**Thermoelectrically-Cooled InAs**/**GaSb Type-II Superlattice Detectors as an Alternative to HgCdTe in a Real-Time Mid-Infrared Backscattering Spectroscopy System**

#### **Raphael Müller \*, Marko Haertelt, Jasmin Niemasz, Klaus Schwarz, Volker Daumer, Yuri V. Flores, Ralf Ostendorf and Robert Rehm**

Fraunhofer Institute for Applied Solid State Physics IAF, Tullastraße 72, 79108 Freiburg, Germany; marko.haertelt@iaf.fraunhofer.de (M.H.); jasmin.niemasz@iaf.fraunhofer.de (J.N.); klaus.schwarz@iaf.fraunhofer.de (K.S.); volker.daumer@iaf.fraunhofer.de (V.D.); yuri.flores@iaf.fraunhofer.de (Y.V.F.); ralf.ostendorf@iaf.frauhofer.de (R.O.); robert.rehm@iaf.fraunhofer.de (R.R.) **\*** Correspondence: raphael.mueller@iaf.fraunhofer.de

Received: 26 October 2020; Accepted: 14 December 2020; Published: 18 December 2020

**Abstract:** We report on the development of thermoelectrically cooled (TE-cooled) InAs/GaSb type-II superlattice (T2SL) single element infrared (IR) photodetectors and exemplify their applicability for real-time IR spectroscopy in the mid-infrared in a possible application. As the European Union's Restriction of Hazardous Substances (RoHS) threatens the usage of the state-of-the-art detector material mercury cadmium telluride (MCT), RoHS-compatible alternatives to MCT have to be established for IR detection. We use bandgap engineered InAs/GaSb T2SLs to tailor the temperature-dependent bandgap energy for detection throughout the required spectral range. Molecular beam epitaxy of superlattice samples is performed on GaAs substrates with a metamorphic GaAsSb buffer layer. Photolithographic processing yields laterally-operated T2SL photodetectors. Integrated in a TE-cooled IR detector module, such T2SL photodetectors can be an alternative to MCT photodetectors for spectroscopy applications. Here, we exemplify this by exchanging a commercially available MCT-based IR detector module with our T2SL-based IR detector module in a real-time mid-infrared backscattering spectroscopy system for substance identification. The key detector requirements imposed by the spectroscopy system are a MHz-bandwidth, a broad spectral response, and a high signal-to-noise ratio, all of which are covered by the reported T2SL-based IR detector module. Hence, in this paper, we demonstrate the versatility of TE-cooled InAs/GaSb T2SL photodetectors and their applicability in an IR spectroscopy system.

**Keywords:** InAs/GaSb; T2SL; IR; photodetector; TE-cooled; spectroscopy; RoHS; MCT

#### **1. Introduction**

In numerous applications in science and industry, detection of infrared (IR) radiation is indispensable. A wide area of application is IR spectroscopy in the mid-infrared (MIR, 3–12 μm). Since several substances in gaseous, liquid, and solid state of aggregation have their characteristic transitions here, this region, which is sometimes referred to as the "fingerprint region", is therefore clearly relevant for industrial or medical spectroscopy applications and when chemical identification or verification is required [1,2]. For industrial applications, common requirements of the IR detector arise. These can be summarized as: fast response, broadband spectral coverage, linearity, and high signal-to-noise ratio. These requirements can be met by specially designed IR photodetectors.

In an IR photodetector, a signal is generated after photon absorption across the fundamental bandgap of the underlying semiconductor material. The bandgap energy *E*<sup>g</sup> of this material defines the cutoff wavelength of the detector, implying that radiation of longer wavelength cannot be detected. As the performance of IR photodetectors decreases for longer cutoff wavelength, choosing the detector cutoff wavelength based on the requirements of the application is essential. In general, cooling the detector material improves the performance of IR photodetectors. Utmost performance is achievable with expensive cooling with cryogenic liquids or Stirling coolers. However, for most applications, low-cost, small, lightweight IR detector modules are required. In these modules, the detector element is thermoelectrically cooled (TE-cooled) with multistage Peltier elements to a so-called high operating temperature (HOT) in the range between 180 K and 300 K.

So far, the commercialized state-of-the-art material of choice for HOT IR photodetectors is mercury cadmium telluride (HgCdTe or MCT). This is due to MCT featuring both a bandgap energy that is widely tunable in the IR, as well as a top-notch electrooptical performance. By adjusting the cadmium content, MCT allows for the fabrication of IR photodetectors with a cutoff wavelength in and beyond the fingerprint region. Numerous studies dedicated to the development and optimization of HOT MCT IR detectors have been conducted [3,4]. However, the Restriction of Hazardous Substances (RoHS) of the European Union regulates the allowed concentration of mercury and cadmium in electronic devices [5]. It is only due to temporary exemptions that this regulation does not prohibit the use of MCT detectors. Hence, for future applications, alternative, RoHS-compatible detector materials need to be established.

Since III-V semiconductors do not contain RoHS-restricted substances, RoHS-compatible photodetectors can be fabricated from them. For detection in the MIR, bulk III-V semiconductors are only partly suitable. InSb, the binary III-V material with the lowest bandgap energy, can only be utilized for detection up to around 5 μm when cryogenically cooled or up to around 7 micron for uncooled operation, which is insufficient for many applications. The ternary alloy InAs1-xSbx allows for bandgap tuning by modification of the composition. This enables bandgap energies that are smaller than the one of InSb. The limits of the bandgap tuning range for InAs1-xSbx, i.e., the temperature and the composition dependence of the bandgap, were recently re-investigated [6]. As no substrate material exists that allows for lattice-matched growth of InAs1-xSbx, handling the layer strain is inevitable.

We investigate InAs/GaSb type-II superlattices (T2SLs) that are RoHS-compatible, feature a widely tunable bandgap energy and can be grown lattice-matched to GaSb [7–9]. InAs/GaSb T2SLs consist of alternating layers of InAs and GaSb that are usually grown by molecular beam epitaxy (MBE). Each individual layer is just a few atomic monolayers wide and acts as a quantum well for charge carriers. By quantum mechanical coupling of neighboring quantum well states, electron, and hole minibands are created, respectively. The fundamental bandgap of this artificial bandgap material opens between the lowest electron miniband and the highest hole miniband (see Figure 1a). It can be tuned by altering the width of the InAs and GaSb sublayers. Due to the peculiar type-IIb band alignment between InAs and GaSb, the superlattice bandgap energy can be engineered flexibly for a spectral range roughly corresponding to 3–20 μm, which is equivalent to photon energies from about 60 to 400 meV.

To illustrate the bandgap tuning in InAs/GaSb T2SLs, in Figure 1b a calculation of the bandgap energy in dependence of the superlattice composition based on the superlattice empirical pseudopotential method is shown [10]. Apparently, the InAs sublayer width has the main impact on the bandgap energy. Commonly, the superlattice composition is given in dependence of the sublayer width of InAs and GaSb, which is calculated based on calibrated growth rates and the MBE shutter sequence during T2SL growth. However, in Figure 1b, an As content of 17% is indicated for the GaSb sublayer, which was determined by X-ray diffraction. During the growth of a GaSb sublayer, the chamber atmosphere still contains As due to previously grown InAs sublayers. Since As is the group V component that is preferably incorporated into the layer, this leads to a non-negligible As content in the nominal GaSb sublayer. Details on the MBE growth procedure and the method for the determination of the As content in the GaSb sublayers are given in [10].

**Figure 1.** (**a**) Schematic of the type-IIb band alignment between InAs and GaSb and an InAs/GaSb type-II superlattice with lowest electron miniband and highest hole miniband. (**b**) Bandgap energy in dependence of the InAs/GaSb superlattice composition, calculated by the superlattice empirical pseudopotential method [10].

After the proposal to use InAs/GaSb T2SLs for IR detection [11], fundamental research on this material system [12,13] and development of single element detectors [14–16] and detector arrays [17] has intensified in the last decades. Important developmental steps in the field are reviewed in [18].

Activities in research and development of InAs/GaSb T2SLs have mainly focused on high-performance applications at low operating temperatures that require cryogenic cooling. As a result, for low operating temperatures, InAs/GaSb T2SLs emerge as a viable alternative to MCT for IR detectors and IR cameras. For the HOT range, IR detection with InAs/GaSb T2SLs in the longwave infrared was demonstrated [19–22], but dedicated device development and commercialization were never conducted. Now, mainly due to the RoHS, there is renewed interest in InAs/GaSb T2SL IR photodetectors for HOT applications.

Within the last few years, we have worked on the development of InAs/GaSb T2SL single element detectors for the HOT range and demonstrated that they can be combined with the immersion lens technology of VIGO system [23–25]. In this paper, we briefly describe the layout of the detector as a laterally-operated photoconductor, the superlattice and buffer layer growth as well as the detector processing. Then, after the detector is integrated into an IR detector module with a four-stage TE-cooler, which allows for operation at 200 K, we focus on a possible spectroscopy application in which an MCT-based IR detector module could be replaced by a T2SL-based IR detector module.

In addition to the cutoff wavelength, two more detector figures of merit are crucial for the content of this paper. The first is the specific detectivity *D*∗ , which describes the signal-to-noise ratio:

$$D^\*(\lambda, f) = \frac{R(\lambda)}{I\_{\mathbb{R}}(f)} \sqrt{A\_{\vartheta} \Delta f}. \tag{1}$$

*D*<sup>∗</sup> depends on the spectral responsivity *R*(λ), the noise current *In*(*f*) and the bandwidth Δ*f*. It is normalized to the optical detector area *Ao*. By using a lens to focus incoming radiation, *Ao* can be increased significantly. The increase depends on the form of the lens and its refractive index *n*. A hyperhemispheric lens can increase *Ao* by a factor of *n*<sup>4</sup> [26]. For backside-illuminated detectors, the lens can be immersed into the substrate material beneath the detector. The second figure of merit is the detector bandwidth that relates to the detection speed. For the device concept under study, the detector bandwidth is inversely proportional to the carrier recombination time. However,

the responsivity is proportional to the carrier recombination time. Therefore, there is a trade-off between photosignal and detection speed in photoconductor optimization.

#### **2. Design, Growth, Processing and Module Integration of an IR Detector**

The InAs/GaSb T2SL discussed in this paper was grown by molecular beam epitaxy on a 3 inch, n-type, (100)-oriented, 1100 μm thick GaAs substrate after careful calibration of shutter sequences and growth rates. Figure 2a shows the epitaxial layer structure. It consists of two buffer layers, the superlattice absorber layer and a thin superlattice contact layer. The first buffer layer is a metamorphic GaAsSb buffer, in which Sb gradually replaces As over 2 μm layer width. This results in a strain relaxed GaSb-like growth template for the subsequent layers [23]. The second buffer layer consists of 10 μm GaSb. This layer is followed by the superlattice absorber layer, which comprises 750 non-intentionally doped superlattice periods (residually n-type). Each of these periods features 14 monolayers (ML) InAs and 7 ML GaSb. InSb-like interfacial layers were realized between the individual InAs and GaSb sublayers to minimize the relative lattice mismatch to the underlying substrate. In the end, the heavily n-type doped contact layer was grown on top.

**Figure 2.** (**a**) Epitaxial layer structure for the fabrication of laterally-operated InAs/GaSb type-II superlattice (T2SL) detectors on GaAs substrate. (**b**) Schematic of a processed InAs/GaSb T2SL detector that is backside-illuminated through an immersion lens (not to scale).

After growth, standard superlattice layer characterization was performed. A superlattice period length of 7.0 nm was determined by high-resolution X-ray diffraction, which was also used to verify the negligible relative lattice mismatch to the GaSb buffer. Spectral photoluminescence was measured to confirm the intended bandgap energy. At 10 K, a bandgap energy of 143 meV (corresponding to a wavelength of 8.7 μm) was obtained. As the bandgap shrinks for rising temperature, which can be described with the Varshni model [25], the corresponding cutoff wavelength of a detector increases. Hence, this superlattice can absorb radiation throughout a large fraction of the MIR at high operating temperatures.

Photolithographic processing was used to fabricate laterally-operated photoconductors (see Figure 2b). Unlike in most T2SL-based detector concepts, in which the current flows parallel to the superlattice growth direction, in this concept, the current flows mainly perpendicular to the growth direction between two ohmic metal contacts and requires external bias voltage for operation. When radiation of suitable wavelength enters the absorber layer, an additional photoconductivity is generated. The processing steps for detector fabrication included dry etching for structuring of the contact and the absorber layer, dielectric passivation, selective opening of the passivation layer and metalization. In the last step, the mesa front was also metalized. This metalized area acts as a mirror facilitating a double pass of the radiation incident from the backside, which increases the quantum efficiency. The processing sequence has been presented in more detail before [24].

A differing lattice constant of layer and substrate, which is the case for InAs/GaSb T2SLs lattice matched to GaSb on GaAs substrates, may result in an increased density of defects, growth inhomogeneities and a reduced device yield. Our wafer-level device characterization at 200 K suggested that device drop out due to material- or processing-related defects is negligible [24]. As device performance proved to be homogeneous across the wafer, a large device yield would be expected for manufacturing purposes. To allow for immersion of hyperhemispheric microlenses into the substrate, the detectors were processed with a horizontal and vertical pitch of 1480 μm. In this way, more than 1000 detectors could be fabricated per 3 inch wafer—the wafer size used in our study. Assuming the increasingly common 4 inch and 6 inch GaSb substrate diameters, the number of devices per wafer would scale according to the wafer area. For fabrication of detectors without substrate microlenses, the number of detectors per wafer depends on the intended detector size and can be significantly higher.

After processing and the characterization of the T2SLs and the fabricated detectors, the fully processed 3 inch wafers were diced into single element detectors. The module integration of the detector elements was completed in cooperation with VIGO System. In these modules, a T2SL detector 50 μm × 50 μm in size is mounted on top of a four-stage TE-cooler. The detectors feature a hyperhemispheric lens that was immersed into the GaAs substrate. As *n*GaAs ≈ 3.3, the lens increases *Ao* for backside incident radiation by about two orders of magnitude and *D*<sup>∗</sup> by one order of magnitude when compared to detector elements without such an immersion lens. Furthermore, the IR detector modules also comprise standard electronics from VIGO System: a fast preamplifier and a TE-cooler controller. These TE-cooled T2SL-based IR detector modules constitute RoHS-compatible turnkey systems.

#### **3. Comparison to MCT**

To benchmark the performance of these detectors, we compare the detectivity of MCT-based and T2SL-based photoconductors without immersion lens. They are operated at 210 K with the noise current taken at 20 kHz. In Figure 3, we show the mean value of the detectivity of InAs/GaSb T2SL photoconductors, which we deduced from measurements that were already discussed in [24]. Here, we compare this mean value with specified detectivities of commercial MCT photoconductors from VIGO System for different cutoff wavelengths from 9–13 μm [27]. For detectors with the same cutoff wavelength of 10.6 μm, the detectivity of the MCT photoconductor is less than a factor of two higher than the detectivity of the T2SL photoconductors. Given the brief development of HOT InAs/GaSb T2SL photodetectors in comparison to the longstanding heritage of MCT photodetectors, this is a highly promising result. Doping optimization [25] and increasing the quantum efficiency are expected to further enhance the T2SL detector performance and increase its competitiveness.

**Figure 3.** Detectivity of InAs/GaSb T2SL photoconductors (mean value) and commercial mercury cadmium telluride (MCT) photoconductors from VIGO System (guaranteed values) for different cutoff wavelengths at 210 K and 20 kHz [24,27].

As a longer cutoff wavelength implies a lower bandgap and an increased carrier generation, which leads to an increased noise level, the peak detectivity of InAs/GaSb T2SL detectors is expected to drop for longer cutoff wavelengths as it is the case for MCT-based detectors (see Figure 3). As the cutoff of an MCT detector crucially depends on the Cd content in the composition, which becomes more challenging to control precisely and homogeneously towards longer cutoff wavelength, for more elaborate device concepts the device yield drops and in turn the detector price rises. This drawback does not exist for InAs/GaSb T2SLs.

#### **4. Real-Time MIR Backscattering Spectroscopy System**

In addition to our development of HOT InAs/GaSb T2SL IR detectors, we realized a demonstrator system for MIR backscattering spectroscopy. The operation principle of the demonstrator exploits the characteristic spectral diffuse reflection of solid chemical substances in the MIR that can be utilized for substance identification. Using a fast spectrally tunable quantum cascade laser (QCL) as the illumination source and a fast photodetector, the system is able to record IR spectra over more than 250 cm−<sup>1</sup> at rates of 1 kHz and therefore real-time spectroscopy. The high spectral scan speed of the system is ideal for fast changing scenarios or handheld operation as was demonstrated before. Here, we go beyond previous lab demonstrations of the measurement principle [28] as the system can run constantly without user intervention for several hours.

#### *4.1. Setup of the Demonstrator System*

The first core component of the system, the IR light source, is an agile wavelength-tunable external cavity quantum cascade laser (EC-QCL) developed by Fraunhofer IAF and Fraunhofer IPMS [28–30]. Its emission wavelength is defined by the deflection of a resonant micro-opto-electro-mechanical system (MOEMS) diffraction grating in Littrow-configuration, which is driven close to the resonance frequency of ~1 kHz (i.e., it harmonically oscillates around its zero-deflection position). Synchronized with the MOEMS oscillation, the EC-QCL is operated in pulsed mode with a pulse length of 100 ns and a repetition rate of about 500 kHz. Due to the resonant nature of the MOEMS scanner, the laser wavelength is continuously tuned and the full spectral range between 1060 cm–1 and 1350 cm−<sup>1</sup> provided by the QCL chip can be scanned in only half a MOEMS period, i.e., ~500 μs. However, typically the IR spectra are constructed from a full MOEMS period, as this increases the spectral resolution. For the parameters mentioned above, one achieves a typical spectral resolution of about 2 cm−<sup>1</sup> and a spectral broadening per pulse (i.e., per emission wavelength) also of <2 cm−<sup>1</sup> [27]. These performance parameters allow for spectroscopy on a number of solids and liquids with characteristic bands within the IR fingerprint region. The laser module itself is very compact, as can be seen in Figure 4a. Fraunhofer IAF and IPMS have also developed a non-resonant MOEMS EC-QCL, which allows addressing individual wavelength or (arbitrary) trajectories with scan frequencies of up to several ten hertz in an identical footprint [31].

**Figure 4.** Photographs showing (**a**) the compact design*s* of the external cavity quantum cascade laser (EC-QCL) with micro-opto-electro-mechanical system (MOEMS) diffraction grating and (**b**) the high operating temperature (HOT) T2SL IR detector from Fraunhofer IAF in a detector module from VIGO System.

The second core component of the system, a fast IR photodetector, detects the QCL radiation after it is diffusely reflected by the substance under investigation. The detector was chosen to meet the requirements set by the laser system. These were a MHz-bandwidth, to resolve each individual laser pulse, and a sufficiently long cutoff wavelength, to cover the required spectral range. As in diffuse reflection typical signal intensities are small, a high *D\** is also necessary. To achieve a portable and compact system, only TE-cooled detectors were considered. Up until now, only MCT detectors met these requirements and hence an MCT-based IR detector module was initially selected. Its specifications will be presented later, alongside those of the T2SL-based IR detector module. It differs from an MCT-based IR detector module from VIGO only in terms of the employed detector chip. Figure 4b demonstrates the small size of the detector module.

A picture of the demonstrator system and a simplified schematic showing its interior are presented in Figure 5. During the operation of the system, the QCL beam impinges on a continuously rotating sample platform. On this platform, several substances in the form of pills, powders or foils are arranged in small sample compartments. The samples are listed in Table 1.

**Figure 5.** Picture and simplified schematic of the demonstrator system. Backscattering IR spectra are continuously recorded using the tunable EC-QCL and the HOT InAs/GaSb T2SL IR detector. As the sample platform rotates, different substances are illuminated and subsequently identified after comparison with the database.

**Table 1.** Samples used in the demonstrator system.


As the sample platform rotates, the different substances are sequentially exposed to the incoming QCL beam. The rotational frequency of the sample platform sets the exposure time per substance. In our case, the sample platform rotates with a speed of ~4 rpm, resulting in an exposure time per substance of around ~1 s. After interaction with the respective substance, the laser radiation diffusely backscatters. In the case of the foils, the transmitted light is diffusely backscattered by a plate located

below them. Then, the collected portion of the backscattered light is deflected and focused to the fast IR detector. Each single laser pulse is detected, and an IR spectrum is constructed. The substance identification occurs by matching the measured fingerprint spectra to the previously acquired database spectra. The realization of the identification process is described in more detail in the following section.

#### *4.2. System Operation and Database Comparison*

Each spectrum measured with the demonstrator system contains a spectral signature, mainly due to the wavelength dependence of the responsivity of the detector. To determine the reflectivity of the different substances under test, the system-dependent spectral signature needs to be corrected for. Hence, at the beginning of the experiment a reference spectrum is acquired with a diffuse scattering plate. It is placed at the same distance as the rotating samples on the sample platform. During operation of the demonstrator system, the measured spectra are always divided by this reference.

With the demonstrator system, IR spectra are continuously recorded at a rate of 1 kHz. Typically, 25 spectra are averaged, corresponding to only 25 ms measurement time. Subsequently, the averaged spectra are compared to a database by using a cross-correlation algorithm that enables substance identification. The database is composed of MIR diffuse reflection spectra (in the case of pills or powders) or transmission spectra (in the case of foils). These spectra were previously acquired with the system itself or a commercial FTIR spectrometer from the same samples. Note that a FTIR measurement takes several minutes in order to achieve a spectral point spacing (~2 cm<sup>−</sup>1) comparable to our MOEMS EC-QCL-based measurement.

The averaged spectra are continuously compared to the database, while new spectra are still acquired within that time. Hence, no time is lost due to the post-processing of data. Regarding the comparison algorithm, we chose a standard cross-correlation comparison algorithm for simplicity, which is explained in more detail below. The idle time of the system between the recordings of two averaged spectra is sufficient to perform a comparison with the database. The database comprises 15 substances in the given case, which enables a comparison in ~10 ms when using this algorithm. The same algorithm could also be employed for an enlarged database; however, it would be at the cost of a slower database comparison.

In detail, we use the following procedure in our analysis. First, we calculate the normalized cross correlation (NCC) of the averaged spectrum to each database entry. It serves as measure for the similarity between two sets of data. Then, the largest cross-correlation (NCCmax) is selected. If NCCmax is larger than a threshold value (NCCth), the substance related to the respective database entry is considered as identified. Thereupon, the name of the substance is displayed on the demonstrator screen together with the averaged and the database spectrum. If NCCmax is smaller than NCCth, no output is returned. It needs to be mentioned that NCCth is an arbitrary yet fixed number. It is chosen based on the measurement conditions at the beginning of the experiment after an initial test run. It is set as high as possible in order to avoid false positives and as low as possible in order to avoid no returns.

The post-processing of data does not need to interfere with the acquisition of further spectra. To save time, this part can be delegated to different sub-systems or processors on demand, i.e., to a distributed computing architecture. This would also allow a more advanced data processing and analysis. In this context, resolving mixtures into their components or an automatized subtraction of spectral fingerprints from their background are typically of interest. Background subtraction becomes particularly important for the analysis of samples that are not bulk-like, e.g., when a potentially hazardous powder sample on an unknown substrate needs to be identified [32,33].

The presented approach for substance identification or discrimination with the demonstrator setup is solely based on matching IR fingerprint spectra to database spectra that were previously measured for known substances. Therefore, no precise knowledge about the specific nature of the vibrational or vibrational-rotational molecular bands is needed for identification. In fact, the set of substances on the sample platform was chosen arbitrarily. A modified set of substances could also be used as long as the corresponding spectra provide sufficient distinction for discrimination in the MIR.

#### *4.3. Detector Module Interchangeability*

To demonstrate the applicability of the T2SL-based IR module for spectroscopy, we replaced the MCT-based IR detector module in the demonstrator system with the T2SL-based IR detector module. The MCT-based module features a two-stage TE-cooled, photovoltaic IR detector that is illuminated via a hemispheric lens resulting in an optical area of 1 mm × 1 mm. This module has been specified with a cutoff wavelength of 10.6 <sup>μ</sup>m, a bandwidth of 100 MHz and a detectivity of 6.8 <sup>×</sup> <sup>10</sup><sup>8</sup> cm <sup>√</sup> Hz/W. The T2SL-based module features a four-stage TE-cooled, photoconductive, 50 μm × 50 μm-sized IR detector that is illuminated via a hyperhemispheric lens, resulting in an optical area of approximately 500 μm × 500 μm. It has been specified with a cutoff wavelength of 9.3 μm, a bandwidth of 10 MHz and a detectivity of 6.7 <sup>×</sup> <sup>109</sup> cm <sup>√</sup> Hz/W. The specifications of both detectors are listed in Table 2.

**Table 2.** Specifications of the two IR detector modules, based on an MCT detector and an InAs/GaSb T2SL detector, respectively.


As both modules feature equal packages and housings from VIGO System, replacing the MCT-based detector module, integrating the T2SL-based IR detector module into the setup and its optical alignment were straightforward. In the following, we report on the operation of the demonstrator system with both IR detector modules.

In Figure 6, exemplary diffuse reflection spectra are shown, which were measured during operation of the demonstrator system on commercial aspirin 500 mg pills and glucose powder with the T2SL-based and MCT-based IR detector module, respectively. The spectra measured with the MCT detector are normalized to their maximum value. The spectra acquired with the T2SL detector are multiplied by a constant, chosen to simplify comparison of the spectra. For both substances, the spectral trends measured with the two IR detectors are well comparable. Clearly, both detectors were able to resolve characteristic spectral features of the aspirin pills and the glucose powder, which allowed for substance identification by database comparison. As all substances on the sample platform (Table 1) have characteristic spectral features in the spectral range coverable with both IR detector modules, the T2SL-based IR detector module was also able to identify them during standard operation after fast comparison with the database in real-time.

**Figure 6.** Diffuse IR reflectance spectra obtained from aspirin 500 mg pills and glucose powder with the two IR detector modules featuring a HOT InAs/GaSb T2SL IR detector and a HOT MCT IR detector during operation of the demonstrator system.

#### *4.4. Long-Term Stabilty*

We studied the long-term stability of the demonstrator system with the T2SL detector. Following the identification procedure described before, we recorded the identified substance and the calculated NCC as a function of time. More than 50,000 measurements were performed in over 16 h of measurement time. Since the rotation speed of the sample platform is not constant, but rather fluctuates constantly in an uncontrolled manner, on average a new substance was identified every 1.1 ± 0.25 s. As there is no synchronization between the platform and our laser system, for each averaged spectrum, a different area was illuminated and used for analysis.

The results of the long-term stability test of the substance identification are presented in Figure 7. The time evolution is encoded in the figure through the color and size of the dots that are used to represent a single result. Substances for which the reference spectrum was obtained by the FTIR spectrometer are labelled accordingly. Overall, no significant drifts can be observed in the data. However, the distribution of the NCC strongly depends on the substance and varies depending on its form, i.e., foil, pill, or powder. In general, the transmission spectra through the foils show narrower distributions, whereas for the pills the distributions are typically wider. The assumption of our simple model—that each substance can be matched to the database using a single database spectrum—does not necessarily hold for the pills. This relates to the difficulties in solid dose manufacturing to achieve good homogeneity in the blending process. This also broadens the distributions of NCC values in our analysis.

**Figure 7.** Investigation of the long-term stability of the demonstrator system using the T2SL detector. The time information is encoded in color and size of the dots, i.e., early results are represented by big blue dots, whereas later results are given by smaller and greener dots. Please note the two groups for loratadine, where the lower ones correspond to false assignments.

The detailed analysis of the results of the long-term stability test with more than 50,000 measurements showed that in total 16 samples could not be identified and 68 measurements have been assigned to the wrong substance. Furthermore, 67 of these events correspond to a false assignment to loratadine. Since the corresponding NCCs of these events are smaller and form a separate group in Figure 7, these events could easily be rejected, if a more complex model were used. The 16 missing hits are attributed to ibuprofen 400 mg and the naproxen-based pill, certainly due to a too high NCCth value, which was chosen to be 1.91 in this experiment. In total, the error rate for missing hits is as low as 0.3‰ with the potential to be improved.

#### **5. Discussion**

The detector development and the presented application show the potential of TE-cooled T2SL-based IR detector modules for substance discrimination and in a broader scope for IR spectroscopy in general. This potential results from several key properties that the IR detector module exhibits. The first key property is a sufficiently long cutoff wavelength, which is tunable for InAs/GaSb T2SL detectors as it is for MCT detectors. The other key properties are a high detectivity and a bandwidth in the MHz range. A meaningful one-to-one comparison of the two IR detector modules used in the demonstrator system is problematic as they differ in several specifications such as size, operation mode, operating temperature, and cutoff wavelength (see Figure 3). The properties of the T2SL-based IR detector module can be slightly altered by changing the cooling power and hence the operating temperature. Rising the operating temperature of the InAs/GaSb T2SL photoconductor increases the cutoff wavelength and the detector bandwidth but reduces the detectivity. Due to the versatility of the InAs/GaSb T2SL material system and mature device processing at hand, detectors operating in more elaborate device concepts and with properties tailored to a particular application could be realized for the HOT range.

#### **6. Summary**

We demonstrated a RoHS-compatible, TE-cooled IR detector module based on an InAs/GaSb T2SL single element detector. For the fabrication of this module, we combined Fraunhofer IAF's expertise in the growth and processing of InAs/GaSb T2SLs with VIGO System's expertise in the fabrication of TE-cooled IR detector modules. This paper shows that this T2SL-based IR detector module and a commercial MCT-based IR detector module can be employed interchangeably in a compact and real-time MIR backscattering spectroscopy system. This system provides a very low error rate of only 0.3‰ in substance differentiation, which can be further improved. Furthermore, we showed that for equal operation mode, operating temperature, cutoff wavelength, and noise frequency, the detectivity of photoconductors based on InAs/GaSb T2SLs and MCT is comparable. This renders InAs/GaSb T2SLs promising for fully RoHS-compatible HOT IR photodetectors.

**Author Contributions:** Conceptualization, growth, and processing of the IR detector, R.M., J.N., V.D. and R.R.; conceptualization of the laser demonstrator, K.S. and M.H.; methodology, R.M. and M.H.; validation, Y.V.F. and K.S.; data curation, R.M.; writing—original draft preparation, R.M.; writing—review and editing, R.R., Y.V.F. and M.H.; visualization, R.M.; supervision, R.R. and R.O.; project administration, R.R. and R.O.; funding acquisition, R.R. and R.O. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Horizon 2020 Research and Innovation program under grant agreement no. 688265.

**Acknowledgments:** The authors thank Peter Holl and Stefan Hugger for developmental work (both with Fraunhofer IAF). We acknowledge the support from VIGO System during the developmental phase of HOT InAs/GaSb T2SL IR photoconductors, including detector characterization (Figure 3). Furthermore, we are grateful to VIGO System for the module integration of our HOT InAs/GaSb T2SL IR detectors.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Manganite Heterojunction Photodetector with Broad Spectral Response Range from 200 nm to 2** μ**m**

**Ru Chen 1, Zhiqing Lu <sup>2</sup> and Kun Zhao 1,\***


Received: 19 December 2019; Accepted: 16 January 2020; Published: 23 January 2020

**Abstract:** In this paper, we investigate the broad spectral photocurrent properties of the La0.67Ca0.33MnO3/Si (LCMO/Si) heterojunction from 200 nm to 2.0 μm, as the temperature increases from 95 to 300 K. We observed the junction's uniform responsivity in the visible range and five absorption peaks at 940 nm, 1180 nm, 1380 nm, 1580 nm, and 1900 nm wavelengths. The temperature showed effective affection to the photocurrents at absorption peaks and the transition point occurred at 216 K, which was also displayed in the temperature dependence of junction resistance. On the basis of the results, we propose a possible model involving the quantum size effect at the junction interface as the mechanism. This understanding of the infrared photodetection properties of oxide heterostructures should open a route for devising future microelectronic devices.

**Keywords:** manganite; heterostructure; photodetector

#### **1. Introduction**

Oxide semiconductor devices based on the perovskite oxide films, whose properties can be controlled by magnetic field, electric field, and light irradiation, have attracted a great deal of interest. Experiments confirmed that manganite-based perovskite-type oxides have excellent ultraviolet (UV) photoresponse characteristics with ultrafast response-time of picosecond and high sensitivity, which makes this class of materials potentially useful for UV sensor applications [1,2]. Furthermore, manganite heterojunctions offer the features of tunability by magnetic and electric fields, high-sensitivity to light illumination and high carrier mobility, suggesting many possible applications and research directions including information storage, optoelectronics information processing, and advanced sample preparation techniques associated with microstructure modulate research [3–7]. In addition, similar to many optical materials with chemical stability, manganite heterojunctions are insensitive to harsh physical environment such as fluctuations of temperature and pressure, suggesting a potential application of manganite heterojunction photodetectors in harsh environments for the need of oil and gas optics [8–12]. Integrating the perovskite-type transition metal oxides with the silicon (Si)-based semiconductor technology would also introduce the possibility for a multifunctional microelectronic device [13–15].

Si photodetectors have already found wide acceptance for visible light applications, while it has small absorption coefficient in near-infrared (NIR) wavelength range because of the cut off wavelength of ~1100 nm. Now most NIR photodetectors were composed of PbS, PbSe, or InGaAs. The toxic precursors, such as Pb, Se, and As, was usually used to synthesize these materials. It is a meaningful thing to find non-toxic and pollution-free material for photodetector working at NIR wavelength range.

The infrared (IR) spectrum has become an important method to study the lattice distortion and been applied to investigate the photoconductive effect in perovskite manganese oxides, where mid-infrared or far-infrared spectra was used to explain the complex physical process in manganites such as the electronic transition, electron-phonon interaction, coupling between lattice, orbital, and spin, etc., [16–22]. In this paper the photocurrent response spectrum between 200 nm and 2 μm of the heterojunction La0.67Ca0.33MnO3/Si (LCMO/Si) is reported. The temperature dependence of the photocurrent response of the sample was investigated to reveal more information related to the photoelectric response, and selective absorption peaks were observed. The mechanism about the results is also discussed in the paper.

#### **2. Materials and Methods**

The LCMO/Si heterojunction was fabricated using the facing target sputtering technique. A 100 nm thickness LCMO layer was grown on a 0.5 mm thick n-type Si (001) wafer. The wafer temperature was kept at 680 ◦C with the oxygen pressure being 60 mTorr during deposition. Immediately after each deposition, the vacuum chamber was back-filled with 1 atm oxygen gas.

The photocurrent of the sample was detected by the spectral response measurement system, as shown in Figure 1. The system was designed to measure the UV and IR spectral responsivity characteristics of samples in low temperature environment. The operation was automatically controlled, and the system maintained good closure during the measurement process. The selected all-reflected-light-route system, UV, visible light or IR, can be switched automatically with maximum light path coupling efficiency. The diameter of the light spot was 3 mm. The light intensity was calibrated using the spectrum of a commercial UV-100L Si photodiode (from OSI Systems Inc., Hawthorne, CA, USA) and the spectral responsivity was measured by a monochromator.

**Figure 1.** Spectral response measurement system.

The LCMO/Si heterojunction for the photoelectric measurements was cut into 5 × 5 mm and two colloidal silver electrodes were prepared on the LCMO film and Si wafer. The sample was placed in an airtight holder with a quartz window and connected with the spectral response measurement system (Figure 2a). The typical current-voltage curves of the LCMO/Si heterojunction, shown in Figure 2b, were measured in the dark by tuning the applied voltage with a pulse-modulated voltage source at 300 and 60 K. The forward bias was defined as the current flowing from the upper LCMO layer to Si substrate. Thus the diodelike rectification characteristic can be ascribed to the presence of LCMO/Si interfacial potential because of the carrier diffusion.

**Figure 2.** (**a**) The setup of La0.67Ca0.33MnO3/Si (LCMO/Si) heterojunction for the spectral response measurement. (**b**) The current-voltage curves of the LCMO/Si heterojunction at 300 and 60 K.

#### **3. Results and Discussions**

The junction resistance *R*<sup>j</sup> in LCMO/Si junction was measured with the temperature. As shown in Figure 3, *R*<sup>j</sup> strongly depends on the bias, e.g., when the bias was turned from 20 μA to −20 μA *R*<sup>j</sup> changed from 166.0 kΩ, 20.0 kΩ, and 15.1 kΩ to 178.1 kΩ, 98.2 kΩ, and 15.6 kΩ at 95 K, 202 K, and 300 K. In addition, taking the bias of 50 μA as an example, *R*<sup>j</sup> decreased slightly from the beginning 95 K to 172 K and had a sharp change from 78 kΩ at 172 K to 8.2 kΩ at 216 K with a corresponding rate of 1.6 kΩ/K. Subsequently *R*<sup>j</sup> maintained small change of about 0.01 kΩ/K till 300 K.

**Figure 3.** The temperature dependence of junction resistance *R*<sup>j</sup> of a LCMO/Si junction under (**a**) the positive current bias and (**b**) the negative current bias.

Figure 4a displays the photocurrent (*PI*) spectrum of the LCMO/Si junction under zero bias in the wavelength range of 200 nm < λ < 2200 nm. The junction's responsivity was spectrally uniform in the visible range, while five absorption peaks P1, P2, P3, P4, and P5 were observed at λ<sup>1</sup> = 1940 nm, λ<sup>2</sup> = 1180 nm, λ<sup>3</sup> = 1380 nm, λ<sup>4</sup> = 1580 nm, and λ<sup>5</sup> = 1900 nm wavelengths in each temperature because of the absorption characteristics of the LCMO/Si junction, and the peak value decreased with the increase of the wavelength. The temperature dependences of the photocurrent response *PI*<sup>P</sup> at the five absorption wavelengths are shown in Figure 4b. *PI*<sup>P</sup> monotonically increased from 0.0065 A/W, 0.012 A/W, 0.0115 A/W, 0.004 A/W, and 0.002 A/W at 108 K to 0.0122 A/W, 0.017 A/W, 0.0155 A/W, 0.006 A/W, and 0.0025 A/W at a turning point of 216 K and then dropped 0.004 A/W, 0.005 A/W, 0.005 A/W, 0.002 A/W, and 0.001 A/W at 300 K for selected wavelengths of λ1, λ2, λ3, λ4, and λ5.

**Figure 4.** Photocurrent (*PI*) spectrum for various temperatures (**a**) and *PI* peaks (**b**) for 940 nm (P1), 1180 nm (P2), 1380 nm (P3), 1580 nm (P4), and 1900 nm (P5) of a LCMO/Si junction.

Si cannot produce strong absorption features. Most of heterojunctions of manganite-based perovskite-type oxides exhibit the properties of p-n junction and quantum size effect can be produced when the thickness of the potential well material is about 50 nm thick. If quantum size effect occurs, the energy is quantized in the direction of the vertical interface, which will lead to the quantization of energy absorption in the material. Since a thin SiO2 layer of 3.6 nm thick exists in the LCMO/Si heterojunction [23], a quantum size effect was expected to occur. The interval of adjacent energy levels in an infinite quantum well is described as:

$$
\Delta E\_{n,n+1} = \pi^2 \eta^2 m\_n^{-1} d\_w^{-2} (n+1/2) \propto (n+1/2) \tag{1}
$$

where *mn* is the electron effective mass and *dw* is the quantum well width. Thus, (Δ*En*,*n*+<sup>1</sup> − Δ*En*+1,*n*<sup>+</sup>2) is independent on *n* and

$$(\Delta E\_{n,n+1} - \Delta E\_{n+1,n+2}) - (\Delta E\_{n+1,n+2} - \Delta E\_{n+2,n+3}) \approx \Delta E\_{n,n+1} - 2\Delta E\_{n+1,n+2} + \Delta E\_{n+2,n+3} \tag{2}$$

As for present five special wavelengths λ<sup>n</sup> (n = 1, 2, 3, 4, and 5),

$$(\lambda\_1^{-1} - \lambda\_2^{-1}) - 2(\lambda\_2^{-1} - \lambda\_3^{-1}) + (\lambda\_3^{-1} - \lambda\_4^{-1}) \approx 0.0034\tag{3}$$

and

$$(\lambda\_2^{-1} - \lambda\_3^{-1}) - 2(\lambda\_3^{-1} - \lambda\_4^{-1}) + (\lambda\_4^{-1} - \lambda\_5^{-1}) \approx 0.0030\tag{4}$$

Here, the above two similar data suggested that the present model involving the quantum size effect was adopted as the mechanism of IR photocurrent in LCMO/Si.

Noise performance is a critical factor for evaluating a detector. The noise current *I*<sup>n</sup> is about 10−<sup>4</sup> A/W in dark and is very low compared to the responsivity PI of the LCMO/Si junction when the light was on. The detectivity *D*\* is determined by the ratio of *PI* and *I*n, and *D*\* = *PI* (*fS*) <sup>1</sup>/2/*I*n, where *f* is the amplifier frequency bandwidth (500 MHz) and *S* is the detector area (~7.065 mm2). Thus *D*\* is estimated to be about 2.38 <sup>×</sup> 103 Hz1/2m, 2.97 <sup>×</sup> 103 Hz1/2m, 2.97 <sup>×</sup> 103 Hz1/2m, 1.19 <sup>×</sup> 103 Hz1/2m, and 0.59 <sup>×</sup> 103 Hz1/2m at 300 K for selected wavelengths of <sup>λ</sup>1, <sup>λ</sup>2, <sup>λ</sup>3, <sup>λ</sup>4, and <sup>λ</sup>5, suggesting that the LCMO/Si junction could be well-suited as an IR detector.

Perovskite-type oxides detectors possess a number of significant characteristics, and are ideally suited to detect small changes in a relatively large background level of incident energy, which can be used over a large spectral bandwidth. Here it has been shown that a specific manganite heterojunction has the ability to be an IR detector since it can produce photocurrent in the IR regime. The devices have a number of important characteristics (low cost, low power, good performance, wide operating range of temperature, a high degree of environmental stability, and reliability) which make them ideal for a

range of applications from consumer and commercial to military requirements. LCMO/Si junction is a new material for photodetector fabrication compared to traditional materials. It is anticipated that manganite heterojunction IR detectors will assume an ever growing importance in our society over the next few years.

#### **4. Conclusions**

In conclusion, we fabricated a manganite-based heterojunction by depositing a LCMO thin film on the Si substrate. The broad spectral photocurrent effect of the junction was systematically studied in a temperature range from 95 to 300 K. The responsivity of LCMO/Si heterojunction was spectrally uniform in the visible range. Five absorption peaks occurred at 940 nm, 1180 nm, 1380 nm, 1580 nm, and 1900 nm in the IR range, which is explained in terms of a quantum size effect model since an interface existed in the present photodetector. However, relative contributions from individual interface are still not clear and further studies is needed to clarify the *PI* mechanisms.

**Author Contributions:** Writing—original draft preparation, R.C. and Z.L.; writing—review and editing, K.Z. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the National Nature Science Foundation of China, grant number 11574401.

**Acknowledgments:** We thank H. K. Wong and Y.C. Kong for the sample preparation in The Chinese University of Hong Kong.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **InAs**/**InAsSb Strained-Layer Superlattice Mid-Wavelength Infrared Detector for High-Temperature Operation**

#### **Gamini Ariyawansa \*, Joshua Duran, Charles Reyner and John Scheihing**

Air Force Research Laboratory, Sensors Directorate, Wright-Patterson Air Force Base, OH 45433, USA; joshua.duran.2@us.af.mil (J.D.); charles.reyner.1@us.af.mil (C.R.); john.scheihing@us.af.mil (J.S.) **\*** Correspondence: Gamini.Ariyawansa.2@us.af.mil

Received: 31 October 2019; Accepted: 19 November 2019; Published: 22 November 2019

**Abstract:** This paper reports an InAs/InAsSb strained-layer superlattice (SLS) mid-wavelength infrared detector and a focal plane array particularly suited for high-temperature operation. Utilizing the *n*B*n* architecture, the detector structure was grown by molecular beam epitaxy and consists of a 5.5 μm thick *n*-type SLS as the infrared-absorbing element. Through detailed characterization, it was found that the detector exhibits a cut-off wavelength of 5.5 um, a peak external quantum efficiency (without anti-reflection coating) of 56%, and a dark current of 3.4 <sup>×</sup> 10−<sup>4</sup> A/cm2, which is a factor of 9 times Rule 07, at 160 K temperature. It was also found that the quantum efficiency increases with temperature and reaches ~56% at 140 K, which is probably due to the diffusion length being shorter than the absorber thickness at temperatures below 140 K. A 320 × 256 focal plane array was also fabricated and tested, revealing noise equivalent temperature difference of ~10 mK at 80 K with f/2.3 optics and 3 ms integration time. The overall performance indicates that these SLS detectors have the potential to reach the performance comparable to InSb detectors at temperatures higher than 80 K, enabling high-temperature operation.

**Keywords:** Infrared detector; strained layer superlattice; InAs/InAsSb; absorption coefficient; barrier detector; high operating temperature

#### **1. Introduction**

Lower cost, size, weight, and power (C-SWaP) have become a requirement for many infrared imaging systems. A great impact on C-SWaP could be achieved through high operating-temperature (HOT) [1] sensors and focal plane arrays (FPAs), which in turn require developing suitable sensor materials exhibiting high uniformity, high stability, and good electrical and optical properties. One class of materials that has the potential to do just that are III–V, antimony-based, strained-layer superlattices [2] (SLSs), which have already shown impressive results. HOT [1,3] capability has been the primary goal in the mid-wavelength infrared (MWIR) band and a few demonstrations [4] have been already reported. Research groups have explored different detector architectures (*n*B*n* [5], *n*B*p* [6], *p*B*p* [7], *X*B*n* [8], CBIRDs [9] etc.) and pixel geometries [5,10] across a multitude of SLS designs [11–16] to mitigate generation-recombination (G-R) current and surface-leakage current, while maximizing electrical/optical properties. Among those, detectors with InAs/InAsSb SLSs incorporated in the *n*B*n* architecture have received special attention due to high carrier lifetime [16–18] and reduced complexity of SLS growth [4]. Since the first demonstration of SLSs for infrared (IR) detectors [12], InAs/Ga(In)Sb SLSs continue to improve [19], while InGaAs/InAsSb SLSs [15,20] have also shown promising results. Researchers are currently addressing the poor hole mobility [21] and carrier localization [22,23] effects in *n*-type SLS to improve the diffusion length. They have also explored *p*-type SLS detectors, but surface passivation leading to surface leakage current [24,25] remains an insurmountable problem. In this paper,

the focus is on a detector utilizing *n*-type InAs/InAsSb SLSs based on the *n*B*n* architecture, with the emphasis on HOT capability. Detector design, material characterization, and detector performance are discussed in detail, while FPA fabrication and testing are briefly discussed.

#### **2. Strained-Layer Superlattice (SLS) Design and Detector Structure**

Group III–V antimony-based SLSs are periodic structures of thin layers of semiconductor materials typically grown on GaSb substrates, which comprise a band-engineered artificial infrared material. In the InAs/InAsSb SLS reported here, the unit cell consists of 16 ML InAs and 6 ML InAsSb layers and the total unit cell thickness was then adjusted to achieve the desired bandgap and spectral response. The band structure of the superlattice was calculated using NRLMultiband software [26] and the band parameters such as the bandgap and carrier effective masses as well as material properties such as absorption coefficient were obtained. Figure 1a illustrates the conduction band (CB) and valence band (VB) profile of the bulk constituents of the superlattice along with superlattice minibands (HH1 and C1) and its bandgap (*Eg*). The electron and hole probability distributions are also shown, indicating nearly free electrons in the C1 miniband and heavily confined holes in the HH1 miniband. This is a typical feature of InAs/InAsSb SLSs and one can optimize the design [15] to maximize the electron and hole wave function overlap in order to maximize the absorption coefficient and vertical hole mobility. The designed value of *Eg* is 234 meV (5.3 μm) at 80 K in order to cover the entire 3–5 μm atmospheric band (MWIR).

**Figure 1.** (**a**) InAs/InGaAs strained-layer superlattice (SLS) design showing the conduction and valence band profile for the bulk material as well as the valence and conduction minibands of the superlattice. The superlattice bandgap is also indicated as *Eg*. (**b**) Structure of the *n*B*n* detector consisting of SLS layers as the absorber and contact layers and a bulk AlGaAsSb layer as the electron barrier.

The SLS shown in Figure 1a was incorporated into an *n*B*n* architecture in order to build a detector. Figure 1b shows the complete structure of the *n*B*n* detector grown on GaSb substrate by molecular beam epitaxy at a commercial foundry. The active elements in the structure include a 5.5 μm thick *n*-type SLS absorber, a 0.2 μm thick AlGaAsSb electron barrier layer, and a 0.2 μm thick *n*-type SLS top contact layer. For single element device characterization, mesas were fabricated using standard photolithography, wet chemical etching, and metallization processes. A fully fabricated mesa device is illustrated in Figure 1b. In addition, a metallic mirror was also deposited on the backside of the wafer; this mirror provides a double pass optical geometry under front-side illumination, which approximates the performance of a backside-illuminated FPA.

#### **3. Characteristics of Detectors**

Material and device characterization was performed at a range of temperatures from 78 to 300 K. The absorption coefficient (α) was determined from transmission and reflection measurements using a

Fourier transform infrared (FTIR) spectrometer; the details of the method are discussed elsewhere [27]. The absorption coefficient spectra for the superlattice reported here is shown in Figure 2 at 78 and 300 K. Also indicated in the Figure 2 is the cut-off wavelength (~5.2 μm) corresponding to the inflection point of the spectrum (also the edge of the Urbach [28] tail). This value is very close to the designed bandgap of the superlattice (5.3 μm). The value of α at 4 μm and 300 K is 3081 cm−<sup>1</sup> and the average α in the 3–5 μm band is 3461 cm−1. Furthermore, the cut-off wavelength shifts to about 6.5 μm at 300 K, as expected. These spectra were used for calculating the absorption efficiency in the absorber for comparison against the measured quantum efficiency (QE) of the detector, which will be discussed later.

**Figure 2.** Measured absorption coefficient spectra of the InAs/InAsSb SLS at 78 and 300 K.

The fully processed detectors were packaged in leadless chip carriers and wire-bonded to the chip carrier leads to make electrical contacts. Dark current and photocurrent measurements were carried out after mounting the packaged devices in a liquid nitrogen pour-filled dewar. The dark current-voltage-temperature (IVT) characteristics measured in the 80–240 K temperature range are shown in Figure 3a. Here, the bias polarity is defined as negative (positive) when a negative (positive) voltage is applied on the top contact. The *n*B*n* detector is operated under negative bias where majority electrons flowing from the top contact to the bottom contact are blocked by the electron barrier, while photo-generated minority holes are collected at the top contact. Figure 3b shows the variation of the dark current density (*Jd*) under –0.2 V bias with temperature (*T*) with a linear fitting to the experimental data at temperatures higher than 160 K. Based on the slope of *Jd*/*T*<sup>3</sup> vs. 1/*T* plot, the activation energy was calculated to be approximately 203 meV. This value is very close to the bandgap of the SLS at ~200 K (confirmed by the spectral cut-off discussed later). It also confirms that the dark current at T > 160 K is diffusion limited. At lower temperatures, the dark current of SLS-based *n*B*n* detectors is typically limited by generation/recombination (G-R) current which is characterized by an activation energy of approximately half the bandgap. This is confirmed by an activation energy of ~115 meV, obtained from the slope of *Jd*/*T*<sup>3</sup> vs. 1/*T* plot for T < 100 K, which is approximately half of the bandgap.

Photocurrent was measured using a calibrated blackbody and a set of notch filters at a few specific wavelengths. Then, the quantum efficiency was calculated through radiometric analysis. The resulting quantum efficiency of the detector and its variation with bias at 80 K and 3.4 μm are shown in Figure 4. It should be noted that this is the external quantum efficiency of the detector measured without using an antireflection (AR) coating. From Figure 4, it appears that the detector reaches 90% of max QE (i.e. the turn-on voltage) at a bias of ~ –0.2 V and QE increases slowly when the bias voltage magnitude is increased further. Unlike for a homojunction diode, *n*B*n* detectors are not expected to operate at 0 V, as there is no built-in field in the structure. However, pushing the operating bias voltage near 0 V is preferred, which can be done through optimization of the barrier band alignment and doping. The turn-on voltage at 80 K is reasonably small (200 mV), which decreases with increasing temperature [20].

**Figure 3.** (**a**) Dark current-voltage-temperature (IVT) characteristics of a 400 × 400 μm size *n*B*n* detector and (**b**) Arrhenius plot at a bias voltage of –0.2 V. The linear fit to the current at T > 160 K yields an activation energy of 203 meV.

**Figure 4.** Variation of the external quantum efficiency (no AR coating) of the detector with bias at 80 K and 3.4 μm.

Another important detector characteristic is the spectral response, which is typically measured using a spectrometer. An FTIR spectrometer was used to measure the relative spectral response of this device, which was scaled to spectral QE using the calibrated blackbody measurements. The spectral QE and its variation with temperature under a bias of −0.2 V is shown in Figure 5a. As observed, the cut-off wavelength (inflection point) at 80 K is approximately 5.25 μm (236 meV), which is extremely close to the designed bandgap of the superlattice (234 meV). If the 50% of peak QE is considered, the corresponding wavelength at the band edge is approximately 5.12 μm. This value will be considered as the cut-off for a comparison of the dark current against that of mercury cadmium telluride (MCT) detectors described by Rule 07 [29], as discussed in the next section. As the temperature is increased, the cut-off wavelength increases as expected, but the increase in peak QE is not ideal. As shown in Figure 5b, the magnitude of the peak QE (at 4.2 μm) increases with temperature in the 80–140 K range. At 140 K and beyond, QE saturates, indicating that the QE is likely absorption-limited at these temperatures. This value of QE will be compared with the maximum theoretical value, equal to the absorption efficiency, in the following section. The overall result indicates that this detector's QE peaks at T > 140 K, making it suitable for HOT detectors.

**Figure 5.** (**a**) Spectral quantum efficiency of the detector at –0.2 V at various temperatures and (**b**) variation of the peak quantum efficiency with temperature at 4.2 μm under –0.2 V.

#### **4. Discussion**

Using the experimentally determined absorption coefficient spectra shown in Figure 2, the total absorption in the 5.5 μm thick absorber was calculated [27] using the transfer matrix method under frontside illumination. This calculated absorption efficiency corresponds to the maximum theoretical quantum efficiency of the detector. A comparison of the absorption at 78 and 300 K is compared with the detector QE measured at 80 and 240 K, as shown in Figure 6. It was not possible to measure both absorption efficiency and the quantum efficiency at the same temperature, therefore, close values for the temperature were chosen for this comparison. It is clear that the value of QE at high temperatures is in good agreement with the absorption efficiency. The minor discrepancy observed in the overall spectral shape and the band edge could be due to two reasons: (i) the temperature difference (77 K vs. 80 K and 240 K vs. 300 K), impacting the bandgap and the cut-off wavelength, and (ii) optical resonant effects in the structure, which are very sensitive to the refractive index of the layers in the structure. Currently, the model considers real refractive index values reported in the literature, however, the actual values of the refractive index for the layers in the structure should be measured at corresponding temperatures and used in the model in order to further improve the simulated results.

Figure 6 also confirms that the maximum QE, shown in Figure 5b, is very close to the maximum theoretical QE, i.e. collection efficiency is near unity. However, for T < 140 K, the QE decreases as the temperature is decreased, indicating collection efficiency is less than 1 at these temperatures. Assuming that the diffusion length, *L*, is lower than the absorber thickness (= 5.5 μm), the absorption in a portion of the absorber with a thickness equal to *L* measured from the barrier/absorber interface, as indicated in Figure 1b, was calculated. Then, the value of *L* that gives the best fit between the absorption and QE was determined. At *T* ~80 K, this value was found to be ~4.8 μm. In other words, QE measured at 80 K corresponds to the collection of carriers generated within a ~4.8 μm region of the absorber. In Figure 6, the absorption efficiency spectra at 78 K correspond to *L* = 4.8 and 5.5 μm and the spectrum when *L* = 4.8 μm fits reasonably well with the quantum efficiency spectrum measured at 80 K. Moreover, when the temperature is increased from 80 to 140 K, *L* increases from 4.8 μm to 5.5 μm, respectively. While this is one straightforward way to explain the QE dependence on temperature, there could be other effects leading to the same observation such as variation in the barrier band alignment to the absorber with temperature and recombination of trapped holes at the interfaces [4].

As of today, MCT technology is still the leading technology for HOT detectors, while SLS technology has become a viable competitor. Therefore, it is worthwhile comparing the dark current between state-of-the-art MCT detectors described by Rule 07 [29] and the SLS detector reported in this paper. Defining the cut-off wavelength as the wavelength near the band edge corresponding to 50% of the peak QE (see Figure 5a), the cut-off wavelength values at different temperatures were determined and the Rule 07 dark current corresponding to those cut-off values and temperatures were calculated. It was then found that the dark current of this SLS detector is approximately a factor of 9, 4, and 3 higher than that of Rule 07 at 160, 180, and 200 K temperatures, respectively.

**Figure 6.** Comparison of the quantum efficiency measured at 80 and 240 K against the absorption efficiency calculated at 78 K (for *L* = 4.8 and 5.5 μm) and 300 K (for *L* = 5.5 μm), indicating a good agreement in the quantum efficiency values as well as the spectral shape. The highest temperature for quantum efficiency data available is 240 K and it was chosen to compare against the absorption at 300 K.

The dark current and quantum efficiency of the detector reported in this paper are comparable to similar InAs/InAsSb SLS detectors recently reported in the literature. Ting et al. [4] have reported an InAs/InAsSb SLS *<sup>n</sup>*B*<sup>n</sup>* detector with a quantum efficiency of ~52% and dark current of 9.6 <sup>×</sup> <sup>10</sup>−<sup>5</sup> <sup>A</sup>/cm2 (a factor of ~4.5 higher than Rule 07) at ~157 K. With a detailed analysis of dark current characteristics, Rhiger et al. [30] have reported a similar InAs/InAsSb SLS *n*B*n* detector exhibiting a dark current 5 times higher than Rule 07. Furthermore, a comprehensive review of antimony-based detectors has been reported by Rogalski et al. [2] With this level of dark current performance and external quantum efficiency >50%, it can be predicted that SLS detectors are well within the reach of performance of InSb detectors but at high temperatures, promising as a candidate for HOT detectors.

To demonstrate the imaging performance, a 320 × 256 detector array with 30 μm pitch was fabricated, flip-chip bonded to a commercial readout integrated circuit chip (FLIR ISC9705), and tested to obtain performance metrics. As shown in Figure 7, the FPA exhibits promising results, including a median noise-equivalent temperature difference (NEDT) of 10 mK. This FPA also showed good uniformity and image quality up to about 140 K. Furthermore, these performance metrics agree with the characteristics measured at the single element detector level, discussed earlier in this paper.

**Figure 7.** (**a**) Noise-equivalent temperature difference (NEDT) histogram of a 320 × 256 focal plane array (FPA) at 80 K; (**b**) NEDT operability map; and (**c**) an image taken at 80 K with f/2.3 optics and a 3 ms integration time.

#### **5. Conclusions**

A MWIR *n*B*n* detector designed using InAs/InAsSb SLS was reported. Detector characteristics were measured and analyzed with an emphasis on high temperature operation. At 160 K, this detector exhibits dark current of 9 times Rule 07 and peak quantum efficiency of 56% (~84% of internal quantum efficiency). The turn ON voltage is at or below −200 mV over the full temperature range. It was estimated that the diffusion length of the SLS is approximately 4.8 μm at 80 K, which increases to a value comparable to the absorber thickness (5.5 μm) when the temperature is increased to 140 K. Comparing the calculated absorption efficiency and the measured detector quantum efficiency, it was possible to conclude that the detector exhibits nearly 100% collection at temperatures higher than 140 K. While the performance metrics reported here do not meet those of InSb detectors yet, SLS technology continues to improve with the promise that it has the potential to deliver future HOT detectors required for many applications.

**Author Contributions:** Conceptualization, G.A., J.D., C.R. and J.S.; methodology, G.A., J.D., C.R. and J.S.; software, G.A. and J.D.; validation, G.A., J.D., C.R. and J.S.; formal analysis, G.A. and J.D.; investigation, G.A. and J.D.; resources, G.A., J.D., C.R. and J.S.; data curation, G.A. and J.D.; writing—original draft preparation, G.A.; writing—review and editing, J.D., C.R. and J.S.; supervision, C.R. and J.S.; project administration, J.S.; funding acquisition, J.S.

**Funding:** This work was funded by the Air Force Research Laboratory, Sensors Directorate under project "III–V Focal Plane Array Development Using Novel Superlattices".

**Acknowledgments:** The authors would like to acknowledge the support from the scientists/engineers at the Naval Research Laboratory (NRL) by providing us with the NRL MULTIBANDS software.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
