The measurement of thermal radiation from ambient-temperature objects using short-wave infrared detectors and regular glass optics is described. The detectors are chosen to operate in the 2.0 µm to 2.5 µm atmospheric window. Selection of detectors with high shunt resistance along with the 4-stage thermo-electric cooling of the detectors to -85 °C results in detectivity, D*, of 4×1013 cm Hz1/2/W which is near the background limited performance at 295 K. Furthermore, the use of regular-glass commercial optics to collect the thermal radiation results in diffraction-limited imaging. The use of a radiation thermometer constructed with these elements for the measurement of a blackbody from 20 °C to 50 °C results in noise-equivalent temperature difference (NETD) of <3 mK at 50 °C. The operation at shorter wavelengths than traditional thermal sensors also leads to lower sensitivity to the emissivity of the object in determining the temperature of the object. These elements are used to construct a calibrator for an infrared collimator, and such a system demonstrates noise-equivalent irradiances of <5 fW/cm2. These results indicate that radiometers using short-wave infrared sensors could be constructed utilizing commercial glass optics with possible better performance and lower NETD than existing radiometers using cryogenically-cooled mid-infrared or thermal infrared detectors.
©2008 Optical Society of America
The needs for measurement of thermal radiation from ambient-temperature and higher-temperature objects are vast and wide ranging . These needs arise from requirements of surveillance and non-destructive testing to early and accurate detection of hot objects at far distances for defense and other applications. At present, these needs are met by the use of cryogenically-cooled quantum detector materials such as indium antimonide (InSb) or mercury-cadmium telluride (HgCdTe) detectors or by the use of thermal detectors such as thermopiles, pyroelectrics or bolometers . Although the quantum detectors have sufficient sensitivities to be able to measure thermal radiation with low noise-equivalent temperature difference (NETD), the need for cryogens or cryo-coolers and the accompanying vacuum jacket for low-noise operations limit their versatility in many applications. Furthermore, the thermal cycling and eventual ice build-up inside the cryostat lead to poor long-term stability of the responsivity and require routine maintenance of the system. Instruments utilizing thermal detectors have relatively low signal-to-noise ratios resulting from the low-power responsivities of the thermal detectors which lead to high NETD. Also, the use of thermal detectors can result in drift of the measured signal from any change in the surrounding ambient temperatures. For collection optics, these traditional detectors also require the use of either reflective mirrors or refractive lenses which transmit in the mid-infrared and long-infrared wavelength such as those constructed using Si, Ge or ZnSe.
In the past, the use of detectors in the short-wave infrared (SWIR) wavelength region has been limited due to the poor signal-to-noise performances of the existing detectors. The commonly utilized detector in this wavelength region is the photoconductive PbS which requires operation with dry ice at 193 K for optimum operation. However, photoconductive PbS was found to be limited by its high noise-equivalent power, poor long-term stability of response and possible signal nonlinearity. Other SWIR detectors such as photoconductive MCT, PbSe and InAs detectors have also been used. Photoconductive InSb detectors with cryogenic cooling and a cooled cut-off filter can be used for these wavelength regions.
Due to the rapid improvements in InGaAs material fabrication driven by the optical fiber communications, InGaAs photodiodes are the preferred detectors in the 900 nm to 1700 nm range. With modification of the fabrication process, InGaAs can be made to be sensitive to infrared radiation from 900 nm to 2500 nm, with the long-wavelength bandgap occurring at around 2500 nm. However, the extension of the responsivity to longer wavelengths results in extended-InGaAs (ex-InGaAs) diodes having lower shunt resistances than the regular InGaAs photodiodes. The shunt resistances of the ex-InGaAs diodes can be increased by cooling the diode with 1-stage to 4-stage thermo-electric (TE) coolers. Similarly, photovoltaic (pv) MCT photodiodes can also be used. MCT photodiodes can be fabricated for sensitivity from 2.0 µm to 2.5 µm with long-wavelength cutoffs near 2.5 µm, distinct from the usual 10 µm to 12 µm operation of these devices. The shunt resistances of the SWIR MCT detectors can also be increased by thermoelectric (TE) cooling, and such detectors can be substituted for ex-InGaAs.
In this work, the use of TE-cooled, extended-InGaAs and short-wave, pv MCT diodes for the measurement of thermal radiation from ambient-temperature objects is described. The work is motivated by the examination of the thermal emission and the radiance ratios between the object and the background radiation. The SWIR radiometers can be designed to operate in a clear, atmospheric window between 2.0 µm and 2.5 µm. For this wavelength region, in contrast to other infrared (IR) bands, off-the-shelf glass optics can be utilized for the collection of the emitted radiation. Optical modeling using glass lenses indicates that diffraction-limited performance is achieved. As examples, we demonstrate, using a TE-cooled ex-InGaAs radiation thermometer and a radiometer for measuring collimated sources, that such radiometers can achieve significantly better irradiance and radiance sensitivity than radiometers utilizing cryogenically-cooled InSb.
2. Planck radiance analysis
The choice of the optimum wavelengths for the detection of Planck radiation depends upon many factors. Although the obvious choice of the wavelength would be at the peak of the Planck radiance of the emitting body, other considerations can mitigate the signal advantages from measuring at the peak of the spectral distribution. Often, the object of interest will be in the presence of ambient background blackbody radiation which is also detected by the sensors leading to a large contaminating, background signal. The background radiation could come from objects in the field-of-view of the radiometer or from the thermal emission from body of the radiometer. Much of the unwanted background signal can be eliminated by the use of lock-in detection and the placement of the modulator. However, outside of the laboratory environment, the optimum modulation, which is directly at the source, is not feasible.
In many defense-related applications, the detection of thermal radiation from an object which is hotter than the ambient background is desired. Many forward-looking infrared (FLIR) systems with mid-infrared or thermal infrared sensitive InSb or MCT sensors are used in these applications. In such cases, the advantages of using the shortest wavelength while having sufficient signals from the target area can be seen from Fig 1.
The Planck radiances for the temperatures of interest are plotted, and for comparisons, the radiance differences between the ambient background at 22 °C and the target at 300 °C at 2.4 µm and at 4.0 µm can be compared. At 4.0 µm, the spectral radiance ratio from the room temperature 22 °C background and a blackbody at 300 °C is
However, at a center wavelength of 2.4 µm, the Planck radiance ratio is much larger at
The radiance from the 22 °C ambient at 2.4 µm is lower than at 4.0 µm by a factor of 260 which will result in lower interfering radiances from the background at the shorter wavelengths. This simple analysis indicates that if sufficiently sensitive detectors with low noise and high responsivity can be developed, then such detectors will be less influenced by the background radiation while still being able to detect the radiation source. It is critical that the band-gap of the detectors cut-off is at or near 2.5 µm so that the sensor is less sensitive to the ambient background radiation. The Planck radiances also indicate that the dynamic range of the signals (the useful-signal and background-signal ratios) for a source at 300 °C and at 1000 °C will be greater at 2.4 µm than at 4.0 µm, thus leading to greater contrast ratios between the objects.
3. SWIR detectivity or D*
The relative performances of detectors can be compared independently from the detector element sizes using the criterion of D * or detectivity (sometimes called Jones) with units of cm Hz1/2/W. The D * of the detector can be increased by decreasing the noise-equivalent power (NEP). A decrease in the NEP can be achieved by increasing the shunt resistance, RS, of the detector which will decrease the noise at the output of the photodiode trans-impedance amplifier. The RS can be increased by several techniques. Since the shunt resistance is typically due to the areal density of the electrical imperfections of the detector, it depends on the surface area of the detector. A detector with a smaller surface area will have higher shunt resistance resulting in a lower NEP and a higher D *.
The shunt resistances of semiconductor photodiodes also depend on the temperature of the diode. The shunt resistances of 3 mm and 1 mm diameter ex-InGaAs and short-wave MCT (2.8 µm bandgap) detectors were measured as a function of diode temperature. The temperatures were varied using built-in 4-stage thermoelectric coolers with thermistor feedback controls. The shunt resistances shown in Fig. 2 were determined from the inverse slope of the current-voltage (I–V) curves. All the detectors are found to have a semilogarithmic dependence on diode temperature. These results indicate that the shunt impedances for the ex-InGaAs detector and the sw-MCT can be increased by a factor >1000, resulting in 3×106 Ω with the 3 mm diameter detector cooled to -85 °C (188 K). The 3 mm diameter diode has a lower shunt resistance than the 1 mm diameter diode which is inversely proportional to the area of the diode. This indicates that the defects which limit the shunt resistance are uniformly distributed across the diode area.
The D* or the detectivity of various detectors at the peak wavelength of their spectral responsivities are shown in Fig. 3 along with the background-limited performance (BLIP) at 295 K. BLIP limits the performances of detectors operating in typical laboratory environments since the fluctuations in the arrival rate of photons from the ambient background will limit the noise floor. One can observe that much higher detectivities, without being limited by BLIP, are possible as the detection wavelengths are made shorter than the peak of the 295 K Planck radiances at 10 µm. With cooling of the ex-InGaAs diode, the measured D* is >3×1013 cm Hz1/2/W, which is almost a factor of 100 greater than that of liquid-nitrogen cooled InSb detector at ~5×1011 cm Hz1/2/W. These detectivities at long wavelengths would be unusable in ambient environments because of the limitations of BLIP, but at the shorter wavelengths, the reduction in BLIP allows even higher D*.
4. Spectral responsivity and atmospheric transmission
The advantages of higher detectivities for these detectors cannot be utilized if the spectral responsivities of the detectors are in regions of atmospheric absorption bands. The spectral irradiance responsivity of a 2-stage cooled ex-InGaAs photodiode is shown in Fig. 4. The peak of the responsivity can be tuned by selecting the material composition such that the band-gap energy of the InGaAs or HgCdTe diodes can be made to occur at a range of wavelengths from 1.7 µm to 2.6 µm. Since these are direct band-gap materials, the quantum efficiency of the photon-to-electron conversion is also high.
In Fig. 5, the atmospheric transmission calculated using high-resolution transmission molecular absorption database (HITRAN)  as a function of wavelength is plotted showing that such spectrally selected radiometers can operate in one of the atmospherically clear transmission regions. Although at longer path lengths, the atmospheric transmittance in Fig. 5 will change due to environmental conditions and further attenuation, and the atmospheric window between 2.0 µm to 2.5 µm become narrower, much of the atmospheric window will remain even at much longer distances. To avoid fluctuations due to atmospheric changes in the transmission or absorption, ex-InGaAs radiometers can be constructed with cut-on filters to further restrict the spectral region from 2.0 µm to 2.5 µm.
5. Detector signals from ambient-temperature objects
The results of initial feasibility studies of using SWIR detectors for measurements of ambient temperature objects are shown in Table 1. A 3 mm diameter ex-InGaAs detector was
TE cooled to -85 °C and the photocurrents were converted to voltages using a current-to-voltage preamplifier at a gain of 107 V/A. Various objects with differing temperatures and emissivities were placed to fill the field-of-view of the detector, and thus the room temperature background did not substantially affect the measurements. Although the detector was placed over the mouth of a liquid-nitrogen (LN2) dewar to reduce the ambient background signal, the field-of-view could not be completely filled with a 77 K environment. The signals from the various objects are listed in Column 2 of Table 1. The lower temperature background of the LN2 dewar results in lower signals, and the low emissivity of the aluminum foil also leads to low signals. The difference in the signals with the changing temperature and the emissivity of the objects at room temperature and human body temperatures were observed. These results clearly indicate that SWIR detectors have the sensitivity to measure objects at ambient temperatures.
6. Use of traditional glass refractive optics
One of the greatest advantages of operating in this atmospheric window is that unlike the traditional mid- or thermal-infrared regions where the choices of refractive lens materials are limited, many optical glasses transmit to 2.5 µm. The transmittances of 10 mm thick BK7 and SF5 glasses , which are used in typical achromat lens combinations, are shown in Fig. 6.
These off-the-shelf achromats are used to collect the SWIR radiation. Since these glasses also transmit in the visible radiation, the optical alignment can be performed using visible radiation and then refocused to optimize the thermal infrared signal. Such lenses could be easily optimized for diffraction-limited optical performance as shown later by optical modeling using commercial software. In addition, if the detectors are designed so that the back substrate is thinned, then the spectral responsivities can be made to range from the visible to 2.5 µm. For such a detector, these optical glass lenses can be refocused to use a single lens system for visible to thermal measurements.
7. Imaging performance of traditional glass optics in the SWIR
Optical performances of lenses used in the visible wavelength region can be assessed for the SWIR by using optical modeling software. In Table 2, typical achromat specifications are
shown with the surface curvatures and thicknesses along with the glass types. The achromat specifications are for an off-the-shelf achromat which has been optimized for infinite conjugate in the visible wavelength region. For these achromats, if the optics has been optimized at shorter wavelengths, then the performance at longer wavelengths (with refocus to account for the chromatic focal shift) will be better than at the optimized wavelengths. This is partially due to the wavelength dependence of the diffraction limit. The specifications in Table 2 for a 200 mm focal length lens were used to model the optical performance shown in Table 3 at 2.45 µm. Table 3 shows the root-mean-squared radius as compared to the
diffraction-limited radius, indicating that the lens is diffraction-limited with close-to-optimum Strehl ratio of 0.9883 for an infinite conjugate operation at 2.45 µm. Although the optical performance will suffer if infinite conjugate optimized lenses are used in finite conjugate (1:1) imaging, custom lenses with computer-optimized curvatures and thicknesses can be fabricated if better performance is desired.
8. SWIR detector performance in radiation thermometers
To compare the performance of SWIR detectors against traditional cryogenically-cooled detectors, the SWIR detectors were constructed into experimental, prototype radiation thermometers. Since optical glasses transmit at these wavelengths, regular achromats were used as objective lenses in the prototype shown in Fig. 7. A variable-temperature blackbody set to room temperature and higher was used as the source of radiation. A 6 mm diameter target spot was imaged by the radiation thermometer at a distance of 50 cm between the source aperture and the front of the objective lens. The radiation was imaged onto a field stop with a chopper wheel placed close to the field stop. The radiation was collimated and focused onto the 3 mm diameter detector with achromats. The ex-InGaAs detector was cooled to - 85 °C with thermoelectric cooling. The Lyot stop  serves as the aperture stop behind the field stop. Due to the low scatter objective, the Lyot stop and the internal field-of-view limiter, this radiation thermometer design has out-of-field rejection which enables internal chopping. A novel chopper using a mirrored surface was used to have the TE-cooled diode view itself using the self-staring or “Narcissus” effect. The reference for the chopped radiation then becomes the radiation from the temperature-stabilized detector leading to long-term stability of the chopped signal.
The results of the measurements with the blackbody at 50 °C can be seen in Fig. 8. A noise-equivalent temperature difference (NETD) of <3 mK can be observed from the standard deviation of the temperature measurements taken with an electrical bandwidth of 0.16 Hz. A longer time sequence of temperatures showed changes due to the fluctuations from the heater controller of the blackbody which are not shown in the short time sequence in Fig. 8.
The NETD plotted as a function of the blackbody temperature is shown in Fig. 9. The NETD increases rapidly as the target temperature becomes comparable to the chopper-wheel temperature, but at the human body temperature of 36 °C, the NETD remains <10 mK. These results demonstrate that the NETD is comparable to that of InSb detectors for blackbody temperatures >25 °C without the need for cryogenic cooling.
9. Relationship between emissivity and temperature
For the temperature region up to about 1200 K, the relationship between the temperatures, radiance and emittance can be determined using the derivative of the Wien approximation. Since the spectral responsivity of the SWIR radiation thermometer peaks in a wavelength region where the Wien approximation to the Planck radiation law is applicable, the uncertainty of the temperature from the uncertainty of radiance is found from the derivative of the Wien approximation,
where L is the radiance, λ is the centroid wavelength, T is the temperature and c2 is the second radiation constant. The spectral emissivity of an object specifies the deviation of the thermal radiation from the object from a perfect Planckian radiator and is directly proportional to the radiance. Thus the relationship between the uncertainty in the temperature to an uncertainty in the emissivity is also given by Eq. 3, with the radiance, L, replaced by emittance, ε. Eq. 3 clearly shows that the uncertainty in the temperature to a fixed uncertainty in the radiance or the emissivity is linearly dependent on the wavelength. The operation in the SWIR compared with the mid-infrared wavelength will result in approximately a factor of 2 reduction in the temperature uncertainty with the same uncertainty in the emissivity.
10. Noise-equivalent irradiance
To test the limits of detection for irradiance, a blackbody at 300 °C was placed at the entrance to a 1.24 m focal length collimator as shown in Fig. 10. An aperture wheel was placed in front of the blackbody, and a chopper wheel (Chopper 1) was placed between the blackbody and the aperture wheel. From the geometry of the collimator, the calculated irradiance at the 40 mm diameter entrance pupil of the radiometer can be determined. Another alternatively used chopper wheel (Chopper 2) was placed inside the radiometer for “down-stream” chopping at the radiometer. The stray-light rejection due to the diffraction-limited performances of the optics enabled <1% difference in the signal between the “up-stream” and the “down-stream” modulated signals at the lowest measured irradiance shown in Table 4. The measurements of the noise-equivalent irradiance (NEI) with the setup in Fig. 10 are shown in Table 4.
11. Limitations and applications of SWIR sensors
One major issue of using the SWIR wavelength region for outdoor applications is the presence of substantial solar radiation at these wavelengths. For FLIR sensing applications, the reflected solar radiation from the target will need to be considered and mitigated for proper operation of the sensor. In other applications such as in surveillance systems, the reflected solar radiation during the daylight hours could be used to further enhance the object discrimination. Under moon-lit or clear-sky conditions, the radiation from the object will be a combination of the thermal emitted radiation and the reflected radiation from the night-glow or the moon. However, if SWIR detectors could be utilized in imaging applications, it is possible that a single focal-plane array could be used for both day-light and night-time applications.
A major advantage of using the SWIR radiometers is the transparency of glass in buildings, automobiles and other equipment at these wavelengths. If SWIR thermal imagers with low NETD could be developed at these wavelengths, then it would be possible to perform thermal imaging for human detection through the automotive glass and building windows. Such thermal imaging is not possible for the currently used long-wavelength infrared detectors. In addition, thermal imagers are used for diagnostics of electrical power switches where the operator must be protected from potentially lethal voltages with a protective cover. Often, these covers require special infrared transmitting glasses for thermal imaging in order to determine which switches are overheating and about to fail. Regular window glass covers could be used in these applications. There are also a large number of military applications, especially in target discrimination at long distances, where the traditional InSb and MCT detectors are not sensitive enough for accurate sensing.
We demonstrated that SWIR detectors with a 2.5 µm wavelength cut-off can be used as new generation of sensors for IR radiometers and radiation thermometers. Although the cut-on at 2.0 µm can be accomplished with a filter, the cut-off at around 2.5 µm must be intrinsic to the detector material from the fabrication process to avoid large background currents. These detectors can be used in a spectral wavelength region where the atmosphere has a clear window from 2.0 µm to 2.5 µm, leading to the use of refractive-glass optics without central obscurations. Furthermore, equivalent or better infrared detection capability than the traditionally used detectors can be achieved in a compact, low-maintenance design with the capabilities of measuring human body temperatures.
# Certain commercial equipment, instruments, or materials are identified in this paper to foster understanding. Such identification does not imply recommendation or endorsement by the National Institute of Standards and Technology, nor does it imply that the materials or equipment are necessarily the best available for the purpose.
References and links
1. A. Rogalski and K. Chrzanowski, “Infrared devices and technique,” Opto-Electron. Rev. 10, 111(2002).
2. D.G. Crowe, P.R. Norton, T. Limperis, and J. Mudar, “Detectors,” in Electro-Optical Components, W. D. Rogatto ed., Infrared Information Analysis Center, Michigan, 1993.
3. L.S. Rothman, et al., “The HITRAN 2004 molecular spectroscopic database,” J. Quantitative Spectroscopy and Radiative Transfer 96, 139–204 (2005). [CrossRef]
4. Schott glass designations.
5. H.W. Yoon, D.W. Allen, and R.D. Saunders, “Methods to reduce the size-of-source effect in radiometers,” Metrologia 42, 89–96 (2005). [CrossRef]