A superlattice-based resonant cavity-enhanced photodetector operating in the long-wavelength infrared

III–V semiconductor-based photodetectors present a particularly attractive approach to achieving high responsivity and selectivity for gas, security, and environmental sensing purposes, owing to their optical sensitivity spanning the short, medium, and long-wavelength infrared (LWIR). For the latter two regions, which include absorption features from a large number of environmentally and industrially significant gases,¹ antimonide compounds emerge as the most appropriate due to their particularly low bandgaps.² The most common approach to chemical detection has been to utilize either filtered broadband or dual-band structures,^1,3 whose relatively limited spectral selectivity is not an optimal feature for discriminating the narrow-band fingerprint characteristics of chemicals. An alternative approach arises in the form of a resonant cavity-enhanced photodetector (RCE PD). This design is inherently spectrally selective owing to its use of distributed Bragg reflectors (DBRs) to create an optical cavity, which in turn is designed so that the maximum of the reflected electric field is centered at the very thin absorber. This dramatic enhancement of the optical signal at the design wavelength and rejection of off-resonance absorption enables a large reduction in the absorber volume; and in turn, this should allow for a commensurate reduction in Auger diffusion and defect-induced dark currents due to their dependence on the absorber thickness.⁴ 

Research into III–V semiconductor-based RCE PDs has mostly focused on telecommunication applications at or near 1550 nm with AlGaAs/GaAs being a particularly popular choice for the DBR mirrors,⁵ and the use of InGaAs absorbers extending the resonant range to 1.55 μm.⁶ To make absorption possible in the MWIR regime, however, other material systems were necessary. Green et al.⁷ utilized an InAs bulk absorber and a set of GaAs/AlAs mirrors to achieve the resonant response near the absorption fingerprint of methane, at 3.3 μm; however, in order to enable absorption at wavelengths corresponding to gases such as CO₂ (4.3 μm), N₂O (4.5 μm), and CO (4.6 μm) and beyond, it has been necessary to use bulk antimonides⁸ and finally Sb-based quantum structures, such as superlattices.^9,10

In this work, an RCE PD that functions in the LWIR regime is demonstrated as a proof of concept. This wavelength range is of particular interest for security applications, as it includes absorption fingerprints from a number of gases and chemicals commonly found in explosives, such as pentaerythritol tetranitrate (PETN) (∼7.8 μm),¹¹ acetic acid (7.2 and 7.8 μm),¹¹ cyclonite and TNT,¹² and a number of other compounds,¹³ and where development of spectrally selective detectors could help the realization of stand-off threat detection systems, which currently rely on broadband detection. In addition, this wavelength range is relevant in health diagnostics for detection of compounds such as breath acetone (7–9 μm),¹¹ which is a biomarker for diabetes.¹⁴ 

This particular RCE PD, intended to achieve resonance at 7.8 μm, has been grown on a (001) GaSb substrate using the design shown in Fig. 1(a). The bottom DBR consists of a set of 12 pairs of $AlAs0.08Sb0.92$/GaSb, with thicknesses of 638 and 523 nm, respectively, and which forms an optical cavity around the design resonant wavelength of 7.8 μm with reflectivity R₂ approaching 1. The top DBR, whose optimal reflectivity $R1∼0.9$ was estimated using the condition for maximum resonant quantum efficiency,⁵ was achieved with the use of five pairs of the same system and an additional GaSb cap. Because no lattice-matched bulk III–V alloy exists that allows absorption at these wavelengths, instead 20 repetitions of the type-II strained-layer superlattice (T2SL) InAs/$InAs0.70Sb0.30$ (24 monolayers/11 monolayers) are grown to a total of 220 nm to form the absorber layer. In addition to their tunable bandgap, type-II superlattices are also theoretically predicted to suppress Auger-induced dark current mechanisms.^15–17 Finally, the inclusion of $xGa=0.29$ in the topmost layer of the cavity, as well as the five repetitions (totalling 53.4 nm) of the same superlattice above it, forms an nBn structure, which, in broadband detectors, has been shown to substantially reduce Shockley–Read–Hall (SRH) dark currents compared to a standard diode structure.^18–20 The optical losses in the top contact superlattice are negligible due to its position at the node of the optical field in the cavity, leading to suppression of absorption. The total thickness of the cavity (including the 456 nm of AlAsSb below the absorber and 487 nm of the AlGaAsSb layer above it) is half the resonant wavelength, which maximizes the reflected optical field at the absorber position as represented by the yellow line in the inset of Fig. 1(a).

The RCE PD was grown using a Veeco GENXplor molecular beam epitaxy reactor equipped with SUMO cells for group III materials, as well as valved cracker cells providing As₂ and Sb₂ fluxes. The speed of deposition was kept to ∼1 ML/s; all the mirror and spacer layers were grown at $505 °$C, while the superlattices were grown at $410 °$C. The absorber and contact layers were n-doped with Te at 1 × 10¹⁷ cm⁻³; in addition, following a series of calibration growths, a compensation doping scheme was applied to the ordinarily p-type,^21,22 barrier-like $Al0.71Ga0.29As0.08Sb0.92$ layer in order to suppress band bending at its interface with the n-doped absorber.⁸ Following growth, structural characterisation and modeling at the wafer level were carried out with the use of a Bruker D8 Discover x-ray diffractometer and Bede RADS software, and the resultant coupled scan spectrum can be seen in Fig. 1(b). Periodic satellite peaks due to the very thin superlattice layers can still be discerned despite the 6.3 μm of DBR material deposited on top of the cavity, pointing to good crystallinity and well defined interfaces of the T2SL layers.

A Bruker Vertex70 Fourier Transform Infrared (FTIR) spectrometer was employed to characterize the optical transmission of the long-wavelength cavity as a function of temperature, and the results are presented in Fig. 2. As expected, the magnitude of the absorption increases with temperature in line with the rise of the absorption coefficient of the absorber at the resonant wavelength [Fig. 2(a)]. The shift of the resonant wavelength to higher values occurs due to a convolution of the lattice constant and refractive index temperature dependencies (the latter, however, is not well-documented in the literature, especially for complex quaternaries like AlGaAsSb). The corresponding temperature coefficient [Fig. 2(b)] displays a strongly linear behavior and is found to be 0.52 nm K⁻¹, which is representative of the high thermal stability of RCE PDs when compared to broadband detectors.

In order to obtain the optoelectronic characteristics of the structure, the wafer was processed into photodetectors as detailed in the diagram in Fig. 1(a) using standard cleanroom methods of photolithography, wet etching, and thermal deposition of Ti/Au contacts. The nBn structure of the cavity was realized by a wet etch of the top DBR stack, terminating on the wide-bandgap $Al0.71Ga0.29As0.08Sb0.92$ layer, while the Ohmic contacts were deposited on the top superlattice layer, entirely bypassing the undoped top DBR stack.

Dark current characteristics were measured by mounting the device in an LN₂-cooled Lakeshore TTPX probe station with a cold shield in place. The resultant temperature and voltage dependence of the current density (J_d), as well as the derived Arrhenius plot at the operational voltage of −0.4 V, chosen as it corresponds to the highest specific detectivity in the measurement range, can be seen in Figs. 3(b) and 3(a), respectively. At 77 K, J_d equals 2.0 × 10⁻⁴ A cm⁻², which is broadly of the same order of magnitude as comparable broadband LWIR InAs/InAsSb nBn detectors at their operational voltages.^23,24 The 0% cut-off wavelength of 9.2 μm, obtained at 77 K from a reference broadband nBn device with an InAs/$InAs0.70Sb0.30$ (22 ML/11 ML) absorber with a total thickness of 3.67 μm and the same level of doping in the absorber, was also used to determine Rule 07. Due to the order-of-magnitude reduction in the absorber thickness, the dark current density of the RCE PD equalizes with and then falls below the Rule 07 curve at 180 K and above, which raises the prospect that suppression or elimination of the SRH component from the dark current will bring the J_d values at operational temperatures close to or below Rule 07 as well.

Further analysis of the Arrhenius plot reveals an activation energy (E_a), which decreases monotonically with increasing voltage from 78 meV at −0.1 V to 38 meV at −1 V in all temperature regions, fitted for the high-doped regime.²⁵ Moreover, by analyzing the cut-off wavelength of the broadband reference nBn device, the low-temperature bandgap of the absorber is found to be ∼135 meV. Because E_a of the RCE PD is found to be roughly half that at lower voltages, this points to the SRH mechanism dominating the dark current behavior. This is consistent with the presence of a non-negligible turn-on voltage required to achieve photoresponse and indicates that band bending at interfaces of the thin absorber is not fully eliminated with the amount of compensation doping present in the AlGaAsSb layer. However, because this current component originates at a heterojunction and is not volume-dependent, it is not a fundamental limitation and can be addressed with further growths aimed at doping optimization. Meanwhile, elimination of any crystal imperfections arising due to the inherently complex and long growth process (e.g., via in situ growth rate monitoring systems) would minimize any SRH currents generated in the quasi-neutral region of the T2SL, helping to further realize the advantage of the reduction in the absorber volume.

Temperature-dependent responsivity of the RCE PD was determined using the FTIR spectrometer and a Vigo CMT detector with a known responsivity as the reference. The resonant response at ∼7.75 μm was detectable between 77 and 160 K, reaching a maximum value of 0.91 A/W at 77 K as plotted in Fig. 4 at a constant voltage of −0.8 V. In addition, the full voltage and temperature-dependent spectra of the peak external quantum efficiency (EQE), recalculated from the responsivity data, are plotted in Fig. 5(a) showing maximum values of ∼15% at 77 and 100 K, a good result for an active layer one order of magnitude thinner than the typical broadband absorber.^23,24

The Beer–Lambert law offers a simple estimate of the resonant enhancement in the RCE PD: the theoretical maximum external quantum efficiency of a broadband 220 nm InAs/$InAs0.70Sb0.30$T2SL absorber with absorption coefficient α = 1750 cm⁻¹ (obtained from a reference epilayer sample at 77 K) is calculated to be 3.7%, pointing to a fourfold resonant enhancement in the quantum efficiency of the RCE PD.

The rising-then-falling behavior of EQE as bias increases is attributed to two factors. At low voltages, the band bending at the absorber-AlGaAsSb interface necessitates an application of potential in order to extract the photocarriers before they recombine. The bias value at 77 K required to reach a plateau in the signal is notably larger than that at higher temperatures; this could be explained by the thermal component of the carrier energy being too low to help promote them across the barrier, requiring a higher electrical field. Then, after responsivity peaks between −0.6 and −0.7 V, the fall could be due to a decrease in the inversely voltage-dependent Shockley–Read–Hall lifetime. Similarly, the inverse relation between the SRH carrier lifetime and temperature, which has its origin in the phonon scattering process,²⁶ explains the decrease in responsivity seen in Figs. 4 and 5(a).

The optical line shapes of the resonant peaks were characterized by their full-widths-at-half-maximum (FWHM) and quality (Q) factors, both of which can be seen in Fig. 5(b). The FWHM values span a range between 92 and 101 nm, with a pronounced plateau at the lowest three temperatures, which indicates the most favorable combination of temperature-dependent parameters of the mirrors and the cavity is present at these temperatures. The resultant Q factors fall in the range of 76–86, a result that compares very well with similar MWIR structures found in the literature^7–10 and indicates that very high quality of mirror interfaces has been achieved despite their combined epitaxial thickness reaching almost 19 μm. In addition, the bypassing of the top DBR by the current path, achieved by the deposition of the Ohmic contacts directly on the top T2SL layer [see Fig. 1(a)], eliminates the need to n-dope the top DBR, which in turn prevents optical losses due to free carriers in the stack and so improves the Q factor.

Specific detectivity (⁠$D*$⁠) was calculated in the shot and Johnson noise-limited regime by combining peak external quantum efficiency values from Fig. 5(a) and the dark current densities at the corresponding voltages. The resultant $D*$ data, which are plotted as a function of temperature and voltage in Fig. 6, rise to a peak plateau in the range of −0.3 to −0.7 V (with the maximum values occurring at −0.4 V) where the bias applied is high enough to overcome the inbuilt valence band offset, but subsequently drops after that mark due to the contribution of the voltage-activated SRH dark current and its associated noise. At 77 K and −0.4 V, $D*$ reaches 3.7 × 10¹⁰ cm $Hz1/2$ W⁻¹; as a way of comparison, the theoretical limit on the background-limited $D*$ of an ideal photovoltaic detector with a 0% cut-off wavelength of 9 μm (which is approximately the 0% cut-off wavelength of the reference broadband nBn device at 77 K) is calculated by Rogalski²⁷ to be 5.5 × 10¹⁰ cm $Hz1/2$ W⁻¹. Achieving this BLIP performance for the RCE PD in the shot and Johnson-limited regime would require a twofold reduction in dark current density, which would easily be achievable for future iterations of this RCE PD as the SRH dark currents are minimized via optimization of the compensation doping scheme.

To summarize, by employing distributed Bragg reflectors of $AlAs0.08Sb0.92$/GaSb and an InAs/$InAs0.70Sb0.30$ superlattice absorber, we have demonstrated a resonant cavity-enhanced photodetector operating in the long-wavelength infrared regime, with the resonant response occurring at ∼7.75 μm. Temperature-dependent spectral measurements of transmission indicate a highly stable temperature coefficient of 0.52 nm K⁻¹. Dark current measurements reached 2.0 × 10⁻⁴ A cm⁻² at 77 K and −0.4 V, while the derived Arrhenius plot characteristics show low activation energies, suggesting that Shockley-Read-Hall is the dominant current mechanism, and which is most likely due to band bending at the interface of the very thin absorber layer and the barrier and the resultant electric field. External quantum efficiency, meanwhile, was measured to reach a maximum of ∼15% at 77 and 100 K, with generally the highest values reached at and above −0.5 V. Meanwhile, the quality factor of the cavity was found to fall in the range of 76–86, a very good result indicating that high quality mirror interfaces were achieved despite the very thick epitaxial growth, as well as negligible optical losses outside the absorber. The corresponding $D*$ values, calculated in the shot and Johnson-limited regimes, show a generally stable response in the −0.3 to −0.7 V bias range, with a maximum of 3.7 × 10¹⁰ cm $Hz1/2$ W⁻¹ at 77 K and −0.4 V. The SRH-limited dark current noise of this particular device is likely due to band bending present at the interfaces of the very thin absorber—an issue that can be easily addressed by carrying out consecutive optimization growths, promising an easy pathway to improvement of this structure.

See the supplementary material for the structural and optical parameters of the reference broadband nBn device and the reference epilayer sample.

The authors would like to thank UK MOD and DSTL for funding under Grant No. DSTLX:1000116341, and the EPSRC for the funding under Grant No. EP/M506369/1.