1. Introduction

The mid-infrared (MIR) spectral region offers the advantages of selectivity and sensitivity in chemical sensing by targeting the fundamental vibrational bands of chemical species. This is particularly important for the quantification of chemicals in complex gas phase mixtures for a diverse number of commercial and industrial applications that include medical breath analysis, measurement of atmospheric pollutants, monitoring of combustion processes, or in-line monitoring of industrial effluents [1–4 ]. Chemical species of interest include both small molecules with resolved rovibrational transitions consisting of narrow isolated absorption lines, and also larger molecules with unresolved rotational structure consisting of broad, but distinctive absorption features.

In terms of detecting, identifying, and quantifying chemical species in mixtures based on spectral measurements, broad wavelength coverage is crucial to measure enough of each chemical’s spectrum to allow differentiation from other chemicals. Spectral resolution is also important so that the distinctive spectral features can be measured accurately without instrumental broadening and therefore differentiated from other species with overlapping spectra. For atmospheric pressure gases, the narrowest spectral features which need to be resolved are typically >0.1 cm⁻¹. Another important attribute for chemical detection is the spectral acquisition rate. The necessary acquisition rate is application dependent, but needs to be faster than the timescale of changes in the chemical species concentrations such that distortion of the absorption spectrum is minimized. There are many demanding applications where the spectral acquisition rate may need to be faster than 1 Hz. These applications include atmospheric monitoring of transient chemical releases, inline measurements in smokestacks, or measurements in turbulent combustion environments.

External-cavity quantum cascade lasers (ECQCLs) are excellent sources for MIR spectroscopy of complex chemical mixtures due to their high spectral brightness, narrow spectral linewidth, and broad wavelength coverage (Δλ/λ > 10%). In particular, the broad wavelength coverage provided by ECQCLs is beneficial when analyzing complex mixtures where significant spectral congestion can make quantification of individual chemicals challenging. Broad wavelength coverage also enables measurement of absorption bands of large chemical species with broad unresolved rotational structure — a capability that cannot be achieved with the limited tuning range of distributed feedback (DFB)-QCLs or DFB diode lasers. In addition to their broad wavelength coverage, ECQCLs also exhibit a narrow linewidth, enabling high-resolution spectral measurements [5–7 ]. However, due to design and engineering constraints on wavelength tuning mechanisms, high-resolution measurements with ECQCLs often may only be performed over selected narrow tuning ranges within the overall tuning range of the ECQCL. In this case, an ECQCL is used similarly to a DFB-QCL but with additional flexibility to select tuning windows over a much larger overall wavelength range. While small tuning windows may be “stitched” together sequentially to achieve a larger overall tuning range, this process is slow and often cumbersome making it unsuitable for many applications requiring fast time response.

Most current ECQCLs lack the ability to perform rapid measurements over large spectral ranges combined with simultaneous high spectral resolution. Demonstration of ECQCL sources used in slow scanning, or step scanning, over the entire wavelength range at times of 1 second or greater, with various spectral resolutions, have been reported in a number of studies [8–13 ]. These ECQCL systems are most relevant to offline sample analysis or chemical mixtures where the species concentrations are constant or changing on timescales greater than 1 second. Rapidly-swept ECQCL sources have been developed that incorporate intra-cavity microelectromechanical (MEMS) elements or acousto-optic modulators that can be swept over > 100 cm⁻¹ in times less than 1 ms [14–16 ]; however, the spectral resolution of these systems is > 1 cm⁻¹, making them most suitable for detection of chemicals with broad absorption features (solids, liquids, or large gas-phase molecules). Fast tuning rates of 5 kHz with a 0.4 cm⁻¹ spectral resolution have been achieved using a piezo-actuated mirror with a grating-tuned ECQCL, but over a limited tuning range of 7 cm⁻¹ [17]. Commercial ECQCL systems are available which provide 250 cm⁻¹ of tuning range at 100 Hz scan rates with a spectral resolution of 2 cm⁻¹ [18].

In this manuscript, we demonstrate an ECQCL system that can simultaneously achieve a rapid tuning rate (200 Hz), a large tuning range (110 cm⁻¹), and a high spectral resolution (0.2 cm⁻¹). The narrow spectral resolution of the ECQCL system enables rapid measurements of small chemical species relevant to atmospheric sensing while still preserving the broad spectral coverage needed for simultaneous measurement of multiple chemicals, which includes large chemical species with broad absorption spectra. This is a new capability that has not been realized using other rapidly-swept ECQCL sources [14–18 ]. We previously published a detailed study of the performance of a trace-gas sensor that used an ECQCL source swept at 20 Hz over a tuning range of 120 cm⁻¹ with a spectral resolution of ~0.2-0.3 cm⁻¹ to record broadband absorption spectra of fluorocarbons [19]. In this work, we increase the spectral acquisition rates up to 200 Hz for an ECQCL operating from 7.19 µm (1390 cm⁻¹) to 7.82 µm (1278 cm⁻¹), and use the ECQCL to measure absorption spectra of a representative large molecule (1,1,1,2-tetrafluoroethane: F134A) and a small molecule (nitrous oxide: N₂O), in the presence of strong spectral interference due to water vapor absorption. The ECQCL was used in combination with an open-air Herriott cell to measure concentration changes with a 5 ms time resolution in transient plumes of F134A and N₂O near the open path Herriot cell. For 200 Hz sweeps over a 110 cm⁻¹ range, the spectral resolution of the ECQCL was measured to be 0.2 cm⁻¹, which was sufficient to resolve the rotational structure in N₂O and reduce the effect of water vapor absorption lines. At this 200 Hz acquisition rate, noise-equivalent concentrations (NECs) for F134A and N₂O were determined to be 16 ppb and 70 ppb, respectively. We also present a detailed characterization of absorbance noise, spectral resolution, scan wavelength repeatability, and noise equivalent concentrations as a function of spectral sweep rate. The conclusions from this analysis show minimal degradation in the performance of the rapidly swept ECQCL as the spectral acquisition is increased from 40 to 200 Hz. Noise equivalent column density (NECD) values of 0.64 and 0.25 $ppmv\times m$in 1 sec are obtained for N₂O and F134A, respectively, which are compared with other DFB-QCL and ECQCL-based sensors.

2. Experimental details

2.1 Experimental layout

The experimental configuration used in this work is shown in Fig. 1(a) . Wavelength tunable MIR light from 7.19 µm (1390 cm⁻¹) to 7.82 µm (1278 cm⁻¹) was provided by an ECQCL with a design based on a Littman-Metcalf cavity, which has been described in prior work where it was used for trace gas sensing [19,20 ] and interrogation of condensed phase materials [21–23 ]. The ECQCL wavelength was tuned by adjusting the angle of an intra-cavity mirror that is attached to a galvanometer. Current was sourced to the laser using a custom low-noise current controller [24], and the laser current was amplitude-modulated by applying a square voltage waveform to the modulation input of the current controller. Details regarding acquisition of data with the ECQCL will be provided in the following section.

 

Fig. 1 Panel (a) A diagram of the experimental layout used in this work. Acronyms for the labels in the figure are defined as follows: ECQCL: External cavity quantum cascade laser, SIG DET: liquid nitrogen cooled PV-MCT detector used for measuring transmitted light from the multi-pass cell, MPC: Herriott multi-pass cell, RP: release point for the simulated N₂O and F134A plumes, DAQ: digital acquisition system, POL: wire grid polarizer used to attenuate the transmitted laser beam from the MPC. Panel (b) A plot of the transmitted intensity through the Herriott cell as the laser was swept over its full tuning range is compared to a simulated absorption spectrum through 127 meters of the atmosphere that was generated using parameters from the HITRAN spectral database [36]. Panels (c) and (d) are plots of the NWIR library spectra for F134A and N₂O used in the WLS fitting analysis [25].

Download Full Size | PDF

The light from the ECQCL was sent through a 2-lens telescope (2x magnification, 25.4 mm and 50.8 mm focal length ZnSe lenses) to facilitate mode matching of the output beam from the ECQCL to an open-path Herriott cell. The Herriot cell was composed of two 6-inch diameter protected gold-coated concave mirrors that were mounted on a dovetail rail. The mirrors have a focal length of 500 mm, and were approximately separated by 0.75 meters. A 4-mm diameter hole in one mirror was used for coupling the laser beam in and out of the Herriott cell. A 127-m optical path was determined in this configuration by measuring the delay time in the arrival of light pulses from the ECQCL between photodetectors placed before and after the Herriott cell. The beam path through the Herriott cell was completely open to the room air at ambient pressure.

As shown in Fig. 1(a), two 120-mm DC fans were used to draw air in on the side of the Herriott cell where plumes of N₂O (Matheson Tri-Gas, 99.99% purity) and F134A were generated. The fans were used to simulate the rapid passage of a plume through the Herriott cell, making it possible to test the time response of the system to quick changes in chemical species concentrations. N₂O and F134A were released six inches from the fans. The N₂O release was controlled by leaking the gas through a needle valve, while the F134A was released manually from a commercial air duster canister.

Light exiting the Herriott cell was passed through a wire grid polarizer and then focused onto a liquid nitrogen cooled PV-MCT detector (SIG DET in Fig. 1(a)) (Fermionics Corp., PV-12-1) with a custom low-noise transimpedance amplifier. The wire-grid polarizer was used to attenuate the laser beam to avoid saturation of the detector. The voltage output from the amplifier was read by a 16-bit digital acquisition board (DAQ) (National Instruments, USB-6366) at 2 MS/s. An example of the transmitted intensity spectrum from the Herriott cell observed by the signal detector is shown in Fig. 1(b). The transmission spectrum through the Herriott cell highlights the pervasive interference from water absorption over the tuning window of the ECQCL. Processing of the voltage signal acquired by the signal detector was handled by a custom LabView program that will be described in the following section.

2.2 Data acquisition and post-processing

LabView was used to synchronize control of the laser with acquisition of spectral data. The ECQCL wavelength was swept by applying a sine wave drive to the galvanometer controlling the intra-cavity mirror. In the work presented here, sinusoidal drive frequencies of 20, 50, and 100 Hz were used, corresponding to sweeps of the ECQCL wavelength across its full tuning range at 40, 100, and 200 Hz (using both the forward and backward wavelength sweeps). The current applied to the QCL was modulated from 0 to 580 mA using a square waveform with a 50% duty cycle and a repetition rate set to 100 kHz with the 20 Hz galvanometer drive and 200 kHz otherwise. Increasing the current repetition rate to 200 kHz was necessary to increase the number of wavelength points sampled as the galvanometer is swept. As demonstrated previously, amplitude modulation of the QCL current reduces the impact of low-frequency noise and the effect of external cavity (EC) mode-hops on the spectrum [19].

Figures 2(a) and 2(b) illustrate the collection and processing of simulated data during spectral acquisition. The galvanometer position is recorded by measuring the signal from an angular encoder integrated with the galvanometer, and this is used for wavelength calibration of the scan. The intensity transmitted through the Herriott cell was obtained by recording the voltage from the signal detector, which is amplitude modulated due to the modulation of the laser current. The modulated detector signal was processed digitally in software to provide a single value for each current modulation period, corresponding to the difference between the average of the high (I_on) and low detector signals (I_off) as shown in Fig. 2(b), and provided high-frequency subtraction of any background signals on the detector not due to the ECQCL. Due to the 2 MS/s sampling rate of the DAQ board, 20 intensity readings were acquired for each modulation period of the laser current at a current modulation frequency of 100 kHz. For a current modulation frequency of 200 kHz this dropped to 10 intensity readings per current modulation period. Of the 20 intensity readings acquired at a current modulation frequency of 100 kHz, average values for I_on and I_off were calculated using 8 intensity readings each. At a current modulation frequency of 200 kHz, the average values for I_on and I_off were calculated using 4 intensity readings each. The value from the angular encoder assigned to the value of I_on-off is the average angular encoder signal when a high detector signal was recorded (I_on).

 

Fig. 2 Panel (a) A simulated plot of the time-dependent encoder signal (red trace) and light intensity transmitted (black trace) by the Herriott cell during a single spectral acquisition. For this example a galvanometer frequency and intensity modulation frequency of 20 Hz and 100 kHz were used, respectively. Panel (b) An expanded view of the intensity modulated signal around 0.01 seconds. I_on,avg and I_off,avg correspond to the average intensities transmitted by the Herriott cell when the laser is on and off respectively. The value of I_on-off is calculated by subtracting I_on,avg from I_off,avg.

Download Full Size | PDF

Following collection of the spectral data, offline post-processing was carried out to calculate the wavelength axis from the encoder voltage. The initial step in the post-processing procedure was to split each galvanometer modulation period into separate spectra for the forward and backwards sweep of the ECQCL wavelength. Next, a subset of spectra acquired before the release of N₂O and F134A were averaged to create a background intensity spectrum ($I_0(V_{encoder})$). To convert the encoder voltage to wavelength, the encoder voltage positions for 9 water lines in the background spectrum were matched to their wavelength positions using a simulated absorption spectrum generated from the HITRAN database (see Fig. 1(b)). The wavelength-versus-encoder voltage data was fit to the grating equation formula by adjusting coefficients that relate the encoder voltage to the grating angle. The coefficients extracted from this fit are used to calibrate the wavelength for the background spectrum and all the spectra in the data set. Due to small differences in the scanning behavior of the galvanometer, the forward and backward sweeps were wavelength calibrated independently and then mapped to a common wavelength axis, resulting in sequential calibrated spectra at twice the galvanometer modulation sweep rate. Following wavelength calibration, the background spectrum was used to convert the intensity spectra in the data set to base-10 absorbance, $A_{10}=-\log \left(I\left(\tilde{v}\right)/I_0\left(\tilde{v}\right)\right)$.

Species concentrations were extracted from the calculated absorbance spectra by applying a weighted least squares (WLS) fitting algorithm to the data in LabView. Four basis vectors were used in the WLS fit: 1) an N₂O library spectrum, 2) an F134A library spectrum, 3) an offset, and 4) a slope. The library spectra for N₂O and F134A were taken from the NWIR database [25]. Both library spectra are presented in Figs. 1(c) and 1(d) respectively. The NWIR library spectra were broadened using a Gaussian function to account for the instrumental scan resolution of the ECQCL. Further information regarding the selection of the Gaussian FWHM used to account for the instrumental broadening is provided in Section 3.2. Spectral features from atmospheric H₂O and CH₄ that overlapped with spectral features from N₂O and F134A were present in the spectra collected; however, inclusion of library spectra in the WLS for H₂O and CH₄ resulted in a negligible change in the extracted N₂O and F134A concentrations. Because of this, they were not added to the set of basis vectors. The square of the transmitted intensity for each spectrum was used as the weighting factor in the analysis to minimize the influence of regions in the spectrum with low transmitted light intensity due to strong absorption from H₂O.

3. Results and discussion

3.1 Demonstration of swept-wavelength detections of N₂O and F134A

To demonstrate the response time of the current ECQCL system to changes in chemical species concentrations, plumes of N₂O and F134A were generated near the open-path Herriott cell. As described in Section 2.1, the plumes were forced through the Herriott cell using DC fans to simulate a rapidly moving plume. The ECQCL was swept across its full tuning range at 200 Hz rates, acquiring a full spectral sweep every 5 ms. Spectra were continuously acquired over a 30 minute time interval. Within this time interval there were multiple releases of N₂O and F134A, and for each individual release 20 seconds of pre-release spectra were used to generate the background spectrum.

Figures 3(a) and 3(c) show single frames taken from movies of absorbance spectra recorded with the ECQCL system at an acquisition rate of 200 Hz/spectrum for N₂O and F134A (Visualization 1 and Visualization 2), respectively. Note that to show better detail in the absorption spectra, the spectral ranges displayed are subsets of the total spectral range acquired by the sensor. Also shown are the best-fit spectra from the WLS algorithm and the determined concentrations. The animations are displayed in slow motion at a 20 Hz rate to allow visualization at video display rates. The results demonstrate acquisition of full spectra at 200 Hz rates, which is sufficient to track the rapid changes in chemical concentrations present from the plume passing through the open sensing region. It is also apparent from the WLS fits that the spectra are recorded with minimal distortion which might arise from concentration changes during a single wavelength sweep, i.e. the concentration is nearly constant during each sweep of the ECQCL wavelength. Regions of high absorbance noise correspond to strong atmospheric water vapor absorption, and low light levels on the detector. These localized regions have a minimal effect on the spectral fitting results due to the low weight placed on these regions by the WLS algorithm. In addition, by using the full absorbance spectrum in the analysis, encompassing multiple lines or spectral features for each chemical, localized regions of random or systematic noise has a minimal effect on the overall WLS fit.

 

Fig. 3 Panels (a) and (c) Example spectra for N₂O and F134A collected at an acquisition rate of 200 Hz. In both panels the result of the WLS fit (red trace) is overlaid on the experimental spectrum (black trace). Panels (b) and (d) The time dependent N₂O and F134A concentration profiles obtained from WLS fitting of the spectral data are provided in panels (a) and (c). For the WLS of the N₂O data, the NWIR library spectrum was convolved with a Gaussian function with a FWHM of 0.20 cm⁻¹. The time dependent concentration profiles for N₂O and F134A extracted from WLS fitting of individual spectra over 2 seconds are provided as movies in Visualization 1 and Visualization 2 respectively.

Download Full Size | PDF

As demonstrated in Fig. 3, a high spectral acquisition rate is needed to follow the rapidly changing chemical concentrations. To quantitatively characterize the timescales for the species concentration changes, a fast Fourier transform (FFT) is applied separately to the pre-release and post-release N₂O concentration data collected. The pre-release and post-release data from the analysis are shown in Figs. 4(a) and 4(b) respectively. The resulting FFT spectra are provided in Fig. 4(c). The FFT amplitude spectrum from the post-release data shows a pronounced increase in signal at lower frequencies (< 40 Hz) that approaches a baseline around 50 Hz. This indicates that around a spectral acquisition rate of 50 Hz the largest changes in the N₂O concentration are being captured. The relatively flat FFT spectrum for the pre-release data supports the conclusion that the low-frequency amplitude component in the post-release data arises solely from N₂O concentration changes in the plume.

 

Fig. 4 Panels (a) and (b) 20 seconds of N₂O concentration data recorded at a 200 Hz acquisition rate for pre and post-release conditions, respectively. Panel (c) An FFT of the time-dependent concentration traces from panels (a) and (b). In panel (c), “Background FFT” and “Signal FFT” were generated from the data shown in panels (a) and (b) respectively. The “Smoothed Signal” and “Smoothed Background” traces in panel (c) serve as a guide for the trend in the FFT amplitudes of the signal and background for the time-dependent concentration data.

Download Full Size | PDF

3.2 Characterization of swept-wavelength performance

The results in Section 3.1 demonstrate the capability for rapid measurement of chemical species using broad spectral ranges. In this section we present additional characterization to understand how the performance of the ECQCL varies as a function of the acquisition rate. Specifically, we analyze the spectral resolution, absorbance noise, and the scan wavelength repeatability as functions of spectral acquisition rate. These results in turn provide insight into potential trade-offs between measurement sensitivity and response time.

One factor that can potentially degrade performance at higher acquisition rates is a loss of spectral resolution. Figure 5 shows absorbance spectra for N₂O at different sweep rates, where each spectrum is averaged from 20 sequential measurements. Based on the averaged N₂O absorbance spectra shown in Fig. 5, there is not a significant increase in the linewidth for the N₂O spectral features as the acquisition rate is increased from 40 to 200 Hz. The instrumental resolution is determined by convolving a Gaussian function with the NWIR library spectrum. Best fits of the library to the experimental spectra are found for a Gaussian width of 0.12 cm⁻¹ (FWHM) for acquisition rates of 40 and 100 Hz, and 0.20 cm⁻¹ for 200 Hz. In Fig. 5 the agreement with broadened NWIR library spectra from the WLS analysis of the average absorbance spectra is provided. The 0.20 cm⁻¹ spectral resolution at 200 Hz is smaller than what has been reported for other rapidly swept ECQCL sources [14–18 ]. The narrow spectral resolution can be attributed to the use of a Littman-Metcalf rather than a Littrow EC configuration, and operation of the ECQCL with amplitude modulated current [19]. Maintaining a narrow spectral resolution for spectral acquisition rates greater than 200 Hz requires a concomitant increase in the current modulation frequency beyond 200 kHz. Due to the 2 MS/s sampling rate of the DAQ board, under sampling of the modulated intensity signal becomes problematic beyond 200 kHz, so acquisition rates greater than 200 Hz are not explored in this work.

 

Fig. 5 Average N₂O spectra acquired at acquisition rates from 40 to 200 Hz. For each acquisition rate 20 spectra are averaged. The resulting WLS fit using the broadened NWIR library spectrum is provided for each average spectrum (red trace).

Download Full Size | PDF

Another important parameter of the swept-wavelength ECQCL is the repeatability of scans, especially given the manner in which absorbance spectra are calculated using background spectra acquired at prior times. The scan wavelength repeatability of the current ECQCL system is characterized by monitoring the peak position of an H₂O spectral feature at 1308.2 cm⁻¹. This spectral feature is shown in Fig. 6(a) , and is one of the nine H₂O spectral features used in the wavelength calibration procedure. The wavelength repeatability for the selected H₂O spectral feature is evaluated using 1 hour data sets recorded in the absence of the release of N₂O or F134A. The first 360 seconds of spectra recorded are used to generate a background spectrum for wavelength calibration. The peak position of the water line is then determined for each scan acquired in the remaining 54 minutes of the data set. Figure 6(b) presents Allan deviation plots for the peak position at different spectral acquisition rates. For the smallest values of $\tau$in Fig. 6(b) at acquisition rates of 40, 100, and 200 Hz, we observe a ~0.03 cm⁻¹ uncertainty in the line position. This is half of the ~0.06 cm⁻¹ separation between the external cavity modes of the ECQCL, and represents the limit in the precision for determining the peak location of the spectral line due to external cavity mode-hops in a single scan. As shown in Fig. 6(b), the fluctuations in the position of the peak are white-noise limited out to ~30s, indicating that the effects of external cavity mode hops on scan precision can be reduced by averaging over multiple scans. Based on these results, it may be advantageous to collect data at a higher acquisition rate than is needed to meet the time response requirements for a particular measurement because averaging will improve the spectral precision. This is helpful for measurements of small molecules, where the spectral features are on the order of the ECQCL scan resolution. For measurements that allow averaging on timescales of seconds or longer the precision approaches 0.001 cm⁻¹, which is much smaller than linewidths of gas-phase molecules at atmospheric pressure that are typically > 0.1 cm⁻¹ even for small molecules.

 

Fig. 6 (a) Transmission spectrum from 1305 – 1310 cm⁻¹ acquired at a spectral acquisition rate of 200 Hz. The H₂O spectral feature used for the scan frequency reproducibility studies is highlighted by a vertical arrow. (b) Allan plots for the stability of the center position of the H₂O spectral feature at 1308.20 cm⁻¹ for acquisition rates of 40, 100, and 200 Hz.

Download Full Size | PDF

The spectral absorbance noise is often used to characterize performance of spectroscopic-based sensors, and affects the precision of spectral fits used to determine species concentrations. We define the absorbance noise as the per spectral point standard deviation in the absorbance: $A_e(\tilde{v},t)=-\ln \left(I(\tilde{v},t)/I_0(\tilde{v})\right)$. $I_0(\tilde{v})$ is calculated using the first 360 seconds in one-hour data sets that were collected in the absence of N₂O and F134A releases. Example absorbance spectra for acquisition rates from 40 to 200 Hz are shown in Fig. 7 . Large spikes in the absorbance spectra in Fig. 7 are associated with spectral features from ambient H₂O and to a lesser extent CH₄. The strong absorption near these spectral lines results in low average transmitted intensity on the detector, leading to high absorbance noise. In addition, changes in H₂O and CH₄ concentrations and small drifts in the wavelength repeatability from scan to scan result in high absorbance noise near these strong absorption lines. Therefore, to marginalize the impact of H₂O and CH₄ absorption features on the estimated baseline absorbance noise, the absorbance noise $\left({\sigma }_{Ae}\right)$is calculated by taking the standard deviation of the absorbance in a 1297.5 to 1301.7 cm⁻¹ spectral window as shown in Fig. 7.

 

Fig. 7 Example absorbance spectra from 1290 to 1310 cm⁻¹ acquired at acquisition rates of 40, 100, and 200 Hz corresponding to plots (a), (b), and (c) respectively. The 1297.5 to 1301.7 cm⁻¹ region for each absorbance spectrum used to calculate the absorbance noise is highlighted using dotted vertical lines.

Download Full Size | PDF

Table 1 summarizes the results of the scan characterization for the different acquisition rates evaluated in this work. The detection bandwidths provided in Table 1 are calculated based on the amplitude modulation frequency of the QCL current. The bandwidth normalized noise equivalent absorbance shows no significant degradation in spectral performance as the spectral acquisition rate is increased. Also shown in Table 1 is the noise equivalent absorption coefficient (NEAC) that is calculated by division of the absorbance noise by the 127-m optical path in the current experiment. The observed NEAC values found in this study are comparable to the value of 4.8x10⁻⁹ cm⁻¹ Hz^-1/2 measured using a similar rapidly swept ECQCL system operated at an acquisition rate of 20 Hz [19].

Table 1. Scanning characteristics for the ECQCL at the spectral acquisition rates explored in this work. The detection bandwidth is set by the current modulation frequency of the laser. The instrumental scan resolution is taken from the FWHM of the Gaussian function convolved with the NWIR library spectra to match the experimentally observed spectral resolution.

View Table | View all tables in this article

3.3 Dependence of detection limits for F134A and N₂O on acquisition rate

The detection limits for F134A and N₂O as a function of the spectral acquisition rate are evaluated via statistical analysis of repeated measurements of spectra with nominally zero analyte. This method of analysis determines variations in the concentrations of the targeted chemical species after all spectral analysis is performed, and accounts for both spectral noise and drifts. For this analysis the same one-hour data sets that are used in the absorbance noise and scan wavelength repeatability analysis in Section 3.2 are evaluated. During the one-hour acquisition time there were no intentional releases of N₂O or F134A, and 360 seconds of spectra are used to calculate the average background spectrum ($I_0\left(\tilde{v}\right)$). Time-dependent species concentrations are extracted by applying the WLS fitting algorithm to the absorbance spectra collected over the remaining 54-minute period.

Allan plots for N₂O and F134A concentrations are provided in Figs. 8(a) and 8(c) respectively. The Allan plots for the 100-Hz data are similar at early times (<10 seconds) to the 40 Hz and 200 Hz data, and are not shown in Fig. 8 for the purpose of clarity. The corresponding time dependent concentration profiles are provided in Figs. 8(b) and 8(d). Based on the Allan plots for data collected at 200 Hz, for the shortest measurement times of 5 ms, the NECs for N₂O and F134A are 70 and 16 ppbv respectively. For longer integration times out to 100 seconds, the 1-σ NEC detection limits for N₂O and F134A are down to the single ppbv level. Only small differences are seen between the 40 Hz and 200 Hz data, which is significant in that it implies that operating at the higher sweep rate does not reduce performance for detection of these chemicals. The minima in the Allan plots occur at ~100 s integration time, demonstrating the high stability of the ECQCL-based sensor over 4 orders of magnitude in time scales. The small drift in the Allan plot for the F134A data collected at 200 Hz may be due to small (~ppb) changes in the F134A concentration in the room air during the 1-hour measurement (although nominally zero, uncontrolled background sources may have been present).

 

Fig. 8 Allan deviation plots for N₂O (panel (a)) and F134A (panel (c)) calculated from time dependent concentration data acquired at 40 and 200 Hz acquisition rates. The time dependent concentrations for the data collected at 200 Hz are provided for N₂O and F134A in panels (b) and (d), respectively.

Download Full Size | PDF

Because the high spectral resolution and large tuning range of the ECQCL-based sensor presented here enables detection of both large and small molecules, it is possible to compare results to both DFB-QCL and ECQCL-based sensors. To facilitate comparison of concentration detection limits with other sensors, it is convenient to work in units of NECD. NECDs are calculated from the NEC values shown in the Allan plots in Fig. 8 through multiplication by the 127-meter optical path used in this work. For data collected at 200 Hz, NECD values of 8.9 and 2.0 $ppmv\times m$are estimated for N₂O and F134A respectively for a 5-ms integration time. At an acquisition rate of 200 Hz and an effective integration time of 1 sec, the NECD values for N₂O and F134A are reduced to 0.64 and 0.26 $ppmv\times m$ respectively.

In Table 2 , the 1-sec NECD value obtained for N₂O in this work is compared to other QCL-based sensors. The 1-sec NECD values are provided for N₂O sensors operated in a closed (less than atmospheric pressure in an enclosed sample cell), open-path configuration (atmospheric pressure multi-pass cell with no enclosed beam path), or standoff (atmospheric pressure with propagation to a retroreflector or hard target). It should be emphasized that the classification of the system type in Table 2 is intended to provide a general comparison with the limited number of QCL-based systems used for trace-gas measurements of N₂O. As such, it does not fully account for differences in the experimental conditions or optical techniques important in determining NECDs.

Table 2. A summary of the noise equivalent column densities (NECD) and sweep rates reported for closed path, open path, and standoff measurements of N₂O. The value provided for the current work is taken from the 1 sec NEC value provided in Fig. 8 at a 200 Hz acquisition rate. The 1 sec NECD values presented in the table for other measurements were calculated after bandwidth normalization and accounting for the effective optical path. The system type “Closed” describes N₂O measurements made in partially evacuated sample cells at pressures less than atmospheric. An “Open” system type describes an open-path measurement made at atmospheric pressure using a multi-pass cell without an enclosed beam path. A “Standoff” measurement denotes measurement over an open path to a retroreflector or hard target.

View Table | View all tables in this article

In Table 2 there are three NECD values reported for DFB QCL-based sensors operated in a closed configuration. Except for the lowest NECD value reported by Nelson et al. made using direct absorption spectroscopy [26], the other two reported NECD values achieved using frequency modulation spectroscopy [27,28 ] are comparable to the open-path NECD obtained from this work. This is in spite of the advantage of operating at reduced pressures (< 100 Torr). In addition to the current work, there is only one entry in Table 2 of measurement of N₂O using an ECQCL. For this open path ECQCL measurement an NECD of 17.8 $ppmv\times m$ was reported using the active coherent laser absorption spectroscopy technique (ACLAS) [29]. When using the ACLAS detection method with a DFB QCL the same group achieved an NECD of 1.9 $ppmv\times m$ [30]. Additional open path measurements of N₂O have been reported using DFB QCLs, yielding 1 sec NECD values of 1.2 $ppmv\times m$ [31] using chirped laser dispersion spectroscopy and 0.016 $ppmv\times m$using frequency-modulation spectroscopy [32]. It should be noted that the two smallest NECD values reported by Nelson et al. [26] and Tao et al. [32] were made targeting the v₃ vibrational band of N₂O that has a peak absorbance 6 times larger than the v₁ vibrational band that was measured in this work. Based on these reported literature values, the performance of the rapidly swept ECQCL is comparable to closed- and open-path measurements made using DFB-QCLs. This performance is achieved while also having a larger sweep rate than all but one of the DFB-QCL based systems.

In Table 3 , the 1-sec NECD value obtained for F134A in this work is compared to other ECQCL-based sensors. In this case, because F134A has broad absorption features there is no comparison to DFB-QCLs and all measurements were performed at atmospheric pressure. All of the F134A measurements reported in Table 3 targeted vibrational bands centered at 1188 cm⁻¹ or 1297 cm⁻¹ which have equivalent peak absorbance. The 1-sec NECD value from the current work is greater than the NECD value of 0.07 $ppmv\times m$reported in our previous work using a similar ECQCL source swept at 20 Hz [19]. An important difference between the previous work and the current study is that a custom lock-in amplifier was used for the demodulation of the modulated intensity signal. The bandwidth normalized noise equivalent absorbance from the prior work is 4.8x10⁻⁷ Hz^-1/2, and this is comparable to the value of 6.7x10⁻⁷ Hz^-1/2 obtained in the current study using a software-based demodulation approach. A likely source of the disparity between the two reported NECDs is connected to drift present in the Allan plot at 1-sec (see Fig. 8(c)) that was not present in the prior work. This drift may be due to an uncontrolled background source of F134A in the NEC studies. In addition to the open path measurement, there are two standoff measurements of F134A reported where 1s NECDs of 2.00 and 10.5 $ppmv\times m$were obtained for sweep rates of 20 and 0.03 Hz, respectively [29,33 ]. Overall, the 1-sec NECD value for F134A from this study demonstrates that the rapidly swept ECQCL can be used for sensitive measurements of species with broad absorption profiles, and this is a capability that DFB-QCL based sensors cannot provide.

Table 3. A summary of the noise equivalent column densities (NECD) and sweep rates reported for open path and standoff measurements of F134A. The value provided for the current work is taken from the 1 sec NEC value provided in Fig. 8 at a 200 Hz acquisition rate. The 1 sec NECD values presented in the table for other measurements were calculated after bandwidth normalization and accounting for the effective optical path.

View Table | View all tables in this article

4. Conclusions

The capability of a rapidly swept ECQCL in an open-path configuration for measurement of chemical species in transient plumes is demonstrated in simulated release studies with N₂O and F134A. The system was used to measure N₂O and F134A concentrations at a 200 Hz acquisition rate. Wavelength sweeps over the 112 cm⁻¹ scan range of the ECQCL were performed in 5 ms, corresponding to a tuning rate of 22,400 cm⁻¹/s. The release studies demonstrate the capability of the current system to measure concentration changes on millisecond timescales. While these demonstrations were limited to a single species release, the broad spectral coverage provided by the ECQCL can be used for rapid simultaneous multi-species measurements in complex gas phase mixtures.

The rapidly swept ECQCL was characterized as a function of the spectral acquisition rate to evaluate the absorbance noise, spectral resolution, wavelength scan repeatability, and noise equivalent concentrations of N₂O and F134A as a function of acquisition rate. For all three properties, the difference in performance from 40- to 200-Hz acquisition rates was marginal. It was found that the behavior in the wavelength scan repeatability is ultimately limited by the EC mode-hops of the ECQCL for a single spectral acquisition, but improved with averaging over multiple scans. The performance of the current ECQCL source was not evaluated beyond a 200 Hz acquisition rate due to an increase in the spectral resolution associated with the 2 MS/s sampling rate of the DAQ board. A narrow spectral resolution could be maintained for acquisition rates > 200 Hz by using a DAQ board with a higher sampling rate, or by reducing the size of the acquired spectral window.

For the shortest acquisition time of 5 ms, NECs of 70 ppbv for N₂O and 16 ppbv for F134A were obtained. Single ppbv detection limits were obtained with averaging to 1 sec for N₂O and F134A for acquisition rates from 40 to 200 Hz. A comparison of the NECD value achieved for N₂O at a 200-Hz acquisition rate to other reported NECD values using QCL-based instruments demonstrates that the current approach is competitive with DFB-QCL-based closed cell and open-path sensors. While being competitive with DFB-QCL based approaches for detection of N₂O, the ECQCL can be used for simultaneous measurements of multiple chemical species given the ~100 cm⁻¹ spectral window [34]. This is a capability that is difficult to achieve with DFB QCLs even for small molecular species, and impossible for larger molecular species with broad absorption spectra (i.e. F134A).

It is also important to note that the low detection limits for N₂O and F134A in this work were obtained even though significant water interference was present over the 1278 – 1390 cm⁻¹ spectral window used in this work. This illustrates that it is possible to perform sensitive open-path detection in the MIR in spectral regions where absorption from atmospheric water is significant. This is a direct result of the combination of broad wavelength coverage with high spectral resolution of the ECQCL, permitting measurement of multiple transmission micro-windows during spectral acquisition. This capability is advantageous in remote sensing where the optical paths can easily exceed 100 meters, or in applications where trace gases must be measured in the presence of other background species as in breath analysis and combustion diagnostics.

Overall, the results from this work demonstrate that the swept ECQCL is capable of rapid, sensitive, multi-species measurements of transient plumes in an open-path configuration. While the current studies were performed indoors, future work will focus on outdoor remote open-path measurements. In this situation, the rapid spectral acquisition capability and broad tuning range of the ECQCL should help mitigate the impact of atmospheric turbulence and water absorption that will be present for long optical paths (> 100 m).