A neural network-based system designed for automated detection of concealed items using a postal terahertz scanner is presented, with system optimization provided. A dataset of objects scanned by a THz scanner is introduced. A convolutional neural network is trained on this dataset of terahertz images to classify and detect whether an image contains a prohibited item or not. The system is tested using real-world samples, achieving an accuracy of 95.5% mAP@0.5. The results demonstrate the effectiveness of employing a neural network in postal terahertz scanners and its potential for use in security and surveillance applications.
Issues associated with the development and exploitation of infrared (IR) and terahertz (THz) radiation detectors based on a narrow-gap “HgCdTe” semiconductor have been discussed. This mercury–cadmium–telluride (MCT) semiconductor can be applied for two-color detector operation in IR and sub-THz spectral ranges. Two-color uncooled and cooled down to 78 K narrow-gap MCT semiconductor thin layers grown using the liquid phase epitaxy or molecular beam epitaxy methods on high-resistive “CdZnTe” or “GaAs” substrates, with bow-type antennas, have been considered both as sub-THz direct detection bolometers and 3 to 10 μm IR photoconductors. Their room temperature noise equivalent power at the frequency ν≈140 GHz and signal-to-noise ratio at the spectral sensitivity maximum under monochromatic (spectral resolution ∼0.1 μm) globar illumination reached the following values; ∼4.5×10−10 W/Hz1/2 and ∼50, respectively. Aspheric lenses used for obtaining the images in the sub-THz spectral region were designed and manufactured. With these detectors, about 140 and 270 GHz imaging data have been demonstrated.
Development of infrared and sub-terahertz radiation detectors at the same sensitive elements on the base of mercurycadmium- telluride (MCT) is reported. Two-color un-cooled and cooled to 78 K narrow-gap MCT semiconductor thin layers, grown by liquid phase epitaxy or molecular beam epitaxy method on high resistivity CdZnTe or GaAs substrates, with bow-type antennas were considered both as sub-terahertz direct detection bolometers and 3 to 10 μm infrared photoconductors. Their room temperature noise equivalent power (NEP) at frequency ~ 140 GHz and signal-to-noise ratio (S/N) in the spectral sensitivity maximum under the monochromatic (spectral resolution of ~0.1 μm) globar illumination were reached NEP ~4.5*10-10 W/Hz1/2 and S/N~50, respectively.
Field effect transistors are promising detectors of THz radiation. They operate at room temperatures have high responsivity, low noise equivalent power, and fast response time. However, their linearity (dynamic range) and possibility of their application in the domain of high power radiation has not been yet sufficiently studied. We have investigated room temperature field effect transistors, detection at frequencies from 0.3 to 3 THz with power up to 100 kW/cm2. Several types of HEMTs and MOSFETs operating in the broadband non resonant detection regime, have been investigated. To provide a wide range of incident THz radiation intensities we used continuous-wave and pulsed sources: backward oscillators, CO2 pumped methanol laser, free electron laser, NH3, D2O, and CH3F lasers. We find that the photoresponse of HEMTs and MOSFETs is linear in radiation intensity up to a several kW/cm2 and then it saturates. The onset of the saturation depends on the radiation frequency and the transistor type. The observed saturation behavior can not be explained by the existing theoretical model which predict a square root like dependence of the photoresponse. We tentatively attribute the unusual features of the photoresponse saturation observed at high intensities considering high electric field transport phenomena, e.g., electron heating and electron velocity saturation.
It is shown that electron heating by electromagnetic radiation in mercury-cadmium-telluride (MCT) layers can be used for designing of uncooled terahertz (THz)/sub-THz detectors with appropriate for active imaging characteristics (noise equivalent power ∼2.6×10 −10 W/Hz 1/2 at ν∼140 GHz ) and these detectors can be manufactured within well established MCT technologies. This narrow-gap semiconductor can be considered as a material for THz/sub-THz detectors with possibility to be assembled into arrays. The characteristics of those detectors can be controlled and improved by selection of parameters of initial layers, substrate properties, and antenna configuration. For field effect transistor detectors, even for transistors with rather long channels (∼1 μm ), rather similar characteristics at ν∼140 GHz can also be obtained.
It is shown that electron heating by electromagnetic radiation in MCT layers can be used for designing of uncooled
THz/sub-THz detectors with appropriate for active imaging characteristics (NEP ~2.6•10-10 W/Hz1/2 at ν ~ 140 GHz)
and these detectors can be manufactured within well established MCT technologies. This narrow-gap semiconductor
can be considered as a material for THz/sub-THz detectors with possibility to be assembled into arrays. The
characteristics of those detectors can be controlled and improved by selection of parameters of initial layers, substrate
properties and antennas configuration. For FET detectors, even for transistors with rather long channels (~ 1 μm) rather
similar characteristics at ν ~ 140 GHz can be obtained too.
The concept of THz detection based on excitation of plasma waves in two-dimensional electron gas in Si FETs is
one of the most attractive ones, as it makes possible the development of the large-scale integrated devices based
on a conventional microelectronic technology including on-chip antennas and readout devices integration. In this
work we report on investigations of Terahertz detectors based on low-cost silicon technology field effect
transistors and asymmetric unit cell double grating gate field effect transistor. Double-grating-gate field-effect
transistors have a great potential as terahertz detectors. This is because the double grating gate serves not only for
carrier density tuning but also as an efficient THz radiation coupler. In this paper, we present characterization of
these transistors using high magnetic fields. Low and high magnetic field data are used to determine the electron
mobility and electron concentration, respectively, in different parts of the transistor channel. We show that
detectors, consisting of a coupling antenna and a n-MOS field effect transistor as rectifying element, are efficient
for THz detection and imaging. We demonstrate that in the atmospheric window around 300 GHz, these detectors
can achieve a record noise equivalent power below 10 pW/Hz0.5 and responsivity above 90 kV/W once integrated
with on-chip amplifier. We show also that they can be used in a very wide frequency range: from ~0.2 THz up to
1.1 THz. THz detection by Si FETs paves the way towards high sensitivity silicon technology based focal plane
arrays for THz imaging.
Current-voltage characteristics and differential resistance of n+ -p Hg1-xCdxTe (x=0.2236±0.0015) LWIR (long wavelength infrared) photodiodes forming 128-element array were measured in the temperature range 77-95 K. Experimentally obtained characteristics were fitted by the special nonlinear fitting program based on the carrier- balance equation method which interconnects two processes of transport through the trap level in the band gap: trap-assisted tunneling and thermal Shockley-Reed-Hall generation-recombination process. Other essential current mechanisms (bulk diffusion, band-to-band tunneling, etc.) were included in the model as independent and additive. By the fitting procedure we determine the concentration of donor, acceptor and trap centers, carrier lifetimes, and trap level energy position in the band gap. A good agreement with the experimental data in the whole temperature range of measurements was found assuming that the energy of traps is Et= 0.75 eV above the top of the valence band and does not depend on temperature, unlike the band gap energy. This level seems to be a donor metal vacancies nature of traps in HgCdTe.
Phase states of the lead nanoinclusions in stoichiometric lead telluride were investigated at a magnetic field varying from 0 to 1 kOe in a temperature range T = 1.27 ÷ 6.5 K which is lower compared to superconductivity transition temperature for lead. It has been shown that the nanoinclusions of lead are type-I superconductors. Size of superconductor nanoinclusions at 6.5 K can be more than 300 nm. The model of superconductor intermediate state depended on the inclusions geometry has been suggested to explain experimental data
4×288 MCT LWIR linear arrays with 28X25 μm diodes and silicon ROICs were designed, manufactured and tested. MCT layers were grown by MBE technology on (013) GaAs substrates with CdTe/ZnTe buffer layers and λco = 11.2±0.15 μm at T = 78 K. CCD and CMOS “hybrid” technology for design and manufacture of silicon ROICs was used. The design rules 2.5 μm for CCD technology and 2.0 μm design rules for CMOS technology happened to be sufficient to realize most of the functions for 288×4 MCT TDI array. Analog functions were realized by CCD elements. An amplification of the output signals is realized by CMOS buffer amplifier. Decoding and deselection code storing functions are accomplished by digital CMOS elements. 288 information channels were attached to 4 analog outputs operating in the frequency range f≤4 MHz clock. Total consumption power measured is 50 mW at T = 298 K and 70 mW at T = 78 K. Before hybridization the parameters of MCT linear arrays and Si readouts were tested separately. With aperture 280×640 the detectivities Dλ ≈ 1.8.1011 cm.Hz1/2/W were achieved (λco ≈ 11.2 μm, λmax 10.0 μm) with standard deviation about 15 % and operability close to 100 %.
MCT 2×64 and 4×288 linear arrays with silicon readouts were designed, manufactured and tested. (013) MCT MBE layers were grown on GaAs substrates with ZnTe and CdTe buffer layers. 2×64 arrays were also manufactured on the base of LPE layers on CdZnTe (111) substrates. 50×55 and ≈30×30 μm area n-p-type photodiodes were formed by 50 ÷ 120 keV boron implantation. The dark currents at V ≈ 100 mV reversed biased diodes used in arrays with cutoff wavelength λco ≈ 10.0 - 12.2 μm were within 15 - 50 nA and zero bias resistance-area products were within R0A ≈ 5 ÷ 20 Ohm×cm2. Designed silicon readouts with skimming and partitioning functions were manufactured by n-channel MOS technology with buried or surface channel CCD register. For achievement with the silicon readouts the deselection function, the “composite” technology approach was considered. In this case both the technology of n-channel CCD and CMOS technology were applied, which allow to weaken considerably the technological design rules for realization of 288×4 readouts with deselection of “dead” elements. It is shown that 2.5 μm design rules for CCD and 2.0 design rules for CMOS technologies allow to realize most of the functions needed for 288×4 MCT array operation with deselection function. Before hybridisation the parameters of MCT linear arrays and Si readouts were tested separately. HgCdTe arrays and Si readouts were hybridised by cold welding In bumps technology. In dependence of FOV with skimming mode used for integration time of 8 - 20 μs detectivities within D*λ (0.4 - 1.7)×1011 cm×Hz1/2/W were achieved in dependence of the array format. Dark carrier transport mechanisms in MCT diodes were calculated and compared with experimental data.
For 288x4 mercury-cadmium telluride (MCT) diode array silicon readouts with deselection function, the "composite" technology approach, which simplifies the technology of their manufacturing, is considered. Both technology of n-channel CCD devices and the CMOS technology are applied, which allow to weaken considerably the technological requirements for realization of 288x4 readouts with deselection of "dead" elements (generally the 0.8 micron design rules technology is applied). It is shown that the design rules 2.5 μm for CCD technology and 2.0 design rules for CMOS technology are sufficient to realize most of functions needed for 288x4 MCT array design and manufacture. All analog functions (including TDI as the most complex function for realization in CMOS basis) are realized by CCD elements. Four-phase TDI register was realized using semi-buried channel by phosphorus ion implantation. An amplification of the output signals is realized by CMOS buffer amplifier. Decoding and deselection code storing functions are realized by digital CMOS elements. The parameters of the 288x4 silicon readout device: direct injection input circuits, 4 elements TDI function, 4 outputs; 4 MHz maximum information output frequency; 2 MHz maximum clock frequency; 3 V swing output voltage; not less than 1.6 pC maximum charge capacity per each input; 3.0 pC maximum charge capacity at multiplexor input; 75 dB dynamic band; 28 output pins.
x4×288 heteroepitaxial mercury-cadmium telluride (MCT) linear arrays for long wavelength infrared (LWIR) applications with 28×25 micron diodes and charge coupled devices (CCD) silicon readouts were designed, manufactured and tested. MCT heteroepitaxial layers were grown by MBE technology on (013) GaAs substrates with CdZnTe buffer layers and have cutoff wavelength λco ≈ 11.8 μm at T = 78 K. To decrease the surface influence of the carriers recombination processes the layers with composition changes and its increase both toward the surface and HgCdTe/CdZnTe boundary were grown. Silicon read-outs with CCD multiplexers with input direct injection circuits were designed, manufactured and tested. The testing procedure to qualify read-out integrated circuits (ROICs) on wafer level at T = 300 K was worked out. The silicon read-outs for 4×288 arrays, with skimming and partitioning functions included were manufactured by n-channel MOS technology with buried or surface channel CCD register. Designed CCD readouts are driven with four- or two-phase clock pulses. The HgCdTe arrays and Si CCD readouts were hybridized by cold welding indium bumps technology. With skimming mode used for 4×288 MCT n-p-junctions, the detectivity was about (formula available in paper) for background temperature Tb = 295 K.
Investigations of performance of mercury cadmium telluride (MCT) multipixel arrays at T≈80 K are considered. MCT hybrid arrays for long-wavelength infrared (LWIR) applications with n+-p-diodes and n-channel charged coupled devices (CCD) silicon readouts were designed, manufactured and tested. For testing procedure the measurements of noise and signal-to-noise ratio (SNR) are the key issues to determine performance parameters to characterize IR-sensors. That puts certain requirements to the registration system and methods of measuring used. The noise of the signals from LWIR-photodiodes with CCD readouts or CCD readouts itself was measured using several different techniques. To find out and eliminate noise sources the spectral noise power of signals was analyzed. It allowed the possibility to implement actions for reducing of the registration system noise, and to define the software noise filters to be used. The testing procedure of FPA performance characteristics includes the measurements of detectivity D*, noise equivalent temperature difference NETD, cut-off wavelength and some other parameters of the arrays.
Dark carrier transport mechanisms in narrow-gap Hg1-xCdxTe multilayer structures and Pb1-zSnzTe/PbTe1-yS(Se)y heterojunctions at T~80 K for applications in IR arrays are analyzed and compared with homojunction mercury-cadmium telluride (MCT) photodiode characteristics in the temperature range T~70-150 K. In the analysis procedure two major current mechanisms were included into the current balance equations: trap-assisted tunneling (TAT) and Shockley-Reed-Hall (SRH) generation-recombination processes for a defect trap level. Other current mechanisms (e.g., band-to-band tunneling, bulk diffusion) were taken into account as additive contributions. For TAT the tunneling rate characteristics were calculated within the k-p-approximation. Using donor and acceptor concentrations, trap level energies and concentrations, and in-junction trap level lifetimes as fitting parameters, good agreement with experimental data for HgCdTe and PbSnTe heterojunction and homojunction diodes was obtained, which allows one to predict the diode parameters from the known material characteristics. Photodiode or array parameters itself, or with CCD readouts, or CCD readouts separately were tested to study the influence of readout cascade on the diodes' properties.
Mercury-Cadmium-Telluride (MCT) 2 X 64 linear arrays with silicon readouts were designed, manufactured and tested. NCT layers were grown by MBE method on (103) GaAs substrates with CdZnTe buffer layers. 50 X 50 mm n-p-type photodiodes were formed by 80 divided by 120 keV boron implantation. The dark current at 100 mV reversed biased diodes was within 15*30 nA and zero bias resistance-area product was within R0 approximately equals 20 divided by 50 Ohm X cm2. Silicon read-out circuits were designed, manufactured and tested. Read-outs with skimming and partitioning functions were manufactured by n-channel MOS technology with buried or surface channel CCD register. The parameters of LWIR MCT linear arrays with cutoff wavelength (lambda) co 10.0 divided by 12.2 micrometers and Si readouts were tested separately before hybridization. The HgCdTe arrays and Si readouts were hybridized by cold welding In bumps technology. With skimming mode used for integration time of 24-30 ms for such MCT n-p-junctions, the detectivity D*(lambda ) approximately equals 4 X 1010 cmXHz1/2/W. Dark carrier transport mechanisms in these diodes were calculated and compared with experimental data. Two major current mechanisms were included into the current balance equations: trap-assisted tunneling and Shockley-Reed-Hall generation-recombination processes via a defect trap level in the gap. Other current mechanisms (band-to-band tunneling, bulk diffusion) were taken into account as additive contributions. Tunneling rate characteristics were calculated within k-p approximation with the constant barrier electric field. Relatively good agreement with experimental data for diodes with large zero resistance-area products (R0A > 10 OhmXcm2) was obtained.
KEYWORDS: Semiconductors, Group IV-VI semiconductors, Crystals, Magnetism, Metals, V band, Lead, Magnetic semiconductors, Laser crystals, Temperature metrology
The role of deep defect states in recombination processes of narrow-gap IV-VI semiconductors is discussed. Earlier several defect states in the gap of IV-VI semiconductors were revealed. Estimations have shown that defect levels nearer to c-band presumably are connected with metal-rich microinclusions by diameter about 1300 angstrom. These metal- rich inclusions are the result of low metal vacancies enthalpy formation ((Delta) v approximately equals 0.35 eV). These defect states seem are arising at the inclusions and semiconductor matrix boundaries. The density and the diameter of these intrinsic metal microinclusions were calculated from the results of magnetic susceptibility by SQUID-magnetometry experiments in the 1.7 - 20 K temperature region. The levels above the v- band, with the density of states about 3 times lower compared to density of states below c-band, seem to be connected with Te-rich microclusters. The enthalpy of Te-vacancies formation is about 0.45 eV. Experiments and estimations fulfilled on defect levels position and concentrations allow to explain the experiments on band-to-band recombination, photoconductivity and relaxation measurements, stimulated recombination between defect level and v-band, and some nonlinear effects in these semiconductors.
Two X sixty-four linear photodiode arrays on the base of HgCdTe MBE grown layers with CCD silicon readouts were designed, fabricated and tested. It is shown that detectivity for the given arrays even with skimming mode used for long integration times that is need for large square n-p-junctions used and cut-off wavelength of 12.2 micrometer was near the ultimate performance limit.
Nonlinear optical transmission in single Cd0.2Hg0.8Te crystals in the range of fundamental absorption at high laser intensity level was investigated at 298K. For thick samples the decease of light transmission up to five times was registered at wavelength length (lambda) equals 10.6 micrometers and at laser intensity level 300 kW/cm2. It was explained by new channel of absorption: two-level transition of the electron inter-intra-band increase under the laser pulse influence.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.