A 350 GHz high-resolution high-sensitivity passive video imaging system∗ Daniel Beckera, James Bealla, Hsiao-Mei Choa, William Duncana†, Gene Hiltona, Rob Horanskya, Kent Irwina, Peter Lowella, Michael Niemacka, Nick Paultera, Carl Reintsemaa, Frank Schimaa, Robert Schwalla, Ki Won Yoona, Peter Adeb, Carole Tuckerb, Simon Dickerc, Mark Halpernd aNIST-Boulder bCardiff University cUniversity of Pennslyvania dUniversity of British Columbia ABSTRACT We are developing a 350 GHz cryogenic passive video imaging system. This demonstration system uses 800 photon-noise-limited superconducting transition edge sensor bolometers. It will image a 1 m x 1 m area at a standoff distance of 16 m to a resolution of approximately 1 cm at video frame rates (20 frames per second). High spatial resolution is achieved by the use of an f/2.0 Cassegrain optical system with 1.3 m primary mirror. Preliminary testing of prototype detectors indicates that we can achieve a noise equivalent temperature difference (NETD) of 70 mK for the fully sampled 1 m x 1 m image at 20 frames per second. 1. INTRODUCTION Millimeter-wavelength radiation penetrates clothing well and is non-ionizing, making it a good candidate for identifying threats held beneath clothing.1 Active millimeter-wave systems have been deployed at airports and are commercially available, but require operation at close distances. For identifying threats at larger distances (more than eight meters), passive video imaging systems are available commercially2 but have poor spatial and temperature resolution. Cryogenic passive imaging systems using superconducting transition edge sensors are also being developed. One system with 64 detectors produces video images with a noise equivalent temperature difference (NETD) of 0.5 K at 6 Hz with a resolution of 4 cm over a 2 m x 4 m area.3–6 Another system has much lower noise on a per-detector basis, but currently has only 10 detectors.7 The millimeter and sub-millimeter astronomy community has spent the past several decades working on a very similar problem: high-speed mapping of high resolution images over a large field of view. A commonly used solution has been the use of large-format arrays of over 1000 superconducting transition edge sensor (TES) bolometers, with multiplexed readout using superconducting quantum interference devices (SQUIDs).8–10 The development of this technology provides the opportunity to greatly increase temperature sensitivity of terrestrial passive imaging systems. We are developing a cryogenic passive video imaging system for operation at distances of 16 m - 28 m. The focal plane will contain 800 photon-noise-limited TESs. The system is intended as a test-bed to explore trade-offs between NETD, resolution and video frame rate, with the goal of understanding what specifications are required to achieve the requirements of stand-off imaging applications. The system will operate at a frequency of 350 GHz (850 μm) with 11 % fractional bandwidth. Spatial resolution will be 1 cm over a 1 m x 1 m field of view at distance of 16 m. We expect to generate video images at 20 frames per second. Measurements of prototype detectors indicate that the fully populated focal plane with 800 detectors should achieve NETD of 70 mK. In this paper we describe the optics, detectors, readout, cryogenics and current status of this system. ∗ US Government contribution; not subject to copyright in the United States. † Current Address: Intellectual Ventures, 11235 SE 6th St., Suite A 200, Bellevue, WA 98004 Send correspondence to Daniel Becker: E-mail: beckerd@boulder.nist.gov Passive Millimeter-Wave Imaging Technology XIII, edited by David A. Wikner, Arttu R. Luukanen, Proc. of SPIE Vol. 7670, 76700M · © 2010 SPIE · CCC code: 0277-786X/10/$18 · doi: 10.1117/12.852932 Proc. of SPIE Vol. 7670 76700M-1 Downloaded from SPIE Digital Library on 13 Sep 2011 to 137.82.117.28. Terms of Use: http://spiedl.org/terms 2. DESIGN OVERVIEW 2.1 Optical Frequency Choice The choice of frequency for a passive security imaging system is a tradeoff between the ability to penetrate clothing (favoring low frequencies) and spatial resolution (favoring high frequencies). The 3 dB point for attenuation through clothing ranges from 350 GHz - 1 THz (high frequencies for thin materials such as Nylon, lower for thicker materials such as wool).11 We choose 350 GHz to keep clothing attenuation at or below 3 dB and provide good spatial resolution. 350 GHz is also the center of an atmospheric transmission band (see Figure 1). 300 320 340 360 380 4000 0.2 0.4 0.6 0.8 1 Frequency [GHz] Tr an sm is si on (A rbi tra ry Un its ) Atmospheric & Filter Transmission Atmospheric Transmission Filter Transmission Figure 1. Plot showing measured transmission of our bandpass filter (solid lines) and typical atmospheric transmission (dashed lines). The passband is centered at approximately 350 GHz (850 µm) and has a full width at half-maximum of about 11 %. Figure 2. Schematic of optical system. Labels are (A) 1.3 m primary mirror, (B) secondary mirror with dithering mechanism, (C) cryostat window lens, (D) detector fo- cal plane module, including bandpass filter. For clarity, the cryostat is not shown. Inset is a photograph of the primary and secondary mirror assembly. 2.2 Optics Our optical system is an f/2.0 Cassegrain design. A 1.3 m elliptical primary mirror focuses light onto a hyperbolic secondary, which directs the light back through a central hole in the primary; the light then enters an ultra-high molecular weight (UHMW) polyethylene lens at the mouth of our cryostat. The lens makes the system telecentric, allowing use of a flat focal plane. The lens/cryostat combination can be moved relative to the primary mirror to focus the system at distances between 16 m and 28 m. The optics were simulated with ray-tracing software and shown to be diffraction-limited across the entire focal plane and at all focusing distances between 16 m and 28 m. See Figure 2 for an annotated picture of the optical system. A detailed description of the optics has been published elsewhere.12 The observation band is defined by a metal mesh filter13 cooled to 850 mK, defining an optical bandwidth of 40 GHz (11 % FWHM fractional bandwidth); the measured passband of this filter is shown in Figure 1. Additional filters for thermal blocking are located at the 4 K, 50 K, and 300 K stages.14 Total transmittance of the filter stack plus lens is expected to be 63 %. At the focal plane, light is captured by smooth-walled conical feedhorns, which were chosen for their simplicity and ease of manufacture. To optimize coupling of the feedhorns to the rest of the optical system, we use feedhorns with a diameter of 3.2 mm, or approximately 2Fλ, where F = 2.0 is the f-number of the optics and λ = 850 μm is the optical wavelength. The taper half-angle is 9.4 deg, tapering to 0.6 mm circular waveguide that leads to the detectors. The feedhorn design was modeled with finite-element electromagnetic simulations, predicting a spillover efficiency of 68 %. Proc. of SPIE Vol. 7670 76700M-2 Downloaded from SPIE Digital Library on 13 Sep 2011 to 137.82.117.28. Terms of Use: http://spiedl.org/terms Because the point-spread-function of the optics has size Fλ on the focal plane, the feedhorns will be spaced farther apart than the Nyquist limit. Additionally, the feedhorns will cover only a portion of the diffraction- limited area of the focal plane. To generate fully sampled images, and to use the entire diffraction-limited extent of the optics, the secondary mirror will be moved by a set of actuators. This dithering of the secondary mirror will move the image across the focal plane, allowing fully sampled video images to be generated across the entire 1 m2 target area. To create a 1 m2 image to resolution of 1 cm, 2002 pixels per video frame are required. With an 800 detector system, this leads to a “dithering factor” of 2002/800 = 50, meaning that each detector will image 50 image pixels per video frame. At the back of the 0.6 mm circular waveguide, in front of a quarter-wave backshort, the light is absorbed by a PdAu mesh patterned on a relieved membrane as described in section 2.4. Finite-element electromagnetic models indicate the PdAu mesh will absorb more than 90 % of the optical power coming down the waveguide. The mesh is sensitive to both polarizations. 2.3 Transition Edge Sensors Optical power is detected by voltage-biased superconducting transition edge sensor (TES) bolometers.15, 16 Below a critical temperature Tc, a superconductor loses all resistance to DC electrical current (see Figure 3). The transition from the normal state to superconducting state is very sharp, allowing a superconductor that has been biased into the transition to serve as a very sensitive thermometer. Any signal that can be turned into a temperature change can then be detected as a change in the resistance of the superconductor. Figure 4 schematically depicts the operation of a voltage-biased TES bolometer. Optical power is absorbed by an absorber that is thermally sunk to a superconducting film. Increasing optical power raises the temperature of the film, increasing the resistance of the TES, which decreases the current passing through it. This drop in current is detected by a SQUID. The device operates in a negative-feedback loop, with increasing optical power leading to lower V 2/R Joule heating, allowing the device to operate stably. While operating, the total power dissipated in the bolometer is a constant, Psat, with any change in absorbed optical power compensated by an opposite change in Joule heating. The absorber and film are connected to a low-temperature heat bath by a weak thermal link. The thermal conductance of this link determines the saturation power Psat according to Psat = K(T nc − T nbath), where K and n depend on the geometry of the thermal link and the dominant heat transport mechanism in the temperature regime of interest; for our detectors we expect n ≈ 4.0.17 The differential thermal conductance G is then defined as G = dP/dTc = nKT n−1c ; G is an important parameter for the design of a bolometer. The dominant source of intrinsic noise in our TES bolometers is thermal fluctuations of the temperature of the TES itself, due to random exchanges of energy between the TES and the low-temperature heat bath. This noise is given by NEPG = √ 4kBGT 2c , where kB is Boltzmann’s constant. 17 Photon noise is also a significant source of noise in our system. The expression for photon noise, including the effect of photon bunching, is given by18 NEPph = √ 2hνPopt(1 + ηn̄), (1) where h is Planck’s constant, ν is the central frequency, Popt is the total optical loading, η is the optical efficiency (excluding feedhorn spillover efficiency, because the feedhorn spillover is terminated at 300 K), and n̄ = [exp(hν/kBT )− 1]−1 is the average photon occupation number per mode. For our system observing 300 K radiation, ηn̄ ≈ 10, so the bunching term is important. Total optical power per detector when observing a 300 K target will be 2ηkBTΔν = 190 pW. Use of equation 1 then gives NEPph = 1.0 fW/ √ Hz. Proc. of SPIE Vol. 7670 76700M-3 Downloaded from SPIE Digital Library on 13 Sep 2011 to 137.82.117.28. Terms of Use: http://spiedl.org/terms Figure 3. Illustrative plot of resistance vs temperature for a superconductor. The transition is typically very nar- row, allowing the use of a superconductor as a sensitive thermometer. A TES detector is voltage-biased into the transition, with the bias point changing based on changes in optical power. Figure 4. Schematic depiction of the operation of a TES. Optical power is absorbed onto the TES. Increases in ab- sorbed optical power raises the temperature of the TES, raising its resistance. The dissipated V 2/R Joule power in the TES drops, compensating for the increase in opti- cal power, leaving as constant the total dissipated power flowing out of the TES through the weak thermal link. This negative electro-thermal feedback allows the detec- tor to operate stably. Changes in current passing through the TES are detected by a SQUID. 2.4 Design of Prototype Detectors Figure 5 shows a picture of one of our prototype detectors. The device is fabricated on a relieved SiN membrane approximately 800 μm in diameter. Light is absorbed by a PdAu mesh, with dimensions (2 μm wide lines spaced 85 μm apart) chosen to match the wave impedance of the waveguide. The Al TES is located at the bottom center of the membrane, and has dimensions 63 μm x 63 μm. A 15 μm x 16 μm PdAu heater resistor is located near the center of the relieved membrane for use in testing. We targeted our detectors to have Psat = 1 nW (safety factor of 5). Given this saturation power, an aluminum TES with Tc = 1.2 K, a bath at 850 mK, and a target G of 4.5 nW/K should meet our NETD requirements. The dimensions of the eight SiN legs (11 μm wide x 40 μm long, 0.5 μm thick) were chosen to achieve this G. This is a fairly large thermal conductance, which requires a large heat capacity C to keep the thermal time constant near the 1 ms that is optimal for our readout system. We have achieved this by adding a 2 μm thick gold ring around the outer diameter of the relieved membrane. 2.5 Readout The array is read out by the SQUID-based time-division multiplexing (TDM) system developed at NIST.19 This readout system has been in use for many years in a variety of applications, including millimeter-wave astronomy. The warm electronics will be provided by the Multi-Channel Electronics (MCE) system, which consists of a single crate that can read out all 800 of our detectors.20 The combination of NIST SQUID TDM with the MCE is currently being used to collect data in the field at two different telescopes.8, 10 All hardware for the full 800 pixel array, including cryogenic wiring, is in place and installed in our cryostat. 2.6 Cryogenics To cool our focal plane to an operating bath temperature of 850 mK, we use a two-stage refrigeration system. A commercially available cryogen-free Gifford-McMahon cryocooler allows us to reach 3 K. To cool the focal plane to below 1K we use a sealed 4He sorption fridge. The sorption fridge is a custom-built system based on a design Proc. of SPIE Vol. 7670 76700M-4 Downloaded from SPIE Digital Library on 13 Sep 2011 to 137.82.117.28. Terms of Use: http://spiedl.org/terms Figure 5. A picture of one of our fabricated TES detec- tors. The relieved area is 800 µm in diameter. The dif- ferent components are: (c) PdAu absorbing mesh, lines 2 µm wide spaced 85 µm apart, (d) SiN legs connecting relieved membrane to the bulk Si substrate that serves as low-temperature bath, (e) thin-film Al TES, (f) Au ring providing sufficient heat capacity to slow down thermal response time of device for optimal readout (g) PdAu heater resistor used for testing. 800 900 1000 1100 1200 13000 500 1000 1500 2000 Bath Temperature [mK] To ta l P ow er (T ES + he ate r) [pW ] Saturation Power vs Bath Temp 4309 Data 4309 Fit 4311 Data 4311 Fit 4307 Data 4307 Fit Figure 6. Plot showing saturation power Psat vs bath temperature for three of our prototype devices. Fitting all three parameters n, K, and Tc implies n ≈ 3.6. The fit to n is very weak, so for this paper we assuming n = 4.0 everywhere; this increases our estimated detector noise by a few percent. Detector 4307 was measured in a dif- ferent cryostat with different thermometry, which could explain the temperature offset from detectors 4309 and 4311. that has been proven in the field.21 The system reaches a base temperature of 850 mK with a hold time of 36 hours under no optical load. Operation of this system will be fully automated. 3. CURRENT STATUS We have fabricated and performed dark tests on prototype detectors. Our prototype devices have a transition temperature of approximately 1.23 K with a normal-state resistance of about 3.5 mΩ. Figure 6 shows measured saturation power vs. bath temperature for the three tested devices; the measured G values of 5.0 nW/K - 5.4 nW/K are about 10 % - 25 % higher than expected. These values of G imply NEP of 0.64 fW/ √ Hz - 0.69 fW/ √ Hz. Figure 7 shows the measured noise power spectral density for one of our detectors. The white noise level is about 0.8 fW/ √ Hz, 25 % higher than predicted based on G. Nevertheless our detectors are at the start of the photon-noise-limited level. Adding the detector and expected photon noise in quadrature gives NEPtot = 1.3 fW/ √ Hz. This total NEP can be converted to an NETD through the use of the radiometer equation, NETD = NEPtot 2 √ 2τkBΔνη , (2) where τ is the total integration time per pixel in the processed video frame, Δν is the pre-detection optical bandwidth, and η is the total optical efficiency. For our system, τ = 1 50(20 Hz) (50 is the dithering factor defined in section 2.2), Δν = 40 GHz, and η = 0.38 (0.90 absorber efficiency, 0.63 filter stack efficiency, and 0.68 feedhorn spillover efficiency). Using equation 2, we estimate that our system will have NETD = 70 mK across the full field of view at 20 FPS. Our next steps are to characterize our prototype detectors optically, generate still images with a 4-pixel prototyping setup, and then install the full 800-pixel focal plane. Proc. of SPIE Vol. 7670 76700M-5 Downloaded from SPIE Digital Library on 13 Sep 2011 to 137.82.117.28. Terms of Use: http://spiedl.org/terms 1 100 10000 1 10 Frequency (Hz) N EP (fW /rt− Hz ) TES Noise Power Spectral Density Figure 7. Plot showing measured noise equivalent power (NEP) spectral density for one of our detectors. The white noise level is about 0.8 fW/ √ Hz. The major peaks are at power line frequencies. 4. NEXT-GENERATION SYSTEM The current system was designed for high sensitivity and low noise to facilitate the investigation of the tradeoffs inherent in an operational system. Adapting this system for field deployment will require reduction in cost, weight, and complexity. In addition to obvious modifications such as molded composite optical elements and simplified cryogenics, we are considering a readout system based on gigahertz micro-resonators that could be used to read out focal planes of either TESs or Microwave Kinetic Inductance Devices (MKIDs).22 This readout system will read out hundreds of detectors with a single coaxial cable, greatly reducing the cost and complexity of cryogenic wiring, and allowing larger numbers of detectors to be placed on the focal plane. 5. CONCLUSION We have designed a 350 GHz cryogenic terrestrial passive video imaging system by leveraging technology proven in the millimeter and sub-millimeter astronomy field. The video imager will use 800 TES bolometers to image a 1 m x 1 m field to a resolution of 1 cm at a distance of 16 m at video frame rates. Measurements of our prototype pixels indicate that they are close to being photon-noise-limited, and we should be able to achieve an NETD of 70 mK across the full field of view at 20 FPS. Our next steps are to generate still images with a four-pixel prototyping setup, followed by installation of the full 800-pixel focal plane. REFERENCES [1] R. Appleby, “Passive millimetre–wave imaging and how it differs from terahertz imaging,” Philosophical Transactions A 362(1815), p. 379, 2004. [2] C. Mann, “First demonstration of a vehicle mounted 250GHz real time passive imager,” in Proceedings of SPIE, 7311, p. 73110Q, 2009. [3] A. Luukanen and J. Pekola, “A superconducting antenna-coupled hot-spot microbolometer,” Applied Physics Letters 82, p. 3970, 2003. [4] E. Grossman, C. Dietlein, J. Bjarnason, A. Luukanen, M. Leivo, and J. Pentilla, “THz microbolometers for imaging applications,” in Infrared, Millimeter and Terahertz Waves, 2008. IRMMW-THz 2008. 33rd International Conference on, pp. 1–1, 2008. Proc. of SPIE Vol. 7670 76700M-6 Downloaded from SPIE Digital Library on 13 Sep 2011 to 137.82.117.28. Terms of Use: http://spiedl.org/terms [5] A. Luukanen, P. Helistö, P. Lappalainen, M. Leivo, A. Rautiainen, H. Toivanen, H. Seppä, Z. Taylor, C. Dietlein, and E. Grossman, “Stand-off passive THz imaging at 8-meter stand-off distance: results from a 64-channel real-time imager,” in Proceedings of SPIE, 7309, p. 73090F, 2009. [6] A. R. Luukanen, L. Grnberg, M. Grnholm, P. Lappalainen, M. Leivo, E. N. Grossman, and C. R. Dietlein, “Real-time passive terahertz imaging system for standoff concealed weapons imaging,” in Proceedings of SPIE, 7670, pp. 7670–3, 2010. [7] T. May, G. Zieger, S. Anders, V. Zakosarenko, H. Meyer, M. Schubert, M. Starkloff, M. Rößler, G. Thorwirth, and U. Krause, “Safe VISITOR: visible, infrared, and terahertz object recognition for security screening application,” in Proceedings of SPIE, 7309, p. 73090E, 2009. [8] D. Swetz, P. Ade, C. Allen, M. Amiri, J. Appel, E. Battistelli, B. Burger, J. Chervenak, A. Dahlen, S. Das, et al., “Instrument design and characterization of the millimeter Bolometer Array Camera on the Atacama Cosmology Telescope,” in Proceedings of SPIE, 7020, p. 702008, 2008. [9] J. Ruhl, P. Ade, J. Carlstrom, H. Cho, T. Crawford, M. Dobbs, C. Greer, N. Halverson, W. Holzapfel, T. Lanting, et al., “The south pole telescope,” in Proceedings of SPIE, 5498, pp. 11–29, 2004. [10] W. Holland, W. Duncan, B. Kelly, K. Irwin, A. Walton, P. Ade, and E. Robson, “SCUBA-2: a large-format submillimeter camera on the James Clerk Maxwell Telescope,” in Proceedings of SPIE, 4855, p. 1, 2003. [11] J. Bjarnason, T. Chan, A. Lee, M. Celis, and E. Brown, “Millimeter-wave, terahertz, and mid-infrared transmission through common clothing,” Applied Physics Letters 85(4), pp. 519–521, 2004. [12] W. Duncan, R. Schwall, K. Irwin, J. Beall, C. Reintsema, W. Doriese, H. Cho, B. Estey, G. Chattopadhyay, P. Ade, et al., “An Optical System for Body Imaging from a Distance Using Near-TeraHertz Frequencies,” Journal of Low Temperature Physics 151(3), pp. 777–783, 2008. [13] P. Ade, G. Pisano, C. Tucker, and S. Weaver, “A review of metal mesh filters,” in Proceedings of SPIE, 6275, pp. 62750U–1, 2006. [14] C. Tucker and P. Ade, “Thermal filtering for large aperture cryogenic detector arrays,” in Proceedings of SPIE, 6275, p. 62750T, 2006. [15] K. Irwin, “An application of electrothermal feedback for high resolution cryogenic particle detection,” Applied Physics Letters 66, p. 1998, 1995. [16] A. Lee, P. Richards, S. Nam, B. Cabrera, and K. Irwin, “A superconducting bolometer with strong elec- trothermal feedback,” Applied Physics Letters 69, p. 1801, 1996. [17] K. D. Irwin and G. C. Hilton, Cryogenic Particle Detection, ch. Transition Edge Sensors, pp. 63–149. Springer-Verlag Berlin Heidelberg, 2005. [18] J. Zmuidzinas, “Thermal noise and correlations in photon detection,” Applied optics 42(25), pp. 4989–5008, 2003. [19] P. de Korte, J. Beyer, S. Deiker, G. Hilton, K. Irwin, M. MacIntosh, S. Nam, C. Reintsema, L. Vale, and M. Huber, “Time-division superconducting quantum interference device multiplexer for transition-edge sensors,” Review of Scientific Instruments 74, p. 3807, 2003. [20] E. Battistelli, M. Amiri, B. Burger, M. Halpern, S. Knotek, M. Ellis, X. Gao, D. Kelly, M. Macintosh, K. Irwin, et al., “Functional description of read-out electronics for time-domain multiplexed bolometers for millimeter and sub-millimeter astronomy,” Journal of Low Temperature Physics 151(3), pp. 908–914, 2008. [21] M. Devlin, S. Dicker, J. Klein, and M. Supanich, “A high capacity completely closed-cycle 250 mK 3He refrigeration system based on a pulse tube cooler,” Cryogenics 44(9), pp. 611–616, 2004. [22] B. Mazin, P. Day, K. Irwin, C. Reintsema, and J. Zmuidzinas, “Digital readouts for large microwave low-temperature detector arrays,” Nuclear Instruments and Methods in Physics Research Section A: Accel- erators, Spectrometers, Detectors and Associated Equipment 559(2), pp. 799–801, 2006. Proc. of SPIE Vol. 7670 76700M-7 Downloaded from SPIE Digital Library on 13 Sep 2011 to 137.82.117.28. Terms of Use: http://spiedl.org/terms