**Analysis and Simulation of Forcing the Limits of Thermal Sensing for Microbolometers in CMOS–MEMS Technology**

#### **Hasan Gökta¸s 1,\* and Fikri Serdar Gökhan <sup>2</sup>**


Received: 25 September 2019; Accepted: 25 October 2019; Published: 29 October 2019

**Abstract:** Room-temperature highly sensitive microbolometers are becoming very attractive in infrared (IR) sensing with the increase in demand for the internet of things (IOT), night vision, and medical imaging. Different techniques, such as building extremely small-scale devices (nanotubes, etc.) or using 2D materials, showed promising results in terms of high sensitivity with the cost of challenges in fabrication and low-noise readout circuit. Here, we propose a new and simple technique on the application of joule heating on a clamped–clamped beam without adding any complexity. It provides much better uniformity in temperature distribution in comparison to conventional joule heating, and this results in higher thermal stresses on fixed ends. This consequently brings around 60.5× improvement in the overall temperature sensitivity according to both theory and COMSOL (multiphysics solver). The sensitivity increased with the increase in the stiffness constant, and it was calculated as 134 N/m for a device with a 60.5× improvement. A considerable amount of decrease in the operation temperature (36× below 383 K and 47× below 428 K) was achieved via a new technique. That's why the proposed solution can be used either to build highly reliable long-term devices or to increase the thermal sensitivity.

**Keywords:** microbolometer; infrared sensor; complementary metal-oxide semiconductor (CMOS); high sensitivity; temperature sensor; microresonator; MEMS; clamped–clamped beam; thermal detector

#### **1. Introduction**

MEMS/NEMS (Micro/Nano-Electro-Mechanical Systems) resonators got tremendous attention in the last decades, especially with the increase in the demand for the internet of things (IOT), biosensors, gas sensors, and infrared (IR) sensing applications (night vision, gas detection, medical imaging, etc.). Photon detectors [1,2] and microbolometers [3,4] are the two widely used and well-known competitors in building IR sensors. Photon detectors suffer from the requirement of cryogenic cooling, intrinsic noise, relatively high cost, fabrication complexity, and being bulky and expensive. On the other hand, room-temperature microbolometers not only eliminate all these problems, but can also easily be implemented in widely used CMOS (complementary metal-oxide semiconductor) processes [3]. The working principle of the microbolometer is based on the conversion of incident radiation into heat via a plasmonic absorber, and then conversion of this heat into an electrical signal via a temperature sensor. This electrical signal can be either resistance change (non-resonant) or frequency change (resonant type), depending on the device type. Resonant sensors [5–7], in contrast to non-resonant sensors [8], are the most popular because they offer significant advantages [6], such as high-quality factor of 1 million [9], ultra-low-noise measurement, and highly accurate measurement. The working

principle of MEMS-type resonant-based thermal sensors is based on the resonance frequency shift, with respect to change in the temperature.

Many different techniques have been applied to increase the sensitivity of resonant-type microbolometers, where the thermal sensitivity strongly depends on the temperature coefficient of frequency (TCF). Extremely small devices [10,11] have higher sensitivity and a higher TCF, but they have challenges, such as difficulties in fabrication, high density integration, and low-noise readout circuit implementation. Using NEMS offered better sensitivity and TCF by achieving better thermal isolation [6,12,13], but sacrifices low stiffness constant. High quality factor [9,14,15] was achieved to detect a much smaller frequency shift and consequently enables higher thermal sensitivity and a higher TCF, but this method either requires vacuum environment or fabrication complexity or fabrication compatibility problem with CMOS process. Using phase change materials [16] in building cantilever-type resonators increased their thermal sensitivity and TCF, while bringing the fabrication challenges and CMOS compatibility problem. A fixed–fixed beam-type device was built [17] to drastically increase the TCF in comparison to other well-known resonant-type devices (cantilever, tuning fork, and free-free beam). The |TCF| of 4537 ppm/K was achieved in [5], and 19,500 ppm/K was achieved in [17], with fixed–fixed beam, where CMOS allowed high density integration and low-noise readout circuit. Later, this method was improved, and a possibility of 31× improvement, with a |TCF| of 2,178,946 ppm/K in thermal sensing, was demonstrated [7] by not only combining joule heating with ambient temperature change but also keeping the pull-in voltage as small as possible. However, nonuniform temperature distribution via joule heating prevented this approach from going down to smaller scale and consequently having higher thermal sensitivity due to relatively high thermal stresses.

We present a new type of joule heating technique in this work that allows much better temperature uniformity throughout the beam in comparison to conventional joule heating. In contrast to [7], here, joule heating is applied on sidebars attached on the fixed ends of the clamped–clamped beam. This uniform temperature profile not only increases the thermal stress on the fixed ends but also allows building design with relatively higher stiffness constant, while keeping the temperature limit around 530 K [18]. This increase in return results in the increase in the sensitivity multiplier and TCF. The sensitivity multiplier was computed as 60.5×, with a 30 μm long and 0.9 μm thick device, according to both COMSOL and theory in comparison to the possibility of 31× demonstrated in [7]. The |TCF| for 60.8× improvement was calculated as 3,991,168 ppm/K in comparison to 2,178,946 ppm/K in [7], 19,500 ppm/K in [17], 4537 ppm/K in [5], 30 ppm/K in [6], 548 ppm/K in [10], 86.2 ppm/K in [19], and 29.4 ppm/K in [13]. Furthermore, this uniform heating allows the design to have the same thermal sensitivity, with a conventional joule heating scheme [7], while drastically decreasing the operation temperature and enabling long-term operation. In other words, the proposed design methodology can be used either to increase the thermal sensitivity or to decrease the operation temperature for high reliability.

Using aluminum or composite structure in building a clamped–clamped beam in a CMOS 0.6 process showed almost no difference in the sensitivity multiplier, even at different sizes, where the stiffness constant stayed the same. That's why optimum design with a critical temperature below 530 K can be designed in any fabrication process/technology (CMOS, silicon, Silicon on insulator (SOI), etc.), regardless of the material and size, as long as the stiffness constant is limited to 134 N/m.

#### **2. Theory, Design, and Optimization**

The main mechanism behind the wide-range frequency tuning is the thermal stresses created via joule heating, as demonstrated for the first time in [17]. Another study [7] conducted on CMOS clamped–clamped beam proved that thermal stress is the dominant factor on the temperature sensitivity, and the stresses caused by the increase in the pull-in force adversely affect the sensitivity. The main goal of this study is to increase the temperature uniformity throughout the beam to get further increase in the thermal stress at a relatively low temperature, and consequently get higher thermal sensitivity for microbolometer applications.

#### *2.1. Design for Uniform Heating*

One of the main disadvantages of joule heating is the well-known nonuniform temperature distribution throughout the conventional clamped–clamped beam [17] (Figure 1c). Its temperature profile can be derived by combining thermal conduction and heat generation equations:

$$T\_a - T\_b = \frac{I^2 \rho L^3}{2VAk} \tag{1}$$

where *Ta* is the maximum temperature at the center of the beam, *Tb* is the minimum temperature at the fixed ends, *I* is the current flow, *L* is the length, ρ is the density, *V* is the volume, *A* is the cross-sectional area, and *k* is the thermal conduction constant.

**Figure 1.** (**a**) UniJoule structure built via a CMOS process for performance improvement, and temperature distribution as a result of joule heating application of (**b**) UniJoule structure and (**c**) conventional clamped–clamped beam computed via COMSOL.

The conventional clamped–clamped beam was converted into a new type of structure (called UniJoule (Figure 1b)) by adding sidebars for the sake of better temperature uniformity. The design was built via a CMOS 0.6 um process, and all the details related to the fabrication process were given in [7,17], where CHF3/O2 was used to etch SiO2, and XeF2 was used to etch silicon underneath the beams. Here, the voltage (Vth) and ground (Gnd) applied on the sidebars (Figure 1b) in contrast to the conventional one, where Vth and Gnd applied on the fixed ends (Figure 1c). This enables the maximum temperature locating on fixed ends of the clamped–clamped beam embedded in the UniJoule structure (Figure 1b) and consequently creates a much more uniform temperature distribution.

The COMSOL was used to model and simulate the entire 3-layer composite beam (Figure 1a), including the substrate layer, where the first layer is SiO2, second layer is Polysilicon, and third layer is Aluminum in the 3-layer composite beam. There are two silicon layers in the structure: the first one is the etched one to allow the beam to resonate, and the second one is the substrate that carries all the layers. The substrate thickness was kept small to decrease the simulation time, while the thermal conduction constant of the substrate was recalculated according to thermal resistance (*R* = *H*/*k*, *H* = substrate thickness, *k* = thermal conductivity) for the sake of the accuracy. The substrate bottom was kept as a fixed surface in solid mechanics, and its temperature was kept at 293 K in a heat-transfer module. The conductivity of polysilicon was set to 1.16 <sup>×</sup> 10<sup>5</sup> S/m [17], and the fine mesh with a tetrahedral structure was used to complete the UniJoule structure in COMSOL. Here, an electric current model was used to apply joule heating, a heat transfer module was used to calculate temperature distribution throughout the beam, and a solid mechanics module was used to calculate deflection, mode shapes, and resonance frequency, with respect to temperature. The high-density sweep points

(up to 2.3 <sup>×</sup> 10−<sup>6</sup> V resolution) in the application of joule heating (Vth) were applied to achieve high accuracy in the results. The convective cooling was added to the heat-transfer module for the sake of high accuracy.

#### *2.2. Design for High Thermal Sensitivity*

The detailed study was conducted on the thermal sensitivity of the UniJoule structure (Figure 2a) via COMSOL, and the results were compared to the uniform heating case (Figure 2b), where the sensitivity was derived via the use of resonance frequency, with respect to axial load [20]:

$$f = \frac{4.73^2}{2\pi L^2} \left( 1 + \frac{PL^2}{EI\pi^2} \right)^{\frac{1}{2}} \left( \frac{EI}{m} \right)^{\frac{1}{2}} \tag{2}$$

$$P = \frac{4}{h}(E\_1 I\_1 + E\_2 I\_2) \left(\frac{(\alpha\_1 - \alpha\_2)T}{\frac{h}{2} + \left(\frac{2E\_1 I\_1 + E\_2 I\_2}{hb}\right)\left(\frac{1}{E\_1 t\_1} + \frac{1}{E\_2 t\_2}\right)}\right) \tag{3}$$

where *I* is the moment of inertia, *m* (kg/m) is the mass per unit length, *P* is the total compressive axial load [21], *L* (m) is the length, *h* (m) is the width and *b* (m) is the thickness, and *E* is the Elastic modulus of the 2-layer composite beam in Equation (2). In Equation (3), *t*<sup>1</sup> is the width, α<sup>1</sup> is the thermal expansion constant, *E*1is the Elastic modulus, and *I*<sup>1</sup> is the moment of inertia for the aluminum layer, *t*<sup>2</sup> is the width, α<sup>2</sup> is the thermal expansion constant, *E*<sup>2</sup> is the Elastic modulus, and *I*<sup>2</sup> is the moment of inertia for SiO2 layer. Here, the 3-layer composite beam (Figure 1) was converted into an equivalent 2-layer composite beam via COMSOL, to find the total compressive axial load in Equation (3), where the first layer is SiO2 and the second layer is Aluminum. The Equation (2) was verified and matched with the measurement results in [18], while the increase in the thermal sensitivity with the increase in the joule heating application was verified with measurements in [17,18]. Further studies [7] demonstrated that the decrease in the pull-in force around the beam-bending point increased the thermal sensitivity. However, one of the main problems and limitations for this type of device is the maximum allowable thermal stress and, consequently, the maximum allowable temperature that the resonator can tolerate. This was measured and verified around 530 K for the same type of structure in the CMOS process [18]. The possibility of a 31× improvement in thermal sensitivity was demonstrated in [7] for the conventional 57 μm long clamped–clamped beam with a maximum temperature of 530 K, around the beam bending point. In contrast to [7], here we built a UniJoule structure that allows uniform heating throughout the structure. A 60.8× ((284 + |−428|) kHz/11.7 kHz) improvement was achieved with a 30 μm long device, according to COMSOL (Figure 2a), and it was calculated around 60.5×, according to Equation (2) (Figure 2b), where the temperature (at the beam bending point [7]) was 520 K, according to COMSOL, and 529 K, according to Equation (2). This improvement is attributed to the fact that uniform heating allows for a larger stiffness constant, while keeping the critical temperature below 530 K, according to both COMSOL and Equation (2), and it consequently results in higher thermal sensitivity.

**Figure 2.** Frequency shift (FS = Fr1 − Fr2) with respect to 1 Kelvin change by (**a**) COMSOL for UniJoule structure with Vth application and (**b**) Equation (2) for uniformly heated conventional clamped–clamped beam, when stiffness constant (K) changed from 29 to 338 N/m; here, X represents positive max, and Y represents negative max of FS [22] in inset.

The Frequency shift (FS) was calculated by taking the difference between two resonance frequencies Fr1 and Fr2 (more details given in [7]), where the ambient temperatures were set as 293 and 294 K, respectively. The sensitivity multiplier (total maximum FS/minimum FS) was increased from 37 to 82, according to COMSOL, and from 36 to 82, according to Equation (2), when the stiffness constant increased from 29 to 338 N/m (Figure 2). The optimum structure was selected to be a 30 μm long beam, where the sensitivity multiplier is around 60.5×, with a stiffness constant of 134 N/m, according to both COMSOL and theory. A decent uniformity in temperature distribution for all UniJoule structures (Figure 2a) was achieved where the difference between the maximum and the minimum temperature throughout the beam is less than 6.5%.

More studies were conducted on the effect of the stiffness constant and material configuration on the sensitivity multiplier and maximum allowable temperature (Figure 3). The first design was built on a composite structure (Figure 1) [5], while the second one was built by only using an aluminum layer [22] in the CMOS process. The results suggest that the sensitivity multiplier depends on the stiffness constant, rather than the material type used in building the resonator. It is around 60.5×, with a temperature below 530 K, for the composite structure (according to both COMSOL and theory), and it is 60.2× at 502 K for the aluminum resonator, where the stiffness constant is 127 N/m for the aluminum design and 134 N/m for composite structure (Figure 3a). Although the sensitivity multiplier is around 60× for both aluminum and the composite structure, the aluminum resonator still shows overall better thermal sensitivity (total |FSaluminum| = 1306 vs. total |FScomposite| = 712) (Figure 3a). This is attributed the fact that aluminum has a larger thermal expansion constant (α*Aluminum* <sup>=</sup> 23.1 <sup>×</sup> <sup>10</sup>−<sup>6</sup> <sup>1</sup>/K) in comparison to composite structure, where SiO2 (αSiO2 <sup>=</sup> 0.5 <sup>×</sup> <sup>10</sup>−<sup>6</sup> <sup>1</sup>/K) and Polysilicon (α*Polysilicon* <sup>=</sup> 2.65 <sup>×</sup> <sup>10</sup>−<sup>6</sup> <sup>1</sup>/K) decreases the overall thermal expansion constant.

**Figure 3.** (**a**) Frequency shift with respect to temperature and (**b**) the relationship between stiffness constant, sensitivity multiplier, and maximum allowable temperature for both aluminum and composite structure (Figure 1) derived via COMSOL and Equation (2).

The relationship between the stiffness constant, the sensitivity multiplier, and the maximum temperature around the beam bending point was analyzed. Both the aluminum and composite structure had a relatively sharp increase in the sensitivity multiplier, where the stiffness constant was smaller than 50 (Figure 3b). The sensitivity multiplier and the maximum temperature increases with the increase in the stiffness constant. The maximum allowable temperature was set to 530 K, and that's why the maximum stiffness constant was calculated as 134 N/m, with a maximum sensitivity multiplier of 60.5×. Moreover, this method can be also used to drastically decrease the operation temperature, while keeping the sensitivity multiplier relatively high. The sensitivity multiplier around 36×, with a temperature below 383 K, and 47×, with a temperature below 428 K, can be achieved according to both COMSOL and the theory (Figures 2 and 3). This feature is especially important for long-term reliable operations.

This finding suggests that using different material or different fabrication processes in building a resonator doesn't affect the sensitivity multiplier or the maximum temperature around beam bending at all (Figure 3). The only critical parameter setting the maximum sensitivity multiplier is the stiffness constant.

#### *2.3. Design for Relatively Low Power Consumption*

The power consumption is another critical parameter for the sensor designs, especially with the increase in demand in the internet of things (IOT) applications. The XeF2 process for isotropic etching of silicon layer is a well-known and widely used process, especially in CMOS–MEMS [5,17]. Here, we applied isotropic etching on the clamped–clamped beam (Figure 1b) for the sake of better thermal isolation. Design-1 has a smaller mask opening (Figure 4a inset), while design-2 has a wider mask opening (Figure 4b inset) to allow 13 μm silicon etching on both sides. This etching process decreases the thermal conduction from beam to substrate, and this allows better thermal isolation. This, in return, enables less power consumption to generate the same amount of heat on the beam, according to Equation (1). Thanks to high thermal conductivity of the silicon layer (1.3 Wcm−1◦C<sup>−</sup>1), this technique would allow a noticeable decrease in the power consumption.

**Figure 4.** The power consumption and bending voltage for a UniJoule structure (Figure 1b), where (**a**) there is no isotropic on fixed-ends and (**b**) there is a 13 μm isotropic etching on both fixed-ends.

Both design-1 and design-2 were heated via applied bias voltage (Vth) (Figure 1b), till the devices reached the bending point, where the maximum thermal sensitivity was achieved (Figure 2) [7]. The bending point was reached at Vth = 0.0464 V for design-1 and Vth = 0.0372 V for design-2. This represents around 43.8 mW power consumption for design-1 and 28 mW for design-2. In other words, the power consumption was decreased by 36% via using a wider mask with the XeF2 isotropic silicon etching process.

The effect of isotropic etching was also investigated on the thermal-sensitivity performance. Although the etching process resulted in lower power consumption, both design-1 and design-2 showed the same sensitivity multiplier. This is attributed to the fact that both designs have the same stiffness constant (Figure 3). There is only a slight difference between the two designs. Design-1 reached the bending point around 527 K, while design-2 reached it around 519 K.

#### **3. Conclusions**

A new type of joule heating scheme was demonstrated for the sake of uniform temperature distribution throughout the beam. This, in return, enabled the higher thermal stresses and higher sensitivity multiplier. Around a 60.5× improvement in thermal sensing was achieved and verified via both COMSOL and theory, while keeping the device temperature below 530 K. The very same method can also be used to drastically decrease the operation temperature and consequently enables long-term reliable operation, where 36× with a temperature below 383 K and 47× with a temperature below 428K was achieved according to both COMSOL and theory. The results suggest that the maximum sensitivity multiplier can be achieved when the stiffness constant is around 134 N/m, regardless of the materials or process used in building the devices. The |TCF| was calculated around 3,991,168 ppm/K, where the applied bias voltage (Vth) is 0.0372 V. It should be noted that this improvement (60.5×) in the thermal sensitivity was achieved without any need for a complex and expensive fabrication process or even special layers, such as 2D materials. This would be very crucial and helpful in supporting the studies conducted on medical imaging. Further improvement was achieved via isotropic etching applied on silicon layer. The power consumption decreased from 43.8 to 28 mW with the decrease in the thermal conduction according to COMSOL. This can be very beneficial for applications that requires compact size, low cost, and wireless communications, such as the internet of things (IOT).

**Author Contributions:** Conceptualization, H.G.; Methodology, H.G.; Software, H.G. and F.S.G.; Validation, H.G. and F.S.G.; Formal Analysis, H.G.; Investigation, H.G. and F.S.G.; Resources, H.G. and F.S.G.; Data Curation, H.G. and F.S.G.; Writing-Original Draft Preparation, H.G.; Writing-Review & Editing, H.G. and F.S.G.; Visualization, H.G. and F.S.G.; Supervision, H.G.; Project Administration, H.G.; Funding Acquisition, H.G.

**Funding:** This research was funded by the Scientific Research Project Foundation of Turkey (grant number 18073).

**Acknowledgments:** The author especially wishes to thank COMSOL for their support in setting up the simulation environment accurately for CMOS–MEMS resonator in this study.

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

#### **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/).

## *Review* **InAs**/**InAsSb Type-II Strained-Layer Superlattice Infrared Photodetectors**

#### **David Z. Ting \*, Sir B. Rafol, Arezou Khoshakhlagh, Alexander Soibel, Sam A. Keo, Anita M. Fisher, Brian J. Pepper, Cory J. Hill and Sarath D. Gunapala**

NASA Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109, USA;

Sir.B.Rafol@jpl.nasa.gov (S.B.R.); Arezou.Khoshakhlagh@jpl.nasa.gov (A.K.);

Alexander.Soibel@jpl.nasa.gov (A.S.); Sam.A.Keo@jpl.nasa.gov (S.A.K.); Anita.M.Fisher@jpl.nasa.gov (A.M.F.); Brian.J.Pepper@jpl.nasa.gov (B.J.P.); Cory.J.Hill@jpl.nasa.gov (C.J.H.); Sarath.D.Gunapala@jpl.nasa.gov (S.D.G.) **\*** Correspondence: David.Z.Ting@jpl.nasa.gov

Received: 22 September 2020; Accepted: 21 October 2020; Published: 26 October 2020

**Abstract:** The InAs/InAsSb (Gallium-free) type-II strained-layer superlattice (T2SLS) has emerged in the last decade as a viable infrared detector material with a continuously adjustable band gap capable of accommodating detector cutoff wavelengths ranging from 4 to 15 μm and beyond. When coupled with the unipolar barrier infrared detector architecture, the InAs/InAsSb T2SLS mid-wavelength infrared (MWIR) focal plane array (FPA) has demonstrated a significantly higher operating temperature than InSb FPA, a major incumbent technology. In this brief review paper, we describe the emergence of the InAs/InAsSb T2SLS infrared photodetector technology, point out its advantages and disadvantages, and survey its recent development.

**Keywords:** InAs/InAsSb; type-II superlattice; infrared detector; mid-wavelength infrared (MWIR); unipolar barrier

#### **1. Introduction**

The II-VI semiconductor HgCdTe (MCT) is the most successful infrared photodetector material to date. MCT grown on nearly lattice-matched CdZnTe (CZT) substrate offers continuous cutoff wavelength (λcutoff) coverage from the short-wave infrared (SWIR) to the very long wavelength infrared (VLWIR), while providing a high quantum efficiency (QE) and low dark current for high-performance applications. In general, III-V semiconductors are more robust than their II-VI counterparts due to their stronger, less ionic chemical bonding. III-V semiconductor-based infrared focal plane arrays (FPAs) excel in operability, spatial uniformity, temporal stability, scalability, producibility, and affordability. InGaAs FPAs with λcutoff ~ 1.7 μm perform at near theoretical limit and dominate the SWIR FPA market. InSb FPAs (λcutoff ~ 5.3 μm), despite a significantly lower operating temperature than MCT, lead the mid-wavelength infrared (MWIR) market in volume due to their superior manufacturability and lower cost. The major limitation for traditional bulk III-V semiconductor detectors grown on (nearly) lattice-matched substrates is the lack of pervasive cutoff wavelength adjustability.

#### *1.1. Advances in III-V Infrared Material*

One method to achieve a wide-range cutoff wavelength adjustability in III-V semiconductors is to use bulk InAsSb alloy grown on metamorphic buffers. Recent results show that the InAsSb band gap bowing is significantly larger than previously believed, and up to λcutoff ~ 12.4 μm could be achieved [1]. In addition, nearly lattice-matched or pseudomorphic III-V semiconductor type-II superlattices (T2SLs) can provide a high degree of flexibility in cutoff wavelength. They can have a sufficient absorption strength to attain ample quantum efficiency, are less susceptible to band-to-band tunneling than bulk

semiconductors [2], and are capable of achieving reduced Auger generation-recombination in properly designed structures [3].

III-V semiconductor extended SWIR (eSWIR) detectors are commonly grown on either InP or GaSb substrates. The first group includes the well-known extended InGaAs [4] and lattice-matched [5] and strain-compensated [6] InGaAs/GaAsSb type-II quantum wells. The second includes bulk GaInAsSb [7,8] and InPSb [9], as well as InAs/GaSb [10], InAs/GaSb/AlSb/GaSb [11], InAs/AlSb [12], InAs/InSb/AlSb/InSb [13], and InAs/InAsSb/AlAsSb [14] superlattices.

For MWIR and long wavelength infrared (LWIR) detectors, InAs/GaSb and InAs/InAsSb are the two most common T2SL absorbers used. The former is well-established, and has been described in detail in various review articles [15–17]. The latter [18–20] emerged more recently as an alternative with simpler growth [18], better defect tolerance, and longer minority lifetime [21], but smaller cutoff wavelength range, weaker optical absorption [22,23], and more challenging growth-direction hole transport [24,25]. In Section 3, we will discuss some basic properties of the InAs/InAsSb T2SLS in more detail.

#### *1.2. Unipolar Infrared Detector Architecture*

The unipolar barrier infrared photodetector architecture is now widely recognized as a highly effective platform for developing high-performance infrared photodetectors, as exemplified by the nBn [26], the XBn [27,28], the complementary barrier infrared detector (CBIRD) [29,30], the double heterostructure (DH) [31,32], and the pMp [33]. A unipolar barrier blocks one carrier type (electron or hole) but allows the unimpeded flow of the other. The unipolar barrier photodetector architecture can be used to lower generation-recombination (G-R) dark current by suppressing Shockley–Read–Hall (SRH) processes [26], and has also been used to reduce surface-leakage dark current [26,34,35] in devices with n-type absorbers. This has been especially beneficial for III-V semiconductor-based infrared photodiodes, many of which traditionally tend to suffer from excess depletion dark current and a lack of good surface passivation. Unipolar barrier infrared photodetectors have been successfully implemented for a variety of bulk and superlattice absorbers.

#### *1.3. Antimonide Barrier Infrared Detectors*

Taking advantage of the recent developments in the antimonide bulk and type-II superlattice infrared absorber material and the advances in the unipolar barrier infrared detector architecture, a new generation of infrared detectors has been successfully implemented in a variety of cutoff wavelengths ranging from SWIR to LWIR. In this section, we briefly mention a few examples to illustrate the effectiveness and versatility of this approach.

The 2006 paper by Maimon and Wicks on the nBn infrared detector [26] has been one of the most influential works in the field of infrared photodetectors in recent years. The performance of the nBn is enhanced by a specially constructed barrier that blocks the majority but not the minority carriers. These unipolar barriers, which are heterostructure barriers that can block one carrier type (electron or hole) but allow the un-impeded flow of the other, can be used to suppress the G-R dark current and surface leakage current [26,34,35], which are the two main dark current mechanisms that have plagued III-V semiconductor infrared detectors. The initial nBn devices used either InAs absorber grown on InAs substrate, or lattice-matched InAs0.91Sb0.09 alloy grown on GaSb substrate, with cutoff wavelengths of ~3.2 and ~4 μm, respectively. These nBn detectors could operate at much higher temperatures than InSb-based MWIR detectors, although their spectral responses do not cover the full 3 to 5 μm MWIR atmospheric transmission window like InSb detectors.

The antimonide barrier infrared detector concept has also been successfully implemented in the eSWIR using a GaInAsSb quaternary absorber and an AlGaSb or AlGaAsSb unipolar electron barrier [7,8]. By adjusting the alloy composition of an GaInAsSb infrared absorber, its cutoff wavelength can vary from ~1.8 to ~4 μm while remaining lattice-matched to the GaSb substrate. In addition, for a lattice-matched GaInAsSb absorber of a given composition, a matching AlGaAsSb electron unipolar barrier can be found for building an nBn (or XBn) detector. The left panel of Figure 1 shows an image taken with an eSWIR FPA made from such an nBn detector at the NASA Jet Propulsion Laboratory (JPL), with a cutoff wavelength of ~2.6 μm at 180 K.

The desire to develop high-performance nBn detectors with cutoff wavelength capable of covering the full 3 to 5 μm MWIR atmospheric transmission window (like InSb) motivated the exploration of T2SL nBn detectors. MWIR InAs/GaSb T2SL nBn detectors were demonstrated by Rodriguez et al. in 2007 [36], soon after the original nBn work by Maimon and Wick [26]. More recently, the InAs/InAsSb type-II strained-layer superlattice (T2SLS) has emerged as a highly versatile absorber material. Along with matching AlAsSb or AlGaAsSb electron barriers (typically lightly p-doped), the InAs/InAsSb type-II T2SLS has been used very effectively in implementing MWIR and LWIR unipolar barrier infrared detectors. Figure 1 shows images of FPAs fabricated at JPL from InAs/InAsSb T2SLS unipolar barrier infrared detectors using cutoff lengths of 5.4, 9.5, and 13.3 μm. In the next section, we will describe the InAs/InAsSb T2SLS infrared detectors in more detail.

**Figure 1.** Images from extended-SWIR (λ<sup>c</sup> = 2.6 μm), MWIR (λ<sup>c</sup> = 5.4 μm), LWIR (λ<sup>c</sup> = 9.5 μm), and VLWIR (λ<sup>c</sup> = 13.3 μm) FPAs made from antimonide bulk and T2SL unipolar barrier infrared detectors.

#### **2. The Emergence of InAs**/**InAsSb T2SLS Infrared Photodetectors**

#### *2.1. Historical Background*

Though much less explored than the InAs/Ga(In)Sb T2SLS, the development of InAs/InAsSb/InSb (Gallium-free, or Ga-free) T2SLS for infrared emitter and detector applications has a long and interesting history that predates the InAs/Ga(In)Sb T2SLS detector [37]. The InAsSb/InAsSb T2SLS was originally proposed by Osbourn as a means of achieving in III-V semiconductors a smaller bandgap than bulk InAsSb in order to reach longer wavelengths [38,39]. This is made possible because of the type-II band alignment (which can lead to a superlattice band gap that is smaller than its constituent bulk materials) and strain-induced band-gap reduction. Researchers at Sandia set out to implement this idea using the InAsSb/InSb SLS grown on InSb substrate with an InSb-rich composition-graded InAsSb strain-relief buffer [40]. By 1990, S. R. Kurtz and co-workers reported a number of photodiodes and photoconductors with cutoff wavelengths ranging from 8.7 to beyond 10 μm [41–43], thus validating the concept of the LWIR InAsSb T2SLS detector. In 1995, Zhang reported continuous-wave operation 3.3–3.4 μm midinfrared lasers based on InAs/InAsSb T2SLS emitter grown on InAs substrate [44]. Meanwhile, researchers in the UK also explored the "As-rich" InAs/InAsSb T2SLS grown metamorphically on GaAs substrates. In 1995, Tang and colleagues from the Imperial College reported InAs/InAsSb T2SLSs with a strong photoluminescence (PL) intensity and infrared emission ranging from 4 to 11 μm [45]. Working with other UK research groups, it was found that As-rich InAs/InAsSb T2SLS with band gaps comparable to InSb have substantially suppressed Auger processes when compared to InSb [46]. Because of the room-temperature Auger suppression, it was suggested that this material may be attractive for mid-IR diode laser applications. In 1999, Pullin and co-workers from the Imperial College demonstrated the room-temperature operation of mid-IR light-emitting diodes based on the InAsSb T2SLS [47].

In the decade since 2000, developments in InAs/Ga(In)Sb T2SLS [2,48] flourished while research activities in the Ga-free InAsSb T2SLS were relatively dormant. Signs of renewed interest in InAsSb T2SLS appeared in 2009, when researchers from Simon-Fraser published a paper on the growth and optical characterization of InAs/InAsSb T2SLS structures strain-balanced relative to the GaSb substrate [49]. The interest in this material as an infrared detector absorber grew stronger as the Zhang Group and collaborators reported significantly longer LWIR minority carrier lifetimes in InAs/InAsSb T2SLS than in the InAs/GaSb superlattice [21]. In 2012, the Zhang Group demonstrated an InAs/InAsSb T2SLS LWIR photodetector [19] based on the nBn device design. The Razeghi Group provided further impetus by showing the versatility of the InAs/InAsSb T2SLS, having reported LWIR [20,50,51], very long wavelength infrared (VLWIR) [52], and bias-selectable dual-band mid-wavelength infrared (MWIR)/LWIR dual-band infrared photodetectors [53]. While the Zhang Group focused on LWIR Ga-free T2SLS, the NASA Jet Propulsion Laboratory started working on Ga-free T2SLS in 2008 in an effort to develop mid-wavelength barrier infrared detectors capable of covering the full 3 to 5 μm MWIR atmospheric transmission window [54]. This effort culminated in the demonstration of InAs/InAsSb T2SLS base mid-wavelength high operating temperature barrier infrared detectors in 2010 and FPAs in 2011 [18,54], and with the subsequent development of detectors and FPAs with longer cutoff wavelengths (see Figure 1) [55].

#### *2.2. Mid-Wavelength InAs*/*InAsSb T2SLS Barrier Infrared Detectors*

Perhaps the most significant impact that the InAs/InAsSb T2SLS has had thus far is in the demonstration of MWIR detectors and FPAs. The MWIR FPA market is traditionally dominated in volume by InSb, with a smaller number of MCT FPAs (with higher operating temperature) filling the needs for high-performance applications. According to G. Fulop of Maxtech International, in 2018 InSb accounted for over 50% of the photodetector-based infrared FPA market (in number of units, all cutoff wavelengths). InAs/InAsSb T2SLS unipolar barrier infrared detector-based FPAs have demonstrated that, like MCT, they can operate at a much higher temperature than their InSb counterparts, all while retaining the same III-V semiconductor manufacturability advantages. The concept and initial detector and FPA results of the InAs/InAsSb T2SLS high operating temperature (HOT) barrier infrared detector (BIRD) were first described in patent documents [18]. The details were reported subsequently in the literature [56,57]; here, we briefly mention some key results.

Figure 2a shows the detector dark current density as a function of applied bias for detector temperatures ranging from 89 to 222 K for an nBn device with an InAs/InAsSb T2SL absorber and an AlAsSb electron unipolar barrier grown on GaSb substrate. Figure 2b shows the spectral quantum efficiency of the detector, demonstrating the full coverage of the MWIR transmission window. The inset shows the dependence of the detector 50%-peak-QE cutoff wavelength as a function of temperature. At 150 K, the 50% cutoff wavelength is 5.37 μm, and the quantum efficiency at 4.5 μm is ~52% without anti-reflection coating. In general, the dark current performance is quite good. Under a −0.2 V bias, the dark current density at 157K is 9.6 <sup>×</sup> <sup>10</sup>−<sup>5</sup> <sup>A</sup>/cm2, which a factor ~4.5 higher than that predicted by the MCT Rule 07 [58] for a cutoff wavelength of 5.4 μm. At 150K, the dark-current-limited and the f/2 black-body (300 K background in 3–5 <sup>μ</sup>m band)-specific detectivities are, respectively, 4.6 <sup>×</sup> 10<sup>11</sup> and 3.0 <sup>×</sup> <sup>10</sup><sup>11</sup> cm-Hz1/2/W.

The detector material was used to fabricate 24-μm pitch, 640 × 512 format arrays hybridized to the SBF-193 readout integrated circuit (ROIC). The cutoff wavelength is ~5.4 μm, closely matching that of the InSb FPA. A 160 K two-point corrected image taken with a resulting FPA is shown in Figure 1. At 160 K, the 300 K background, f/2 aperture mean noise equivalent differential temperature (NEDT) is 18.7 mK, with a standard deviation of 9.2 mK and a NEDT operability of 99.7%. The ion-implanted planar InSb FPAs and molecular beam epitaxy (MBE)-grown epi-InSb FPAs typically operate at 80 K and 95–100 K [59], respectively. The InAs/InAsSb T2SLS-based mid-wavelength high operating temperature barrier infrared detector (HOT-BIRD) FPA has demonstrated significant operating temperature advantages over InSb. For 300 K background imaging applications, the mid-wavelength HOT-BIRD essentially combines the higher operating advantage of MCT with the III-V semiconductor

material robustness advantages of InSb, thus firmly establishing the InAs/InAsSb T2SLS as a viable infrared FPA technology.

**Figure 2.** (**a**) Dark current density–voltage characteristics for an InAs/InAsSb type-II strained-layer superlattice nBn detector taken at temperatures ranging from 89 to 222 K. (**b**) Back-side illuminated spectral quantum efficiency (QE) for the same detector without anti-reflection coating, taken under a −0.2 V bias at 150 K. The inset shows the 50% peak QE cutoff wavelength as a function of temperature.

#### **3. Basic Properties of the InAs**/**InAsSb T2SLS**

In this section, we describe some basic properties of the InAs/InAsSb T2SLS and discuss its advantages and disadvantages as an infrared detector absorber material.

#### *3.1. InAs*/*InAsSb T2SLS Electronic Properties*

Figure 3 shows the energy band diagrams for two (*m*,*n*)-InAs/InAs0.5Sb0.5 superlattices (*m* and *n*, respectively, being the number of monolayers of InAs and InAsSb in each superlattice period) that are strain-balanced with respect to the GaSb substrate. For a given layer thickness ratio (*m*/*n*), the InAsSb alloy composition is selected to achieve strain balancing. Varying the superlattice period (*m* + *n*) changes the band gap and the corresponding cutoff wavelength (λcutoff = *hc*/*E*g). In such a strain-balanced superlattice, typically the InAs layer is under slight tensile strain while InAsSb is under a relatively high compressive strain. Therefore, a comparatively thick InAs layer is required for strain balance against the thinner InAsSb layer. The InAs and InAsSb layers are, respectively, electron and hole quantum wells. Because the InAsSb electron barriers separating the InAs electron quantum wells are relatively weak, the c1 (lowest superlattice conduction band state) wavefunction is only weakly confined, as can be seen in the c1 probability density plots in Figure 3. On the other hand, it can also be seen in Figure 3 that the hh1 (heavy-hole 1, the highest superlattice valence band state) wavefunction is substantially localized in the InAsSb valence band quantum wells, which are separated by relatively thick layers of InAs hole barriers. This is reflected in the corresponding band structure, which is discussed next.

Figure 4 shows the energy band structures for the same two superlattices shown in Figure 3: (a) MWIR superlattice (16,4)-InAs/InAs0.5Sb0.5 with *E*<sup>g</sup> = 0.217 eV and λcutoff = 5.7 μm, and (b) LWIR superlattice (28,7)-InAs/InAs0.5Sb0.5 with *E*<sup>g</sup> = 0.116 eV and λcutoff = 10.7 μm. The relatively weak c1 state confinement and the stronger hh1 state confinement are reflected in the band structure, with the c1 subband having a strong dispersion along the growth direction, while the hh1 subband is nearly dispersionless along the growth direction. This indicates that, along the growth direction, electron transport is more favorable than hole transport (even more so than in the typical bulk semiconductors). The splitting between the hh1 and lh1 (light-hole 1) band is favorable for suppressing band-to-band tunneling, which depends on the c1-lh1 gap [2,37]. For the LWIR superlattice in Figure 4b, the hh1-lh1 splitting is actually larger than the c1-hh1 superlattice band gap. This is favorable for Auger-7

suppression [3]. For a more detailed discussion on the effect of superlattice band structure on infrared absorbers, see, for example, Reference [17].

**Figure 3.** Energy band diagrams showing the bulk InAs and InAsSb conduction band edge (*E*c) and valence band edge (*E*hh), the superlattice zone-center c1 and hh1 energy levels (dotted lines), and the corresponding c1 and hh1 state probability densities (solid lines) for (**a**) the (16,4)-InAs/InAs0.5Sb0.5 strained-layer superlattice (SLS), and (**b**) the (28,7)-InAs/InAs0.5Sb0.5 SLS at 100 K. The InAs and InAsSb layers are, respectively, under tensile and compressive strain on the GaSb substrate.

**Figure 4.** Energy band structure along an in-plane direction, and along the growth direction for (**a**) the (16,4)-InAs/InAs0.5Sb0.5 strained-layer superlattice (SLS) (*E*g = 0.217 eV; λcutoff = 5.7 μm), and (**b**) the (28,7)-InAs/InAs0.5Sb0.5 SLS (*E*g = 0.116 eV; λcutoff = 10.7 μm), on GaSb substrate at 100 K.

Figure 5a shows how the calculated c2, c1, hh1, lh1, and hh2 energy levels vary in strain-balanced (*m*,*n*)-InAs/InAs0.5Sb0.5 superlattices, *m*/*n* = 4, as functions of *P*, where *P* = (*m* + *n*) is the superlattice period in units of monolayers. To help visualize the location of the superlattice energy levels, the background of Figure 5 shows a superlattice energy band diagram with the InAs and InAsSb strained conduction, heavy-hole, and light-hole band edges. Note that, since all the superlattices here have the same (*m*/*n*) ratio and therefore have the same energy band diagram except for a horizontal-axis scaling factor, the background band diagram in Figure 5 is shown with an arbitrary horizontal distance scale and can therefore be shared by all the calculated structures. As the superlattice period

increases, the hh1 level rises while the c1 level stays relatively constant at just a few tens of meV above the InAs conduction band edge. The relatively constant c1 level can be exploited for constructing heterostructures with aligned conduction bands. For instance, an MWIR superlattice can be used as a unipolar hole barrier for an LWIR superlattice. As the superlattice period increases, the (c1-hh1) band gap decreases while the hh1-lh1 splitting increases. As mentioned earlier, a hh1-lh1 splitting larger than the c1-hh1 band gap is favorable for Auger-7 suppression. Figure 5b replots the c1 and hh1 levels, now as functions of the cutoff wavelength calculated from the superlattice band gap. To this, we add also the results for a set of strain-balanced (*m*,*n*)-InAs/InAs0.6Sb0.4 superlattices with (*m*/*n*) = 3, and a set of (*m*,7)-InAs/GaSb superlattices. The c1 levels for the two sets of InAs/InAsSb superlattices are both approximately independent of the cutoff wavelength and have approximately the same value. For the (*m*,7)-InAs/GaSb superlattices, the valence band edge remains approximately constant, since we fix the GaSb hole quantum well width at seven monolayers. In general, the conduction band edges of the InAs/InAsSb superlattices are low compared to those for the InAs/GaSb superlattices.

**Figure 5.** (**a**) The calculated c2, c1, hh1, lh1, and hh2 energy levels for a set of (4*n*,*n*)-InAs/InAs0.5Sb0.5 superlattices, plotted as functions of superlattice period in monolayers (MLs). The background shows a relevant superlattice energy band diagram, with an arbitrary horizontal length scale. (**b**) The calculated superlattice conduction band edge (*E*c1, open symbols) and valence band edge (*E*hh1, solid symbols) as functions of the cutoff wavelength for sets of (*m*,7)-InAs/GaSb, (3*n*,*n*)-InAs/InAs0.6Sb0.4 and (4*n*,*n*)-InAs/InAs0.5Sb0.5 superlattices.

#### *3.2. InAs*/*InAsSb T2SLS Advantages*

The InAs/InAsSb T2SLS has some advantages over the more established InAs/GaSb type-II superlattice (T2SL). The InAs/InAsSb T2SLS is easier to grow, has longer minority carrier lifetimes, and appears to have a better defect tolerance. Figure 6 illustrates the molecular beam epitaxy (MBE) shutter sequence used in the growths of InAs/GaSb T2SL and InAs/InAsSb T2SLS. In principle, the growth of the InAs/InAsSb T2SLS involves only opening and closing the Sb shutter, while the In and As shutters can stay open throughout [18]. This compares to the need to the use of four shutters in the InAs/GaSb T2SL. The growth of the InAs/GaSb T2SL is actually considerably more complicated than indicated in the simplified illustration of Figure 6, which does not include the strain-balancing interfaces required to achieve a high material quality. Thus, in general, the InAs/InAsSb T2SLS is simpler to grow.

**Figure 6.** Schematic illustration of mechanical shutter sequences used in growing (**a**) InAs/GaSb and (**b**) InAs/InAsSb superlattices.

InAs/InAsSb T2SLS has demonstrated longer minority carrier lifetimes than the InAs/GaSb T2SL [21,60,61]. For instance, the MW InAs/GaSb T2SL minority carrier lifetime has been reported at ~80 ns [62], while non-intentionally doped MW InAs/InAsSb T2SLS has exhibited minority carrier lifetime values ranging from 1.8 [61] to 9 μs [60], with a Shockley–Read–Hall (SRH) lifetime of ~10 μs [61].

There has also been evidence suggesting that the longer minority carrier lifetime of InAs/InAsSb T2SLS is related to its defect tolerant. Tang et al. pointed out in their 1995 work [45] that, despite the high threading dislocations expected in the InAs/InAsSb SLSs grown with metamorphic buffers on highly lattice-mismatched GaAs substrates, the Shockley–Read contributions to recombination rates were low, as indicated by the strong photoluminescence (PL) intensity observed. It was hypothesized that this is due to the fact that defect state energy levels in the InAs/InAsSb T2SLS are resonant with the conduction band, rather than in the band gap where they could contribute to carrier recombination. The idea that the defect energy levels are in the conduction band was confirmed in recent pressure-dependence PL experiments on an MWIR InAs/InAsSb T2SLS grown on GaSb [63]. The reason for this is that the InAs/InAsSb T2SLS conduction band edge, like that for bulk InAs, is low, as can be seen in the theoretical results in Figure 5b. At JPL, we have seen anecdotal evidence for defect tolerance. One of the earliest MWIR InAs/InAsSb T2SLS nBn detector wafers was grown on a vintage 2011 developmental 4-inch-diameter GaSb substrate. At that time, the 4-inch substrate surface polishing was not nearly as mature as it is today. The fact that we were nevertheless able to obtain reasonable FPA results can be attributed in part to the defect tolerance of the InAs/InAsSb superlattice (the nBn device architecture being another major contributing factor) [54].

#### *3.3. InAs*/*InAsSb T2SLS Challenges*

The disadvantages of the InAs/InAsSb T2SLS compared to the InAs/Ga(In)Sb T2SLS are (1) weaker LWIR absorption, and (2) more challenging LWIR hole transport. Both are the results of the fact that for the InAs/InAsSb T2SLS a longer superlattice period is required in order to reach the same LWIR band gap as the InAs/Ga(In)Sb T2SLS.

Figure 7 shows the calculated cutoff wavelength as a function of superlattice period for the same three sets of superlattices discussed in Figure 5b. The cutoff wavelength is derived from the calculated superlattice band gap using the relationship λcutoff [μm] = 1.24/*E*<sup>g</sup> [eV]. In the MWIR range, the three set of superlattices have comparable periods. As the cutoff wavelength increases, the periodicity advantage of the InAs/GaSb T2SL becomes more pronounced. Comparison between the two set of (*m*,*n*)-InAs/InAsSb superlattices show that the set with higher (*m*/*n*) ratio and higher Sb fraction InAsSb alloy is more favorable. In a type-II superlattice, the band-edge electron and hole wavefunctions are localized in different layers (see Figure 3). A longer superlattice period reduces the electron-hole wavefunction overlap, leading to weaker oscillator strength and smaller absorption coefficient. Early theoretical analysis by Grein et al. [64] showed that, compared to the InAs/GaInSb superlattice, the InAs/InAsSb superlattice requires wider InAs layers to achieve a comparable band gap and therefore produces smaller optical matrix elements; the calculated absorption coefficients for a 11 μm cutoff InAs/GaInSb T2SLS and a 10 μm cutoff InAs/InAsSb T2SLS are 2000 and 1500 cm<sup>−</sup>1, respectively. More recently, Vurgaftman et al. calculated the absorption coefficients for LWIR superlattices with band gaps of ~0.1 eV (corresponding to cutoff wavelengths of λcutoff = 10−12 μm), and showed that

the InAs/InAsSb T2SLS absorption coefficient is approximately half as large as that for the InAs/GaSb T2SL [23]; at λ = 8 μm, the absorption coefficients are ~1250 and ~700 cm−<sup>1</sup> for the InAs/GaSb (70 Å period) and the InAs/InAsSb (125 Å period) superlattices, respectively. Klipstein et al. modeled the dependence of the LWIR superlattice detector spectral quantum efficiency (QE) on the diffusion length, and concluded that, even for a very large hypothetical diffusion length, the InAs/InAsSb T2SLS has a significantly lower QE than the InAs/GaSb T2SL because of its weaker absorption coefficient [22].

**Figure 7.** Calculated cutoff wavelength for the sets of (*m*,7)-InAs/GaSb, (3*n*,*n*)-InAs/InAs0.6Sb0.4, and (4*n*,*n*)-InAs/InAs0.5Sb0.5 superlattices, as functions of the superlattice period in monolayers (MLs).

The longer period of the LWIR InAs/InAsSb T2SLS also results in larger growth-direction hole conductivity effective masses, which in turn lead to lower vertical hole mobility and shorter diffusion length. From the textbook expressions for diffusion length (*<sup>L</sup>* <sup>=</sup> <sup>√</sup> *D*τ*r*), diffusivity (*D* = μ*kBT*/*e*), and mobility (μ = *e*τ*c*/*m*∗∗), we see that the diffusion length depends explicitly on the conductivity effective mass through the expression *Li* = *kBT*/*m*∗∗ *i* τ*r*τ*c*,*<sup>i</sup>* 1/2, where *<sup>m</sup>*∗∗ *<sup>i</sup>* is the conductivity effective mass, τ*<sup>r</sup>* is the minority carrier recombination lifetime, τ*c*,*<sup>i</sup>* is the collision (momentum relaxation) time, and *i* is the direction index. It can be seen that a large growth-direction conductivity effective mass reduces diffusion length, which in turn limits the practical absorber thickness. In the case of LWIR InAs/InAsSb T2SLS, which has s weaker absorption coefficient than the corresponding InAs/GaSb T2SL and bulk absorbers, the inadequate absorber thickness limits the attainable quantum efficiency. Calculations by Klipstein et al. demonstrate from another perspective the strong QE dependence on the hole diffusion length in an LWIR InAs/InAsSb T2SLS XBn detector: for a 9.7 μm cutoff detector with a fixed 5 μm thick n-type absorber, the QE at 8.5 μm is 40% for a 5 μm hole diffusion length but drops to only 10% for a 1 μm hole diffusion length [22].

Figure 8 shows the calculated electron and hole conductivity effective masses along the growth direction (*m*n,z\*\* and *m*p,z\*\*, respectively) as functions of the cutoff wavelengths for the same sets of superlattices discussed in Figure 7. The conductivity effective masses are thermally averaged quantities which take into consideration the anisotropy and non-parabolicity in the superlattice band structure; detail discussions can be found in References [24,25]. Figure 8a shows the calculated growth-direction electron conductivity effective mass *m*n,z\*\*. For the InAs/GaSb and the (*m*/*n*) = 4 InAs/InAs0.5Sb0.5 superlattices, *m*n,z\*\* values are quite small since the c1 wavefunctions are only weakly confined in the relative shallow conduction band quantum wells separated by thin InAsSb barriers (see Figure 3). For (*m*/*n*) = 3 InAs/InAs0.6Sb0.4 superlattices (with lower Sb fraction InAsSb), the electron effective mass *m*n,z\*\* can increase with the cutoff wavelength to rather large values because of the much longer superlattice period; see Table 1 for a comparison of the superlattice period and *m*n,z\*\* for three different superlattices, all with cutoff wavelengths in the 12–13 μm range. Figure 8b shows the calculated growth-direction hole conductivity effective mass *m*p,z\*\* for the same three sets of superlattices. Again, the *m*p,z\*\* values for the three sets of superlattices are very similar in the MWIR. However, as the cutoff wavelength increases, the superlattice periods required to reach the same cutoff wavelength

diverge (see Figure 7), and the *m*p,z\*\* for the three sets of superlattices also diverge significantly (see Table 1 for specific examples). For the InAs/InAsSb superlattices, *m*p,z\*\* can be very large when the InAsSb hole quantum wells are separated by a wider InAs layer (see Figure 3). Here, the higher Sb fraction in the InAsSb alloy decreases the InAs/InAsSb T2SLS period, and is therefore especially helpful in reducing the growth-direction hole conductivity effective masses. Since the growth-direction electron conductivity effective masses *m*n,z\*\* are much smaller than the hole effective masses *m*p,z\*\*, the diffusion length can be much longer in the p-type InAs/InAsSb T2SLS than in the n-type, which would be more favorable for achieving a higher quantum efficiency. However, for reticulated detector structures with p-type InAs/InAsSb T2SLS absorbers, the exposed side-wall surfaces are inverted to degenerate the n-type (like in InAs) and passivation is often needed for reducing the surface related dark currents; for a discussion on surface leakage dark current mechanisms in unipolar barrier detectors, see Reference [34,65–67]. Despite the challenges, (V)LWIR n-type and p-type InAs/InAsSb T2SLS detectors with cutoff wavelengths ranging from 8 to 15 μm and front-side illuminated quantum efficiencies from 2.5% to 40% have been reported [19,20,50–52,68,69].

**Figure 8.** Growth-direction (**a**) electron and (**b**) hole conductivity effective masses for two families of InAs/InAsSb superlattices, and a set of InAs/GaSb superlattices.

**Table 1.** Growth direction electron and hole conductivity effective masses in units of bare electron mass for three superlattices with approximately the same cutoff wavelength.


#### *3.4. Concepts for Addressing LWIR InAs*/*InAsSb T2SLS Challenges*

LWIR and VLWIR InAs/InAsSb T2SLS have relatively long superlattice periods and therefore have relatively weak optical absorptions and short hole diffusion lengths, which are challenging for achieving a high quantum efficiency. We recently explored theoretically some ideas for addressing the challenges for the LWIR InAs/InAsSb superlattices [70]; here, we briefly summarize the results. In comparing the (*m*/*n*) = 3 InAs/InAs0.6Sb0.4 and the (*m*/*n*) = 4 InAs/InAs0.5Sb0.5 superlattices in the discussions above, we found that increasing the Sb fraction in the InAsSb alloy can reduce the InAs/InAsSb superlattice period significantly. In fact, at sufficiently high Sb fraction (~75%), InAs/InAsSb can match InAs/GaSb in terms of the superlattice period required to reach a given cutoff wavelength [70]. However, high Sb fraction InAs/InAsSb superlattices are more prone to Sb segregation [71–74], which can negate the period-reduction benefit of high-fraction Sb. Polytype superlattices [75] such as the "W" [31], "M" [76], and "N" [77] structures have been used for improving the oscillator strength over the basic InAs/GaSb

T2SL. The analogous polytype W, M, and N superlattices formed by inserting thin AlAsSb barrier layers in InAs/InAsSb T2SLS have been considered as a means for increasing electron-hole wavefunction overlap for stronger optical absorption. This strategy turns out to be unfavorable because the presence of the AlAsSb barriers leads to increased band gap, and therefore increases the superlattice period required to reach a given cutoff wavelength [70]. Metamorphic growth on virtual substrates with larger lattice constants than GaSb can decrease the superlattice period needed to reach a specified cutoff wavelength, but this benefit should be weighed against the need for metamorphic buffer growth and the resulting higher defect density [70]. Finally, as mentioned previously, p-type InAs/InAsSb T2SLS has a longer diffusion length than n-type, but reticulated detector pixels with exposed p-type absorber side-wall surface would require passivation to suppress the surface leakage dark current.

#### **4. Recent Development and Outlook**

InAs/InAsSb T2SLS infrared detectors have been under very active development in recent years, with reported results by research groups worldwide in MWIR [18,56,57,78–82], LWIR [19,20,50,51,83,84], and VLWIR [52,85] detectors, as well as in MWIR/LWIR [53] and LWIR/VLWIR [68] bias-switchable dual-band detectors. Microlens [85] and resonant cavity structures [86] have been used to enhance the photo-response of InAs/InAsSb T2SLS infrared detectors. Studies of transport properties [87–90] and defect levels [63] have led to an improved understanding of InAs/InAsSb superlattices. While most of the more recent InAs/InAsSb T2SLS structures have been grown on (100) GaSb substrates by MBE, growths by a variety of modes have also been demonstrated. MWIR and LWIR detectors have been grown by metal organic chemical vapor deposition (MOCVD) [91–94]. Growths on GaAs [84], Si [80,95], Ge-Si [96], and AlSb (via metamorphic buffer on GaSb) [97] substrates have been reported. The growth of MWIR and LWIR detectors on (211)A and B, and (311)A and B GaSb substrates have also been demonstrated [69,98].

The MWIR InAs/InAsSb T2SLS FPA rivals InSb in manufacturability and affordability, but offers a 40 to 50 K higher operating temperature advantage, which can lead to a lower cryocooler size, weight, and power (SWaP). As such, it is poised to replace the InSb FPA, a major incumbent technology, in many imaging applications. The MWIR InAs/InAsSb T2SLS detectors and FPAs have also demonstrated very low dark current densities at lower temperatures [99], and are therefore suitable for more demanding applications such as the CubeSat-based hyperspectral imaging of 300 K scenes while operating in an intermediate temperature range (100–120 K) [100]. Although LWIR InAs/InAsSb T2SLS FPAs have only achieved a moderate quantum efficiency, their demonstrated large-format capability and high uniformity and operability makes them already suitable for applications such as LWIR imaging for Earth remote sensing applications, where photon flux is abundant. Dual-band FPAs are also expected to find applications because of the manufacturability and cost effectiveness of InAs/InAsSb T2SLS FPAs. The further development of InAs/InAsSb T2SLS infrared detectors will continue to benefit from the infrastructure established largely during the the VISTA Program [101,102], including the availability of large-diameter format GaSb substrates [103–106] and multi-wafer growth capability at commercial foundries [107,108].

**Funding:** The research was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration (80NM0018D0004).

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

**Acknowledgments:** The authors would like to thank S. Bandara, A.J. Ciani, R.E. DeWames, D. Forrai, C.H. Grein, L. Höglund, M.A. Kinch, P.C. Klipstein, A.W.K. Liu, D. Lubyshev, S. Maimon, T.S. Pagano, P. Pinsukanjana, D.R. Rhiger, J.N. Schulman, W.E. Tennant, M.Z. Tidrow, Y. Wei, G.W. Wicks, R.Q. Yang, and Y.-H. Zhang for helpful discussions.

#### **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/).

## *Review* **Overview of Ultrasound Detection Technologies for Photoacoustic Imaging**

#### **Rayyan Manwar 1,2, Karl Kratkiewicz <sup>2</sup> and Kamran Avanaki 1,2,3,\***


Received: 12 June 2020; Accepted: 14 July 2020; Published: 17 July 2020

**Abstract:** Ultrasound detection is one of the major components of photoacoustic imaging systems. Advancement in ultrasound transducer technology has a significant impact on the translation of photoacoustic imaging to the clinic. Here, we present an overview on various ultrasound transducer technologies including conventional piezoelectric and micromachined transducers, as well as optical ultrasound detection technology. We explain the core components of each technology, their working principle, and describe their manufacturing process. We then quantitatively compare their performance when they are used in the receive mode of a photoacoustic imaging system.

**Keywords:** ultrasound transducer; photoacoustic imaging; piezoelectric; micromachined; CMUT; PMUT; optical ultrasound detection

#### **1. Introduction**

Ultrasound transducers are devices that convert ultrasound pressure waves into electrical signal. In an ultrasound imaging machine the transducer is a transceiver device: the waves propagated from an ultrasound transducer are backscattered/reflected from an impedance mismatch in the tissue and received by the same transducer; the strength of the received pressure waves is in the range of 0.1~4 MPa [1]. Another modality that directly benefits from ultrasound transducer technology is photoacoustic imaging (PAI). PAI is an emerging modality that uses a combination of optical excitation and acoustic detection for visualizing vascular, functional, and molecular changes within living tissue [2–11]. As opposed to the optical imaging modalities such as optical coherence tomography [12] that employs ballistic photons, PAI uses diffused photons providing significantly deeper penetration. In PAI, thermoelastic expansion of tissue chromophores occurs when irradiated by a nanosecond pulsed laser—resulting in emission of acoustic waves that are then detected by ultrasound transducers for image formation [9,13–17]; in PAI the ultrasound transducer is a receiver device. The strength of the acoustic waves generated from the chromophores is around 800 PA·mK−<sup>1</sup> [4]. The strength of the generated pressure in PAI depends on the absorption coefficient of the chromophores, the light fluence, and the characteristics of the ultrasound transducer. The lower range of the generated pressure waves in PAI compared to ultrasound imaging signifies the importance of an efficient and effective ultrasound detection technology [18]. In PAI, where the optically induced ultrasound pressure is typically weak [17], the primary requirement of the detection unit is to have a high sensitivity and a large acceptance angle over a wide range of spectral bandwidth.

Acoustic and optical detection methods are complementary technologies that together have solved many unmet industrial and clinical needs. The contactless nature and the wavelength selectivity capability to study a particular target in the tissue (e.g., enabling functional sensing) are advantages of

optical sensing over acoustic sensing, and the less penetration depth of optical sensing compared to acoustic detection technology is its disadvantage; although utilizing short wavelengths (such as X-ray), deep sensing applications are possible at a cost of ionizing the imaging target. In photoacoustic sensing, the advantages of both technologies are utilized: wavelength selectivity and adequate penetration depth; to address different unmet needs, photoacoustic technology has been implemented for different wavelengths from X-ray to infrared (IR). Due to the widely used intermediate penetration depth achieved by IR light, higher sensitivity of IR devices, and the fast growing advancement of IR optical components, most of photoacoustic systems are implemented in this regime. Infrared photoacoustic technology has been successfully used in both preclinical (to study small animal brain [19–22], eye [23–26], and skin [27–29]) and clinical (to detect breast cancer [30–32], cervical cancer [33,34], skin melanoma [35,36], and brain tumor [37,38]) applications.

Based on the ultrasound detection mechanism, transducers can be categorized into two main categories: physical ultrasound transducers and optical ultrasound detection. The physical transducers can further be classified as: (i) conventional piezoelectric, and (ii) micromachined (capacitive or piezoelectric).

Several review studies have been conducted on ultrasound transducer technologies [39–46]. However, these studies were primarily focused on ultrasound imaging and specific to one or two technologies. The purpose of this review is to study the effectiveness of various ultrasound transducer technologies, recognize their pros and cons, and learn about their performance when they are used in the receive mode of a photoacoustic imaging system. This review does not cover the ultrasound transducer technologies that are specifically used for industrial or intravascular applications.

The search protocol used for this review study is as follows. A PubMed database search of "transducer" AND "ultrasound" AND "photoacoustic" AND "imaging" yielded 216 results with 196 published in the last ten years. We have narrowed down the search by "piezoelectric" (45 results) and "micromachined" (29 results). Capacitive micromachined ultrasonic transducers (CMUTs) and piezoelectric micromachined ultrasound transducers (PMUTs) have been utilized in 22 and 7 articles, respectively. Moreover, 4 articles were found relevant to photoacoustic imaging among articles on optical ultrasound detection. In this study, we have reviewed a total of 189 articles.

The organization of the manuscript is as follows. First, the general design characteristics of physical ultrasound transducers are discussed. We then investigate the physical ultrasound transducers along with a quantitative analysis of their performance. Next, we discuss optical ultrasound detection technologies and present their corresponding specifications. Finally, we summarize the pros and cons of various ultrasound detection technologies and discuss their performance in photoacoustic imaging applications.

#### **2. Ultrasound Transducer Characteristics**

The design parameters in an ultrasound transducer are classified into two categories: (i) geometric characteristics of layers (width, length, thickness, and specific to arrays including the number of elements, kerf, and pitch size [47]), (ii) material properties (such as coupling coefficient, elastic modulus, Poisson's ratio, density, stress coefficient, stiffness constant, acoustic impedance, and dielectric constant) used for each section. By adjusting these parameters, a transducer with a desired sensitivity, center frequency, and bandwidth is obtained. If cost is a deciding factor, sensitivity and bandwidth of the transducer may be affected. The manufacturing cost of a transducer largely depends on the fabrication process and the number of attempts needed to obtain required specifications [9].

Electromechanical coupling coefficient of the material represents the coupling efficiency of the transducer. In receive mode, this coefficient can be defined as the ratio between the electrical energy induced and the mechanical energy applied to the sensing material [48]. Coupling coefficient is primarily determined based on the inherent material properties of the transducer sensing elements such as stress coefficient, stiffness constant, acoustic impedance, and dielectric constant. Stress coefficient and stiffness constant are two mechanical properties that determine the elasticity of the sensing material. A highly elastic material with thermal stability is desired to build a transducer with a wide frequency range and low mechanical loss. Acoustic impedance determines the compliance of the transducer material to the target tissue material. Acoustic waves can be transmitted efficiently through the propagating medium when there is less acoustic impedance mismatch between the transducer material and the imaging target medium. Reduced acoustic impedance mismatch improves the signal-to-noise ratio (SNR) of the signal converted from the received pressure waves. In addition to acoustic impedance, there is the electrical impedance match between the transducer material and the back-end electronics (i.e., signal routing, data acquisition unit, amplifiers). Electrical impedance also affects the power transmission efficiency and SNR of the converted signal. Such electrical match is achieved using a matching network that can be realized using a material with high dielectric constant [49,50]. Moreover, a high dielectric constant is essential to improve the coupling coefficient [51].

The receive sensitivity of an ultrasound transducer is typically represented as a ratio of the detected electrical signal amplitude (in the range of micro- to milli-volts) and applied acoustic pressure (in the range of pascal to kilopascal). In photoacoustic imaging, the sensitivity is represented by noise equivalent pressure (NEP) [52], a frequency dependant metric with a unit of Pa·Hz−1/2. NEP is defined as the photoacoustic pressure at the imaging target that generates a transducer output equal to the noise amplitude [53]. We used sensitivity (mV/kPa) for quantitative comparison between the physical transducers and NEP (Pa·Hz−1/2) for transducers that work based on optical ultrasound detection methods.

An ideal ultrasound detection device should possess the following attributes: sufficiently high electromechanical coupling coefficient, an acoustic impedance that is close to tissue impedance, a large dielectric constant, low electrical and mechanical losses, low stiffness, and high thermal stability, that all together leads to a transducer with a high sensitivity over a wide spectral bandwidth.

#### **3. Physical Ultrasound Transducer Technologies**

#### *3.1. Piezoelectric Transducers*

Piezoelectric ultrasound transducers are the most widely manufactured and clinically available transducers that are integrated in commercial ultrasound systems [39,54,55]. The main component of a piezoelectric ultrasound transducer is piezo-material that operates based on converse and direct piezoelectric effect. In transmission mode of an ultrasound transducer, the generated acoustic waves are a result of the transient expansion and contraction of a piezo-material when exposed to an alternating electric field across the piezo-electrodes [56]. In receive mode, the incident acoustic pressure waves deform the piezo-material, and are measured in terms of the potential difference across the piezo-electrodes induced by the deformation [57,58]. A cross section of a piezoelectric linear array transducer is shown in Figure 1a. Piezoelectric ultrasound transducer elements are usually manufactured with a matching layer to reduce the impedance mismatch between the imaging target and backing layer to suppress the back scattered ringing effect. Among piezo-materials, naturally occurring crystals (quartz [59]), are seldom used in manufacturing transducers because of their weak piezoelectric performance, low dielectric and elastic properties, and low stability [54]. Engineered single crystals (such as lead magnesium niobate–lead titanate (PMN–PT) [60] and lead zinc niobate–lead titanate (PZN–PT) [61]) exhibit a high coupling coefficient and a large bandwidth that can specifically be valuable to photoacoustic imaging applications, however, the manufacturing process of these transducers is complex, expensive, and time consuming.

**Figure 1.** Geometric characteristics and manufacturing steps of a piezoelectric linear array ultrasound transducer. (**a**) A photograph of a piezoelectric ultrasound imaging probe; the cross-section of the sensing layer is provided in the blue dotted box [39], (**b**) structural difference between 2-2, 1-3 composite material when used in an ultrasound array transducer [62], (**c**) process flow of conventional dice and fill (DF) fabrication method using 1-3 composite and epoxy filling that includes: (i) piezoceramic material, (ii) dice in *x* direction, (iii) dice in *y* direction, (iv) epoxy filling, (v) reverse, (vi) backside dicing in *x* and *y* directions, (vii) 2nd epoxy filling, and (viii) deposit conductive layer. Reprinted with permission from [63].

The most popular piezoelectric materials are piezoceramics (such as barium titanate (BaTiO3) [64], lithium niobite (LiNbO3) [65,66], lead zirconate titanate (PZT) [67], zinc oxide (ZnO) [68]), and polymers (such as polyvinylidene difluoride (PVDF) [69]). Piezoceramics consist of randomly oriented crystallites separated by grain boundaries. They offer strong piezoelectric properties along the polarization axes, and are less expensive than polymers. Polymers such as PVDF as piezo-material alongside their copolymer trifluoroethylene (TrFE) as the thin electrode have also been found to be effective for producing high frequency transducers due to their low stiffness and improved adhesion (compliance) when compared to traditional sputtered thick metal electrodes [70–74]. With these polymers, low acoustic impedance (i.e., close to the tissue impedance) can be achieved at the cost of low energy conversion. To further improve the quality of piezoelectric transducers, composite materials have been developed [75]. The piezoelectric composite consists of a piezoelectric phase (piezo-ceramic) and a polymer phase (epoxy resin), with a certain connection mode, a certain volume or mass ratio, and a certain spatial geometric distribution [76]. Among different materials, PZT-epoxy resin-based composites have been the dominant material to realize the active elements of transducers in piezoelectric transducers [77–80]. Piezo-composites are classified according to respective phase connectivity (0, 1, 2, or 3) through which the phase is continuous. Since, there are two phases in piezo-composites they are referred by 2-digit numbers [76]. The first digit references the piezoelectric phase and the second digit references the polymer phase. Out of 10 conventional different combinations of connectivity [81,82], piezoelectric 1–3 [9,13] and 2–2 [14] composites are commonly used in transducer technology and are proven to exhibit high coupling coefficient with low-acoustic impedance and low stiffness, leading to improved sensitivity compared to monolithic piezo-materials [62,83]. Structural schematics of 1-3 and 2-2 piezo-composites are shown in Figure 1b. The composites are limited to low energy and low temperature applications due to their inherently low mechanical quality factor and thermal conductivity. A quantitative comparison among different types of piezo-material in terms of their determinant properties is provided in Table 1. Speed of sound (SOS) in biological tissues are in the range of 1450–1580 ms<sup>−</sup>1, thus it is desirable to choose a piezo-material with similar SOS. Table 1 shows that piezo-composite materials provide acoustic impedance and SOS similar to those of biological tissues, with a higher coupling coefficient as compared to polymer or ceramic based piezo-materials; that justifies the use of composites as the preferred piezo layer in ultrasound transducers.

**Table 1.** Material properties of widely used piezoelectric materials in manufacturing of ultrasound transducers [65,66,76,84–86].


Piezoelectric transducers can be developed as single elements or aggregated into an array (e.g., linear, convex, arc, ring, and spherical). A list of ultrasound transducer arrays that have been used in different photoacoustic imaging applications is given in Table 2. Conventionally, the arrays are realized through dice and fill method (DF) [75]. The DF method involves making a series of parallel cuts on a piece of bulk piezoelectric material with a mechanical dicing saw (Figure 1c shows the steps of a conventional DF fabrication method using 1-3 composite and epoxy filling). The material is then diced in the perpendicular direction to produce posts with a rectangular cross section. The diced material is backfilled with a polymer, then the base ceramic support is removed by lapping polishing [62,63,76]. For 2-2 composite, step 3 is skipped and the remaining steps are similar to those for 1-3 composite. Other alternative methods to make piezoelectric transducers include the interdigital bonding technique, stacked plates or lamination techniques, fiber processing, and laser machining [75].


**Table 2.** Different configurations of piezoelectric ultrasound transducer arrays that are used in clinical applications of photoacoustic imaging. BW: bandwidth.

NA: not available.

One design constraint in piezoelectric transducer arrays is that the center frequency is inversely proportional to the thickness. Simultaneously, the element length (*l*) and width (*w*) to thickness (*t*) ratio of *l*:*t* ≥ 10 and *w*:*t* ≤ 0.5 must be maintained [104–106]. Despite the simplicity of DF method, a maximum kerf width of 10 to 15 μm can be achieved using this method and hence, manufacturing high frequency transducers (center frequency: >20 MHz) is difficult [107]. Other alternative methods such as interdigital bonding technique, stacked plates or lamination techniques, fiber processing, and laser machining, have a more complex manufacturing process and introduce non-uniformity [75]; for instance, in the laser machining approach, rapid divergence of the tightly focused laser leads to thickness non-uniformity, this inhomogeneity causes interference in the signal generated from the traducer elements. In addition to the challenges in the fabrication process, incorporation of the backing layer in piezoelectric transducers adds manufacturing difficulty to maintain layer thickness uniformity. Since medium range frequencies are commonly used for PAI of biological tissues, the DF method can be utilized to manufacture piezoelectric transducer arrays.

#### *3.2. Micromachined Ultrasonic Transducers*

#### 3.2.1. Capacitive Micromachined Ultrasonic Transducer (CMUT)

Capacitive micromachined ultrasonic transducers (CMUTs) are considered to be the next generation of ultrasound transducers [108]. CMUT is an array of miniaturized capacitors consisting of suspended membranes made of silicon nitride on dielectric posts, made of silicon nitride/oxide, with a conducting layer made of aluminum/gold and a rigid silicon conducting substrate as the base with a cavity in between. Different polymer materials (e.g., bisbenzocyclobutene) have also been used as the dielectric posts and diaphragms of CMUT arrays [109,110]. As opposed to the conventional piezoelectric transducers, CMUTs rely on electrostatic principles for ultrasound wave generation and reception when a superimposed DC bias and AC signal of desired frequency is applied [111] (see Figure 2a).

**Figure 2.** Capacitive micromachined ultrasonic transducer (CMUT) technology. (**a**) Schematic of a cross-section of a CMUT and its working principle, (**b**) steps of sacrificial release process: (i) substrate and insulation layer realization, (ii) sacrificial layer deposition and pattern, (iii) membrane layer deposition, (iv) sacrificial layer release, and (v) top electrode deposition, (**c**) steps of wafer bonding process: (i) thermal oxidation of silicon wafer (substrate), (ii) gap height and shape realization, (iii) bonding between silicon on insulator (SOI) and oxidized silicon wafer, (iv) thick silicon wafer etching, (v) buried oxide layer etching and top electrode realization [112], and (**d**) other CMUT designs fabricated using: (i) local oxidation of silicon (LOCOS) process [113], (ii) thick-buried-oxide process [114–117], (iii) mechanically coupled plate to the membrane [118], (iv) compliant post structure [114]. Reproduced with permission from [46].

Several process flows have been proposed by various research groups to implement CMUT arrays including, surface micromachining, fusion bonding, and adhesive bonding techniques [105]. Among those, the two most common fabrication methods of CMUT arrays are sacrificial release and wafer bonding processes. The basic process flow of the sacrificial release process is as follows (Figure 2b); initially, a sacrificial layer is deposited or grown on the carrier substrate. After membrane material deposition, the sacrificial layer is etched out with an etchant, specifically chosen for sacrificial layer material and not to etch the membrane layer material [115]. Although, sacrificial release process is relatively simple, reliable, and can be achieved at lower maximum processing temperature (250 ◦C) [116], non-uniform effective gap height due to the roughness in the silicon nitride layer causes deviations in device performance [117]. In addition, the diaphragm may induce substantial intrinsic stress that eventually alters the device properties. The basic process flow of the wafer bonding process is as follows (Figure 2c): initially, a highly doped silicon wafer is thermally oxidized to grow a SiO2 layer followed by etching the oxide layer to determine the gap height and shape of the transducer elements. Next, silicon on insulator (SOI) wafer is brought in contact with the oxidized silicon wafer for bonding process [118]. The bulk silicon form SOI is removed by mechanical grinding and the buried oxide layer is etched to expose the Si diaphragm. Finally, a conducting layer such as aluminum is deposited for electrical routing. This process offers better control over gap height and thickness of the diaphragm with less residual stress. However, the wafer bonding process is very sensitive to surface roughness and cleanness that might affect the overall yield. There are a few other less popular fabrication processes such as local oxidation of silicon (LOCOS), thick buried oxide, mechanically coupled plate, and compliant post structure based CMUT arrays. The details of these processes can be

found in [46]. A graphical representation of the final device structure for each method is depicted in Figure 2d.

CMUTs have gained much popularity over the last decade because they consume lower power, provide excellent electrical and thermal stability, and have a wider fractional bandwidth [46,104,105]. Micromachining techniques have advanced to allow batch fabrication of CMUT arrays of different shapes and frequencies on the same wafer with high yield and reduced price. CMUT technology has also enabled realizing densely packed elements in 2D configurations for volumetric imaging (see [119–121] for more details). The limitation of CMUT is that a large DC bias near the collapse voltage is required to achieve adequate sensitivity. This increases the risk of dielectric charging, changes the DC operating point which leads to an early breakdown of the device, hence greatly limiting the biomedical applicability of CMUT [111]. Hitachi Medico, Japan, Vermon, France, Butterfly Inc., USA, Kolo Medical, USA, Philips, USA, and Fraunhofer Institute for Photonic Microsystems (IPMS) are the pioneers in developing and commercializing CMUT technology.

As compared to conventional piezoelectric transducers, capacitive transducers may offer higher sensitivity and wider bandwidth as well as higher acceptance angle. These features are all important in photoacoustic imaging, where the spectral content of PA signals is distributed over a wide frequency range [43,122]. There is extensive literature discussing the use of CMUTs in different photoacoustic applications [43,119,120,123,124]. In Table 3, the existing CMUT probes that have been used in photoacoustic imaging applications are listed.


**Table 3.** Different configurations of CMUT that have been used in photoacoustic imaging applications.

BW: bandwidth; CF: center frequency, NA: not available. \* Clinical application.

#### 3.2.2. Piezoelectric Micromachined Ultrasonic Transducer (PMUT)

Piezoelectric micromachined ultrasound transducers (PMUTs) are low-cost technology with a high sensitivity that follows the principle of piezoelectric effect. In PMUT, an ultrasound wave is generated and detected based on flexural vibration of a diaphragm similar to a thin film on a silicon substrate without any vacuum gap [40] (see Figure 3a). Apart from the classification of sacrificial layer release and reverse wafer bonding methods that are similar to those used in manufacturing CMUT (see Figure 2b,c), there are two other methods to realize PMUT array diaphragms through back and front side etching (see Figure 3b,c) [130].

The manufacturing process of PMUT with circular diaphragms released from the front-side is described in [131]. For front-side etching (depicted in Figure 3b), a silicon wafer with a platinized thermal oxide layer is used as the substrate for the deposition of the piezoelectric and electrode layer. Lithography and reactive ion etch processes are used to pattern the top electrode, etch the thin film PZT layer to expose the bottom electrode followed by the deposition of insulation layer and electrode track fan-out to bonding pads. Then, the SiO2/Ti/Pt/PZT/Pt thin film membranes are released from the Si substrate with XeF2. Finally, the devices are laminated with a 15 μm thick dry film resist to seal the etched chambers and protect the thin film stack. In backside etching [132] (depicted in Figure 3c), the fabrication process flow starts with a Si (100) wafer [29]. As is the case for surface micromachining, the process begins with preparation of the insulator, (e.g., SiO2 or Si3N4) on the silicon. This is then etched from one side of the Si in preparation for boron (B) doping. Boron

diffusion occurs at a specific rate, allowing the control of the junction depth. After doping, the surface is cleaned and coated with low temperature oxide (LTO). Subsequently, standard photolithography is used to pattern the backside etch window. Later, the wafer is etched with an etchant such as ethylenediamine-pyrocatechol-water-pyrazine (EDP). After the back-side etching, a Ti/Pt bottom electrode is deposited by e-beam evaporation, followed by deposition of PZT and the top electrode. Finally, the top electrode and PZT are etched separately to pattern the top electrode and access the bottom electrode.

**Figure 3.** Piezoelectric micromachined ultrasound transducers (PMUT) technology. (**a**) Schematic of a cross-section of a PMUT and its working principle, (**b**) fabrication process flow of PMUTs with diaphragm defined by front-side etching method: (i) deposition of piezoelectric and electrode layer on top of oxidized silicon wafer, (i) SiO2, Ti, Pt, and PZT layer grown on silicon wafer, (ii) pattern top electrode, (iii) insulation pad deposition, (iv) Ti/Pt deposition, (v) etch through layers to realize bias, and (vi) release the front side diaphragm, reprinted with permission from [130], and (**c**) steps of backside etching method: (i) silicon wafer, (ii) wet oxidation, (iii) oxide etching, (iv) boron diffusion, (v) low temperature oxide growth, (vi) oxide etch, (vii) Si etch, (viii) Ti-Pt deposition, (ix) PZT deposition, (x) TiW-Au deposition, (xi) top electrode etch, and (xii) PZT etch. Reprinted with permission from [132].

Since ultrasound transducers become smaller with increasing frequency, the effects of surface damage introduced during composite machining should be taken into account because the damaged layer volume increases in relation to the size of active piezoelectric materials. The use of a micromachining technique resolves the miniaturization issue of conventional piezoelectric transducers by realizing narrow channels or kerfs less than 10 microns, enabling high aspect ratio of piezoelectric elements [62]; this problem has been resolved in PMUT. Since the sensitivity is not limited unlike CMUTs, because they do not have a vacuum gap between the top and bottom electrodes, there is room to improve the coupling coefficient in PMUTs. Attributes of PMUTs such as low-cost with stable operation, established fabrication process, usage of popular materials (similar to conventional piezoelectric transducers) and capability of miniaturization [40,130–133] have made PMUTs a suitable candidate for photoacoustic imaging applications. Table 4 lists the studies where PMUTs have been used for photoacoustic imaging.

**Table 4.** Different configurations of PMUT that have been used in photoacoustic imaging applications.


BW: bandwidth; CF: center frequency.

#### 3.2.3. ASIC Technology in Physical Ultrasound Transducers

In clinical transducer arrays, each element is connected through a long wire to the analog-front-end (AFE) unit which includes transmit and receive beamformer, preamplifier, switches, and analog-digital converters (ADCs). Although this keeps all the electronics in one place, this arrangement causes interferences and reflections along the cable [137,138]. The number of cable connections can be reduced by multiplexing, however that has negative consequences such as limited bandwidth and slower processing [139]. CMOS technology-based application specific integrated circuits (ASIC) [140] is a novel technology, applicable to micromachined transducers, that is capable of integrating the AFE along with preamplifiers immediately after the ultrasound waves are received [141]. Philips, GE, and Siemens have successfully implemented ASICs within their probes (Philips X7-2t [142], GE 6VT-D [143], Siemens Z6M [144]). ASICs are also applicable to CMUTs and PMUTs [123,134]. Recently, Kolo Medical [145] and Butterfly Network [123] have launched commercial CMUT arrays based on SiliconWave™ and CMOS technologies, respectively. Since ASICs are custom designed, they are expensive, and their repair processes are still highly complicated.

#### *3.3. Comparison between Physical Ultrasound Transducer Technologies*

The physical ultrasound transducer technologies including PZT, CMUT, and PMUT are compared in terms of sensitivity, bandwidth, energy conversion and some other technical specifications in Table 5. Quantitative measurements of piezoelectric and CMUT are based on 2.43 and 2.63 MHz transducers, respectively, presented in [146,147] and that of PMUT are based on a 7~9 MHz transducer presented in [134,135].


**Table 5.** Comparison between physical ultrasound transducer technologies. DF: dice and fill, IC: integrated circuit, DC: direct current.

#### **4. Optical Ultrasound Detection Technologies**

The large size and optically opaque design of the widely used piezoelectric ultrasound transducers cause technical difficulties in some of the biomedical applications where optical illumination path and acoustic detection path must be coaxial. The mechanism that optical ultrasound detection methods offer could be a potential solution. This method employs high-finesse optical resonators to detect incident elastic waves. Providing miniaturized and optically transparent ultrasonic detectors [41], this technique yields a high sensitivity over a significantly wide frequency range, that are together ideal for photoacoustic imaging [42]. Wissmeyer et al. and Dong et al. have reviewed different methods with which optical ultrasound detection can be realized [41,42]. Based on different configurations and detection parameters, optical ultrasound detection techniques can be categorized into: (i) interferometric method and (ii) refractometric method [42]. Interferometric detection can be realized using Michelson interferometry (MI) [152,153], Mach-Zehnder interferometry (MZI) [154,155], doppler [156,157], or resonator [158–160]. In MI or MZI, two-beam method is employed where a laser beam passes into two optical paths, one of which is disturbed by the ultrasound wave and the other serves as a reference (see Figure 4a(i),(ii)). The changes in the optical path caused by the received pressure waves cause proportional changes in the intensity of the beam at the interferometer output [42]. In contrast to two-beam interferometers, doppler method senses ultrasound waves by measuring doppler shift (see Figure 4a(iii)). In resonator-based technique, a micron-scale optical resonator detects ultrasound waves (see Figure 4a(iv)); using this technique, miniaturization of the ultrasound detection unit is feasible. The optical resonator geometries that are most frequently used in photoacoustic imaging are Fabry–Pérot interferometers (FP) [161–163], micro-ring resonators (MRRs) [164–166], and π-phase-shifted fiber Bragg gratings (π-FBGs) [167–169]. Refractometric methods can be classified as intensity sensitive, beam deflectometry, and phase sensitive [41,42]. In intensity sensitive method [170,171], when the ultrasound waves pass through the interface of two media with different refractive indices, the intensity of the beam incident on that interface varies (see Figure 4b(i)). In the beam deflectometry method (see Figure 4b(ii)) [172,173], the interaction of the received acoustic waves with the medium alters the refractive index of the medium, which in turn deflects the probe beam that is eventually detected using a position-sensitive detector such as a quadrant photodiode [42]. In phase sensitive method [174], a collimated light beam passes through an acoustic field; the beam is deflected from the original path and perturbed; this beam is then focused through a spatial filter (see Figure 4b(iii)); the resultant beam is collimated and detected by a charge coupled device (CCD) or complementary metal-oxide-semiconductor (CMOS) camera; the image produced by the camera is the intensity map of the acoustic field.

One of the major limitations of optical ultrasound detection techniques is that they are slow. Although the scanning time can be reduced by parallelization [42,175], this would increase both the complexity and cost of the detection unit [41]. Another limitation is that these configurations mostly rely on continuous-wave (CW) lasers. CW interferometry is sensitive to temperature drifts and vibrations [42]. More details about the limitations of optical ultrasound detection techniques are given in [42,176–178]. Performance comparison between different optical ultrasound detectors is summarized in Table 6. The sensitivity of the optical ultrasound detection methods are represented in terms of noise equivalent pressure (NEP) that is a function of frequency [42]. By multiplying the square root of the center frequency, NEP can be presented in terms of pressure unit (Pascal) as shown in [41]; we used this unit in Table 6.

**Figure 4.** Optical ultrasound detection techniques. (**a**) Interferometric methods: (i) Michelson, (ii) Mach–Zehnder, (iii) doppler-based sensing, (iv) resonator-based sensing, and (**b**) refractometric: (i) intensity-sensitive detection of refractive index, (ii) single-beam deflectometry, (iii) phase-sensitive ultrasound detection. AL: acoustic lens, US: ultrasound, BS: beam splitter, D: detector, DM: demodulator, LA: laser, R: reflector, US: ultrasound, CMOS: complementary metal-oxide-semiconductor, FP: Fourier plane, L: lens, P: prism, PD: photodiode, QPD: quadrant photodiode, SB: Schlieren beam, SF: spatial filter. Reprinted with permission from [42].


**Table 6.** Summary of the performances of different optical ultrasound detection techniques. Reproduced from [41,42].

\* Unit: mPA·Hz−1/2, \*\* Diffraction limited, BW: bandwidth; NEP: noise equivalent pressure, MI: Michelson interferometry; MZI: Mach–Zehnder interferometry; DI: doppler- interferometry; RI: resonator- interferometry.

#### **5. Discussion and Conclusions**

During the past several years, photoacoustic imaging technology has advanced in preclinical and clinical applications [7,8,16,18,122,187–193]. The clinical translation of this emerging imaging technology largely depends on the future of laser technology, data acquisition systems, and ultrasound transducer technology [194]. The ideal fabrication flow of a transducer device is as follows: depending on the application requirements such as geometrical restriction, desired penetration depth, and spatial resolution, the type and technology of the transducers are determined; an optimized structural/material design is then obtained by adjusting the geometric characteristics of the transducers' layers and their material properties; finally the transducers are built with a particular fabrication method, complexity of which depends on the budget. Ultrasound transducers with a high sensitivity in a wide spectral bandwidth, if cost is not a deciding factor, are ideal for photoacoustic imaging; a higher sensitivity can help reduce the necessary optical excitation energy and improve the penetration depth.

Among various ultrasound detection technologies, piezoelectric transducers are the most commonly used [195]; they have been made in forms of single element, as well as linear, arc, ring, hemispheric, and 2D matrix arrays. Their main limitations are that they require a matching layer, thermal instability, difficulty in realizing high frequency transducer arrays, and difficulty in miniaturization. As compared to piezoelectric transducers, CMUTs offer a higher sensitivity and a wider bandwidth, as well as a higher acceptance angle that are all important in photoacoustic imaging, where the spectral content of the PA signal is distributed over a wide frequency range [43,122]. In addition, fabrication of miniaturized transparent transducer arrays with desired shape is feasible using CMUTs. PMUT is a more recent technology with an improved bandwidth, higher sensitivity, lower acoustic impedance mismatch, flexible geometry, and the capability of CMOS/ASIC integration. PMUT, although does not outperform CMUT, relies on the established and reliable piezoelectric technology with micromachining capability. There is extensive research focused on improving the performance of PMUT that has led to promising results [40,130,133], therefore, despite better performance of CMUT, PMUT may have faster growth due to existing infrastructure.

In comparison with physical transducers, optical ultrasound detection technologies offer higher sensitivity over a significantly wide frequency range [42]. These technologies also demonstrate the capability of miniaturization and optically transparent transducers, which are both valuable features in biomedical imaging applications where optical illumination and acoustic detection paths must be coaxial for higher efficiency; this is an ideal arrangement of illumination and detection units in a photoacoustic imaging system. The main limitations of optical transducers are high sensitivity to temperature fluctuations and vibrations, as well as system cost. Overall, NEP of the reviewed technologies, (piezoelectric: 2 mPA·Hz−1/<sup>2</sup> [52], CMUT: 1.8~2.3 mPA·Hz−1/<sup>2</sup> [123], PMUT: 0.84~1.3 mPA·Hz−1/<sup>2</sup> [196], optical ultrasound detection: 0.45~486 mPA·Hz−1/<sup>2</sup> [42]) suggest that micromachined transducers (i.e., CMUTs and PMUTs) may be the more suitable transducers for photoacoustic imaging applications.

ASIC is a complimentary technology in transducer manufacturing to integrate the analog-front-end within the probe housing in order to reduce the noise of the transducer signal. ASIC improves the overall performance of existing transducers and therefore could help in facilitating the clinical translation of photoacoustic imaging. According to the technology market analyst projection, the usage of ASIC integrated miniature ultrasound transducer probes based on micromachined technologies will see ~18% annual compound growth rate by 2023 due to the advent of micromachining processes [197]. With the fast-growing ultrasound transducer technology, numerous computational methods have also been studied to further improve the performance of transducers by reducing noise in the transducer signal [198,199]; the result has been higher quality images [5,200,201]. With the advancement of ultrasound technology, more biomedical applications, which are currently performed using optical technologies, can be realized [12,202–208].

**Author Contributions:** Conceptualization, R.M. and K.A.; Methodology, K.A.; Software, R.M.; Validation, R.M., K.K., and K.A.; Formal Analysis, R.M. and K.A.; Investigation, R.M. and K.A.; Resources, K.A.; Writing-Original Draft Preparation, R.M.; Writing-Review & Editing, R.M., K.K., and K.A.; Visualization, R.M.; Supervision, K.A.; Funding Acquisition, K.A. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the National Institutes of Health R01EB027769-01 and R01EB028661-01.

**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/).

## *Review* **Recent Progress on Extended Wavelength and Split-O**ff **Band Heterostructure Infrared Detectors**

**Hemendra Ghimire 1, P. V. V. Jayaweera 2, Divya Somvanshi 3, Yanfeng Lao <sup>4</sup> and A. G. Unil Perera 1,\***


Received: 21 April 2020; Accepted: 20 May 2020; Published: 28 May 2020

**Abstract:** The use of multilayer semiconductor heterojunction structures has shown promise in infrared detector applications. Several heterostructures with innovative compositional and architectural designs have been displayed on emerging infrared technologies. In this review, we aim to illustrate the principles of heterostructure detectors for infrared detection and explore the recent progress on the development of detectors with the split-off band and threshold wavelength extension mechanism. This review article includes an understanding of the compositional and the architectural design of split-off band detectors and to prepare a database of their performances for the wavelength extension mechanism. Preparing a unique database of the compositional or architectural design of structures, their performance, and penetrating the basics of infrared detection mechanisms can lead to significant improvements in the quality of research. The brief outlook of the fundamentals of the infrared detection technique with its appropriateness and limitations for better performance is also provided. The results of the long-term study presented in this review article would be of considerable assistance to those who are focused on the heterostructure infrared detector development.

**Keywords:** heterostructures; split-off band; wavelength extension; device performance

#### **1. Introduction**

Technological advancement in material growth, processing, and characterization setups, furnished with new concepts of device structures and functions, has been widely implemented for the development of new, low-disorder, and often highly engineered material combinations, such as heterostructures [1,2]. The using layers of structurally abrupt interfaces of dissimilar compounds artificially periodic structures can be obtained [3]. Many scientists have thus used refractive indexes, bandgap, effective masses and mobilities of charge carriers, and the electron energy spectrum advantages of semiconductor heterostructures with various material combinations, architectures, and doping densities for the futuristic scientific, technical and biomedical applications [4–10]. The examples include GaAs/AlGaAs heterostructures [11], which have been well-studied for its potential application in high-speed digital and optoelectronic devices [12] including diode lasers [13], light-emitting diodes [14], solar cells [15] and optical detectors [8,16–20]. The material advantage of GaAs/AlGaAs provides excellent uniformity and large arrays. It is because there is a close match between the lattice constant of AlGaAs and GaAs. Herein, AlGaAs has a lattice constant that varies linearly between that of AlAs, 5.661 angstroms, and that of GaAs is 5.653 angstroms, depending on the mole fraction of aluminum [21]. The alternations of the aluminum fraction in the AlGaAs layer modulate the band-gap [22] and hence, one can adjust the

barrier height of the designed device to match the required photon energy. Similar to GaAs/AlGaAs [23], research has been carried out on the number of other semiconductor heterostructures having various, material combinations emitter/barrier architectures and doping densities for infrared (IR) detector development [6,24–28].

The heterojunction detectors offer wavelength flexibility [29–33] and multicolor capability in these regions. The wavelength range includes short-wave infrared (SWIR) of 1–3 μm [29], mid-wave infrared (MWIR) of 3–5 μm [30], long-wave infrared (LWIR) of 5–14 μm [33], very-long-wave infrared (VLWIR) of 14–30 μm [32], far-wave infrared (FWIR) of 30–100 μm or 3–10 THz [31] and so on. Among these wavelength ranges of IR radiations, 3–5 [34] and 8–14 μm [35] are considered the atmospheric windows and are suitable for IR applications.

In this review, we stepwise describe investigations concerning the physics and applications of split-off band semiconductors heterostructures infrared detectors for the threshold wavelength extension mechanism. The description systematically explains the architectural design of the device and assesses current experimental and theoretical understanding used to improve the performance.

#### *1.1. Heterostructure IR Detectors*

The architecture of heterostructures [36] mainly involves alternate thin layers of lattice-matched and band-gap tuned different compound semiconductors. The growth of sandwiches' layers follows appropriate fabrication so that unique (unlike from component semiconductors) electrical and optoelectrical properties of heterostructure can be achieved. Bandgap engineered compact heterostructure is then processed with appropriate conducting rings for characterization. The processing for the fabrication also includes photolithography [37], etching to open an optical window for incidence illumination, lift-off, and others. The typical architecture of the heterostructure detector is shown in Figure 1A. In such a compact structure, each layer of the emitter is sandwiched between two layers of barriers. A sharp interface between the junctions of each layer can be achieved by making it easier for tuning the energy gap between the emitter and barrier. The development of sophisticated controllable epitaxial growth methods, such as Molecular Beam Epitaxy (MBE) and Metal-Organic Chemical Vapor Deposition (MOCVD), has allowed for the fabrication of such ideal heterojunction structures [38–40].

**Figure 1.** Typical architecture of a GaAs/AlGaAs heterostructure and the type depending on band alignments. (**A**) The combination of two dissimilar semiconductors in the heterostructure. The design is a p-GaAs/AlGaAs heterostructure and every layer of the emitter (p-type GaAs) is sandwiched between two layers of the barrier (AlGaAs). (**B**) Type-I heterostructure made of GaAs and AlGaAs. In such heterostructures, bandgap overlaps and Δ*E*g = Δ*E*c + Δ*E*v.

The emitter and barrier regions of heterostructure have different energy bands, and it will have an additional change in the presence of biasing [41]. In such compound heterostructures, quantum-mechanical effects such as potential discontinuity (band off-set) play a crucial role [42]. Based on the alignment of bands producing (step) discontinuity, heterostructures are classified into three types [36], type I, type II, and type III. Figure 1B shows the straddling alignment in a type-I heterostructure, where signs of the band are offset, as the two bands are opposite. In type-I heterostructures, the band gaps of one material entirely overlap with that of another, and potential discontinuities are as Δ*E*c = *E*c1 − *E*c2, Δ*E*v = *E*v1 − *E*v2 and Δ*E*g (*E*g1 − *E*g2) = Δ*E*c + Δ*E*v. Similarly, in a type-II heterostructure, *E*v1 > *E*v2 and Δ*E*c may or may not be larger than *E*g1. Moreover, *E*g2 is every so often smaller than *E*g1 in a type-II heterostructure. Type II staggered, (where Δ*E*c < *E*g1) and type II misaligned (where Δ*E*c > *E*g1) are two subclasses of type-II heterostructures. Type-III heterostructures are formed by combining semimetal with the inverted bands of semiconductors [43]. In this paper, we have mainly used GaAs/AlGaAs to cover type I heterostructures as shown in the above figure.

Along with technological development, there is a growing demand for advanced IR systems with better discrimination and identification. Group-IV, III-V, and II-VI semiconductor heterostructure-based [44, 45] photodetectors have been studied extensively for the IR detection, from near infrared (NIR) to far infrared (FIR) region [46]. SiGe/Si heterojunctions were used to study internal photoemission LWIR detectors [33,47]. Low dimensional II-VI oxides semiconductor structures such as NiO/ZnO were also studied to understand their feasibility for light detection [48]. Similarly, studies of III-V semiconductor alloys such as InGaAs, InAsSb, InGaSb, HgZnTe, HgMnTe, GaAs, AlGaAs and their heterostructures are mainly focusing on the MWIR and LWIR detectors [49,50]. Mercury alloys led tin tellurides and selenides, and other alloy combinations were further explored for their potential as alternative material that can overcome the challenges of important semiconductor IR detector mercury cadmium telluride (HgCdTe or MCT) [27,50–52]. HgCdTe-based IR detectors are the most widely used for high-performance applications [50]. However, the progress has been impeded due to the fundamental material problems, such as high defect density related to growth due to the weak HgTe bond [53]. This is a serious technological problem for the mass production of HgCdTe large-sized Focal Plane Arrays (FPA), and to achieve high FPA pixel uniformity and yield.

On the other hand, Quantum Well IR Photodetectors (QWIPs), based on the GaAs/AlGaAs material system, are considered mostly for the LWIR spectral regime requiring high uniformity. The industrial infrastructure in III–V materials/device growth, processing, and packaging brought about by the utility of GaAs-based devices in the telecommunications industry gives QWIPs a potential advantage in producibility and cost [50]. However, QWIPs are not sensitive to normal incident light due to orthogonality between polarization vector of the incident photons and optical transition dipole moment [54]; therefore, light coupling structures are required for a 45◦ angle incidence of light which adds cost and complexity. Besides, a low operating temperature is typically required due to fundamental limitations associated with intersubband transitions [50]. Studies of Quantum Dot IR Photodetectors (QDIPs) are also making progress towards the development of IR detectors [50,55].

Because of the key pieces of evidence presented above, the studies show number of different structures and material combinations were used. The trend of exploration of new material is still increasing day by day. The superiority of heterojunction structures over the homojunction further allows for the estimation of new materials and engineering compositions in the coming days.

#### *1.2. Performances of IR Detectors and Acceptable Figures of Merits*

Studies have illustrated a wide range of accepted figures of merit for a meaningful comparison of the sensitivity performances of detectors [56]. These include responsivity, spectral response, specific detectivity, dark current, and the current gain. Among these performance matrices, responsivity measures the input-output gain (simply called gain) of the detector. Simply put, it is defined [57] as the ratio of the photocurrent (*I*pc) and incoming optical power (*I*o), i.e., *R* = *I*pc/*I*o = η*q*λ/*h*ν ≈ ηλ/1.24 (A/W), where *h*ν = *hc*/λ is the energy of the incident photon, *h* planks constant, *c* velocity of light, the *q* charge of

an electron, λ is the wavelength of incident light and η is quantum efficiency. Studies have used spectral response [58] to describe the sensitivity (ability to convert light of various wavelengths to electricity) of the photodetectors as a function of photon frequency (or wavelength). The efficiency of detection is also simply expressed in terms of quantum efficiency in various studies [59,60]. Specific detectivity [56] is another important performance matrix, which is defined as the reciprocal of noise equivalent power

(NEP) normalized pre-square root of the sensor's area (*A*) and frequency (*f*) bandwidth, i.e., *D*∗ = √*A f NEP* . Similarly, the current flowing through the detector in the absence of light, known as the dark current, is another important parameter to determine the performance of the detector [61]. The dark current characteristics are important to determine an optimum operating condition. Whilst the lowest possible dark current is desired for the operation of an IR photodetector, a certain amount of bias voltage must be applied to operate photoconductive photodetectors [61]. The study of dark current characteristics offers important insights into the device parameters, such as activation energy (Δ). Dark current in the semiconductor heterostructures is temperature dependent, which can be expressed [62,63] as:

$$I\_{\rm dark} = Ac \frac{\mu F}{\left[1 + \left(\frac{\mu F}{v\_{\rm lat}}\right)^2\right]^{1/2}} 2 \left(\frac{m^\* k\_B T}{2\pi \hbar^2}\right)^{3/2} \exp\left(-\frac{E\_{\rm act}}{k\_B T}\right) \tag{1}$$

where *A* is the electrically active area of the detector, *e* is the electronic charge, *Eact* = Δ − *Af* − *Ef* is the activation energy, μ is the mobility of the holes, *vsat* is the saturation velocity, *m*\* is the effective mass, *kB* is Boltzmann's constant, *T* is the temperature, *h*¯ is the reduced Planck constant, α is a constant parameter that determines effective barrier lowering due to the applied field *F*, and *Ef* is the Fermi level.

#### *1.3. Spin-Orbit Split-O*ff *Band Heterostructures Infrared Detectors*

The use of semiconductor heterostructures for the detection of electromagnetic (EM) spectrum blind to the human eye (UV, IR, and THz) [64] leads towards the exploration of structural concepts, their physics, theory, modeling, and experimental measurements. The study of intraband electronic transition within the valance band (heavy hole (HH), light hole (LH) and spin-orbit/split-off (SO) bands) is one of the primary characteristic features to understand the electronic process in semiconductor alloys [65,66]. Split-off band effects have been observed in the emission of GaAs metal-semiconductor field-effect transistors [67] and have enhanced the response of GaInAsP quantum wells [68]. This aspect of career transition is also implemented to develop heterostructure infrared detectors [69].

A detailed explanation of experimental SO response depending on hole transitions between LH, HH and SO bands has been presented by using GaAs/AlGaAs-based heterojunction interfacial workfunction internal photoemission (HEIWIP) detectors [70–72]. The active region of the basic HEIWIP detector consists of one or more periods, each consisting of a doped emitter and an undoped barrier layer. These multiple emitter/barrier layers are sandwiched between two highly doped contact layers. Depending on the doping required for ohmic contacts, the top contact may also serve as the top emitter layer. Herein, the work function (Δ) is given by Δ = Δ<sup>d</sup> + Δx, where Δ<sup>d</sup> and Δ<sup>x</sup> are the contributions from doping and the Al fraction, respectively [23]. As the Al fraction is reduced, Δ will be limited by Δd, which in turn is a homojunction detector [73,74]. The detection mechanism can be divided into three main processes: (i) the photoabsorption that generates excited carriers, (ii) the escape of the carriers, and (iii) a collection of the escaped carriers.

The schematic of different intervalence and intra-band optical transitions showing the infrared detection mechanism is shown in Figure 2 in terms of energy wave vector diagrams. Figure 2A–C shows the energy wavenumber (E-k) diagram for intrinsic, extrinsic, and quantum well infrared photodetectors. Figure 2D shows Split-off detector threshold mechanisms. Infrared photon excites holes from the light/heavy hole bands to the split-off band. (1) Indirect absorption followed by scattering and escape (threshold energy: *E*ESO − *E*f). (2) Direct absorption followed by scattering and escape (threshold energy: *E*ESOf − *E*f). (3) Indirect absorption followed by escape and some

scattered (threshold energy: *E*BSO − *E*f). The band alignment during split-off band intra-valance transitions is shown in Figure 3F.

**Figure 2.** Schematic of different intervalence and intraband optical transitions showing an infrared (IR) detection mechanism. (**A**–**C**) Energy wavenumber (E-k) diagram for intrinsic, extrinsic, and quantum-well-IR photodetectors, respectively. (**D**) Split-off detector threshold mechanisms, where the IR photon excites holes from the light/heavy hole bands to the split-off band. (1) Indirect absorption followed by scattering and escape (threshold energy: *E*ESO − *E*f). (2) Direct absorption followed by scattering and escape (threshold energy: *E*ESO − *E*f). (3) Indirect absorption followed by escape and some scattered (threshold energy: *E*BSO − *E*f). (**F**) Band alignment using during split-off band intra-valance transitions.

**Figure 3.** Absorption coefficient of p-type (5 <sup>×</sup> 1018 cm<sup>−</sup>3) GaAs0.955Sb0.045, In0.15Ga0.85As and GaAs as discussed [69]. The p-type GaAs0.955Sb0.045 shows the highest absorption coefficient by comparing with the other two materials.

The spectral response of the heterojunction detector is primarily determined by the absorbing properties of the emitter, which is, in turn, determined by the electronic structure of the valence bands (VBs) and the concentration of holes. Utilizing semiconductor heterojunctions for the heterojunction detectors is subject to limitations, such as low absorption coefficient in the 3–5 μm range for p-type GaAs. The p-type semiconductor absorbs extra photons owing to optical transitions among three VBs, i.e., the LH, HH, and SO bands, contributing to an absorption band spanning primarily from 1 to 10 μm for p-type GaAs, as shown in Figure 3. The p-type semiconductors including InGaAs, GaAsSb, and GaAs indicate that the shifting of the absorption peak (between 1 and 5 μm) can be associated with the SO–HH transition. The intra-band free-carrier absorption is typically proportional to λp, which becomes dominant in the very-long-wavelength range, where λ is the wavelength, and p is an exponent. p can be predicted by the Drude theory, and usually equals 1.2, 2.5, and 3.5 correspondings to scatterings by acoustic phonons, polar optical phonons, and ionized impurities, respectively. By comparing with p-type GaAs, p-type GaAsSb shows a higher absorption, which benefits a higher absorption efficiency and hence the detector performance.

Jayaweera et al. studied an uncooled infrared detector for 3–5 μm and beyond [75] by using the concept of a split-off band infrared detection mechanism [70,71]. In their study, they have experimentally demonstrated uncooled infrared detection using intra-valence bands using a set of three p-GaAs/AlGaAs heterostructure. The focus of their study is to demonstrate an uncooled infrared detection using intra-valence bands. The uncooled detection of infrared radiation is important in a wide range of applications in the civilian, industrial, medical, astronomy, and military sectors. The calculated *<sup>D</sup>*\* value in this study is 6.8 <sup>×</sup> 105 Jones at a temperature of 300 K and an SO band offset of 0.31 eV, while, for those with an SO band offset of 0.155 and 0.207 eV, *D*\* values are 2.1 <sup>×</sup> 106 and 1.8 <sup>×</sup> 106 at temperatures of 140 and 190 K, respectively. It is noted that the studies on the high operating temperature that split off the transition in such a heterostructure [70–72] further support the infrared detection mechanism. These studies have introduced intra-valence band transitions, such as LH to heavy-hole HH transitions or HH transitions to SO band in detectors to overcome the selection rule limitations. A band diagram (E-k) of the emitter region of an S-O band detector illustrating the detection mechanism is based on the carrier transitions in the three valence bands is also demonstrated.

In the study by Lao and et al. [76], the development of structure with multi-spectral detection probability is presented. The hole transition from the HH to the LH band of p-GaAs/AlGaAs detectors show a spectral response up to 16.5 m, operating up to a temperature of 330 K where the LH-HH response is superimposed on the free-carrier response. Similarly, in another study by Matsik, and et al. [77] device modeling study of S-O band IR detector was carried out to find optimized conditions for its performance improvement. The study suggested two important device architecture modifications in order to improve device performance. One of the suggestions was to include an offset (δ*E*) between the energy barriers so that the low energy barrier towards the collector side would enhance the collection. The other suggestion was to include a graded barrier on the injector side so that the carrier trapping would be reduced and the injection of the carriers to the absorber/emitter would be improved.

Altogether, the peak response due to intra-valance band transitions [70,71,78] is believed to originate from a build-up of a quasi-equilibrium Fermi level, at a fixed level, irrespective of the variation of the device parameter. Thus, the SO band was found to be the most probable energy level to build-up a quasi-equilibrium Fermi level as a consequence of the hot-phonon bottleneck effect. The study of the dark current characteristics of these IR photodetectors confirmed no compromise in the dark current due to the presence of the extended-wavelength mechanism of the photoresponse. Furthermore, the quasi-equilibrium will be built-up not only in the emitter but also in the bottom and top contact layers as long as there is photoexcitation to the SO band. The net flow of the photo-excited carriers will determine the spectral photoresponse [62,79].

In summary, HEIWIP detectors [80], based on intra-valence band transitions LH to HH transitions or HH transitions to SO band were introduced to overcome the selection rule limitations. The SO band infrared detector is a newly developed, emerging device based on the p-doped GaAs/AlGaAs system [71], which utilizes hole transition in the HH/LH band to the SO band as the detection mechanism [71]. The SO band detectors have shown promising results to be developed as an uncooled IR detector [75].

#### *1.4. E*ff*ect of a Current Blocking Barrier on Heterojunction Infrared Detector*

Studies have shown that incorporating current blocking architectures into detector designs alters their performances. As part of their studies, Wang and et al. [81], Rotella and et al. [82], Wang and et al. [83], Lin and et al. [84], Pal and et al. [85] and Nevou and et al. [86] have presented the use of AlGaAs current blocking layers in quantum dot IR photodetectors (QDIPs) to enhance the performance of detectors. Similarly, Stiff et al. [87], Tang [88], and Chakrabarti [89] have presented the use of a similar concept to achieve higher operating temperatures. In another study by Nguyen et al. [90], hole blocking layers have been implemented in type-II InAs/GaSb superlattice infrared photodetectors. The electron blocking and hole blocking unipolar barriers in complementary barrier infrared detectors [91] and p-type-intrinsic-n-type photodiodes [92] were also studied. The most important goal in these architectures is to increase the performance by lowering the dark current with a relatively small compromise in the photocurrent.

Chauhan et al. have reported the performance of a p-GaAs/AlxGa1xAs heterojunction photovoltaic infrared detector, with graded barriers, operating in the 2–6 μm wavelength range [16]. They found that the incorporation of the current blocking barrier in heterostructure architecture leads to achieving a significant improvement in the specific detectivity (*D*\*). It further reduces the dark current under photoconductive operation and increases the resistance-area product (*R*0*A*, *R*<sup>0</sup> is the resistance at 0 V and *A* is the electrical area) under photovoltaic operation. In blocking barrier studies, measurements across the top and bottom contacts include the current blocking barrier, whilst the middle and bottom contacts measure the same sample (mesa) without the current blocking barrier, as shown in the schematic diagram of the valence band of the heterostructure in Figure 4. Herein, the implementation of a current blocking barrier increases the specific detectivity (*D*\*) under dark conditions by two orders of magnitude to 1.9 <sup>×</sup> 1011 Jones, at 77 K. Furthermore, at zero bias, the resistance-area product (*R*0*A*) attains a five orders enhancement due to the current blocking barrier, with the responsivity reduced by only a factor of 1.5.

**Figure 4.** Schematic of the valence band alignment of the heterostructure under equilibrium showing the connections with and without the current blocking barrier.

#### *1.5. Threshold Wavelength Extension Mechanism*

The study of the S-O band detector devices with the barrier offset (δ*E*) and a graded barrier showed an unprecedented result in terms of the spectral range of the photoresponse: a photoresponse that is far beyond the spectral limit. In general, the standard expected threshold wavelength (λt) of a photodetector is governed by the spectral rule Δ = 1.24/λt, where Δ controls the dark current. The extended wavelength IR photodetectors are a novel class of photodetectors showing spectral

photoresponse far beyond the conventional limit of Δ = *hc*/λt. In extended wavelength detectors, the effective response threshold wavelength (λeff) is governed by a Δ = 1.24/λeff, with an effective activation energy Δ Δ. The valance band diagram used to demonstrate the extended wavelength photoresponse utilized an offset (δ*E*) between the energy barriers in the heterostructure is shown in Figure 5B. A reference photodetector without the offset and cannot show the extended wavelength mechanism is also shown in Figure 5A.

**Figure 5.** Conventional and wavelength extended photodetector. (**A**) Conventional photodetector. The threshold wavelength (λc) of detection is determined by Δ, where λ<sup>c</sup> = *hc*/Δ. (**B**) Wavelength extended photodetector showing barrier offset contributing to wavelength-extended photoresponse.

The mechanism responsible for the extension of threshold wavelength in heterostructure detectors has been analyzed by Somvanshi et al. [93], which is based on the hot carrier effects in the semiconductor heterostructures [94–96]. The hot carrier effect is principally governed by the carrier–carrier and carrier–phonon scattering processes and has been widely studied in the past [96–98]. In general, hot carriers are created in energy states above the band edge, interact with lattice vibrations, and cold carries through carrier–carrier interactions. This leads to quasi-equilibrium Fermi distributions at a temperature much higher than the lattice temperature [97,98]. It is observed that, upon the incidence of light from an external optical source, hot holes have created in the bottom contact, photon absorber, and top contact of heterostructure detectors, as shown in Figure 5. In heterostructure detectors with band offset (Figure 5B), a net flow of hot holes has observed towards the top contact, owing to the difference in barrier heights, Δ*Ev*. However, no such flow of carriers observed in heterostructure detectors without band offset, i.e., conventional detectors (Figure 5A). The dynamics of hot hols and cold holes interaction in the p-GaAs photon absorber can be explained based on hot carrier effects. Upon the interaction of carriers in the photon absorber, the exchange of energy takes place through hole–hole and hole–phonon scattering that leads to the formation of a quasi-Fermi distribution (*E*quasiF) at a hot hole temperature (*T*H) that is significantly greater than the lattice temperature (*T*L). The distribution of hot holes at this quasi-Fermi level will lead to an escape of hot holes across the collector barrier when a long-wavelength photon is absorbed and the extension in threshold wavelength is observed.

Chauhan et al. have studied the photoresponse of extended wavelength IR photodetectors [99] under different bias conditions. It is observed that, with the increase in bias voltages, the photoresponse becomes stronger; however, the spectral threshold remains relatively constant. The dark current and photo-response characteristics of extended wavelength infrared photodetectors [62] studied by Chauhan and et al. show that the measured dark current of extended wavelength detectors was found to agree well with fits obtained from a 3D carrier drift model using the designed value of Δ. In contrast, the spectral photoresponse showed

extended wavelength thresholds corresponding to Δ . Since the dark current in the extended-wavelength IR photodetectors is still limited by Δ as in the conventional photodetectors, the extended wavelength mechanism offers a new avenue for the design and development of the IR photodetectors.

To understand the role of energy offset for threshold wavelength extension, a few heterostructure semiconductors were comparatively analyzed using references [100–102]. Schematic diagrams of the valence band alignment of the detector designs under equilibrium are shown in Figure 6. Detectors without barrier offset, LH1002 is shown in Figure 6A. SP1001 (Figure 6B) consists of a p-GaAs emitter (80 nm), an 80 nm Al0.75Ga0.25As barrier at the bottom, and a 400 nm Al0.57Ga0.43As barrier at the top. These layers are sandwiched between highly doped p-GaAs top and bottom contact layers. SP1001 has an energy offset (δE) of ~0.10 eV between the barriers. In SP1007, 15SP3, GSU17I, GSU17II, and GSU17III, as shown in Figure 6C, the bottom AlxGa1-xAs barrier (80 nm) has a graded potential profile obtained by increasing the Al mole fraction from *X*<sup>1</sup> at the bottom of this layer to *X*<sup>2</sup> at the top. The top AlxGa1-xAs barrier (400 nm) has a constant barrier potential profile with *X*<sup>3</sup> = *X*4. The emitters (80 nm) are thick enough to have a bulk-like distribution of energy states. Changing the *X*<sup>3</sup> causes the δ*E* variation and changing *X*<sup>1</sup> causes the gradient variation, as summarized in Table 1.

In summary, these studies and the devices LH1002, SP1001, SP1007, 15SP3, GSU17I, GSU17II, and GSU17III were grown on semi-insulating GaAs substrate by molecular beam epitaxy. Each heterostructure device consists of an AlxGa1-xAs barrier at the bottom, followed by a p-GaAs emitter, and then another AlxGa1-xAs barrier at the top. The bottom AlxGa1-xAs barrier is graded by increasing the Al mole fraction from a lower value x1 at the bottom of this layer to higher value *X*<sup>2</sup> at the top, except for SP1001 with *X*<sup>1</sup> = *X*<sup>2</sup> to form a constant barrier. These emitter/barrier layers are sandwiched between the bottom and top contact layers of p-type doped GaAs. The emitter and the top and bottom contact layers are degenerately p-type doped at 1 <sup>×</sup> 1019 cm<sup>−</sup>3, whilst the AlxGa1-xAs barriers are undoped. The thickness of the p-GaAs emitter in all devices is sufficiently large to have a bulk-like distribution of energy states. 15SP3, GSU17I, and GSU17II constitute another set with varying values of the gradient ~20.6, 28.9, and 37.1 kV/cm, given by (Δ*E*v(*x*2) − Δ*E*v(*x*1))/*w*<sup>1</sup> for *x*<sup>1</sup> = 0.45, 0.33, and 0.21, respectively. The gradients are the only differences in these three devices. It should be noted that the device 15SP3 is common in both sets with an δ*E* ~0.19 eV and gradient 20.6 kV/cm. The energy values (ΔTDIPS) corresponding to the determined wavelength threshold are also shown in Table 1. Herein, nearly doubling the barrier gradient from 20.6 to 37.1 kV/cm did not show a significant change in the wavelength threshold. These results confirm that the extended wavelength mechanism originating from the quasi-equilibrium Fermi level at a fixed energy level.

**Figure 6.** Schematic diagrams of the valence band alignment of the detector designs under equilibrium. (**A**) Detectors without barrier offset LH1002. (**B**) SP1001 consists of an emitter at the bottom, the barrier at the top, with an energy offset (δ*E*v) of ~0.10 eV between the barriers. (**C**) In SP1007, 15SP3, GSU17I, GSU17II, and GSU17III have graded bottom by tuning the Al mole fraction and barrier offset δEv not equal to zero.

ݔ ݒܧ ݓ ݔ ݒܧ


**Table 1.** Summary of results from the temperature-dependent internal photoemission spectroscopy (TDIPS) fitting method with corresponding device parameters (threshold wavelength is also shown).

The role of barrier parameters is very critical in determining the photoresponse of extended wavelength detectors. In this view, the effect of δ*E*<sup>V</sup> and gradient on the extended threshold wavelength of infrared photodetectors for the temperature range up to 50 K was studied by Somvanshi et al. [103]. In this study, it is observed the δ*E*<sup>V</sup> is critical to obtain extended wavelength response λeff ( λt) in IR detectors at 5.3 K; however, the gradient is needed to obtain λeff at 50 K. A conventional detector shows λ<sup>t</sup> − λeff over operating temperature from 5.3 K and 50 K whereas the flat injector barrier and barrier-energy-offset-detector (SP1001) shows λeff ~36 μm at 5.3 K and 4.1 μm at 50 K. When the injector barrier changes from flat (SP1001) to graded (SP1007), the λeff increases from 4.1 to 8.1 μm at 50 K for a given graded injector barrier, as δ*E*<sup>V</sup> increases from 0.10 to 0.19 eV, the λeff also increases from 8.9 to 13.7 μm. The results of this study clearly indicate that by the optimization of δ*E*<sup>V</sup> and the gradient, infrared detectors with λeff ( λt) can be designed to operate over a wide range of temperature.

For practical applications, to provide an advantage over the conventional detector, an extended wavelength detector should have a lower dark current and specific detectivity (*D*\*) that is comparable or better than the conventional detector. In general, by lowering the operating temperature, the dark current can be reduced; however, this makes detectors' operation more costly. Therefore, lowering the dark current without cooling will be an advantage, especially for longer wavelength detectors for next-generation optoelectronic devices. In extended wavelength detectors, the dark is determined by designed Δ as in the conventional photodetectors. This leads to an advantage in specific detectivity (*D*\*) for extended wavelength detectors even though their responsivity is much lower as compared to conventional detectors. Using this idea, standard threshold semiconductor detectors could be designed to operate as long-wavelength detectors with a higher value of detectivity and reduced dark current (corresponding to the original short-wavelength threshold).

#### **2. Conclusions**

Objects having temperatures higher than absolute zero (*T* > 0 K) radiates the energy in the form of electromagnetic waves and so are the sources of infrared. Thus, infrared detectors can have diverse civilian and military activities applications. Major applications include: estimating heat losses in buildings, roads or any heat-emitting objects of engineering use; imaging and quality control check in biomedical use; security applications for firefighters, night vision, airports; technological applications for electrical circuit manufacturing or identifying faulty connections; application for astronomical studies and much more.

To date, many materials, and physics characteristics have been investigated in IR detection [60,104]. The example includes [105] thermoelectric power (thermocouples) [106], change in electrical conductivity (bolometers) [107], gas expansion (Golay cell) [108], pyroelectricity (pyroelectric detectors) [109], photon drag, Josephson effect (Josephson junctions) [110], internal emission (PtSi Schottky barriers) [111], fundamental absorption (intrinsic photodetectors) [112], impurity absorption (extrinsic photodetectors), low-dimensional solids (superlattice (SL) and quantum well (QW) detectors) [113], and so on. Progress in semiconductor infrared detector technology has been

made in IR detector technology to improve their performance. Furthermore, studies are going on for the improvement of detector performances. In reference to this, potential applications of a heterostructure IR detector working within a wide range of temperatures, including a room-temperature environment, can be anticipated. Of course, innovation in the heterostructure architecture of thoroughly documented semiconductor compounds will create another degree of freedom in recent technology.

This study documented the present studies on the S-O band energy, and H-H/L-H transitions give rise to the photoresponse in heterostructure infrared detectors with a peak in the 3–5 μm regime, with the most important band being an atmospheric window. The incorporation of current blocking barriers to lower the dark current density and to increase the performance of heterostructure semiconductor detectors is also discussed. The importance of the wavelength threshold extension mechanism is also clearly illustrated, with its application for future IR photodetector design and development. Finally, a considerable achievement in the practical application of a new kind (unlike the conventional) of IR photodetectors, with the wavelength threshold determined in the near future, can be anticipated.

Photovoltaic infrared sensor arrays for the 3–5 μm and 8–12 μm atmospheric windows can be fabricated using similar techniques. GaAs/AlGaAs heterostructures can overcome the serious technological problem for the mass production of HgCdTe large-sized Focal Plane Arrays (FPA) [114], and to achieve high FPA pixel uniformity and yield. Large line or area arrays of such photovoltaic infrared detectors can be desired for thermal imaging and spectroscopic applications.

**Author Contributions:** Conceptualization, A.G.U.P.; writing—original draft preparation, H.G.; supervision, A.G.U.P.; writing—review and editing, P.V.V.J., D.S., Y.L., and A.G.U.P. All authors have read and agreed to the published version of the manuscript.

**Funding:** U.S. Army Research Office (ARO) (W911 NF-15-1-0018).

**Acknowledgments:** H. Ghimire acknowledges support from Molecular Basis of Disease (MBD) at GSU.

**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/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Micromachines* Editorial Office E-mail: micromachines@mdpi.com www.mdpi.com/journal/micromachines

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18

www.mdpi.com

ISBN 978-3-0365-3354-4