INFRARED RADIATION IMAGER
Technical Field
The invention is concerned with infrared detection and imaging and, more particularly, with generating an optical output signal in response to infrared input.
Background of the Invention
Infrared (IR) detector devices have been developed using room- temperature arrays of resistive bolometers. Into a focal plane of an IR optical system that projects an IR image, a detector surface is placed which is divided into small, e.g. 20 μm by 20 μm pixels, with each pixel having a pair of electrically conducting leads for sampling its electrical resistance. Each pixel is suspended on a membrane over a vacuum gap separating it from a bulk substrate, with the leads and the membrane serving as a mechanical support for the pixel. When a pixel is heated by the projected IR radiation, its resistance changes. In an imaging camera the number of pixels may reach 105 and more. Resistance changes in individual pixels are sequentially read out via a special electronic circuit called Read-Out Integrated Circuit (ROIC) which is integrated with a device silicon substrate, with some circuit elements placed beneath the suspended sensor membranes. After irradiation, increased pixel temperature and associated resistance change decay and disappear upon heat dissipation by conduction through the support and electrical lead structure and radiation to the surroundings. The longest time constant achievable by minimizing heat loss in known technology is about 30 ms, using vacuum micro-electro-mechanical systems (MEMS) fabrication.
Sequential readout places stringent requirements on the pixel scanning rate and thus on ROIC, for the complete detector readout to be completed within a time frame of 10 ms to 30 ms. The ROIC and electrical connections to each pixel make for a reduced fill factor, in that the fraction of total sensor surface area which has sensing functionality is about 60% to 80%.
One of the materials of choice for resistance-change sequential readout technology is vanadium dioxide, VO2, or, more precisely, a mixture of vanadium oxides containing mainly VO2; it is often designated as VOx. However, after a decade of development effort, serious difficulties are reported as to uniformity, reproducibility and
yield, in part because the VO2 sensor films are deposited on membranes rather then a bulk substrate, and also because of intricate MEMS technology used.
VO2 has been utilized near room temperature, where it is a narrow-gap semiconductor with a temperature coefficient of resistance α = 1/R (dR/dT) in the range of 2% to 4%. In the semiconductor phase, VO2 is essentially transparent to IR radiation with wavelength between 8 mm and 14 mm, i.e. in the window of IR transparency of the atmosphere and the wavelength range of main interest in IR imaging. Thus, such detectors have to rely on auxiliary layers for the absorption of radiation of interest.
Summary of the Invention
Obviating the need for electrical connection to pixel elements, a thermal pattern on a sensor element can be read out optically. A corresponding infrared sensor device has one or several thermally insulated sensor elements such as pixels for exposure to thermal radiation and exposure also to readout illumination. The sensor elements include a material whose refractive index is temperature dependent.
Brief Description of the Drawing
Fig. 1 is a schematic of an imager, image sensor or detector-converter in accordance with a preferred first embodiment of the invention. Fig. 2a is a schematic and Fig. 2b a graph for illustrating applicable optical resonator principles.
Fig. 3 a, 3 b and 3 c are graphs for illustrating applicable hysteretic behavior of refractive index.
Fig. 4a and 4b are graphs for illustrating optical memory based on hysteretic behavior.
Fig. 5 is a graph of irradiative heating power and attendant temperature in a VO2 layer as a function of time.
Fig. 6 is a schematic of an imager in accordance with a preferred second embodiment of the invention. Fig. 7 is a schematic for illustrating applicable spatial Fourier analysis of
an image.
Detailed Description
Optical Conversion and Readout Fig. 1 in its lower part shows IR-sensor layers on a substrate 11 of silicon, for example. The sensor includes a layer 14 of a material such as VO2 typically having a thickness in the range from 50 nm to 100 nm, disposed on an optical mirror layer 13, e.g. a film of aluminum about 50 nm thick. VO2 is suitable as a sensor material on account of a strongly temperature-dependent refractive index, between a metallic phase and a semiconductor/insulator phase. There are many other materials undergoing phase transition with desired change of optical properties, including V2O3, titanium oxides, and magnetite (Fe3O4). VO2 is exemplary here for having a particularly suitable phase transition temperature, near 65 C, and temperature-dependent refractive index in the near- infrared and visible spectral range. The layer 13 is separated from the substrate 11 by a layer 12 of thermal insulation, e.g. of SiO2 aerogel or V2O5 aerogel a few microns thick. The layers 12, 13 and 14 are laterally divided into square "passive pixels" without requiring electrical connectivity, with minimal lithography-limited gaps between them. Alternatively, the layer 12 can be left continuous, with only the layers 12 and 13 being pixellated. Pixel size can be 20 μm by 20 μm, for example, and the gap size 1 μm to 2 μm, yielding a fill factor of approximately 91% to 95%. Such subdividing avoids rapid lateral spreading or "blooming" of temperature variations ΔT to be expected along continuous VO2 and aluminum or similar uninterrupted films.
Fig. 1 in its upper part shows additional elements such as IR optics or lens 15 for focusing an IR signal onto the detector surface, near-IR or visible illumination 16 which is uniform, coherent and polarized, e.g. as from a laser with beam expander, analyzer and Fourier optics elements 17 for optical processing, and a charge-coupled device (CCD) 18 or similar matrix for detecting processed light and converting it into an electrical signal. In device use an IR image is projected onto the VO2 surface 14 which simultaneously is illuminated with coherent light 16 of a shorter wavelength. The
coherent light is subject to interference due to reflection from the two VO2 film surfaces, and the reflected light is optically processed further by analyzer and Fourier components 17. The processed light is read out by the CCD or similar matrix 18.
The sensor layer 14 is placed at the focal plane of the detector. As in resistive bolometers, after incident IR radiation has been given sufficient time, e.g. 1 ms, a spatially dependent incident intensity or irradiance induces corresponding spatial temperature variations ΔT(x,y) over the plane of the sensor film 14. The ΔT(x,y) actually comprises a discreet set of different pixel ΔT's, with each pixel having an essentially uniform temperature. Such discreteness can be indicated by the notation ΔT(i, j), where i, j are integer row and column pixel indices.
Device functioning may be attributed to first-order structural and metal- insulator phase transition in VO2. It is known in the art that single-crystal VO2 undergoes a sharp hysteretic metal -insulator transition as a function of temperature at about 65 C, and that in films the transition is broadened. Transition temperature is sensitive further to the presence of impurities, and doping of vanadium oxide especially with tungsten or titanium can be used for adjustment in an approximate range from 30 C to 90 C. Typically, doping with tungsten lowers the transition temperature, and conversely for doping with titanium.
In the transition region, VO2 exhibits variation of optical properties, specifically of the refractive index and absorption coefficient. A detector-converter of the invention can be viewed as including an optical resonator comprising a VO2 film over a mirror, offering a capability to visualize minute variations of optical coefficients. For at least part of the frame or exposure time, the resonator is at a constant or bias temperature inside the hysteresis loop of VO2. By virtue of the phase transition, the temperature variations ΔT(i, j) produce changes in the VO2 optical refractive index, Δn = n(T0 + ΔT) - n(T0), thus encoding an IR image via Δn(i, j) changes. Simultaneously, the sensor surface is uniformly illuminated by polarized laser light either in the visible, e.g. with wavelength of 0.4 μm to 0.7 μm, or in the near IR, e.g. with wavelength of about 1 μm to 3 μm. Depending on its intensity, such illumination may or may not appreciably contribute to the thermal balance of the sensor surface, so that the
PAGE 5 NOT RECEIVED UPON FILING
further as described below. If both "top" and "bottom" mirror layers on layer 14 are made to be semi-transparent, and the substrate 11 is essentially transparent to illuminating radiation, the structure can operate in a transmission mode, with the analyzer and Fourier components 17 and CCD or similar matrix 18 for receiving transmitted illumination on the opposite side of layer 14.
The spatial resolution of the image is determined by the wavelength of IR radiation or by the pixel size, whichever is smaller. On account of their passive nature, the pixels can be made quite small, e.g. 10 μm by 10 μm, making them comparable to the wavelength of IR radiation of interest. In consequence, the use of passive pixels need not reduce the resolution of the imager.
Fig. 3 a is the same as Fig. 2b, repeated here for convenience of following the temperature-hysteretic behavior of the reflectance R(T) for the λ1-minimum of the semiconducting phase per Fig. 3 b, and for the λ2-minimum of the metallic phase per Fig. 3c. Intensity variations induced by the Δn(i, j) and ΔR(i, j) can be expected as rather small in comparison with the constant intensity of the incident illumination light, so that it is beneficial to process the reflected light to substantially subtract the constant background. For this purpose, for polarized monochromatic light a number of well- developed powerful optical techniques are available. The reflected, processed short- wavelength light is collected in a CCD or similar recording matrix, generating an electrical signal corresponding to the optical signal.
As described, device functionality includes two types of conversion, namely (i) from long- wavelength, 8-14 μm IR light to short- wavelength near-IR or visible light, and (ii) from incoherent, mixed-wavelength light to coherent, monochromatic light. The first type of conversion permits the use of existing, well- developed CCD or similar matrix detector technology with contactless, parallel readout, and the second allows the use of interference and application of powerful techniques of optical signal processing.
Once an image is written, optically recorded and read out as described above, the sensor needs to be refreshed/erased in order to prepare for the next image
frame. Indeed, after a few milliseconds time following the image writing stage, the temperature differences ΔT will have vanished, but in each pixel the optical memory effect preserves the changed refractive index and reflectivity for as long as the sensor temperature remains within the hysteresis loop. Figs. 4a and 4b illustrate an effect of optical memory for the case of a small, short temperature pulse ΔT inside the hysteresis loop R(T) for λ,. Fig. 4a shows two closely positioned curves corresponding to R(λ) at T = T0 and T = T0 + ΔT. At λl5 reflectivity changes by ΔR. In Fig. 4b this changed value of reflectivity R + ΔR is "memorized" or recorded at a fixed bias temperature T0. A written image can be erased by disrupting the energy input that caused heating of the VO2 layer to the bias temperature T0. The thermal decay or relaxation time of the VO2 layer can be chosen sufficiently short for the heat to dissipate into the substrate in a few milliseconds. Indeed, under circumstances of minimal heat conduction, decay times are of the order of 30 ms. With an aerogel layer a few microns thick for insulation, estimated characteristic decay times drop to a few milliseconds. Thus, the temperature will quickly decrease to a value below the hysteresis. At the start of a next cycle, the source of radiation for heating of the VO2 layer is turned on again, and the process repeats.
Fig. 5 shows a solid line of temperature of the VO2 layer as a function of time when exposed to heating radiation having power as shown with a broken line. Peak power in the heating input corresponds to raising the VO2 temperature from about 20 C to about 70 C. Then, the heating power is lowered to obtain a flat portion of T(t), and finally switched off to cool back to a temperature below the phase transition, e.g. 20 C. Numerical values in Fig. 5 are intended for guidance only, as it may be possible to erase a picture at temperatures above 20 C, the time domain numbers may differ from 10 ms, and the indicated power of 5000 W/m2 merely represents an estimate for a typical geometry.
One function of illumination is to convert an IR signal into a shorter- wavelength, matrix-readable signal. Efficacy of such conversion depends on the responsiveness of the index of refraction of the VO2 layer to temperature. Accordingly, the wavelength of illuminating light is chosen to provide for a significant temperature
dependence of the refractive index. Measurements performed on VO2 films have shown that Δn increases steadily from about 1 μm to at least 12 μm. If only for this reason it would be beneficial to use illuminating light of longer wavelength, e.g. at 12 μm. However, readily available detector matrices are sufficiently sensitive at shorter wavelengths only, near 2 μm, and such shorter illumination wavelengths also allow the use of thinner VO2 films which are more responsive to IR radiation. Radiative Heating to the Bias Temperature
For combining functions of illumination to provide coherent light for optical readout and to heat the sensor material to a suitable bias temperature, a fairly powerful laser at a wavelength of about 1-2 μm is called for. For a 1-cm2 sensor the 5000 W/m2 for heating the resonator structure by 50 C in 10 ms amounts to 0.5 W. Required power is reduced further if the picture can be erased by lowering the temperature to just 40 C or possibly 50 C which may be achieved with narrow hysteresis loops. Such lasers can be controlled to produce suitably shaped power pulses, e.g. as shown in Fig. 5 or more intricate still. The laser beam can be spread to the desired area, e.g. 1 cm2 by a cylindrical lens, for heating to result in essentially uniform temperature over that area.
For radiative heating separate from illumination, radiation can be absorbed from a broad-band heating lamp, for example. On a 10 ms time scale, corresponding to a 100 Hz frequency scale, heating power can be controlled either by controlling the actual power delivered by the lamp, or by including means for interrupting power, e.g. a chopper or variable-transparency filter. With such separate heating, laser illumination for pattern readout is reduced substantially.
Fig. 6 illustrates radiative heating by a lamp 61 from the back side of the resonator structure 13, 14, through a suitably transparent substrate 11 and aerogel 12. A layer 62 of "black" below the mirror 13 can be included for enhanced absorption of heating radiation.
Optical Data Processing
Principal goals include (i) reduction/elimination of background light from an illuminating laser while preserving and reading out small IR signal-induced variations, and (ii) elimination of an image of regular square pixels. The first goal can be met by
using polarization filtering, and the second by using spatial Fourier transform techniques.
Polarization filtering depends on the fact that incident illumination is linearly polarized in a certain direction. Upon reflection from the VO2/mirror resonator it will be elliptically polarized, with the major elliptical axis turned at an angle to the incident light. By placing an analyzer so that at the bias temperature T0 the reflected light is largely absorbed, the background can be essentially eliminated. The light reflected at T0 + ΔT, with corresponding different optical constants will predominantly pass. Fig. 7 illustrates spatial Fourier analysis of an image 71, with the diaphragm 73 in the focal or "frequency" plane of a lens 72 blocking out spots along the x- and y-axes corresponding to a square grid of pixels. The useful part of the image 71, represented by the two black rectangles, is reproduced on a screen 74, while the grid has been eliminated from the representation. Thus, the grid can be eliminated by forming a spatial Fourier spectrum of the reflected light, in the focal plane of a lens, and by blocking of the Fourier spectrum side bands or spots resulting from the rectangular grid. Additional optical processing can include spatial differentiation of the image for emphasizing image outlines, as well as other applicable coherent-light techniques.
Optical Responsiveness and Noise
Optical imager responsiveness depends in part on the shape of the reflectivity- versus-temperature curve and on the choice of the working point on that curve as resulting from the bias temperature T0. In good-quality films, reflectivity in an optical resonator as described above can change by about 10% per degree C. Thus, for a 10-mK change in sensor temperature, reflectivity changes by about 0.1%. If the constant background can be subtracted, such a change in reflectivity may correspond to a very large number of photons, with actual numbers depending on the intensity of illuminating light. A sufficient number of photons per 10 mK or other desired temperature variation can be obtained readily for confident detection in the CCD or similar matrix.
For bolometric detectors it is known that, for a small pixel at 300 K, the fundamental thermal noise limit due to exchange of thermal energy with the environment is relatively small compared to other noise sources. By integrating over all frequencies
we have found an upper limit of 0.1 mK for temperature fluctuations for our exemplary pixels. The major sources of noise in resistive bolometers have been identified as Johnson noise and random 1/f noise. Both have their origin in electrical resistance, or in the way electrical current flows through the system. Especially large noise is expected and observed in the resistance near percolation limit, i.e. in the resistive phase transition of VO2. The relevant scale on which local irregularities are averaged in that case is the mean free path, which for VO2 is in the nanometer range.
Among advantages of imagers of the invention is freedom from resistive noise and from concerns relating to percolation of an electrical current between micro- regions of the conducting phase, histead, there is optical reflectance on a typical scale of about 1 -micrometer illumination wavelength. On such a scale the material can be expected to contain a large number of metal and semiconductor micro-domains that switch randomly. Such random switching is due to a small expected energy difference between the metal and semiconductor phase in the phase transition region, resulting in local fluctuations of the optical constants and attendant random telegraph noise on small length scales. On longer length scale of the illuminating wavelength much of this noise should average out.
Alternatives and Variations of Detector Design
Without requiring a separate aerogel layer, thermal isolation can be achieved when the VO2 detector film material is included in aerogel form. Such a detector film may be formed by depositing an aerogel layer of V2O5 as described by J. Livage, "Vanadium Pentoxide Gels", Chem. Mater., Vol. 3, pp. 578-593 (1991), followed by annealing in a reducing atmosphere such as hydrogen for reduction to VO2 in aerogel form. The resulting film has low thermal conductivity, with "blooming" of little or no concern even without pixellation. Thus, limited by diffraction only, the highest possible spatial resolution can be achieved with fill factor of 100%. Such a detector can be fabricated without any lithographic or microfabrication processes, and custom detector sizes up to full wafer size can be produced simply through selective dicing.
For a VO2 detector layer in aerogel form, Formula (2) above is not directly applicable if n(T) is taken to represent the refractive index of bulk VO2. To maintain the
same optical path through VO2 , the required thickness of the VO2 aerogel layer increases in proportion to the ratio of the VO2 bulk density to the areogel density. The resulting thickness may be on the order of 2-3 micrometers.
As a detector layer, a VO2 aerogel layer of a certain porosity p can be used on a SiO2 aerogel thermal insulator layer having a higher porosity p', with the optical mirror then under both such layers.
Without reduction processing, V2O5 aerogel can be used also as a chemically similar thermal insulation material for a bulk, non-porous VO2 detector film. Use of an active sensor layer in aerogel form- thus combining sensing and thermal insulation functions in one layer, can be made independent of readout mode, i.e. equally for devices with resistive readout in bolometric devices. Similar independence applies to use of V2Os aerogel as thermal insulation under more dense VO2, either in non- porous bulk form or at least having lower porosity.
With aerogel layer insulation, typical decay times as short as 1-3 ms may be undesirable in some instances. One related disadvantage lies in an associated loss of sensitivity, and another in shortened IR-radiation writing time. Additionally, aerogel tends to be brittle and micro-porous which may hinder deposition of optically smooth layers.
In lieu of thermal insulation by a layer in aerogel form or otherwise, VO2- mirror pixel combinations can be supported by suspended membranes or micro-bridges over a vacuum gap, with the micro-bridges supported on a substrate. As there is no need for electrical contact to such pixels, fabrication is simplified over resistive bolometer technology.
For illumination, light sources other than a laser can be used. One suitable alternative includes an essentially monochromatic light source, e.g. including a wavelength filter, and a polarizer. Light reflected from the VO2 film then reaches an analyzer filter disposed for eliminating the background. Temperature variations induced by the IR signal will change the angle of polarization, for this part of the reflected signal to pass through the polarizer. Use of a two-mirror interferometer-type structure as described above further enhances change of polarization, allowing for more efficient
subtraction of background.
For positive temperature control of a resonator layer if desired, fast Peltier elements can be included to cycle the temperature between heating and erasing, e.g. with 30 ms frame time. Means for temperature control can be included also for stabilizing the substrate temperature.
For discrimination between the illumination background and the reflected image signal, a technique can be used as described in U.S. Patent No. 5,784,157, "Method and Apparatus for Identifying Fluorophores", issued to Gorfinkel et al. on July 21, 1998 which is incorporated herein by reference. The amplitude of the IR input is modulated in the time domain, e.g. by chopping, while the illumination light is not modulated in this fashion. Then, components of the reflected signal other than those at the chopping frequency can be disregarded at the CCD/matrix level, for example. For enhanced sensitivity, the readout matrix can further include photon-counting light- sensitive elements such as photomultipliers and/or avalanche photodiodes.