Department of Engineering & Public Policy
Carnegie Melon University
I. BackgroundHigh spectral resolution image analysis offers interpretive advantages unavailable with broadband imaging. Although fixed dispersive elements such as gratings have long been used for spectral analysis, acousto-optic tunable filters (AOTF) offer an alternate mechanism. AOTF devices offer low to moderate spectral resolution, with . Because they are solid state, tunable, and random access, these spectral filters offer some unique benefits over other dispersive elements such as diffraction gratings.
Among the attributes offered by an AOTF instrument, the most important may be spectral agility. An AOTF device can switch from one spectral range to another in the time that it takes an acoustic wave to traverse the crystal, (typically tens of microseconds). In addition, the solid state nature of the devices makes them mechanically robust. While the use of AOTF devices for electronic signal processing applications is well established, they have not yet been extensively used for spectral analysis of images. Katzka  provides a partial review of the historical development of these devices.
II. Principles of OperationAn acousto-optic cell is a transparent birefringent crystal excited by a radio frequency transducer. (Because the crystal is birefringent, the index of refraction of the "ordinary" axis differs from that of the "extraordinary" axis). Propagating acoustic waves inside the crystal create regular spatial variations of the refractive index. Under phase-matching conditions, light of a particular linear polarization and wavelength, incident on the crystal at a very specific angle, is diffracted by the moving grating produced by the acoustic wave.
The conditions favoring diffraction are only satisfied for a particular spectral frequency at a particular incident angle and a particular driving frequency, (see Xu  for a more complete discussion). By controlling the transducer frequency, the spectral frequency diffracted can be selected. Moreover, controlling the transducer power allows control of the amount of light diffracted. Typical transducer power is on the order of 1 W. While higher transducer power increases the amount of diffracted light, it also degrades spectral resolution by increasing sidelobes of the center frequency (this effect can be minimized through apodization of the transducer).
By making the AOTF part of an imaging system, and projecting the diffracted light onto a 2-D array, it is possible to form an image extracted from the particular spectral component of the incident light. For example, CCD arrays have been utilized for this purpose.
Because the crystal is birefringent, the polarization of the light incident on the device affects the angle to which the light is diffracted. Diffraction changes the polarization of the incident light so that some of the initially ordinary polarization emerges from the device with extraordinary polarization, and some of the initially extraordinary polarization emerges from the device with ordinary polarization. The result of this sensitivity to polarization is that an AOTF imager might be used to discern polarization information about a scene in addition to spectral information. Because the two polarizations of incident light are diffracted differently, it is possible to use two CCD arrays in order to capture separately the image generated by each polarization.
Although a number of birefringent materials have been used for AOTF devices, TeO2 is a frequent choice due to its high acousto-optic figure of merit, and good transmission in the visible and infrared, (0.4-5 mm). Other materials include TAS, which is transparent to 11 mm, and quartz which is transparent in the UV. Although the bulk material of an AOTF may be transparent across a broad spectral range, a particular device is generally limited to a spectral range of an octave or less due to the range of the transducer . Moreover, transparency of the bulk material is only one factor in the overall spectral range of the device. Other spectral limitations may come from the detector array, or the system optics.
Instrument design requires tradeoffs of spectral resolution, spectral consistency across the image, diffraction efficiency, and angular field of view. Glenar et al.  elaborate about phase matching as a function of polarization, incidence angle, spectral frequency, and angular aperture. Because the physics of phase matching rely on a specific incidence angle, the devices typically have a narrow field of view, and a small optical aperture. For example, the field of view of one device used in a breadboard demonstration was 6-10 degrees . In an imaging application, the effective field of view of the system is even smaller than that of the device (by a factor dictated by the optics of the system). For example, the effective field of view for the breadboard system described above was 1.4-3.5 degrees. Consequently, AOTF imaging spectrometers are expected to be most useful for examination of distant targets, a condition often encountered in remote sensing applications.
III. Imaging ApplicationsOne application that demonstrated some of the strengths of AOTF devices used a Thallium Arsenic Selenide (Tl3AsSe3 or TAS) crystal to measure exhaust composition in a commercial stack . Although the stack analyzer is not an imaging application, it does demonstrate the flexibility of an AOTF device. In this case, the AOTF acting as a spectral filter also eliminated the need for filter wheels, gas cells, diffraction gratings, and mechanical light choppers.
A number of AOTF imaging applications have been investigated in recent years. Although in most cases, the instruments have not yet been deployed, the examples described below demonstrate the wide variety of possible applications.
One proposed use for an AOTF imaging application is for forest fire detection . While this application would not make use of polarization, it would exploit the ability to rapidly change spectral sensitivity. The airplane based device would scan a scene at 3.4 mm until an anomalous signal was detected. The instrument would then re-examine the questionable locations at one or more additional wavelengths. By relying on a sequence of spectral regions for discrimination, rather than a single spectral signature, the instrument might be able to reliably identify even incipient forest fires with a low probability of false alarms. The estimated probability of detection cited exceeded 95% while maintaining a low false alarm rate.
Another suggested future application for this technology is the Mars Rover Sample Return Mission . In this application, the instrument would use a sequence of images at differing spectral bands to identify and classify rock samples by mineral type, and make decisions with some autonomy. The autonomy would be necessitated by the round trip transmission time between Mars and Earth, while the simplicity and mechanical durability of the AOTF would make this approach well suited for the stresses of launch and interplanetary travel.
Another suggested application for AOTF imaging technology is detection of camouflaged objects. Although camouflage might be very effective in the visible, or at a single spectral frequency, it is likely to be much less effective across a number of spectral regions. Consequently, switching among spectral regions might facilitate reliable identification of camouflaged objects (as was the case for the forest fire detection scheme described above). The polarization sensitivity of an AOTF imaging spectrometer might be particularly useful for man-made objects rather than for natural objects because man-made objects are more likely to have smooth faces, and reflected light is consequently more polarized.
Although there are no current plans to put an AOTF imaging spectrometer on an Earth orbiting space platform, an airborne prototype has been investigated. The system would be based on a NASA-Lewis Lear Jet, and have a spatial resolution of 1-2 m from an altitude of 30,000 ft with a spectral range of 0.5-0.8 mm . In addition to facilitating development of the optical system, the prototype would be a test bed for signal acquisition and tracking electronics.
Glenar et al.  provide the following specifications for two instruments selected to demonstrate the range of AOTF applications. The two instruments are a camera, and a 1 m telescope, (both using a 576 x 384 pixel 23 mm pitch CCD in conjunction with an AOTF). The spectral resolution of the camera was approximately 1 nm over a range of 600-900 nm, while the resolution of the telescope was approximately 5 nm over the range 450-700 nm for the telescope. For both instruments, the polarization rejection ratio exceeded 1000:1. As a demonstration, the telescope was used to verify the change in belt-zone contrasts of the planet Jupiter when observed near the 725 nm band of CH4.
As part of a broader discussion of "typical" AOTF devices, Beattie et al.  describe an AOTF test facility integrating optics and detector electronics for characterizing AOTF devices. The device examined by them had a spectral resolution of approximately 1-2 nm, and a field of view of 4 degrees. This device had an optical aperture of 10 mm x10 mm, and was designed to operate over a range of 0.45-0.8 nm.
IV. SummaryIn conclusion, AOTF imaging instruments have unique attributes that make them well suited to many remote sensing applications. Prototype instruments have been constructed, and development continues. Particularly promising applications include spectral or polarimetric analysis requiring only a narrow field of view.
REFERENCES Robert L. Nelson. Role of a TAS AOTF in a commercial stack analyzer. In Acousto-Optic, Electro-Optic, and Magneto-Optic Devices and Applications, volume 753, pages 103-113. Society of Photo-optical Instrumentation Engineers, 1987.
 P. Katzka. AOTF overview: Past, present, and future. In Acousto-Optic, Electro-Optic, and Magneto-Optic Devices and Applications, volume 753, pages 22-28. Society of Photo-optical Instrumentation Engineers, 1987.
 Jieping Xu and Robert Stroud. Acousto-Optic Devices: Principles, Design, and Applications, chapter 7, pages 413-424. John Wiley and Sons, Inc., 1992.
 Li-Jen Cheng, Tien-Hsin Chao, and Mack Dowdy et al. Multispectral imaging systems using acousto-optic tunable filter. In Infrared and Millimeter-Wave Engineering, volume 1874, pages 224-231. Society of Photo-optical Instrumentation Engineers, 1993.
 Paul J. Thomas, Allan B. Hollinger, and R. H. Wiens. Adaptive infrared forest fire sensor. In Infrared Imaging Systems: Design, Analysis, Modeling, and Testing IV, volume 1969, pages 370-381. Society of Photo-optical Instrumentation Engineers, April 1993.
 Jeffry Yu, Tien Hsin Chao, and Li-Jen Cheng. Acousto-optic tunable filter (AOTF) imaging spectrometer for NASA applications: System issues. In Optical Information Processing Systems and Architectures II, volume 1347, pages 644--654. Society of Photo-optical Instrumentation Engineers, 1990.
 David A. Glenar, John J. Hillman, Babak Saif, and Jay Bergstrahl. Acousto-optic imaging spectropolarimetry for remote sensing. Applied Optics, 33(31):7412--7424, November 1994.
 Mark E. Beattie and David C. Harrison. Imaging spectrometer based on a acousto-optic tunable filter. In Aerial Surveillance Sensing Including Obscured and Underground Object Detection, volume 2217, pages 388--402. Society of Photo-optical Instrumentation Engineers, April 1994.