Open Access
Add to BibSonomy
Abstract
We report a resonant cavity infrared detector (RCID) with an InAsSb/InAs superlattice absorber with a thickness of only ≈ 100 nm, a 33-period GaAs/Al0.92Ga0.08As distributed Bragg reflector bottom mirror, and a Ge/SiO2/Ge top mirror. At a low bias voltage of 150 mV, the external quantum efficiency (EQE) reaches 58% at the resonance wavelength λres ≈ 4.6 µm, with linewidth δλ = 19-27 nm. The thermal background current for a realistic system scenario with f/4 optic that views a 300 K scene is estimated by integrating the photocurrent generated by background spanning the entire mid-IR spectral band (3-5 µm). The resulting specific detectivity is a factor of 3 lower than for a state-of-the-art broadband HgCdTe device at 300 K, where dark current dominates the noise. However, at 125 K where the suppression of background noise becomes critical, the estimated specific detectivity D* of 5.5 × 1012 cm Hz½/W is more than 3× higher. This occurs despite a non-optimal absorber cut-off that causes the EQE to decrease rapidly with decreasing temperature, e.g., to 33% at 125 K. The present RCID’s advantage over the broadband device depends critically on its low EQE at non-resonance wavelengths: ≤ 1% in the range 3.9-5.5 µm. Simulations using NRL MULTIBANDS indicate that impact ionization in the bottom contact and absorber layers dominates the dark current at near ambient temperatures. We expect future design modifications to substantially enhance D* throughout the investigated temperature range of 100-300 K.
© 2023 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Interband detectors operating in the midwave infrared (MWIR, defined here as λ = 3-5 µm) generate photocurrent when incident light is absorbed in a layer with bandgap smaller than the photon energy. For a conventional photovoltaic architecture that provides at most two passes through the device following possible reflection from a metallized contact, the absorber must be several microns thick to produce high external quantum efficiency (EQE). However, a resonant cavity infrared detector (RCID) can maintain high EQE in a narrow spectral bandwidth even when the absorber is very thin, by imposing multiple passes between top and bottom mirrors [1]. This can be advantageous if the light to be detected is from a laser rather than broadband thermal radiation, or when limited bandwidth is desired as in chemical sensing spectroscopy or hyperspectral imaging. To minimize noise currents associated with the broadband background, the RCID should not only have a narrow spectral bandwidth, but must also minimize the detection of wavelengths outside the narrow band of interest. The RCID architecture also provides enhanced frequency response, since photogenerated carriers can be collected more rapidly from a thin absorber.
The RCID concept has been developed extensively at shorter wavelengths [2,3], where resonant cavity photodiodes with enhanced frequency response are relatively mature. However, until recently no MWIR RCIDs had combined low dark current with high peak quantum efficiency. In 2003, a group at Imperial College reported an RCID with InAs absorber grown on a 10-period GaAs/AlGaAs distributed Bragg reflector (DBR) [4]. The cavity was completed by a CrAu top mirror that also provided electrical contact. The photocurrent was highest at T = 205 K, where λres ≈ 3.3 µm and the full-width at half maximum (FWHM) of the spectral bandwidth (δλ) was ≈ 150 nm under 3 V bias. No dark current (Jdark) or specific detectivity (D*) was reported. In the period 2004-2012, groups at U. Linz. [5], ETH Zurich [6], and MIT [7] reported the incorporation of various lead-salt absorbers into RCID cavities defined by combining grown lead-salt bottom mirrors with metal [5,6] or lead-salt top mirrors [7]. Peak wavelengths ranged from 3.45 µm to 8.4 µm, with resonance linewidths typically 36-90 nm. ETH Zurich tuned the resonance wavelength by shifting the position of an external top metal mirror, either piezoelectrically [8] or with a microelectromechanical system (MEMS) [9]. EQEs up to 90% were reported for the lead-salt RCIDs, although accompanied by relatively low resistance-area products (R0A). Consequently, D* was always at least 2 orders of magnitude lower than the Rule ‘07 value [10] for a state-of-the-art (as of 2007) broadband HgCdTe photodiode with cut-off equal to the resonance wavelength (λres) at the same operating temperature.
In 2017, U. Rochester and Amethyst Research reported an RCID with 100-nm-thick bulk InAs absorber surrounded by grown top and bottom GaAsSb/AlAsSb DBRs [11]. The resonance wavelength was λres = 2.91 µm, with spectral linewidth δλ ≈ 45 nm. Although the dark current density of 0.4 mA/cm2 at 298 K was slightly below Rule ‘07, no EQE or D* was reported.
Two years later, NRL reported the first RCID to exhibit high performance at a mid-IR wavelength beyond 3 µm [12]. At λres = 4.0 µm, the device with grown GaSb/AlAsSb bottom DBR, silver top mirror, and absorber thickness only 50 nm attained external quantum efficiency EQE = 34%, with linewidth δλ = 46 nm and D* = 7 × 109 cm Hz½/W at room temperature. If we define ΓD* = D*RCID/D*Rule07, where the denominator is obtained by combining Rule ‘07 dark current with EQE = 85%, the value for this device was ΓD* ≈ 0.18. Also in 2019, a team led by U. Lancaster reported an RCID with 96-nm-thick InAsSb absorber that displayed 62% EQE at 300 K for λres ≈ 3.7 µm and δλ = 42 nm [13]. At 250 K, the detectivity D* = 8 × 1010 cm Hz½/W yields ΓD* ≈ 0.4. U. Lancaster also reported an RCID with InAs-InAsSb superlattice absorber that operated with λres ≈ 4.4 µm, δλ = 50 nm, 80% EQE, and D* = 2.5 × 1010 cm Hz½/W at 240 K (ΓD* ≈ 0.2) [14]. The same group demonstrated RCIDs with λres ≈ 7.7 µm, although at 77-140 K the ΓD* values were < 0.01 [15]. In 2021, they reported an RCID with bulk InAsSb absorber, λres ≈ 3.7 µm, and δλ ≈ 40 nm that attained ΓD* ≈ 0.4-0.5 at temperatures from 200 to 300 K [16]. They also reported non-dynamic wavelength tuning, from λres ≈ 1.98 to 2.08 µm, by fabricating a series of cavity thicknesses on the same chip [17].
University of Texas recently employed an alternative approach to forming the resonant cavity, in which a high-contrast grating etched into the top of the semiconductor epilayer couples IR input into in-plane waveguide modes [18]. Light is confined to the waveguide by a heavily-doped semiconductor layer grown below the absorber. Devices with λres ≈ 4.4 µm and 4.7 µm yielded up to 60% EQE with ≈ 60 nm linewidth. The detectivity was as high as ΓD* ≈ 0.5 at room temperature, although riding on a large non-resonant background and for only one polarization of the incoming light relative to the grating orientation. An LWIR RCID with λres ≈ 10.8 µm was also demonstrated [19].
Apart from the U. Texas devices that did not employ DBRs, all of the MWIR RCIDs discussed above employed bottom DBRs and absorbers that were grown or deposited as a single structure. For the higher-performance RCIDs grown by III-V molecular beam epitaxy (MBE) on GaSb or InAs substrates [11–17], this requires the challenging growth of thick GaSb/AlAsSb or GaAsSb/AlAsSb DBRs while maintaining precise control over each layer thickness. The roughly linear scaling with wavelength requires a bottom mirror thickness of 14 µm for the device with λres ≈ 7.7 µm [15].
The present investigation has considerably simplified the GaSb-based growth, by heterogeneously bonding a relatively-thin nBn detector structure to a commercially-purchased GaAs/Al0.92Ga0.08As DBR with reflectivity > 99%. This approach can potentially combine a higher cavity quality factor (Q) with higher growth and processing yield. For high frequency response, a further advantage is much lower capacitance of the semi-insulating GaAs compared to GaSb or InAs, for which no semi-insulating substrates are available.
Below we report fabrication and characterization results for an RCID fabricated in this manner with λres ≈ 4.6 µm, which operates at a bias voltage of only 150 mV and combines EQE nearly as high with resonance linewidth slightly narrower than any reported previously in the MWIR. As has been the case for all the previously-reported RCIDs discussed above (ΓD* ≤ 0.01 for lead-salt devices and ≤ 0.5 for III-V devices), our RCID displays lower specific detectivity based on dark current density alone (ΓD* ≤ 0.32) than a state-of-the-art HgCdTe detector. However, this comparison neglects the RCID’s primary advantage, namely its suppression of background current at wavelengths outside the narrow band of interest. To quantify this advantage, we calculate D* for a realistic cooled system in which an f/4 optic views a room-temperature scene. The noise current induced by background radiation is determined by integrating over all MWIR wavelengths spanning 3-5 µm. The RCID’s D* is then compared to that for a state-of-the-art broadband HgCdTe detector with Rule ‘07 dark current density and EQE = 85% at all MWIR wavelengths. It will be seen that while the present RCID has lower D* at ambient temperature where dark current determines the noise, it becomes advantageous at lower temperatures where background photocurrent strongly dominates. We also find that while narrow linewidth and high EQE at the resonance peak are important figures of merit, it is equally important to minimize the background photocurrent at other MWIR wavelengths outside the resonance band. Our device is very effective in this regard, since EQE ≤ 1% at all non-resonance wavelengths within ± 700 nm of λres. The non-resonance EQEs of all previously-demonstrated MWIR RCIDs are more typically ≥ 5% at λ < λres, which would degrade the maximum attainable D*.
We also present simulations of the band structure under bias, electron and hole density profiles, and thermal generation rate profiles in the present nBn structure. The modeling indicates that redesign of the structure should lead to significant further reduction of the dark current and enhancement of D* at both high and low temperatures.
2. Design, MBE growth, and material characterization
The nBn detector structure was grown on a 2” heavily-n-doped GaSb substrate in a Riber Compact 21 T molecular beam epitaxy (MBE) system [20]. To begin, the native oxide layer was removed in situ by heating the substrate slowly under cracked antimony flux to 550°C and dwelling for 15 minutes. This was followed by cool down to 495°C, where a 500-nm-thick GaSb buffer layer was grown at 1 ML/s. Next the wafer was cooled to 450°C and a lattice-matched InAs0.91Sb0.09 etch stop layer and n-type GaSb sacrificial layer were grown. These layers aided the eventual removal of the GaSb substrate after bonding to the GaAs/Al0.92Ga0.08As bottom DBR. The nBn detector structure, which was also grown at ∼ 450°C, consisted of a 300-nm-thick n-type InAs/InAs0.6Sb0.4 superlattice top contact layer, 200-nm-thick unintentionally-doped InAs/AlSb superlattice barrier layer, unintentionally-doped InAs/InAs0.6Sb0.4 superlattice absorber layer with 20 periods and total thickness 103 nm, 250-nm-thick n-type InAs/InAs0.6Sb0.4 superlattice bottom contact layer, and 20-nm-thick n + -InAs bottom contact capping layer (see the band diagram in Fig. 14 below). Prior to growth of the device wafer, high-resolution x-ray diffraction and photoluminescence (PL) measurements were performed on separately-grown samples with thicker (≈ 0.5 µm) absorbers, in order to calibrate the lattice matching, measure the carrier lifetime, calibrate the detector cut-off wavelength, and assess the surface morphology. Figure 1 shows the high-resolution θ-2θ x-ray scan for the wafer from which the devices were processed.
Fig. 1. High resolution θ-2θ x-ray diffraction scan for the wafer from which the devices were processed (BC = bottom contact layer / absorber layers, B = barrier layer, TC = top contact layer).
The minority-carrier lifetime in a PL structure with similar absorber superlattice was determined vs. photoexcited carrier concentration by measuring the optical response roll-off as a function of the modulation frequency of a 1064 nm pump laser. Figure 2 shows that the recombination lifetime at low injected concentrations was ≈ 300 ns at 300 K and 0.7-1 µs at 150-250 K. This contrasts a much shorter lifetime of ≈ 50 ns in the Ga-containing “W” structure grown for NRL’s earlier RCID demonstration [12].
Fig. 2. Lifetime vs. photoexcited carrier concentration at a series of temperatures in a PL structure with similar absorber superlattice, as determined by measuring the optical response roll-off as a function of the modulation frequency of a 1064 nm pump laser.
To evaluate the underlying nBn detector performance, devices without top and bottom mirrors were first processed from the same wafer material. Optical lithography with a two-step wet etch [21] was used to fabricate circular mesas using processing that was described previously [12]. At the RCID operating biases of -150 mV, a device with mesa diameter 200 µm had dark current densities of 0.29 A/cm2 at 300 K, 0.012 A/cm2 at 250 K, 1.7 × 10−4 A/cm2 at 200 K, and 6.1 × 10−7 A/cm2 at 150 K.
Figure 3 shows EQE spectra at a series of temperatures for one of these detectors. The spectrum at 300 K indicates a cut-off wavelength only slightly longer than the target of ≈ 5.1 µm. This EQE was obtained for a mesa with fully-metallized top contact and illumination through the n-doped GaSb substrate, which was not anti-reflection coated. While a full comparison of the theoretical and experimental absorption spectra would require an accurate characterization of the metal reflectivity and substrate absorption, the shape and magnitude of the theoretical absorbance per pass at 300 K from an 8-band k·p calculation (dashed orange curve) are similar to those of the measured EQE.
Fig. 3. External quantum efficiency spectra at a series of temperatures for a non-resonant detector processed from the same wafer material as the RCID. The dashed orange curve shows the theoretical absorbance per pass at 300 K from an 8-band k·p calculation.
3. RCID fabrication
To process RCID devices, an 11 mm × 11 mm chip from the nBn detector wafer was bonded to a commercially-purchased GaAs/Al0.92Ga0.08As mirror that was grown on a semi-insulating GaAs substrate. A total of 33 DBR periods was obtained by bonding two mirror segments together, which provided > 99% reflectance in the 4.4-4.7 µm spectral band. Bonding of the GaSb-based chip began with deposition of a 100Å-thick interfacial dielectric layer at the bond interface, followed by plasma activation and low-temperature bonding below 250°C. After bonding, the GaSb substrate was removed by a chromic-acid-based etch that stopped on the InAsSb etch stop layer. The etch stop and sacrificial GaSb layers were then removed by chemical etching before the detectors were fabricated by lithographic processing.
The first step in fabricating the reticulated shallow etched mesa isolation (RSEMI) detector structures [21] was to define an alignment layer. Then the top ring contact metallization was deposited, followed by a shallow mesa etch that stopped just below the top contact layer. Next a second deep etch proceeded to the bottom contact layer, followed by deposition by plasma enhanced chemical vapor deposition (PECVD) of an SiO2 passivation layer on the sidewalls. A further etch proceeded through the bottom contact, stopping at the top of the semi-insulating GaAs/Al0.92Ga0.08As mirror. This was followed by another SiO2 deposition. Top and bottom contact openings were defined through the SiO2, followed by the deposition of Ti/Pt/Au contact metal and ground signal ground (GSG) pads to enable high speed operation. Devices with active diameters ranging from 15 to 100 µm were fabricated in a single run.
To shift the cavity length and λres, Ge spacers of varying thickness were deposited on top of the nBn detector structure. Then quarter-wave SiO2 and Ge layers were deposited to complete the Ge/SiO2/Ge top mirror. Figure 4 schematically illustrates side and top views of the full structure for a single device.
Fig. 4. Schematic of the RCID layer structure and patterned device geometry, from the side (lower) and top (upper).
The planned Ge thicknesses of 148, 153, and 158 nm were intended to tune the resonance wavelength from 4.60 to 4.62 µm. However, unexpected variation of the deposition thickness led to a much greater shift of λres. The results below were obtained for a device with mesa diameter 93 µm and nominal Ge thickness 158 nm, for which the resonance wavelength was close to the projected value of 4.62 µm. However, the devices with 148 and 153 nm spacer thicknesses had shorter-than-expected resonance wavelengths of 4.56 and 4.59 µm, respectively. The maximum EQE also decreased with decreasing spacer thickness, probably due to higher absorption loss in the bottom superlattice contact layer at shorter wavelengths.
4. Experimental results – J-V characteristics
The device was mounted on a 68-pin LCC and loaded into a liquid nitrogen flow dewar with an internal cold shield held at T ≈ 78 K. The dark J-V characteristics at temperatures between 100 K and 300 K were taken with an HP4145A parameter analyzer.
Figure 5 shows the dark current density (Jd) vs. voltage at a series of temperatures for the RCID device with absorber thickness 103 nm, mesa diameter 93 µm, and Ge spacer thickness 158 nm. The vertical dashed line indicates the bias voltage Vb = -150 mV at which the EQE dependence on voltage is nearly saturated. At that voltage, the dark current densities are 1.3-3× higher than for the detectors processed without mirrors.
Fig. 5. Dark current density vs. voltage at a series of temperatures for the present RCID device with absorber thickness 103 nm, mesa diameter 93 µm, and top Ge spacer thickness 158 nm. The vertical dashed line represents the bias voltage of -150 mV employed in the detection experiments.
Figure 6 plots the dark current density, scaled by T-2.5, vs. inverse temperature for the RCID operating at Vb = -150 mV. The temperature scaling accounts for the carrier densities of states that are approximately 3D for electrons and 2D for holes. The data fit a single slope (dashed line) over the entire temperature range from 125 to 300 K. The corresponding activation energy of ≈ 217 meV is a little smaller than the nominal bandgap of ≈ 250-275 meV deduced from the EQE spectra in Fig. 3. Figure 5 plots the resistance-area product (RA = dV/dJ) vs. inverse temperature at the operating bias, which will enter into the evaluation of D* in Eq. (1) below. All of the devices with different Ge spacer layer thicknesses displayed similar J-V characteristics.
Fig. 6. Dark current density scaled by T-2.5 vs. inverse temperature for the RCID operating at a fixed reverse bias of -150 mV. The fit to a linear dependence indicates an activation energy of ≈ 217 meV.
5. Experimental results – external quantum efficiency and D*
The blue curve in Fig. 8 illustrates the transmissivity spectrum of the nBn detector chip bonded to the GaAs/Al0.92Ga0.08As bottom DBR, following removal of the GaSb substrate but before deposition of the top mirror. As expected, the transmission is negligible within the region λ = 4.4-4.7 µm where the DBR’s reflectivity exceeds 99%. The reflectivity of the fully-processed RCID is given by the red curve. The sharp minimum at λ = 4.618 µm coincides with the cavity resonance, as will be seen below.
Fig. 8. Transmissivity spectrum of the nBn detector chip bonded to the GaAs/Al0.92Ga0.08As DBR, following removal of the GaSb substrate but before deposition of the top mirror (blue curve), and reflectivity spectrum of the fully-processed RCID (red curve).
Optical measurements were taken in the same dewar as the J-V characterization, but with an aperture replacing a blank in the cold shield. The responsivity spectrum was measured with an IS50 Fourier Transform IR (FTIR) spectrometer. The IR source beam was focused to a diameter of ≈ 80 µm centered on the RCID mesa, and the RCID’s photocurrent fed back into the FTIR to obtain a raw spectrum. The source spectrum was then measured at the position of the RCID with a DTGS detector corrected for its frequency response. The RCID responsivity R(λ) is proportional to the raw spectrum divided by the source spectrum, and the EQE spectrum is proportional to R(λ)/λ.
In order to normalize the EQE spectrum, the absolute EQE at resonance was measured using a black body source at 1300 K, with a filter centered near the RCID resonance frequency. The black body output was chopped at 37 Hz, and the photocurrent determined with a lock-in amplifier. The optical power incident at the RCID position was measured with a calibrated pyroelectric detector. The beam was focused to underfill the RCID mesa, and the beam diameter was varied in the range ≈ 15-64 µm. The detector response was determined from the slope of the photocurrent vs. area, and the absolute EQE finally calculated from the detector response and previously-determined spectrum.
Figure 9 illustrates the resulting EQE spectra for the RCID at T = 125 K (green) and 300 K (magenta), for wavelengths spanning the entire MWIR spectral band. The inset shows the same data on a semi-log scale, to emphasize that the EQE remains quite small at off-resonance wavelengths spanning 3.8 µm < λ < 5.5 µm.
Fig. 9. External quantum efficiency at T = 125 K (green) and 300 K (magenta), as measured with the FTIR over a broad spectral band. The insert shows the same data on a semi-log scale. For wavelengths near resonance, the detector responsivity is ≈ 3.7 × EQE, which corresponds to a peak value of ≈ 2 A/W at 300 K.
The magenta curve in Fig. 10 expands the EQE spectrum at 300 K near the resonance wavelength of 4.618 µm. The FWHM linewidth of 27 nm agrees well with the simulated value 26 nm, although the peak EQE of 58% is a little smaller than the theoretical projection of ≈ 70%.
Fig. 10. EQE at 300 K over a much narrower spectral bandwidth near the resonance wavelength, as measured by three different experiments. The magenta curve was obtained by the FTIR as in Fig. 9, whereas the points were derived from measurements of the photocurrent excited by a quantum cascade laser with calibrated intensity at NRL (magenta) and Intraband (blue).
The EQE spectrum for the same device was also quantified over a limited range of wavelengths by measurements at both NRL and Intraband of the photocurrents induced by quantum cascade lasers (QCLs) with known incident powers. The experiment at NRL employed a focused QCL with center wavelength 4.7 µm and beam size 2.3 × 3.5 mm. At a drive current of 1 A, temperature tuning produced three discrete wavelengths of 4.613 (15°C), 4.628 (18°C), and 4.641 µm (21°C). The beam with nominal power < 10 mW passed first through a Lasnix metal grid step attenuator (ND filter), and then through an Andover bandpass filter with bandwidth 150 nm centered on 4.58 µm. A 55/45 beam splitter diverted part of the beam to a power meter, which provided in-situ monitoring of any optical feedback effects on the QCL. Transmission through the beam splitter next passed through optics that expanded the beam by 3× and then focused it to a nominal spot size of < 100 µm, which was small enough to underfill the RCID diameter. The RCID chip was placed in a Senseeker dewar test unit (DTU) with silicon window, although the dewar was operated at 300 K for all the measurements. The RCID signal observed under -150 mV bias was sent through a current preamplifier and lock-in amplifier modulated at 200 Hz. The results, shown as the magenta points in Fig. 10, are nominally consistent with those obtained from the FTIR characterization.
The experiment at Intraband employed a single-mode distributed feedback (DFB) QCL fabricated at U. Wisconsin - Madison [22]. Its wavelength could be varied from 4.6035 to 4.6244 µm through a combination of current and temperature tuning. An uncoated glass slide with strong absorption in the MWIR attenuated the laser power from its nominal output of 20 mW to ≈ 400 µW. This was monitored via reflection from the glass slide to a power meter. An MWIR-coated beam splitter directed half of the power to an FTIR for wavelength monitoring. A CaF2 lens with estimated transmission 99% and focal length 40 mm collimated the beam for coupling to the RCID. Any optical feedback effect on the QCL output could be monitored by the in-situ power and wavelength measurements. None were observed in this case, likely due to the strong attenuation by the glass slide. The input signal was incident through the 88-µm-diameter opening of the RCID’s top annular contact. The EQE at each wavelength was determined by subtracting the dark current of ≈ 50 µA from the photocurrent measured at -800 mV operating bias, and then dividing by the photon density corresponding to the calibrated excitation power. The resulting EQEs are shown as the blue points in Fig. 10. The slightly-longer peak wavelength (4.621 µm) compared to that from the FTIR characterization (4.618 µm) may be due to uncertainty in the temperature calibrations of the two experiments. The peak EQE of 71%, as compared to 58% from the FTIR experiment, may be attributed in part to the use of a larger bias voltage of -800 mV, as compared to -150 mV in both experiments performed at NRL. At 250 K, the photocurrent measured at NRL was 7% higher at -800 mV than at -150 mV. Other potential sources of discrepancy include interference with scattered light in the FTIR experiment (the probable source of the oscillations at shorter wavelengths in Fig. 9), and accumulated small errors that could total ±10% in the QCL experiment. The largest uncertainty in the simulation is parasitic absorption in the superlattice top contact and dielectric top mirror layers.
Figure 11 illustrates the EQE spectra near resonance at a series of temperatures in the range 100-300 K, as measured by the FTIR experiment. The resonance wavelength is seen to decrease gradually with temperature, as expected due to the decreasing refractive index of the III-V semiconductor materials in the cavity. Also apparent is a substantial decrease of the maximum EQE with decreasing temperature below 200 K. This may be due primarily to weaker absorption in the active quantum wells as the cut-off wavelength becomes shorter than λres (see Fig. 3). In future designs, this can be addressed by either a longer cut-off/shorter resonance wavelength (to increase the low-temperature absorption at the cavity resonance) or else by a higher cavity Q (more passes through the absorber induced by a more reflective top mirror, to compensate for weaker absorption per pass). The lower EQE naturally reduces the maximum D* that is attainable at low T.
Fig. 11. EQE spectra near resonance, as determined by the FTIR characterization at a series of temperatures between 100 K and 300 K. All measurements were performed at a fixed bias voltage of -150 mV.
The specific detectivity may be determined from the expression:
A primary advantage of the RCID approach is its potential to strongly suppress thermal background noise at wavelengths outside the narrow spectral resonance. However, this advantage is only realized when EQE outside the resonance bandwidth remains quite small. Otherwise, background photocurrent can still easily dominate the net noise current. To illustrate, we consider the present RCID’s potential to suppress background noise in a realistic scenario.
For a given background temperature, T, the differential power per unit area and solid angle, Lint(λ,T), received by the RCID within a given differential wavelength band is [23]:
The net background current density is finally obtained by integrating the differential contribution over all incident wavelengths:
Having measured the present device’s EQE spectra at a series of operating temperatures (TD), we can perform this integral for any assumed background temperature (Tb) by summing over the fine mesh of wavelengths provided by the FTIR (resolution ≈ 1 nm). We can also determine how much of Jb is due to photoexcitation within the resonance linewidth vs. how much is parasitic photocurrent generated at wavelengths outside this band. Combining with the measured J-V characteristics, we can then use Eq. (1) to determine D*(Tb, TD) at the series of RCID operating temperatures for which EQE data are available.
We will compare with the corresponding D* expected for a state-of-the-art HgCdTe detector operating at the same TD, assuming EQE = 85% at all wavelengths and dark current given by Rule ‘07 when the cut-off wavelength is 5 µm. In both cases, we integrate Eq. (4) over the entire MWIR spectral band (λ1 = 3 µm, λ2 = 5 µm). To simulate a realistic system scenario, we assume that the detector is housed in a dewar with cold shielding, such that most incident angles view a background at the detector operating temperature of TD. Only the input optic exposes the detector to background radiation at the scene temperature Ts, which is assumed to be 300 K. In particular, we will consider an f/4 optic, for which only F ≈ 0.6% of the background is at Ts while 99.4% is at TD.
Figure 12 plots the ratio of background photocurrent, as derived from Eq. (4), to total noise current corresponding to all three contributions in the denominator of Eq. (1). This ratio is plotted as a function of temperature for the broadband device (red points) and an RCID with the characteristics shown in Figs. 5–11 (blue points). The background photocurrent is relatively unimportant at higher temperatures where the dark current in both devices is large. However, Jd decreases by orders of magnitude with decreasing temperature, so the background photocurrent from the scene comes to dominate even though the f/4 optic exposes only the small fraction F of the field of view to the scene. The temperature at which this crossover occurs varies considerably with device type, and whether the f/4 optic is employed (filled points) or the entire field of view (F = 1) is exposed to background at the scene temperature (no optic, open points). Because the RCID strongly suppresses background irradiation outside the narrow resonance bandwidth, its background photocurrent doesn’t become dominant until a lower TD. When the detector’s entire field of view observes the background, the associated noise naturally becomes dominant sooner (at a higher temperature). It follows that the RCID’s strong suppression of background noise is even more advantageous in that case. In the opposite limit of an optic with larger f/#, for which a smaller angular fraction of the input is exposed to the scene background, the cross-over to background-dominance shifts to lower temperature.
Fig. 12. Ratio of background photocurrent to total noise current for a state-of-the-art broadband HgCdTe device vs. an RCID with the EQE(λ,TD) and Jd(TD) characteristics shown in Figs. 5–11 (blue points). The filled points assume an f/4 optic that exposes only F ≈ 0.6% of the field of view to background at the scene temperature, whereas the open points assume the entire field of view observes background at Ts (F = 1, no optic).
In Fig. 13, D* vs. detector operating temperature from Eq. (1), assuming the f/4 optic, for the broadband HgCdTe device (red points) is compared to that of the RCID (blue points). D* for the RCID is lower at TD ≥ 200 K due to its higher (non-optimized) dark current. In this limit the RCID’s substantial suppression of background photocurrent does not play a significant role.
Fig. 13. Specific detectivity vs. detector operating temperature TD for the scenario with an f/4 optic. The RCID performance (blue points), based on the measured dark currents and EQE spectra at each temperature shown in Figs. 5–11, is compared to that for the state-of-the-art broadband HgCdTe device (red points).
However, we saw in Fig. 12 that with decreasing temperature the background photocurrent eventually comes to dominate (e.g., by orders of magnitude in the broadband device at TD < 150 K). At that point D* no longer increases with decreasing TD. But since background current is strongly suppressed in the RCID, its D* continues to increase to lower TD below 150 K. Only at 125 K does the scene background become strong enough that D* no longer increases. In fact, it decreases at TD = 100 K due to the rapid fall-off of peak EQE seen in Fig. 11. Nonetheless, despite the RCID’s much lower EQE peaks of 43% at 150 K, 33% at 125 K, and 23% at 100 K, D* for the RCID exceeds that of the broadband device at all TD ≤ 150 K for which the RCID was characterized. At 125 K, the estimated D* of ≈ 5.5 × 1012 cm Hz½/W is 3.3× higher.
We saw in Fig. 12 that the RCID’s delay of background dominance becomes more pronounced when a larger fraction of the field of view observes the background (smaller f/# or no optic). The RCID’s D* will then become higher at a higher TD, and by a larger maximum factor with decreasing temperature. Of course, the details of this analysis change if the scene is hotter or colder than 300 K. We also note that for both devices in this system scenario, the cold-shielded background at TD never plays a significant role.
Although it would be beneficial in such scenarios to insert a narrow-bandpass filter into the broadband detector’s field of view, the filter itself may contribute substantial background unless it is cooled. For this reason cold filters, and even cold filter wheels, are sometimes employed in conjunction with broadband detectors and imagers [24]. However, for MWIR wavelengths beyond 4.5 µm the minimum bandwidth for such a filter is typically at least ≈ 100 nm [25]. Thus besides introducing 10-20% transmission loss, the background current generated near the wavelength of interest is 5× higher than for an RCID with δλ ≈ 20 nm (and narrower RCID linewidths are anticipated in the future). For applications requiring rapid tuning of the central wavelength, a mechanical cold filter wheel will be much slower and less flexible than an RCID with MEMS or piezoelectric dynamic tuning capability that will be discussed briefly in Section 7 below.
The RCID’s higher D* at low temperatures depends critically on its strong suppression of the scene background at non-resonant wavelengths. While the background flux density at 300 K increases substantially with increasing wavelength, the contribution to Jb from λ > λres + 3δλ is only ≈ 1% of the total because EQE(λ) is quite small in that region as we see in Fig. 9. The shorter-wavelength range λ < λres - 3δλ contributes 24% of the total Jb, while the remaining 75% is due to background near the resonance (λres - 3δλ < λ < λres + 3λ). Despite the RCID’s higher EQE at λ < 3.5 µm, black body emission is weak in that range so the shorter-wavelength contribution comes mostly from wavelengths in the 4.2-4.5 µm range where the shoulder of the EQE’s resonance peak is somewhat softer. It will be especially beneficial if future designs with higher cavity Q can produce a narrower δλ and even smaller off-resonance EQE(λ).
It was noted in the Introduction that some of the previous RCID demonstrations have produced similar EQEs, δλ almost as narrow, and slightly higher D* relative to Rule ‘07 when background currents are ignored. For example, ΓD* ≈ 0.48 and δλ = 25 nm for λres = 3.7 µm at 200 K in Ref. [16], and ΓD* ≈ 0.46 and δλ ≈ 50 nm for λres = 4.4 µm at 300 K (TE polarization) in Ref. [18], as compared to ΓD* ≈ 0.32 and δλ = 19 nm for λres = 4.6 µm at 150 K with the present device. However, all of the non-resonant EQEs reported for earlier devices were much larger than those illustrated in Fig. 9. The resonance shoulders have typically been at the 5-10% level at shorter wavelengths and ≥ 3% at longer wavelengths, or much higher, as compared to ≈ 1% and < 0.5%, respectively, for the present RCID. Although for those devices we do not have the detailed data required to repeat the D* analysis discussed above, it appears their noise photocurrents induced by the scene background at non-resonance wavelengths would have much greater effect on the resulting D*.
6. nBn performance simulation
To investigate the future prospects for increasing the quantum efficiency and reducing the dark current, we have simulated various aspects of the RCID performance. This was done using NRL MULTIBANDS [26], a software package that was co-developed by some of the present authors and is based on physical concepts and mathematical formalisms described in Ref. [27].
Figure 14 illustrates the calculated conduction (blue) and valence (red) band profiles, and respective Fermi levels (dashed curves), for the present nBn detector structure at 300 K under -150 mV bias.
Fig. 14. Simulated conduction (blue) and valence (red) band profiles in the nBn detector structure at 300 K when operated at -150 mV bias. The respective Fermi levels are shown as dashed curves.
Figure 15 plots the corresponding profiles for electron (blue) and hole (red) carrier concentrations in the structure without illumination. The smooth variation for holes is consistent with unimpeded collection of the minority carriers, whereas the barrier clearly blocks electron transport from the top contact. The applied field depletes electrons from the absorber region closest to the barrier. However, spillover from the doped bottom contact layer increases the density at the opposite end of the absorber, to a value much higher than would normally be present in an isolated unintentionally-doped InAs/InAsSb superlattice. Similarly, spillover from the heavily-doped n + -InAs capping layer increases the electron concentration in the adjacent region of the bottom contact.
Fig. 15. Simulated electron (blue) and hole (red) concentrations vs. position in the nBn detector structure of Fig. 14 without illumination.
The Auger decay rate rAug corresponding to the carrier concentrations at a given position is:
Here Cn and Cp are the Auger coefficients for events involving two electrons and one hole vs. those involving two holes and one electron, respectively, while n0 and p0 are the electron and hole populations in thermal equilibrium. Since the hole concentration is quite small everywhere in the present structure, we can ignore the term proportional to Cp. By fitting the observed dark current density in test detector structures with the same absorber, assuming a Shockley-Read-Hall (SRH) lifetime of 1 µs in the absorber, and using the wavelength dependence for Cn in Ref. [28], Cn is ≈ 5 × 10−27 cm6/s in the wider-gap bottom contact layer and ≈ 8 × 10−27 cm6/s in the absorber layer at 300 K. These values are about an order of magnitude larger than in type-II InAs/Ga(In)Sb superlattices with the same energy gaps, which is consistent with the available literature [29,30]. The alternative assumption that the lifetime is dominated by SRH recombination requires an unrealistically short lifetime of a few 10’s of ns.
Figure 16 plots the resulting spatial profile of the Auger recombination rate. Electrons and holes recombine in the top contact layer (blue portion of the curve), where np - n0p0 > 0 because holes are collected and the electrons cannot escape. However, np - n0p0 < 0 in the absorber and bottom contact layers (red portions of the curve), where the applied field depletes carriers and p << p0 even though n > n0 at some positions. The resulting negative Auger rate corresponds to impact ionization, which dominates the thermal generation of additional electron-hole pairs in the present structure at temperatures near ambient.
Fig. 16. Simulated profile of the Auger recombination rate in the nBn detector structure of Fig. 14. A negative Auger rate (red portion of the curve) corresponds to impact ionization that generates additional electron-hole pairs.
The simulation projects that for room temperature operation, there is no appreciable advantage to reducing the thickness of the absorber. However, lower doping of the bottom contact layer may potentially reduce the dark current due to impact ionization. Analogous simulations find that at lower temperatures, the dark current is dominated by inverse SRH processes, assuming the Auger coefficient remains the same. In that limit, it may be possible to reduce the dark current by reducing the absorber thickness.
The experimental EQE of 58% (from the FTIR characterization) at 300 K is already relatively favorable. Nonetheless, further improvement may be possible by increasing the reflectivity of the top mirror and minimizing parasitic absorption in the top contact layer. At lower temperatures, the rapid decline of the peak EQE in Fig. 11 is attributable primarily to the shift of absorber cut-off wavelength relative to the resonance wavelength (see Fig. 3). Future designs intended for low-temperature operation can compensate by employing a longer cut-off and/or increasing the cavity Q.
7. Conclusions
We have demonstrated a resonant cavity infrared detector that strongly suppresses background noise outside its narrow bandwidth centered on λres ≈ 4.6 µm. It was fabricated by heterogeneously bonding a GaSb-based nBn detector chip to a GaAs/Al0.92Ga0.08As DBR. This bottom mirror provides higher reflectivity (> 99%) and is more straightforward to fabricate than the grown GaSb- or InAs-based DBRs used in previous work, besides providing the potential for dynamic MEMS or piezoelectric tuning of the cavity length and hence the resonance wavelength [31]. Following epi-down bonding to the GaAs/Al0.92Ga0.08As bottom mirror, the GaSb substrate was removed by chemical etching, mesas with a range of diameters were formed, a Ge spacer layer was deposited, and quarter-wave SiO2 and Ge layers were deposited to complete the top mirror. The reported measurements were performed on a device with diameter 93 µm and nominal thickness 158 nm of the Ge spacer layer.
Even though the InAs/InAsSb absorber superlattice was only 103 nm thick, at room temperature the external quantum efficiency measured by the FTIR experiment reached 58%. This compares to EQE ≈ 2% for detectors processed from the same wafer material without top and bottom mirrors. The EQE was also characterized at both NRL and Intraband by photoexciting with a quantum cascade laser, with the experiment at Intraband yielding EQE(λres) ≈ 71%. However, the peak value declined rapidly at T ≤ 200 K, for example to 23% at 100 K, due to the temperature shift of the absorber cut-off wavelength to a value longer than λres.
At higher temperatures where dark current dominates the noise, the specific detectivity is proportional to EQE divided by square root of the dark current. At 300 K, the present device has D* only 32% of that for a state-of-the-art broadband HgCdTe device with cut-off wavelength equal to λres and EQE = 60%. Similarly, no previous RCID has exceeded the HgCdTe D* when compared in this way (ΓD* ≤ 0.46 at 300 K). However, the RCID’s primary advantage is its strong suppression of current generated by thermal background radiation at wavelengths outside the narrow bandwidth of interest, which comes to dominate at lower temperatures. We quantified the effect by assuming a realistic system with detector placed inside a dewar at operating temperature TD, and which views a scene (including both a narrow-band signal of interest and 300 K background) through an f/4 optic. The thermal background current was integrated over the entire MWIR and compared to that for a state-of-the-art broadband HgCdTe device with 5 µm cut-off, Rule ‘07 dark current, and EQE = 85% at all wavelengths. Although D* is larger for the broadband HgCdTe device at higher temperatures where the background contribution is negligible compared to dark current, at temperature below ≈ 175 K the RCID becomes advantageous because the background from the scene dominates. At T = 125 K, D* ≈ 5.5 × 1012 cm Hz½/W for the RCID is 3.3× higher than for the broadband HgCdTe. However, to realize this advantage it is critical that a narrow resonance bandwidth and high EQE at λres be combined with very low EQE at non-resonance wavelengths.
Simulations using NRL MULTIBANDS indicate that impact ionization in the bottom contact and absorber layers dominates the dark current at temperatures near ambient. The device is not diffusion limited, therefore, and we would not expect the dark current to decrease appreciably if the absorber thickness were reduced. However, it may be possible to substantially reduce the impact ionization rate by decreasing the doping concentrations in the bottom contact and n + -InAs capping layers. It should also be possible to increase the EQE at 300 K, beyond its current value of 58-71%, by increasing the reflectivity of the top mirror and minimizing parasitic losses due to interband and free carrier absorption in the top contact layer. It was mentioned above that simply red-shifting the absorber cut-off wavelength would substantially improve the EQE and D* of the present device at T < 200 K. Increasing the top mirror reflectivity for higher cavity Q should also be beneficial at all temperatures.
The slopes of the J-V characteristics indicate that the RCID may not be diffusion limited at lower temperatures either. However, thermal generation due to the inverse SRH processes appears to dominate the dark current (see Fig. 2), so reducing the absorber thickness should be beneficial as long as the cavity Q is high enough to maintain high EQE. These modifications should not only increase D*, but also allow it to continue increasing down to T < 125 K. It is anticipated that substantially higher D* will be possible throughout the 100-300 K temperature range.
A further advantage of the RCID approach is faster frequency response because minority carriers can be collected more rapidly from the very thin absorber. Preliminary optical heterodyne measurements at Intraband in collaboration with the University of Wisconsin - Madison, which were performed on smaller-diameter (for lower capacitance) sister devices to the one reported here, have demonstrated frequency response at least 5 GHz. Those experiments will be reported elsewhere.
Funding
Office of Naval Research; Army Research Office (W911NF21C0011).
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time, but may be obtained from the authors upon reasonable request.
References
1. J. R. Meyer, I. Vurgaftman, C. L. Canedy, W. W. Bewley, C. S. Kim, C. D. Merritt, M. V. Warren, and M. Kim, “Resonant-Cavity Infrared Photodetectors with Fully-Depleted Absorbers,” U.S. Patent #10,062,794, 28 August 2018.
2. M. S. Ünlü and S. Strite, “Resonant cavity enhanced photonic devices,” J. Appl. Phys. 78(2), 607–639 (1995). [CrossRef]
3. M. Guan and C. Chang-Hasnain, “Resonant-cavity-enhanced p-i-n photodetector using a high-contrast-grating for 940 nm,” Opt. Express 30(6), 9298–9306 (2022). [CrossRef]
4. A. M. Green, D. G. Gevaux, C. Roberts, P. N. Stavrinou, and C. C. Phillips, “λ ≈ 3 µm InAs resonant-cavity-enhanced photodetector,” Semicond. Sci. Technol. 18(11), 964–967 (2003). [CrossRef]
5. M. Böberl, T. Fromherz, T. Schwarzl, G. Springholz, and W. Heiss, “IV–VI resonant-cavity enhanced photodetectors for the mid-infrared,” Semicond. Sci. Technol. 19(12), L115–L117 (2004). [CrossRef]
6. M. Arnold, D. Zimin, and H. Zogg, “Resonant-cavity-enhanced photodetectors for the mid-infrared,” Appl. Phys. Lett. 87(14), 141103 (2005). [CrossRef]
7. J. Wang, J. Hu, P. Becla, A. M. Agarwal, and L. C. Kimerling, “Resonant-cavity-enhanced mid-infrared photodetector on a silicon platform,” Opt. Express 18(12), 12890–12896 (2010). [CrossRef]
8. F. Felder, M. Arnold, M. Rahim, C. Ebneter, and H. Zogg, “Tunable lead-chalcogenide on Si resonant cavity enhanced midinfrared detector,” Appl. Phys. Lett. 91(10), 101102 (2007). [CrossRef]
9. N. Quack, S. Blunier, J. Dual, F. Felder, M. Arnold, and H. Zogg, “Mid-infrared tunable resonant cavity enhanced detectors,” Sensors 8(9), 5466–5478 (2008). [CrossRef]
10. W. E. Tennant, ““Rule 07” revisited: Still a good heuristic predictor of p/n HgCdTe photodiode performance?” J. Electron. Mater. 39(7), 1030–1035 (2010). [CrossRef]
11. T. A. O’Loughlin, G. R. Savich, D. E. Sidor, B. T. Marozas, T. D. Golding, K. D. Jamison, L. Fredin, B. Fowler, W. Priyantha, and G. W. Wicks, “Mid-IR resonant cavity detectors,” J. Vac. Sci. Technol., B: Nanotechnol. Microelectron.: Mater., Process., Meas., Phenom. 35(2), 02B111 (2017). [CrossRef]
12. C. L. Canedy, W. W. Bewley, C. D. Merritt, C. S. Kim, M. Kim, M. V. Warren, E. M. Jackson, J. A. Nolde, C. A. Affouda, E. H. Aifer, I. Vurgaftman, and J. R. Meyer, “Resonant-cavity infrared detector with five-quantum-well absorber and 34% external quantum efficiency at 4 µm,” Opt. Express 27(3), 3771–3781 (2019). [CrossRef]
13. A. P. Craig, F. Al-Saymari, M. Jain, A. Bainbridge, G. R. Savich, T. Golding, A. Krier, G. W. Wicks, and A. R. Marshall, “Resonant cavity enhanced photodiodes on GaSb or the mid-wave infrared,” Appl. Phys. Lett. 114(15), 151107 (2019). [CrossRef]
14. V. Letka, A. Bainbridge, A. P. Craig, F. Al-Saymari, and A. R. J. Marshall, “Resonant cavity-enhanced photodetector incorporating a type-II superlattice to extend MWIR sensitivity,” Opt. Express 27(17), 23970–23980 (2019). [CrossRef]
15. V. Letka, A. P. Craig, A. Bainbridge, and A. R. J. Marshall, “A superlattice-based resonant cavity-enhanced photodetector operating in the long-wavelength infrared,” Appl. Phys. Lett. 117(7), 073503 (2020). [CrossRef]
16. A. Bainbridge, A. P. Craig, F. Al-Saymari, A. Krier, and A. R. J. Marshall, “Resonant Cavity–Enhanced Photodiodes for Spectroscopy of C–H Bonds,” phys. stat. sol. (a) 218(17), 2100056 (2021). [CrossRef]
17. A. Bainbridge, L. A. Hanks, A. P. Craig, and A. R. J. Marshall, “Resonant cavity-enhanced photodiode array for miniaturised spectroscopic sensing,” Opt. Express 30(3), 3230–3237 (2022). [CrossRef]
18. A. Kamboj, L. Nordin, A. J. Muhowski, D. Woolf, and D. Wasserman, “Room-Temperature Mid-Wave Infrared Guided-Mode Resonance Detectors,” IEEE Photonics Technol. Lett. 34(11), 615–618 (2022). [CrossRef]
19. P. Petluru, A. J. Muhowski, A. Kamboj, N. C. Mansfield, M. Bergthold, E. A. Shaner, J. F. Klem, and D. Wasserman, “All-epitaxial resonant cavity enhanced long-wave infrared detectors for focal plane arrays,” Appl. Phys. Lett. 122(2), 021101 (2023). [CrossRef]
20. C. L. Canedy, C. S. Kim, M. Kim, D. C. Larrabee, J. A. Nolde, W. W. Bewley, I. Vurgaftman, and J. R. Meyer, “High-Power, Narrow-Ridge, Mid-Infrared Interband Cascade Lasers,” J. Vac. Sci. Technol., B: Nanotechnol. Microelectron.: Mater., Process., Meas., Phenom. 26(3), 1160–1162 (2008). [CrossRef]
21. J. A. Nolde, E. M. Jackson, M. F. Bennett, C. A. Affouda, E. R. Cleveland, C. L. Canedy, I. Vurgaftman, G. G. Jernigan, J. R. Meyer, and E. H. Aifer, “Reticulated Shallow Etch Mesa Isolation (RSEMI) for Controlling Surface Leakage in GaSb-Based Infrared Detectors,” Appl. Phys. Lett. 111(5), 051102 (2017). [CrossRef]
22. M. Turville-Heitz, J. Ryu, J. D. Kirch, S. Jacobs, R. Marsland, T. Earles, S. Ruder, K. Oresick, D. Botez, and L. J. Mawst, “High-Power Mid-Infrared Quantum Cascade Lasers for Free-Space Communications,” Photonics Conf. (IEEE, 2022).
23. R. W. Boyd, Radiometry and the Detection of Optical Radiation (Wiley, 1983).
24. N. Gat, J. Zhang, M. D. Li, L. Chen, and H. Gurrola, “Variable Cold Stop for Matching IR Cameras to Multiple f-number Optics,” Proc. SPIE 6542, 65420Y (2007). [CrossRef]
25. edmundoptics.com/p/467m-cwl-250mm-diameter-015-fwhm-ir-bandpass-filter/27182/; thorlabs.com/thorproduct.cfm?partnumber = FB5068-100.
26. M. P. Lumb, I. Vurgaftman, C. A. Affouda, J. R. Meyer, E. H. Aifer, and R. J. Walters, “Quantum Wells and Superlattices for III-V Photovoltaics and Photodetectors,” Proc. SPIE 8471, 84710A (2012). [CrossRef]
27. I. Vurgaftman, M. P. Lumb, and J. R. Meyer, Bands and Photons in III-V Semiconductor Quantum Structures (Oxford University Press, 2021).
28. J. R. Meyer, C. L. Canedy, M. Kim, C. S. Kim, C. D. Merritt, W. W. Bewley, and I. Vurgaftman, “Comparison of Auger Coefficients in Type I and Type II Quantum Well Midwave Infrared Lasers,” IEEE J. Quantum Electron. 57(5), 1–10 (2021). [CrossRef]
29. L. Höoglund, D. Z. Ting, A. Soibel, A. Fisher, A. Khoshakhlagh, C. J. Hill, S. Keo, and S. D. Gunapala, “Minority carrier lifetime in mid-wavelength infrared InAs/InAsSb superlattices: Photon recycling and the role of radiative and Shockley-Read-Hall recombination mechanisms,” Appl. Phys. Lett. 105(19), 193510 (2014). [CrossRef]
30. R. E. DeWames, J. Schuster, E. A. DeCuir Jr, and N. K. Dhar, “Recombination Processes in InAs/InAsSb Type II Strained Layer Superlattice MWIR nBn Detectors,” Proc. SPIE 11002, 110020W (2019). [CrossRef]
31. V. Jayaraman, B. Kolasa, C. Lindblad, A. Cazabat, C. Burgner, S. Segal, K. Lascola, F. Towner, and F. Xie, “Tunable room-temperature continuous-wave mid-infrared VCSELs,” Proc. SPIE 11300, 113000M (2020). [CrossRef]
