A versatile mid-infrared hyperspectral imaging system is demonstrated by combining a broadly tunable external cavity quantum cascade laser and a microbolometer focal plane array. The tunable midinfrared laser provided high brightness illumination over a tuning range from 985 cm-1 to 1075 cm-1 (9.30–10.15 µm). Hypercubes containing images at 300 wavelengths separated by 0.3 cm-1 were obtained in 12 s. High spectral resolution chemical imaging of methanol vapor was demonstrated for both static and dynamic systems. The system was also used to image and characterize multiple component liquid and solid samples.
©2008 Optical Society of America
Hyperspectral imaging, combining both spatial and spectral information, is a powerful technique for visualizing and analyzing the spatial distribution of the constituents in complex objects. The information in the mid-infrared (MIR) spectral region from 5–25 µm arising from fundamental molecular vibrational transitions is particularly valuable for chemical identification. In MIR hyperspectral imaging, a high spectral resolution is desirable to improve the sensitivity and selectivity for imaging gas-phase molecules with narrow absorption features. A wide spectral coverage adds capability for investigating multiple molecular species, or for probing broader absorption features. In this paper, we demonstrate using a broadly tunable external cavity quantum cascade laser (ECQCL) to provide active illumination for a MIR hyperspectral imaging system with high spectral resolution, broad tuning range, and fast data acquisition rates.
The quantum cascade (QC) laser  provides an excellent illumination source for active hyperspectral imaging in the MIR spectral region. QC lasers are available at wavelengths spanning the MIR spectral region, are compact and robust devices, and can supply high output powers at room temperature operation [2–8]. By using a QC laser in an external cavity configuration it is possible to obtain continuous wavelength coverage (or quasi-continuous with mode-hops) over a portion of the MIR spectral region limited by the gain bandwidth of the device [9, 10]. Tuning ranges up to 10% of the central wavelength are typical, and even larger tuning ranges are possible with specific device engineering . ECQCL systems have been used successfully for gas-phase laser spectroscopy of molecules with both broad and narrow absorption features [12–15]. The ECQCL system also presents numerous improvements over tunable Pb-salt diode lasers for imaging applications , including higher output power and room-temperature operation. Although QC lasers have been previously used as illumination sources in imaging applications [17–20], their broad tuning capabilities have not been exploited for hyperspectral imaging.
Hyperspectral imaging systems operating in the MIR often detect light originating from a broadband source, either from thermal background radiation or from an instrument-supplied blackbody source. Fourier transform infrared (FTIR) hyperspectral imaging systems are widely used for microscopic chemical imaging [21–23]. Similarly, dispersive instruments have been used for passive remote sensing in the MIR [24, 25]. The broadband illumination provides excellent coverage over the MIR spectral region; however, the performance of these instruments is coupled to the limited spectral radiance of the thermal source. To achieve fast acquisition speeds and good signal to noise ratio (SNR), they often acquire data at a low spectral resolution. Although the spectral resolution is adequate for condensed-phase samples, the narrow absorption features of simple gas-phase molecules are not well-resolved. While synchrotron sources have been used to provide higher spectral brightness for FTIRbased systems [26, 27], they are impractical for many applications.
On the other hand, laser-based illumination provides both higher spectral radiance and better spectral resolution, but at the expense of broad wavelength coverage. The high spectral resolution improves sensitivity and selectivity for gas-phase imaging, but the lack of wavelength tuning limits these systems to investigating pre-selected molecules with absorption features overlapping the chosen laser wavelengths. Most MIR imaging systems with active laser illumination have been limited to single wavelength or multi-spectral operation [16–19, 28], although there have been demonstrations using line-tunable CO2 lasers to provide over 100 wavelengths .
To bridge the gap between the broad spectral coverage of FTIR or dispersive systems, and the high spectral resolution systems using MIR laser illumination, we have developed a versatile system for MIR active hyperspectral imaging using a broadly tunable ECQCL as the illumination source. The ECQCL system used in this demonstration was centered at a wavelength of 9.7 µm, provided a spectral resolution better than 0.2 cm-1, and had a tuning range of 90 cm-1 (850 nm). For the results presented in this paper, hyperspectral images were obtained by scanning the ECQCL from 985–1075 cm-1 in 0.3 cm-1 steps, yielding a hypercube with 300 wavelength values. Images were recorded using an uncooled microbolometer focal plane array (FPA) with 320×240 pixels acquiring data at a frame rate of 25 Hz, allowing each hypercube to be obtained in 12 seconds at a signal-to-noise ratio (SNR) up to 400. Furthermore, the entire system was cryogen-free.
The laser-based illumination enabled high spectral resolution imaging of gas-phase samples, while the broad wavelength coverage permitted hyperspectral imaging of solid and liquid samples as well. The system was used to perform hyperspectral imaging of methanol vapor, which has both broad and narrow absorption features, and to map concentration variations in the image. By fixing the wavelength of the ECQCL, the system was used to image dynamic concentration changes from the evaporation of methanol at a frame rate of 25 Hz. Chemical imaging of multi-component liquid and solid samples was also performed, demonstrating the versatility of the hyperspectral imaging system for measuring chemicals with broad absorption features. The system could reliably distinguish methanol and ethanol liquids in a mixture, and map relative concentrations. For an example of a potential industrial application, the system was used to distinguish four different polymer samples in an image based on differences in their absorption spectra.
2. Experimental setup
The ECQCL system used in the experiments reported here was based on a previous design [12, 30] but with a different QC laser device designed to operate in the 9–10 µm wavelength region. This ECQCL design, in which the laser is pulsed at a high duty cycle and the wavelength is tuned with no control over external cavity mode-hops, was shown to perform well for infrared spectroscopy of gas-phase molecules. Additional wavelength regions can be accessed by changing the QC laser device. For the experiments described here, the ECQCL system was first characterized apart from the imaging system using single element mercury-cadmium-zinc-telluride (MCZT) or thermopile detectors.
The QC laser device, provided by Maxion Technologies with an antireflective-coated front facet, was held at 15°C using a thermoelectric cooler. The laser current was modulated at a frequency of 10 kHz with pulses of 50 µs duration and 1.9A peak current. The laser output wavelength was scanned by rotating a tuning mirror using a stepper motor actuator. The average output power, plotted in Fig. 1(c), ranged between 1–7 mW over this scan range due to the gain profile of the QC device. The scan resolution was verified to be better than 0.2 cm-1 by analyzing the transmission through an air-spaced etalon . Calibration of the stepper motor position to absolute laser wavelength was accomplished using the absorption features of methanol vapor contained in a gas cell. The resulting peak positions were fit using a third-order polynomial to an accuracy of 0.09 cm-1.
The FPA was contained in a commercial thermal imaging camera from FLIR Systems (Thermovision A40). The detector element in this camera was a 320×240 uncooled microbolometer FPA with a pixel pitch of 47 microns, and specified spectral coverage from 7.5–13 µm. Digital 16-bit images captured by the camera were recorded by a personal computer. Hyperspectral images were obtained by scanning the laser wavelength while acquiring images with the FPA. The motion of the stepper motor controlling the ECQCL tuning was synchronized with the FPA acquisition using TTL-level pulses generated by the computer system controlling the acquisition. The scan was completed in 12 seconds, giving a hypercube with 300 images separated by 0.3 cm-1 steps. These scan parameters were chosen to highlight the ability of the system to obtain high-resolution spectral scans in a short time period.
A schematic of the imaging setup is shown in Fig. 1(a). The imaging system was configured in a transmission geometry, in which the laser beam illumination was directed through the sample and onto the FPA. To provide a more uniform illumination of the sample plane and to avoid saturating the infrared camera, the ECQCL output was expanded using a 50 mm focal length lens placed 780 mm in front of the sample plane. This yielded an illumination of approximately 130 mm diameter in the sample plane. The lens system of the thermal imaging camera (50 mm diameter and 1.0 f/#) was designed for large object distances; therefore, the sample was first imaged using a 50 mm diameter and 100 mm focal length ZnSe lens placed 100 mm from the sample. The image spatial calibration was performed by imaging an object of known dimensions. It was determined that each pixel of the FPA corresponded to an area of 125×125 µm in the object plane. The field of view imaged by the system as configured was 40×30 mm.
Absolute absorbance A (base-10) at a given pixel (m,n) and frequency was calculated according to:
where Isample is the recorded signal of the sample under investigation, Ilaser is the recorded signal of the laser illumination without the sample, and Ibackground is the recorded background signal without the laser illumination. Note that the calculation of absorbance requires additional scans to determine the laser illumination and background, adding to the acquisition time; however, these scans need not be repeated for every image. The resulting hypercube data consisted of images of the absorbance at the 300 wavelengths provided by the ECQCL scan. Figure 1(b) shows an example of a single image from a hypercube, showing the absorbance at a single wavelength. The hypercube absorbance data could be processed further using standard chemometric techniques to provide chemical identification and concentration information. Note that requiring only a simple ratio to obtain absorbance instead of a Fourier transform, as in the case of FTIR imaging, reduces the computation time to process the hyperspectral data.
3. Experimental results
The spectral resolution of the system was demonstrated by imaging methanol vapor, which has both broad and narrow absorption features. Figure 2(a) shows a reference absorption spectrum for methanol vapor from the Northwest Infrared (NWIR) spectral library , which provides the absorbance for a 1 ppm concentration and a pathlength of 1 m, at a spectral resolution of 0.1 cm-1. A sample with a spatially varying vapor concentration was prepared by allowing methanol to evaporate from liquid in a small open container. To minimize the effects of air currents and create a quasi-static image, an enclosure was constructed from low-density polyethylene (LDPE) film, which has high transmission and is spectrally flat over the tuning range of the system. Figure 2(b) shows an experimental absorption spectrum recorded at a single pixel of the hyperspectral image, demonstrating excellent agreement with the library spectrum. Only the sharpest methanol absorption peaks are not fully resolved in the experimental spectrum. The offset in the spectrum in Fig. 2(b) is due to absorption by the LDPE film of the enclosure.
The experimental absorption spectrum at each pixel was expressed as a linear combination of the absorption spectra of methanol vapor and the LDPE film (taken to be a constant value over this spectral range). The coefficients were determined by a least-squares fitting routine, using the NWIR library spectrum to obtain a concentration-pathlength calibration. The results of this analysis are displayed in Fig. 2(c), which shows the methanol vapor concentration-pathlength in ppm·m. The concentration gradient above the liquid is clearly visible and demonstrates the dynamic range of the chemical imaging system.
By tuning the ECQCL wavelength to the peak absorption of methanol at 1033.5 cm-1, the system was used to measure a dynamic system at the acquisition rate of 25 Hz. For this example, the LDPE enclosure was removed from around the methanol liquid container, allowing air currents to disperse the methanol vapor. The absorbance was calculated as before, but using only the single point from the absorption spectrum. Figure 3 shows a video playing in real-time showing the relative concentration of methanol vapor as it evaporates from the container.
To demonstrate chemical imaging of a sample with multiple components, we prepared a sample consisting of a droplet of methanol liquid and a droplet of ethanol liquid, flattened between two LPDE sheets. The droplets were positioned so that they touched and mixed with each other at an interface region. Figure 4(a) shows experimental absorption spectra obtained from three different points in the image located in the methanol region, ethanol region, and mixing region. The data was acquired at a resolution of 0.3 cm-1; however, to remove low-amplitude optical fringing superposed on the spectra caused by the varying thickness of the liquid between the LDPE films, the data was smoothed using a running average filter. The resulting resolution of the filtered data was 6 cm-1, which was small enough to avoid broadening of the methanol and ethanol liquid absorption spectra peaks. Figure 4(b) shows reference absorption spectra for methanol and ethanol liquids, obtained separately using the ECQCL system and a point detector. As before, the hyperspectral image was analyzed using a least-squares fitting routine, but this time three component spectra were used: methanol, ethanol, and the LDPE film. Figure 4(c) and 4(d) show the results of this analysis, where the resulting fit coefficients displayed in false color show the relative concentrations of (c) methanol and (d) ethanol. The air bubbles and wrinkles caused by the LDPE film introduce varying concentrations of the liquids which are imaged by the hyperspectral system. The images show the ability of the system to distinguish the two components of the mixture based on differences in their absorption spectra. In addition, the system resolves the mixing of the methanol and ethanol at their boundary, and provides relative concentration information.
Although the ECQCL tuning range covers only a fraction of the MIR spectrum, the system can nevertheless be used to investigate solid-phase samples with broad absorption features. A sample was prepared containing thin strips of four different polymer films: polypropylene, vinyl, polystyrene, and LDPE. The absorption spectra obtained from the hyperspectral ECQCL data at pixels within each polymer region, shown in Fig. 5(a), allow the four polymers to be easily distinguished. As with the liquid sample, the data was acquired at a resolution of 0.3 cm-1 and filtered to yield an effective resolution of 6 cm-1. For additional analysis of the hyperspectral image, a principal component analysis (PCA) was performed on the raw data and used to color-code the image, as shown in Fig. 5(b). The gray-scale background image shows the average absorbance over the wavelength scan range, and the color coding resulting from the PCA is superposed on top of this image. The PCA was used to reduce the dimensionality of the spectrum at every pixel to a single dimension; this data was then clustered in five regions attributable to the four different polymer films and the air background. The clusters were color-coded and superimposed on the background image in Fig. 5 (b), where the different colors correspond to the absorption spectra shown in Fig. 5(a).
The experimental noise for an absorbance spectrum at a given pixel was characterized by imaging the laser illumination for two successive scans and then applying Eq. (1). In the absence of noise, these scans should be identical and yield an absorbance equal to zero. The actual absorbance calculated from this data had an average noise level of 0.0058 absorbance units (a.u.) for pixels in the central region of the image, determining the noise equivalent absorbance (NEA). This value corresponds to the data at the spectral resolution of 0.3 cm-1 with no signal averaging in either the spatial or spectral domain. For broad absorption features, the noise can be reduced by filtering the spectral data to change the effective resolution, as was done in Figs. 4(a) and 5(a). For a running average filter with a window of 6 cm-1, the average noise level was decreased to 0.0013 a.u., while still providing adequate spectral resolution for the liquid and solid samples investigated.
The SNR of the imaging system was further characterized by measuring both the noise and the signal levels recorded by the FPA. The noise was measured by recording a series of 60 images at 25 Hz without laser illumination and calculating the variations in signal per pixel. The RMS noise measured in this manner was 16 ADC counts, which is slightly higher than the specified 80 mK sensitivity for the IR camera, which corresponds to 11 ADC counts. The signal levels recorded with laser illumination, averaged over the image and varying illumination power as the ECQCL was tuned, corresponded to a SNR of 200. For the highest illumination levels in the center of the image, and center of the laser tuning range, the SNR reached values as high as 400. Averaging multiple frames to reduce the spectral resolution increased the peak SNR to over 1000.
Specifications for the microbolometer FPA responsivity and NEP were not available; therefore, we performed an experimental calibration with the FPA in the configuration used in the hyperspectral imaging experiments reported here. The responsivity of the FPA in the system as configured was characterized by imaging a beam from the ECQCL which was resized to fit entirely within the field of view of the system. The change in total ADC counts summed over the entire array was measured versus the total incident optical power measured using a thermopile. A linear fit to this data yielded a differential responsivity of 2.8×1011 ADC counts/W. For the measured noise of 16 ADC counts and using a 25 Hz bandwidth corresponding to the frame rate of the experiments reported here, the calculated NEP was 11 pW/Hz1/2. This value represents a NEP realized for this particular experimental configuration at this frame rate. However, the actual microbolometer bandwidth is more accurately derived from its thermal response time, which for this FPA was specified at 7 ms. This value results in a NEP of 9 pW/Hz1/2, which more closely characterizes the limits of the microbolometer array detector.
The results presented here demonstrate the utility of an active hyperspectral imaging system using a broadly tunable ECQCL as the illumination source. Due to its high spectral resolution, the system was capable of resolving the narrow absorption features from methanol vapor. A high spectral resolution is important for investigating gas-phase samples, both for increased sensitivity and for distinguishing target molecules from interferents. While the resolution of the ECQCL system is currently limited to ~0.2 cm-1, it is adequate for resolving atmospherically broadened lines of most gas-phase molecules.
The hyperspectral imaging system was used as a fixed wavelength illumination system to investigate a dynamic system of methanol evaporation. This capability was simple to implement with the ECQCL illumination source and could be performed at any wavelength within its tuning range. We obtained results at a 25 Hz frame rate, but this could be increased to at least 60 Hz with improvements in data acquisition from the infrared camera used here and to even higher rates by using a faster FPA. A higher frame rate would also allow the sensitivity of these measurements to be increased by implementing a differential absorption scheme, where the laser is alternately tuned on- and off-resonance with the absorption feature in synchronization with the frame rate.
The tuning range of the ECQCL was sufficient to investigate the broad absorption features of liquid and solid samples. Although the ECQCL tuning range covers only a fraction of the MIR spectrum, it is possible to distinguish many molecules based on their spectral signatures within this range. Further improvements in QC laser technology will enable larger tuning ranges in ECQCL systems , extending the versatility for hyperspectral imaging applications. In addition, systems could be implemented using multiple ECQCL systems to target different spectral regions. An ECQCL system provides the flexibility to choose an illumination source with high spectral and radiometric brightness which is well-suited for a particular sample.
The data acquisition parameters used for the experiments here represent a compromise between spectral resolution, acquisition time, and SNR. An advantage of the ECQCL system is that its wavelength scan parameters can be reconfigured to be optimum for a given sample. The ECQCL scan range can be adjusted to investigate particular spectral features within its overall tuning range. The spectral resolution can be adjusted to match the spectral features of interest by varying the scan rate, limited by the maximum speed of the stepper motor actuator controlling the scan (we have performed scans in times as short as 1 s). For example, the broad absorption features of the liquid and solid samples of Figs. 4 and 5 could be sampled at a 3 cm-1 spectral resolution, in which case the scan could be completed in 1.2 seconds at the same SNR. Alternately, as was done for the experiments here, the data can be collected at a high spectral resolution and then averaged over multiple frames to reduce the effective resolution while increasing the SNR.
The FPA used in these experiments was designed for thermal imaging, and was not fully optimized for the hyperspectral imaging application. It would clearly be advantageous to sample at a higher frame rate, within the limits of the microbolometer response time, to reduce the total acquisition time or increase the SNR through averaging. A higher frame rate would also be amenable to sampling schemes where the ECQCL illumination is turned on and off at the frame rate to perform active background subtraction. Other commercially available FPA systems can acquire images at rates higher than 100 Hz, enabling the design of a very fast instrument.
The ECQCL active hyperspectral imaging system suffered from a large amount of thermal background radiation reaching the FPA. The sensitivity and dynamic range of the system could be increased by using spectral filters and cold shielding to reduce the amount of background radiation reaching the FPA. This would allow higher illumination powers to be used without saturating the FPA pixels. The optical power of the ECQCL was sufficient for the experiments presented here, but increased illumination power would allow additional beam shaping for more uniform illumination. Higher illumination power would also be beneficial for microscopy applications, investigating optically thick samples, or performing standoff imaging in a reflection geometry.
An active MIR hyperspectral imaging system consisting of an ECQCL and a microbolometer FPA was demonstrated. The ECQCL provided MIR illumination with high spectral brightness, permitting high spectral resolution chemical imaging of gas phase samples. Hyperspectral images with a spectral resolution of 0.3 cm-1 over a tuning range from 985 cm-1 to 1075 cm-1 were obtained in 12 seconds. The scan time was limited by a 25 Hz frame rate for triggered acquisition of the FPA, and could be reduced further with improved FPA systems. Without averaging or filtering the data, SNRs between 200 and 400 were achieved. For imaging liquid and solid samples, averaging was used to produce absorption spectra with a SNR greater than 1000 at 6 cm-1 spectral resolution.
The high spectral resolution provided by the ECQCL is particularly beneficial for gas-phase hyperspectral imaging. While the experiments presented here were performed in transmission mode, the system could also be configured in a reflective or backscatter mode. Such a system could be useful for imaging of toxic gas leaks from a safe standoff distance, for example. The high brightness provided by laser illumination would also benefit reflection-mode measurements, or allow optically thick samples to be probed in infrared microscopy applications.
The Pacific Northwest National Laboratory is operated for the U.S. Department of Energy (DOE) by the Battelle Memorial Institute under Contract No. DE-AC05-76RL01830. This work was supported by the DOE Office of Nonproliferation Research and Development (NA- 22).
References and links
2. M. Beck, D. Hofstetter, T. Aellen, J. Faist, U. Oesterle, M. Ilegems, E. Gini, and H. Melchior, “Continuous wave operation of a mid-infrared semiconductor laser at room temperature,” Science 295, 301–305 (2002). [CrossRef] [PubMed]
3. A. Evans, J. S. Yu, J. David, L. Doris, K. Mi, S. Slivken, and M. Razeghi, “High-temperature, high-power, continuous-wave operation of buried heterostructure quantum-cascade lasers,” Appl. Phys. Lett. 84, 314–316 (2004). [CrossRef]
4. C. Gmachl, F. Capasso, D. L. Sivco, and A. Y. Cho, “Recent progress in quantum cascade lasers and applications,” Rep. Prog. Phys. 64, 1533–1601 (2001). [CrossRef]
5. X. J. Wang, J. Y. Fan, T. Tanbun-Ek, and F. S. Choa, “Low threshold quantum-cascade lasers of room temperature continuous-wave operation grown by metal-organic chemical-vapor deposition,” Appl. Phys. Lett. 90, 211103 (2007). [CrossRef]
6. A. Evans, S. R. Darvish, S. Slivken, J. Nguyen, Y. Bai, and M. Razeghi, “Buried heterostructure quantum cascade lasers with high continuous-wave wall plug efficiency,” Appl. Phys. Lett. 91, 071101 (2007). [CrossRef]
7. L. Diehl, D. Bour, S. Corzine, J. Zhu, G. Hofler, M. Loncar, M. Troccoli, and F. Capasso, “High-power quantum cascade lasers grown by low-pressure metal organic vapor-phase epitaxy operating in continuous wave above 400 K,” Appl. Phys. Lett. 88, 201115 (2006). [CrossRef]
8. A. Wittmann, M. Giovannini, J. Faist, L. Hvozdara, S. Blaser, D. Hofstetter, and E. Gini, “Room temperature, continuous wave operation of distributed feedback quantum cascade lasers with widely spaced operation frequencies,” Appl. Phys. Lett. 89, 141116 (2006). [CrossRef]
9. G. P. Luo, C. Peng, H. Q. Le, S. S. Pei, W. Y. Hwang, B. Ishaug, J. Um, J. N. Baillargeon, and C. H. Lin, “Grating-tuned external-cavity quantum-cascade semiconductor lasers,” Appl. Phys. Lett. 78, 2834–2836 (2001). [CrossRef]
10. G. Totschnig, F. Winter, V. Pustogov, J. Faist, and A. Muller, “Mid-infrared external-cavity quantum-cascade laser,” Opt. Lett. 27, 1788–1790 (2002). [CrossRef]
11. R. Maulini, A. Mohan, M. Giovannini, J. Faist, and E. Gini, “External cavity quantum-cascade laser tunable from 8.2 to 10.4 mu m using a gain element with a heterogeneous cascade,” Appl. Phys. Lett. 88, 201113 (2006). [CrossRef]
12. M. C. Phillips, T. L. Myers, M. D. Wojcik, and B. D. Cannon, “External cavity quantum cascade laser for quartz tuning fork photoacoustic spectroscopy of broad absorption features,” Opt. Lett. 32, 1177–1179 (2007). [CrossRef] [PubMed]
13. M. Pushkarsky, A. Tsekoun, I. G. Dunayevskiy, R. Go, and C. K. N. Patel, “Sub-parts-per-billion level detection of NO2 using room-temperature quantum cascade lasers,” Proc. Natl. Acad. Sci. USA 103, 10846–10849 (2006). [CrossRef] [PubMed]
14. G. Wysocki, R. F. Curl, F. K. Tittel, R. Maulini, J. M. Bulliard, and J. Faist, “Widely tunable mode-hop free external cavity quantum cascade laser for high resolution spectroscopic applications,” Appl. Phys. B 81, 769–777 (2005). [CrossRef]
15. R. Lewicki, G. Wysocki, A. A. Kosterev, and F. K. Tittel, “QEPAS based detection of broadband absorbing molecules using a widely tunable, cw quantum cascade laser at 8.4 mu m,” Opt. Express 15, 7357–7366 (2007). [CrossRef] [PubMed]
16. J. A. Bailey, R. B. Dyer, D. K. Graff, and J. R. Schoonover, “High Spatial Resolution for IR Imaging Using an IR Diode Laser,” Appl. Spectrosc. 54, 159–163 (2000). [CrossRef]
18. Y. Wang, C. Peng, H. L. Zhang, and H. Q. Le, “Wavelength modulation imaging with tunable mid-infrared semiconductor laser: spectroscopic and geometrical effects,” Opt. Express 12, 5243–5257 (2004). [CrossRef] [PubMed]
20. A. W. M. Lee, B. S. Williams, S. Kumar, Q. Hu, and J. L. Reno, “Real-time imaging using a 4.3-THz quantum cascade laser and a 320×240 microbolometer focal-plane array,” IEEE Photon. Technol. Lett. 18, 1415–1417 (2006). [CrossRef]
21. E. N. Lewis, P. J. Treado, R. C. Reeder, G. M. Story, A. E. Dowrey, C. Marcott, and I. W. Levin, “Fourier-Transform Spectroscopic Imaging Using an Infrared Focal-Plane Array Detector,” Anal. Chem. 67, 3377–3381 (1995). [CrossRef] [PubMed]
22. R. G. Messerschmidt and M. A. Harthcock, eds., Infrared microspectroscopy. Theory and applications (Marcel Dekker, New York, 1988).
23. I. W. Levin and R. Bhargava, “Fourier transform infrared vibrational spectroscopic imaging: Integrating microscopy and molecular recognition,” Annu. Rev. Phys. Chem. 56, 429–474 (2005). [CrossRef] [PubMed]
24. J. A. Hackwell, D. W. Warren, R. P. Bongiovi, S. J. Hansel, T. L. Hayhurst, D. J. Mabry, M. G. Sivjee, and J. W. Skinner, “LWIR/MWIR imaging hyperspectral sensor for airborne and ground-based remote sensing,” in SPIE Imaging Spectrometry II102–107 (1996).
25. P. G. Lucey, T. J. Williams, J. L. Hinrichs, M. E. Winter, D. Steutel, and E. M. Winter, “Three years of operation of AHI: the University of Hawaii’s Airborne Hyperspectral Imager,” in SPIE Infrared Technology and Applications XXVII112–120 (2001).
26. L. M. Miller and P. Dumas, “Chemical imaging of biological tissue with synchrotron infrared light,” Biochem. Biphys. Acta 1758, 846–857 (2006). [CrossRef]
27. J. A. Reffner, P. A. Martoglio, and G. P. Williams, “Fourier-Transform Infrared Microscopic Analysis with Synchrotron-Radiation - the Microscope Optics and System Performance,” Rev. Sci. Instrum. 66, 1298–1302 (1995). [CrossRef]
28. B. J. Guo, Y. Wang, C. Peng, H. L. Zhang, G. P. Luo, H. Q. Le, C. Gmachl, D. L. Sivco, M. L. Peabody, and A. Y. Cho, “Laser-based mid-infrared reflectance imaging of biological tissues,” Opt. Express 12, 208–219 (2004). [CrossRef] [PubMed]
29. T. J. Cudahy, L. B. Whitbourn, P. M. Conner, P. Mason, and R. N. Phillips, “Mapping surface mineralogy and scattering behavior using backscattered reflectance from a hyperspectral midinfrared airborne CO2 laser system (MIRACO(2)LAS),” IEEE Trans. Geosci. Remote Sens. 37, 2019–2034 (1999). [CrossRef]
30. M. C. Phillips, T. L. Myers, M. D. Wojcik, B. D. Cannon, M. S. Taubman, and D. C. Scott, “Measurement of broad absorption features using a tunable external cavity quantum cascade laser,” in SPIE Infrared, Mid-IR, and Terahertz Technologies for Health and the Environment II676003–676011 (2007).
31. S. W. Sharpe, T. J. Johnson, R. L. Sams, P. M. Chu, G. C. Rhoderick, and P. A. Johnson, “Gas-phase databases for quantitative infrared spectroscopy,” Appl. Spectrosc. 58, 1452–1461 (2004). [CrossRef] [PubMed]