APPLIED PHYSICS LETTERS 102, 011108 (2013) Demonstration of high performance bias-selectable dual-band short-/mid-wavelength infrared photodetectors based on type-II InAs/GaSb/AlSb superlattices A M Hoang, G Chen, A Haddadi, and M Razeghia) Center for Quantum Devices, Department of Electrical Engineering and Computer Science, Northwestern University, Evanston, Illinois 60208, USA (Received 12 October 2012; accepted 14 December 2012; published online January 2013) High performance bias-selectable dual-band short-/mid-wavelength infrared photodetector based on InAs/GaSb/AlSb type-II superlattice with designed cut-off wavelengths of lm and lm was demonstrated At 150 K, the short-wave channel exhibited a quantum efficiency of 55%, a dark current density of 1.0  10À9 A/cm2 at À50 mV bias voltage, providing an associated shot noise detectivity of 3.0  1013 Jones The mid-wavelength channel exhibited a quantum efficiency of 33% and a dark current density of 2.6  10À5 A/cm2 at 300 mV bias voltage, resulting in a detectivity of 4.0  1011 Jones The spectral cross-talk between the two channels was also discussed for further C 2013 American Institute of Physics [http://dx.doi.org/10.1063/1.4773593] optimization V High performance infrared detectors in the short-wave infrared (SWIR) and mid-wave infrared (MWIR) spectral bands are highly needed in a number of tracking and reconnaissance missions A multi-color imager sensing the SWIR and MWIR in a spatially co-incident fashion and all-in-one package allows improved target identification, and detection of chemical signatures specific to a wave-band (e.g., CO2 in the MWIR) that are otherwise not possible with single color imagers In addition, the combination of the SWIR and MWIR provides the flexibility to perform active and passive imaging in a single camera The reflective nature of SWIR light allows detailed and high contrast objects similar to the visible spectrum that can be obtained by illuminating a light source or under night sky radiance known as “night-glow.” Under low light conditions or situations where use of a light source is prohibitive, the MWIR may prove to be more useful without requiring active illumination Bulk semiconductor materials are often limited to a particular detection window near the semiconductors’ bandgap and not suitable for multi-color detection For example, InGaAs compound is the mature technology for the SWIR regime, but InGaAs is incapable to extend its detection limit beyond 2.6 lm.1 InSb only covers the MWIR regimes HgCdTe is a special family which, by changing the molar fraction of Cd, can tailor the cutoff from SWIR to very long wavelength infrared (VLWIR) The current state-of-the-art MWIR/SWIR dual-band detectors are HgCdTe technology However, HgCdTe which is a II-VI based technology is reported to suffer from toxicity and high cost.2 Type-II InAs/GaSb/AlSb superlattice3 (T2SL) has emerged as a candidate highly suitable for multi-spectral detection due to its versatility in band-gap engineering, while retaining lattice-matched conditions Based on the stability and robustness of the mature III–V compound technology, T2SL has demonstrated the feasibility of covering a large infrared detection range, from SWIR to VLWIR,4–7 and the capability of growing complex devices with different supera) Email: razeghi@eecs.northwestern.edu 0003-6951/2013/102(1)/011108/4/$30.00 lattice structures such as W-structure8 and M-structure.9 T2SL material system has demonstrated high performance SWIR4 and MWIR5 photodetectors, as well as the dual-band LWIR-LWIR, MWIR-LWIR, MWIR-MWIR photodetectors10–12 and imaging.13–15 However, up to date, no dualband SWIR-MWIR photodetector performance has been reported yet We present, in this letter, the demonstration of high performance dual-band SWIR-MWIR photodetectors, introducing T2SL InAs/GaSb/AlSb a competitor in this active/passive imaging area Several device architectures of stacked-detectors have been implemented for the dual-band detection, including simultaneous and nearly simultaneous bias-selectable detectors.2 Simultaneous detectors require additional contact and therefore are more complicated in device fabrication In this work, we use the structure of two back-to-back p-i-n-n-i-p photodiodes in which the channels can be addressed alternatively by changing the polarity of applied bias voltage The schematic diagram of device structure and the band alignment of superlattices for absorption regions are shown in Fig The SWIR channel is grown on top of the MWIR channel in order to have the response from both channels while the SWIR channel will also play a role as a low-pass filter for the underneath MWIR channel in front-side illuminated measurement This structure takes advantage of the simplicity of bipolar stacked photodiodes without additional middle contact and keeps each channel quite independent of each other The first challenge in the realization of two-color SWIR/MWIR type-II photodetectors requires developing separate materials sensitive to the SWIR and MWIR with high performance After that, the integration of co-located photodetectors requires taking different priority order of optimizing optical or electrical performance in different channels Indeed, it is essential to enhance the quantum efficiency (QE) of the SWIR channel and at the same time, improve the electrical performance of the MWIR channel On one hand, the higher the quantum efficiency of SWIR is, the less light can go into the MWIR channel and thus less cross-talk On the other hand, given the difference in band 102, 011108-1 C 2013 American Institute of Physics V Downloaded 04 Jan 2013 to 129.105.126.15 Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions 011108-2 Hoang et al Appl Phys Lett 102, 011108 (2013) using the state-of-the-art molecular beam epitaxy (MBE) equipped with group III SUMOV cells and group V valved crackers We introduce silicon (Si) in InAs and beryllium (Be) in GaSb as n-type and p-type dopants, respectively After the growth, the sample was structurally characterized using the atomic force microscopy (AFM) and high resolution X-ray diffraction (HR XRD) The AFM showed a stand˚ over a ard morphology with rms roughness of 1.3 A 10  10 lm2 area The satellite peaks in the high resolution x-ray diffraction rocking curves show the thicknesses of ˚ and 45 A ˚ for each period of MWIR and SWIR active 62 A regions, respectively The photodiodes were then fabricated with linear sizes ranging from 100 to 400 lm The processing technique was described thoroughly elsewhere.16 The photodiodes were left unpassivated for optical characterization but care was paid in order to minimize the surface leakage After the processing, the photodiodes were wire-bonded onto a leadless ceramic chip carrier (LCCC) and loaded into a cryostat for optical and electrical characterizations Shown in Figures and are the electrical and optical performances of the device at 150 K The current-voltage characteristic exhibits the behavior of a SWIR photodiode in the bias range from negative to positive 200 mV Above 250 mV, the I-V curve starts to behave like a MWIR photodiode when the differential resistance of MWIR becomes greater than SWIR The sample exhibits an absolute dark current density of 1.0  10À9 A/cm2 at À50 mV The value of R0A is around  107 X cm2 At zero or negative bias, the diode shows photo-response at the SWIR channel and the MWIR operation is completely off The SWIR response is still present even when we apply a positive 200 mV bias while the MWIR channel starts showing significant response from 270 mV The quantum efficiency of MWIR channel quickly saturates with 33% at peak responsivity for a lm active region at around 300 mV bias This bias for saturated quantum efficiency is small enough to be suitable for focal plane array application At this bias, the measured dark current density is 2.6  10À5 A/cm2 Between the bias values of 100 mV and 300 mV, we switch from saturated SWIR optical signal to saturated MWIR optical signal In the middle of this range, the polarity of the photocurrent depends on the R FIG Schematic diagram of a dual-band SWIR-MWIR back-to-back p-in-n-i-p photodiode structure and schematic band alignment of superlattices in two absorption layers The colored rectangles in the insets represent the forbidden gap of component materials Dotted lines represent the effective band gaps of superlattices gap, the electrical performance of SWIR is several orders of magnitude better than the MWIR channel; a higher operating temperature is dependent on the electrical performance of the MWIR channel The strategies of improving individual channel performance are thus different for each channel; however, they affect profitably to each other and improve the overall dual-band photodiode performance The structure of SWIR channel was previously described in detail in Ref It is a homojunction p-i-n photodiode with superlattice design of 6/1/5/1 monolayers (MLs) of InAs/GaSb/AlSb/GaSb This design gives a 50% cut-off wavelength of lm at 150 K Single color SWIR diodes were grown to verify the performance as well as theirs dynamics in response towards the applied bias Under illumination, the samples showed photoresponse at reverse and zero bias When we applied slight forward bias, the photocurrent was still present and it was only completely off when the forward bias, called open-circuit voltage, reached a certain value This will later explain the spectrum of the dual-band SWIR-MWIR detectors For the MWIR channel, we use the same superlattice structure as in Ref The absorption region of the MWIR channel is designed with 6.5/12 MLs of InAs/GaSb to have a 50% cut-off wavelength of lm at 150 K Unlike standard p-i-n homojunction, in p-p-M-N device architecture, the M barrier was inserted to block the tunneling current After that, the active region p was p-doped heavily to suppress the dark current The M barrier was also carefully engineered to eliminate the misalignment in the conduction band which can result in bias dependence of quantum efficiency.5 The dual-band SWIR-MWIR photodiodes were designed to consist of 0.5 lm thick bottom p-contact, lm thick p-doped active region in MWIR, 0.5 lm thick Mbarrier, lm thick common n-contact (0.5 lm MWIR superlattice and 0.5 lm SWIR superlattice), lm undoped SWIR active region, and 0.5 lm p-contact at the end Prior to the detector, a 0.5 lm thick p-doped InAsSb lattice-matched to the GaSb substrate was grown as an etch-stop layer Our photodetectors were grown on n-type (001) GaSb substrate FIG Dark current and differential resistance-area product vs applied bias of the diode at 150 K SWIR and MWIR arrows represent the operation bias which shows the dominant behavior of each channel Downloaded 04 Jan 2013 to 129.105.126.15 Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions 011108-3 Hoang et al Appl Phys Lett 102, 011108 (2013) The built-in voltage in dark condition is given by kT Na Nd ln ; VbuiltÀin ¼ q n2i FIG Quantum efficiency spectrum of the photodiode at 150 K as function of applied bias The SWIR signal starts to attenuate at 200 mV and the MWIR signal saturated at 300 mV wavelength The photocurrent may be positive for certain wavelengths and negative for others, depending on which diode is generating the higher current at a given wavelength This QE bias dependence of the dual-band photodetectors can be explained by the difference of built-in voltages and the dynamics of generated photocurrent in the two p-n junctions The schematic band alignment of a dual-band SWIR-MWIR p-i-n-n-i-p structure is shown in Fig It is essential to remind that each optimized individual photodiode did not exhibit QE bias dependence when operated separately Therefore, the origin of QE bias dependence in this dual-band detector is not from the band misalignment between different regions that could block the photocurrent The only band misalignment in this structure is from the junction in SWIR and MWIR n-contacts However, these common contacts were heavily n-doped to form an Ohmic contact, which permits electrons to easily tunnel through a thin barrier In this device architecture, the QE bias dependence of the channels comes from the competition of built-in voltages which are the result of the formation of p–n junctions Since the diodes are positioned back to back, the builtin voltage in one channel creates a potential barrier to block the photo-current of the other channel FIG Schematic band diagram at zero bias and under dark condition VS and VM are the built-in voltages of SWIR and MWIR channels, respectively Ec, Ev, and EF are the conduction band, valence band, and Fermi level at zero bias (1) where T is the temperature, q is the electron charge, Na and Nd are the concentrations of acceptors and donors, ni is the intrinsic carrier concentration (ni2 / exp(ÀEg/kT) and is decided by the operating temperature and the band gap of the material, which is fixed for specific requirement of cutoff wavelength A rough estimation shows that if the doping of Na and Nd is almost the same in the two channels, the discrepancy in built-in voltages of the two channels is the difference in band gap Under illumination or applied bias, the generation and redistribution of mobile charges will either reduce or increase the potential barrier At negative or small positive bias (