossRef]
30. F. Ganikhanov, T. Caughey, and K. L. Vodopyanov, “Narrow-linewidth middle-infrared ZnGeP2 optical parametric oscillator,” J. Opt. Soc. Am. B 18, 818–822 (2001). [CrossRef]
31. G. Mennerat and P. Kupecek, “High-energy narrow-linewidth tunable source in the mid infrared,” in OSA TOPS on Advanced Solid State Lasers (OSA, 1998), Vol. 19, paper FC13.
32. A. Kabacinski, J. Armougom, J.-M. Melkonian, M. Raybaut, J.-B. Dherbecourt, A. Godard, R. Vasilyeu, and V. Smirnov, “Wavelength tunable, single-longitudinal-mode optical parametric oscillator with a transversally chirped volume Bragg grating,” Opt. Lett. 45, 607–610 (2020). [CrossRef]
33. Q. Berthomé, A. Grisard, B. Faure, G. Souhaité, E. Lallier, J.-M. Melkonian, and A. Godard, “Actively Q-switched tunable single-longitudinal-mode 2.05 µm Tm:YAP laser using a transversally chirped volume Bragg grating,” Opt. Express 28, 5013–5021 (2020). [CrossRef]
34. J. H. Marquardt, R. W. Mackes, and D. D. Smith, “Single-mode, tunable output from a midwave-infrared-seeded optical parametric oscillator system,” Appl. Opt. 41, 1163–1168 (2002). [CrossRef]
35. A. Godard, M. Raybaut, and M. Lefebvre, “Nested cavity optical parametric oscillators—a tunable frequency synthesizer for gas sensing,” in Encyclopedia of Analytical Chemistry (Wiley, 2006).
36. J. Barrientos Barria, D. Mammez, E. Cadiou, J. B. Dherbecourt, M. Raybaut, T. Schmid, A. Bresson, J. M. Melkonian, A. Godard, J. Pelon, and M. Lefebvre, “Multispecies high-energy emitter for CO2, CH4, and H2O monitoring in the 2 μm range,” Opt. Lett. 39, 6719−6722 (2014). [CrossRef]
37. E. Cadiou, D. Mammez, J.-B. Dherbecourt, G. Gorju, J. Pelon, J.-M. Melkonian, A. Godard, and M. Raybaut, “Atmospheric boundary layer CO2 remote sensing with a direct detection LIDAR instrument based on a widely tunable optical parametric source,” Opt. Lett. 42, 4044–4047 (2017). [CrossRef]
38. J. Armougom, J.-M. Melkonian, J.-B. Dherbecourt, M. Raybaut, A. Grisard, E. Lallier, B. Gérard, B. Faure, G. Souhaité, B. Boulanger, and A. Godard, “A narrowband infrared source based on orientation-patterned GaAs for standoff detection of chemicals,” Appl. Phys. B 124, 133 (2018). [CrossRef]
39. J. Armougom, J.-M. Melkonian, J.-B. Dherbercourt, M. Raybaut, A. Godard, R. S. Coetzee, A. Zukauskas, and V. Pasiskevicius, “70 mJ single-frequency parametric source tunable between 1.87-1.93 µm and 2.37-2.47 µm for difference frequency generation in the LWIR,” in European Conference on Lasers and Electro-Optics and European Quantum Electronics Conference (2017), paper CD_P_32.
40. M. W. Haakestad, G. Arisholm, E. Lippert, S. Nicolas, G. Rustad, and K. Stenersen, “High-pulse-energy 8 µm laser source based on optical parametric amplification in ZnGeP2,” Proc. SPIE 6998, 699812 (2008) [CrossRef]
41. J. Saikawa, M. Miyazaki, M. Fujii, H. Ishizuki, and T. Taira, “High-energy, broadly tunable, narrow-bandwidth mid-infrared optical parametric system pumped by quasi-phase-matched devices,” Opt. Lett. 33, 1699−1701 (2008). [CrossRef]
42. A. Zukauskas, N. Thilmann, V. Pasiskevicius, F. Laurell, and C. Canalias, “5 mm thick periodically poled Rb-doped KTP for high energy optical parametric frequency conversion,” Opt. Mater. Express 1, 201−206 (2011). [CrossRef]
43. J. A. Fox, C. R. Gautier, and J. L. Ahl, “Practical considerations for the design of CO2 lidar systems,” Appl. Opt. 27, 847–855 (1988). [CrossRef]
44. N. Menyuk and D. K. Killinger, “Temporal correlation measurements of pulsed dual CO2 lidar returns,” Opt. Lett. 6, 301–303 (1981). [CrossRef]
45. R. E. Warren, “Effect of pulse-pair correlation on differential absorption lidar,” Appl. Opt. 24, 3472–3475 (1985). [CrossRef]
46. W. Peterson, T. Birdsall, and W. Fox, “The theory of signal detectability,” Trans. IRE Prof. Group Inf. Theory 4, 171–212 (1954). [CrossRef]
47. H. L. V. Trees, Detection, Estimation, and Modulation Theory, Part I (Wiley, 2004).
48. J. Haus, A. Pandey, and P. Powers, “Boosting quantum efficiency using multi-stage parametric amplification,” Opt. Commun. 269, 378−384 (2007). [CrossRef]
49. A. Singh, V. Srivastav, and R. Pal, “HgCdTe avalanche photodiodes: a review,” Opt. Laser Technol. 43, 1358–1370 (2011). [CrossRef]
50. Y.-P. Tseng, C. Pedersen, and P. Tidemand-Lichtenberg, “Upconversion detection of long-wave infrared radiation from a quantum cascade laser,” Opt. Mater. Express 8, 1313–1321 (2018). [CrossRef]
