Anal. Chem. 1996, 68, 3987-3993
Analytical 4D Infrared Tomography Using an InSb Focal Plane Array Sensor. 1. 3D Infrared Tomography (Single-Wavelength Approach) D. Wienke*,† and K. Cammann‡
SONY Germany GmbH, Stuttgart Technology Center, Environmental Center Europe, Stuttgarter Strasse 106, D-70736 Stuttgart-Fellbach, Germany, and Institute for Analytical Chemistry, Wilhelms-University of Mu¨ nster, Mendel-Strasse 8, 48149 Mu¨ nster, Germany
A combined mid-IR and near-IR camera with focal plane InSb detector array has been applied to remote, noninvasive, infrared tomographic analysis of turbid, threedimensional macroscopic polymer bodies. Tomographic infrared images were recorded under distinct, equidistant angular views of samples mounted at a rotating sample holder. Hidden metallic, nonmetallic, and liquid samples inside the turbid hollow polymer body have been threedimensionally reconstructed by using the tomographic back-projection computer algorithm. Radiation penetrated samples up to 2 × 1.5 mm thickness. Potential chemical-analytical applications of quantitative infrared tomography as a new technology to noninvasive 3D analysis of polymers, glasses, tablets and drugs, and turbid liquids were discussed theoretically. Focal plane diode array detectors (FPAs) with many thousands of single-element, infrared light-sensitive semiconductor elements (InSb, InGaAs, HgCdTe, PtSi) recently found their first applications as advanced analytical laboratory instruments. After longterm use as military infrared imaging devices and in technical thermography, now the analytical chemists have discovered the excellent potentials of FPAs for rapid spectroscopic infrared imaging of sample surfaces. The mentioned detector materials allow optical spectroscopic image analysis from the near-infrared (InGaAs, 800-2200 nm) via the mid-infrared (InSb, PtSi, 10005000 nm) up to the far-infrared optical region (HgCdTe, until 12 000 nm). Lewis and Levin1 and Treado et al.,2 for example, used a FPA in combination with an acousto-optical filter (AOTF) as a rapid spectroscopic camera in an infrared microscope for sample observations at distinct wavelengths. Significant increase of image acquisition rate and the opportunity to analyze samples until the mid-IR region were the most obvious gains compared to classical scanning near-IR videcons such as earlier used by Bertrand,3 McClure,4 and Geladi.5 A similar microscopic application, but with modulation of the excitation source by using an interferometer to scan the wavelength range in place of using †
SONY. Wilhelms-University of Mu ¨ nster. (1) Lewis, E. N.; Levin, I. A. Appl. Spectrosc. 1995, 49 (5), 673-678. (2) Treado, P. J.; Levin, I. W.; Levis, E. N. Appl. Spectrosc. 1994, 48 (5), 607615. (3) Robert, P.; Bertrand, D.; Devaux, M. F. NIR News 1991, 2 (2), 9-10. (4) McClure, W. F. NIR News 1991, 2 (2), 8. (5) Geladi, P.; Grahn, H.; Lindgren, F. In Applied Multivariate Analysis in SAR and Environmental Studies; Devillers, J., Karcher W., Eds.; Kluwer Scientific Publishers: Amsterdam, 1991; pp 447-478. ‡
S0003-2700(96)00120-5 CCC: $12.00
© 1996 American Chemical Society
filters in front of the FPA, was recently published by the same authors.6 We recently published a completely different analytical application of a FPA7-11 for remote infrared sensing over distances up to 2 m. The main hardware problem in remote sensing mode (or macroscopy) was the homogeneous large-scale illumination of the sample by powerful infrared excitation sources. Simultaneously we proposed to consider the 3D stack of recorded FPA data neither as a simple collection of multiwavelength images nor as a conventional collection of multipixel spectra as the former authors1,2,6 in their FPA study did. More-dimensional data analysis methods such as singular value decomposition and multivariate image rank analysis (MIRA) were proposed instead and demonstrated for analysis of three-dimensional spectroscopic image data stacks.9-11 The present work continues to explore the future analytical potential of FPA technology. However, this study moves a step ahead from the present two-dimensional analytical world of “spectroscopic infrared imaging”. It is explored, if an FPA camera is suitable for noninvasive remote “spectroscopic mid-infrared tomography” of three-dimensional visually nontransparent samples. INFRARED SPECTROSCOPIC TOMOGRAPHY VERSUS INFRARED SPECTROSCOPY AND INFRARED SPECTROSCOPIC IMAGING The relation among IR spectroscopy, IR spectroscopic imaging, and the proposed new analytical method of IR spectroscopic tomography shows a clear hierarchy from the point of view of dimensionality of sampling spot and of recorded data block per analyzed sample (Figure 1). “Infrared tomography” samples an object in its three spatial dimensions simultaneously. The extension of the method, called “spectroscopic infrared tomography” (SIT) or “4D tomography” samples an object not only at its three spatial dimensions but additionally and simultaneously at several infrared wavelengths. This high-end approach (6) Lewis, E. N.; Treado, P. J.; Reeder, R. C.; Story, G. M.; Dowrey, A. E.; Marcott, C.; Levin, I. W. Anal. Chem. 1995, 67, 3377-3381. (7) Wienke, D.; v. d. Broek, W.; Melssen, W.; Buydens, L. Proceedings, INCOM’95 (Instrumentalized Analytical Chemistry and Computer Technology Conference); Dusseldorf, Germany, March 1995; p 479. (8) v. d. Broek, W.; Wienke, D.; Melssen, W.; Buydens, L. Proceedings of the NIR 95 conference, The Future Waves; 7th International Conference on NearInfrared Spectroscopy; Montreal, Canada, 6-11 August 1995; p 42. (9) v. d. Broek, W.; Wienke, D.; de Crom, K.; Melssen, W.; Buydens, L. Anal. Chem. 1995, 67, 3753-3759. (10) Wienke, D.; v. d. Broek, W.; Buydens, L. Anal. Chem. 1995, 67, 37603766. (11) Wienke, D.; van den Broek, W.; Huth-Fehre, T.; Kantimm, T.; Feldhoff, R.; Winter, F.; Cammann, K.; Buydens, L. Fresenius J. Anal. Chem. 1996, 354, 823-826.
Analytical Chemistry, Vol. 68, No. 22, November 15, 1996 3987
Figure 1. Evolution from classical IR spectroscopy (top) via spectroscopic IR imaging (center) to spectroscopic IR tomography (bottom). Note the spatial 2D measurement of “spectroscopic” IR imaging a 3D data block per sample provides. Note further, if the spatial 3D measurement of an IR tomograph, realized by a moving FPA (and/or a moving sample), is performed at distinct infrared wavelengths, a 4D data set per sample (“IR spectroscopic” 4D tomography) will be obtained.
thus provides not only three-dimensional but four-dimensional data sets for a single analyzed sample. If moving objects or other timedependent events are observed by SIT, then a fifth dimension (time) adds to the data (“spatiotemporal spectroscopic tomography”) or 5D IR tomography. The new idea of spectroscopic mid-infrared tomography generally classifies into the group of optical tomography methods. In 1928 and 1929 Ewing12 and Cutler13 discovered that common visual light from a powerful lamp is able to penetrate parts of the human body. Hidden objects such as bones or tumors could be visualized as vague blurred shadows at normal photographic film paper. Profio et al.14 reviewed this history of medical tomography in the visible and lower near-IR optical region from the point of view of breast cancer detection. Optical tomography provides, similar to X-ray-tomography, shadow projections of samples hidden in a three-dimensional object onto its outer surface. A relative mutual rotation of sample and/or detector allows the required generation of a series of projections under distinct angles of view. From this sequence of projections, the inner threedimensional sample structure can be calculated by tomographic reconstruction algorithms at powerful computers.15-17,38 (12) Ewing J. Neoplastic Disease; Saunders: Philadelphia, PA, 1928. (13) Cutler M. J. Surg. Gyneacol. Obstetr. 1929, 48, 721-728. (14) Profio, A. E.; Navarro, G. A.; Sartorius, O. W. Med. Phys. 1989, 16, 60-65. (15) Herman, G. T. Image Reconstruction from Projections; Academic Press: New York, 1980. (16) Singer, J. R.; Gru ¨ nbaum, F. A.; Kohn, P.; Zubelli, J. P. Science 1990, 248, 990-993. (17) Verhoeven, D. Appl. Opt. 1993, 32, 3736-3754. (18) Wison, B. C.; Sevick, E. M.; Patterson, M. S.; Chance, B. Proc. IEEE 1992, 80 (6), 918-930. (19) Sevick, E. M.; Lakowicz, J. R.; Scmacinski, H.; Nowaczyk, K.; Johnson, M. L. J. Photochem. Photobiol.; B: Biol. 1992, 16, 169-185. (20) Arridge, S. R.; Schweiger, M.; Hiraoka, M.; Delpy, D. T. Proc. SPIE-Int. Soc. Opt. Eng. 1992, 1888, 360-371. (21) Gratton, E.; Mantulin, W. W.; vande Ven, M.; Fishkin, B. J.; Maris, M. B.; Chance, B. Bioimaging 1993, 1, 40-46. (22) Chance, B.; Kong, K.; Sevick, E. Opt. Photonics 1993, 10, 9-11.
3988
Analytical Chemistry, Vol. 68, No. 22, November 15, 1996
In fact, most of the studies dealing with optical tomography still make use of the medically interesting optical window between 600 (visible light) and 1300 nm (lower near-IR). Radiation of this wavelength range easily passes into the human body over distances of several centimeters up to decimeters providing shadow images for a tomographic reconstruction of hidden body parts. But its lower radiation energy causes optical photons to be more scattered than X-radiation by turbid material structures, giving much more blurred images. To overcome these problems, several schools of optical infrared tomography were established in the scientific world with striking solutions.32 The first school represents the already mentioned classical approach, making use of powerful continuous light sources. The present work will also make use of this well-established approach, but in the mid-infrared optical range. The second school, established in Alfano’s laboratory,24-26,28,30 introduced ultrafast spectroscopy for the detection of very early passing “ballistic” and “snake light” photons that penetrate turbid samples through a straight path. Scattered photons need longer time to pass and can be excluded from detection using ultrafast spectroscopy. Alfano’s laboratory is able to take spectacular looks through turbid media several centimeters thick. Pulsed picosecond near-IR lasers are used as excitation sources. The third school in optical tomography is represented by Benaron’s medical physics laboratory.23,27 They invented the time-of-flight and absorbance analysis (TOFA) of photons to get spectacular transparent images of small animals in the lower optical near-infrared region. The fourth and fifth schools can be found in the laboratories of Sevick-Muraca19,29,31 and Chance.18,21,22 Their pioneering approaches are amplitude and phase modulation of excitation lasers, the heat diffusion theory applied to optical radiation transfer through turbid media, and use of the method of diffuse wave spectrometry. Pulsed as well modulated near-IR lasers are used as excitation sources. A methodology called “near-infrared topography” is used by Maki, Watanabe, and Yamashita from Kozumei’s laboratory.35-37 Two near-IR lasers (787 and 827 nm) in combination with a network of fiber optical detectors were fixed at the head skin to get a low-resolution image of a patient’s brain. The technique is similar to those ones reviewed earlier by Profio et al.14 The laser’s theoretical penetration depth in combination with a future higher spatial resolution and some reconstruction algorithms promise that this “topography” evolves toward real “tomography”. (23) Benaron, D. A.; Stevenson, D. K. Science 1993, 259, 1463-1466. (24) Kalpaxis, L. L.; Wang, L. M.; Galland, P.; Liang, X.; Ho, P. P.; Alfano, R. R. Opt. Lett. 1993, 18, 1691-1693. (25) Das, B. B.; Yoo, K. M.; Alfano, R. R. Opt. Lett. 1993, 18, 1092-1094. (26) Anderson, G. E.; Liu, F.; Alfano, R. R. Opt. Lett. 1994, 19, 981-994. (27) Benaron, D. A. Laser Focus World 1994, 1, 79-87. (28) Dolne, J. J.; Yoo, K. M.; Alfano, R. R. Lasers Life Sci. 1994, 6, 131-141. (29) Sevick-Muraca, M. E.; Burch, C. L. Opt. Lett. 1994, 23, 1928-1930. (30) Alfano, R. R.; Liang, X.; Wang, L.; Ho, P. P. Science 1994, 264, 1913-1914. (31) Sevick, E. M.; Frisoli, J. K.; Burch, C. L.; Lakowicz, J. R. Appl. Opt. 1994, 33, 3562-3570. (32) Leutwyler, K. Sci. Am. 1994, 1, 130-131. (33) Otoki, Y.; Watanabe, M.; Inada, T.; Kuma, S. J. Cryst. Growth 1990, 103, 85-90. (34) Kuma, S.; Otoki, Y. Institute of Physics Conference Series 135; IOP Publishing Ltd.: London, 1994; Chapter 4, pp 117-126. (35) Maki, A.; Yamashita Y.; Ito, Y.; Watanabe, E.; Mayanagi, Y.; Koizumi, H. Med. Phys. 1995, 22 (12), 1997-2005. (36) Watanabe, E.; Yamashita, Y.; Maki, A.; Ito, Y.; Koizumi, H. Neurosci. Lett. 1996, 205, 41-44. (37) Yamashita, Y.; Maki, A.; Koizumi, H. Rev. Sci. Instrum. 1996, 67 (3), 730732. (38) Monnig, C. A.; Marshall, K. A.; Rayson, G. D.; Hieftje, G. M. Spectrochim. Acta 1988, 43B (9-11), 1217-1233.
Outside of medicine and medical physics, only few publications dealt with the idea of optical tomography for chemical applications. Monning et al.38,39 equipped an inductive coupled plasma (ICP) spectrometer with a CCD matrix camera. The authors observed the image of the emission plasma under distinct viewing angles. Through tomographic reconstruction, a better understanding of temperature distribution has been reached. Their study is very readable to each analytical chemist with regard to the tomographic measurement conditions such as required pixel resolution of the matrix camera, number of angles of view around the object, and optimal resolution of the calculated tomographic 3D grid. A total different chemical application has been reported by Kuma and Otoki33,34 for tomographic characterization of GaAs semiconductor materials by thermal emission tomography. Unfortunately, no details about the tomographic reconstruction were reported. Another third chemical application was reported by Todd and Ramachandran.40,41 They analyzed the 3D distribution of indoor air components in a single room equipped with a single FT-IR spectrometer but with many single reflectors mounted on the walls. Moving the spectrometer beam through the room provided many sections through the indoor air under different angles of view. The FT-IR data could then be used to reconstruct the air pollution situation in that room as a computed 3D profile (“infrared tomogram”). The aim of the present work is to move away from the medical-biological visual optical window (600-1300 nm) to a part of the rather chemically and analytically attractive optical nearIR/mid-IR tomography window (1100-4600 nm). The second aim is to demonstrate the usefulness of FPA detectors in place of classical spectrometers or single-wavelength lasers. We want to explore the usefulness of the FPA camera for spectroscopic midIR tomography. Chemists are always interested in having noninvasive methods for the study of three-dimensional samples such as polymers, glasses, drugs, emulsions, powders, liquids, and gases. The noninvasive three-dimensional analysis of turbid or opaque (visually nontransparent) materials continues to be a dream of many analytical chemists. The present work will show the principle opportunity to analyze several millimeter thick, visually nontransparent polymer samples with included hidden objects. To reach this aim, modern infrared FPA technology is combined with enhanced computer power and mathematical algorithms for tomographic 3D image reconstruction. To keep this initial midinfrared tomographic study still readable for the analytical chemist and for the infrared spectroscopist, we first focus on the singlewavelength approach (respective wavelength range). This method we call spectroscopic 3D infrared tomography. The following paper will demonstrate the generalization for simultaneous multiwavelength measurements (as illustrated in Figure 1). This generalized method is called spectroscopic 4D infrared tomography. EXPERIMENTAL SECTION Experimental Setup for Mid-IR Tomography. An InSb focal plane diode array camera IRC-64 (Cincinnati Electronics Inc., (39) Monnig, C. A.; Gebhardt, B. D.; Marshall, K. A.; Hieftje, G. M. Spectrochim. Acta 1990, 45B (3), 261-270. (40) Todd, L.; Ramachandran, G. Am. Ind. Hyg. Assoc. J. 1994, 55 (5), 403417. (41) Todd, L.; Ramachandran, G. Am. Ind. Hyg. Assoc. J. 1994, 55 (12), 11331143.
Figure 2. Design of the used experimental setup for spectroscopic mid-infrared tomography. Distinct angles of view are obtained by a rotable sample holder (FPA camera fixed). Wavelength selection is performed by interference filters in front of the FPA lens.
Cincinnati, OH) was fixed at an optical bench equipped with a triggered filter wheel, rotable sample holder, radiation diffuser, and infrared illumination source (Figure 2). The sample holder can be circularly rotated in 64 equidistant angular positions to get the necessary angular projections of each sample onto the focal plane detector array. The sample is illuminated with an 1800 W optical broad-band infrared source (glow bar) that is controllable via a high-performance dimmer. A thin white paper diffuser helped to generate a homogeneous sample background illumination. The high-speed 16-bit digital interface S16D (EDT Inc.) allows real-time data acquisition of 52 camera images/s by a SUN Spark 10 (Unix workstation). The real-time measurement is done under the KHOROS imaging software package (University of New Mexico, Albuquerque, NM). Communication software between the S16D card and KHOROS and the electronic installation has been realized by Starling Consultancy (Hengelo, The Netherlands). Complete infrared movies can be recorded by the IRC64 and then presented by KHOROS by real-time visualization of the movie out of the workstation’s RAM. IRC-64 images can be saved on disk in several formats by KHOROS. In this way, the images can be read into other imaging software, such as MATLAB, and into self-developed programs under the Unix operating system. The cold shield inside the camera and the thermography lenses were replaced. The new ones allowed the acquisition of IR tomographic data in the optical region from the near-IR range to the mid-IR range between 1100 and 4600 nm. Software and Computations. Recorded infrared images from different angles of view were saved on hard disk as matrices in flat ASCII code and further processed by the back-projection algorithm.15 The back-projection algorithm has been implemented by the author in MATLAB code. For practical understanding and limitations of the different reconstruction methods, it is suggested that the analytical chemist read the clearly written work of Monning et al.38 Materials and Methods. Sample collection covered four hollow plastic containers of turbid polymer material [a white polystyrene beaker (PS), a white polyethene beaker (PE), a dark blue polypropene cap (PP), and a dark-green plastic lighter]. Diameters varied between 1 and 10 cm. Wall thicknesses varied from 0.5 to 2 mm. All four containers were absolutely nontransAnalytical Chemistry, Vol. 68, No. 22, November 15, 1996
3989
Figure 3. Infrared optical transparency of a turbid, white PE beaker (wall thickness 2 × 1.5 mm) after excitation with a mid-infrared source and by observation using an InSb FPA matrix camera. Hidden objects are clearly visible by their shadow projections.
parent to the human eye in the visual optical region even using a powerful visual light source. The turbid PS, PE, and PP beakers were filled with air. The lighter partially contained a liquid and a gaseous phase of a propane-butene mixture. Several other materials such as a metal screw, a plastic screw, a metallic spoon, and a piece of paper were hidden fixed inside the turbid plastic containers. Initial experiments served to demonstrate the possibility of making visually turbid media optically transparent by mid-infrared radiation. Significant high absorption coefficients were expected to be the most critical problem for penetration with mid-IR radiation, compared, for example, with visible or near-IR radiation. The second basic experiment of this paper will demonstrate the possibility of three-dimensional mid-infrared tomographic reconstruction of hidden objects inside a turbid polymer container. For this experiment, 16 angles of view were chosen to get a satisfying computer-assisted 3D reconstruction. The reconstructed 3D image consisted of a vertical stack of 64 calculated, horizontal tomographic slices of individual size of 91 × 91 pixels. The height of 64 slices for this calculated vertical 3D stack was directly determined by the vertical observation height of 64 pixels of the FPA InSb matrix detector. The squared dimension of 91 × 91 pixels per individual calculated tomographic slice is related to the circular superposition of the 16 original images with a width of 64 pixels. This width is directly related to the horizontal observation width of the FPA of 64 pixels. The maximum geometric extension of 91 × 91 calculated pixels originates from the domination of the image diagonal of length d ) (642 + 642)1/2 ) 90.51 during the back-projection procedure. It has been rounded to a 91 full pixels. Back-projection of the 16 infrared images (each of resolution 64 × 64 pixels) onto all 91 × 91 × 64 ) 529984 voxels (“volume pixels”) took a significant amount of computation time on a SUN Spark 10 processor workstation (stand alone) with 32 Mbyte of RAM memory. Each one of the 64 vertical slices required 128 s of computation time. To get the fully reconstructed 3D infrared tomogram took in total 64 × 128 s ) 8192 s (more than 2 h of CPU time). RESULTS Figures 3-5 demonstrate how infrared radiation is able to penetrate turbid polymer samples of significant wall thickness of several millimeters. Hidden objects such as pieces of metal and 3990 Analytical Chemistry, Vol. 68, No. 22, November 15, 1996
Figure 4. Infrared optically transparent turbid, dark-blue PP cap (wall thickness 2 × 1 mm). Hidden objects are clearly visible by their shadow projections.
Figure 5. View through a turbid, dark-green plastic lighter using mid-IR radiation. A FPA camera allows the actual quantitative determination of the remaining volume content of liquid gas filling.
plastic became visible as shadow projections of distinct gray value levels (Figure 3). Even a look “through” a dark blue polymer cap (from a shampoo bottle) of 2 × 1.0 mm wall thickness is no problem (Figure 4). The positions of a plastic and a metallic screw are clearly visible. More exciting is the visualization of the liquid gas phase inside a visually nontransparent dark-green plastic lighter (Figure 5) of significant wall thickness. These four initial imaging experiments (Figures 3-5) formed the basis for the following tomographic study in the mid-IR wavelength region. A piece of paper and a metallic screw (Al) were fixed inside a visually nontransparent PS beaker. Four
equals an equidistant step width of tomographic sample rotation of 360/16 ) 22.5°. After reconstruction by back-projection, 64 horizontal slices of dimension 91 × 91 pixels were obtained. These horizontal slices of distinct gray values form artificial, calculated horizontal sections through the PS blister at distinct heights (Figure 7). Within such slices, the location of a hidden object at this height becomes visible as a bride spot. Combination of all 64 horizontal slices together into a vertical stack of 91 × 91 × 64 voxels yields the fully reconstructed three-dimensional infrared tomographic image of the PS beaker (respectively of its hidden content) [Figure 8 for one particular chosen mid-IR wavelength (respectively one-wavelength range)].
Figure 6. A turbid, visually nontransparent PS beaker (wall thickness 2 × 0.2 mm) becoming completely transparent under mid-IR radiation. The FPA camera allows the visualization of the shadow projection of a hidden Al screw (above) and a piece of paper (below) fixed inside the beaker (4 distinct tomographic angular views).
tomographic example images under distinct angles of view (Figure 6) show that the beaker becomes optically completely transparent again if mid-IR radiation is used for illumination in combination with an InSb FPA camera as infrared tomographic image detector. The PS beaker has been observed under 16 viewing angles, which
DISCUSSION The developed experimental setup allowed us to perform a first proof of principle for infrared-spectroscopic 3D tomography using an InSb infrared FPA camera. The enormous speed of the FPA with 52 images/s enables data acquisition from fast rotating samples in short measuring times. A semiquantitative determination of hidden object volumes seems practically possible unless the images are still blurred. Examples for volume determinations were the visualization of the remaining liquid gas volume (Figure 5) in the lighter and the tomographically reconstructed shapes and volumes of the hidden screw and paper inside the turbid PS beaker (Figure 8). In this sense, a potential application of 3D infrared tomography in industry, for example, could be the quantitative volume control of the content of plastic pipes. However, an improvement of the proposed optical setup for infrared tomography is required in future to overcome, for
Figure 7. Eight of the 64 tomographic horizontal slices through the PS beaker (from Figure 6), predicted by the tomographic back-projection algorithm. (Note: height ) 1 corresponds to the top of the beaker, height ) 64 to its foot.) White spots show the predicted objects (screw, paper) inside the turbid beaker as partial sections. Calculation artifacts due to the limited pixel resolution of the FPA camera and the angleof-view stepwidth as well due to the algorithmic limits of the used back-projection algorithm are also visible (star pattern).
Analytical Chemistry, Vol. 68, No. 22, November 15, 1996
3991
Figure 8. Objects found inside the turbid PS beaker by putting the calculated 64 tomographic horizontal slices from Figure 6 together into one vertical stack. The figure thus presents the three-dimensional tomographic computer reconstruction of the PS beaker in the mid-IR wavelength region out of 16 two-dimensional FPA images (from Figure 6). For visual simplicity, only the beaker’s content (screw, piece of paper) is graphically presented.
example, the blurred images. Such possible improvements are the use of a higher pixel resolution FPA (at the moment, only 4096 single-detector elements), the development of more advanced tomographic reconstruction algorithms,15-17,20,38,40 a choice of optimal mid-IR wavelengths with respect to chemical interesting absorbtion bands, and a move toward alternative excitation sources. Another necessary study in future is the maximum reachable penetration depth of radiation for distinct technical and analytically and chemically attractive samples. The use of ideas from medical-optical visible/near-IR tomography research32 such as ultrafast photon detection,24-26,28,30 TOFA,23,27 and light modulation18,19,21,22 as they were designed by the other scientific schools offers an additional potential for further improvement of chemicalanalytical mid-IR tomography and will be considered in future. Possible analytical applications are the tomographic analysis of drugs and tablets, drops of turbid liquids, emulsions, pieces of polymers and medical-biological materials according to chemical and physical homogeneity, for heat and material flow, for volume texture analysis, and for inner volume stress.
QUANTITATIVE ANALYTICAL INFRARED TOMOGRAPHY The chosen experimental setup (Figure 2) works in the absorption mode via the attenuation of the penetrating radiation emitted by the infrared source. Hidden samples are absorbing according their concentration, thickness (for a considered fixed angle of tomographic view), and absorption coefficient. This corresponds to a Beer’s law approach. The lighter example (Figure 5) illustrates the future quantitative possibilites of infrared tomography. Not only the sample volume of the hidden liquid gas can be reconstructed according to the strategy demonstrated in Figures 7 and 8. Additionally, via the known absorption coefficients, given by the FPA’s pixel gray values, or via calibration, the concentration ratio of the liquid gas mixture could be estimated. Note, that this probably requires new chemometrical approaches in multivariate calibration because the sample thickness (and in this way the gray value and finally the absorption) became now a function of locally and spatially resolved tomo3992
Analytical Chemistry, Vol. 68, No. 22, November 15, 1996
graphic measurements. This is quite different from quantitative optical spectroscopy, which usually integrates and averages over a fixed sample volume. The inhomogeneity in gray value level of the liquid gas shadow (Figure 5) illustrates this new question of spatially dependent absorption. The different absorption coefficients and material thicknesses can be presented in the reconstructed infrared tomogram by distinct colors or gray value intensities (Figure 8). In the present study, the distinct absorption coefficients and sample thicknesses of a nontransperent strong optical absorber (Al screw, Figure 8) is symbolized by a darker gray value level compared to a more transparent material (paper, Figure 8), which is plotted by a brighter gray value level. This illustrates, in principle, the future possible way of quantitative analytical infrared tomography. The following paper will further discuss this quantitative analytical aspect with another example and with respect to the spectral absorption coefficients of distinct materials at distinct infrared tomographic wavelengths.
CONCLUSIONS An analytical methodology called mid-infrared spectroscopic 3D tomography has been introduced on the basis of a rapid-scan focal plane diode array matrix camera with InSb detector. Experimentally it has been demonstrated that turbid materials such as opaque or dark-colored, visually nontransparent polymer samples of several millimeters thickness can be analyzed by 3D mid-IR tomography in a noninvasive remote way. A focal plane array detector, known from military infrared imaging and from technical thermography applications, offers the possibility of visualizing hidden objects inside turbid polymer samples in the wavelength region between 2200 and 4600 nm by calculated tomographic reconstruction. After many decades of infrared spectroscopy and after only a few recent years of infrared imaging spectroscopy, a step toward the third spatial dimension is proposed now by a method called a method called spectroscopical midinfrared tomography. Required optical and computational improvements of the setup form an exciting research field for IR spectroscopists and analytical chemists. Many visually turbid three-dimensional sample types such as polymers, special glasses, tablets, turbid liquids, and emulsions are of potential technical and chemical analytical interest to be studied at a selected wavelength by the proposed spectroscopic 3D infrared tomography. In contrast to a classical IR spectrometer, which integrates over the entire sample volume, the proposed tomographic approach provides the volume pixel resolution of the sample in its three dimensions. This is also an advantage of tomography over IR imaging spectroscopy, which usually is only able to provide two-dimensional information about sample surface instead over the full sample volume. Specific properties of 3D IR tomography compared to alternative 3D analytical methods are its potential applicability to microscopic as well as macroscopic samples, the opportunity for noninvasive and remote analysis, and the enormous measuring speed. Further advantageous properties are the relative simplicity and robustness of the experimental setup if compared, for example, with an NMR tomograph and the absolute safety of infrared radiation if compared, for example, with an X-ray tomograph. However, additional critical experimental evaluation and further studies are required to find the optimal chemicalanalytical field of application of mid-IR spectroscopic tomography
and to determine the limitations for this new proposed analytical method. ACKNOWLEDGMENT Authors thank both Prof. Heinz Siesler and Prof. Bernhard Schrader (both at the University of Essen, Essen, Germany) for their valuable comments before manuscript submission. Thanks to the Laboratory of Analytical Chemistry (University of Nijmegen, The Netherlands) for technical assistance and to the European
Commission for financial support. Dedicated to the 100th anniversary of Conrad Ro¨ntgen’s discovery of X-rayss1895-1995.
Received for review February 6, 1996. Accepted July 29, 1996.X AC960120I X
Abstract published in Advance ACS Abstracts, September 15, 1996.
Analytical Chemistry, Vol. 68, No. 22, November 15, 1996
3993