This Perspective seeks to understand and assess why ultrawide bandgap (UWBG) semiconductor-based deep-UV photodetectors have not yet found any noticeable presence in real-world applications despite riding on more than two decades of extensive materials and devices’ research. Keeping the discussion confined to photodetectors based on epitaxial AlGaN and Ga2O3, a broad assessment of the device performance in terms of its various parameters is done vis-à-vis the dependence on the material quality. We introduce a new comprehensive figure of merit (CFOM) to benchmark photodetectors by accounting for their three most critical performance parameters, i.e., gain, noise, and bandwidth. We infer from CFOM that purely from the point of view of device performance, AlGaN detectors do not have any serious shortcoming that is holding them back from entering the market. We try to identify the gaps that exist in the research landscape of AlGaN and Ga2O3 solar-blind photodetectors and also argue that merely improving the material/structural quality and device performance would not help in making this technology transition from the academic realm. Instead of providing a review, this Perspective asks the hard question on whether UWBG solar-blind detectors will ever find real-world applications in a noticeable way and whether these devices will be ever used in space-borne platforms for deep-space imaging, for instance.
The ozone layer fully absorbs solar radiation for λ < 290 nm; thus, there exists no natural source of light on the ground at these wavelengths. Solar-blind detectors, by definition, thus, reject the visible part of the spectrum and find applications1 in missile plume detection, flame monitoring, non-line of sight communication, and environment monitoring besides other industrial, bio-medical applications. Outside the atmosphere, solar-blind sensors aboard spaceships are employed to image astrophysical sources2 and phenomena in the UV, and these sensors are required to be insensitive to the visible and IR spectra. Silicon-based charge coupled devices (CCDs),3 complementary metal oxide semiconductor (CMOS) devices, and UV-enhanced silicon photodiodes are most widely used for this purpose, but they need optical filters4 to reject the visible/IR wavelengths, adding to complexity and weight. Photomultiplier Tubes (PMTs)5 and Microchannel Plates (MCTs)6,7 are traditionally used for boosting the gain via carrier multiplication, which also enables single photon counting, but these are quite bulky and require large voltage (∼kV) to operate.
Ultrawide bandgap (UWBG) semiconductors, such as AlxGa1−xN, Ga2O3 and its alloys, diamond,8 hBN,9 and MgxZn1−xO,10 are quite attractive for solar-blind imaging or sensing applications in the deep-UV (λ < 290 nm) regime. These materials, by virtue of their large bandgaps, intrinsically absorb radiation only in the deep-UV, thereby naturally rejecting visible light by several orders of magnitude, which is expected to eliminate the need for external filters. UWBG devices are also radiation-hard, lightweight, and require no cooling for reducing the dark current, which make them promising candidates for high-efficiency and compact solar-blind imaging. Yet, in the growing UV sensors’ market, which is estimated to reach USD 2.7 billion11 by 2024, detectors based on UWBG semiconductors have barely found any presence in real-world systems and imagers that employ UV sensing. It must be mentioned, however, that “UV sensors” would also encompass detectors that operate in the UVA and UVB regimes spanning wavelengths in the range of 290–400 nm; although III-nitride (AlGaN and InGaN) detectors of suitable alloy compositions would detect UV radiation in this range, they are not solar-blind UV detectors in the strict sense. This Perspective would confine its discussion to deep-UV or solar-blind UV detection only which is where UWBG materials and their heterostructures hold immense promise.
What are the reasons why UWBG semiconductor-based solar-blind UV detectors have barely found any presence in real-world systems such as UV imager aboard spaceships or flame monitors? Indeed, CCDs and CMOS-based UV sensors have been dominating this landscape despite their complexities with additional filters, lower efficiency, and weight. This has kept UWBG UV detector research mostly confined to academia. What would it require to take UWBG detectors from academic labs to real-world applications in space, industry, and healthcare?
Figure 1 seeks to capture the essence of various aspects pertaining to UWBG detector technology. For instance, while there has been a staggering volume of literature reports on all aspects of materials’ growth and device development for UWBG photodetectors, yet the studies of focal plane arrays (FPAs) and their integration with Readout Integrated Circuit (RoIC) are relatively much less explored. There are even fewer reports on the investigation of reliability and yield, age/climatic age testing, radiation hardness, and failure mechanisms of such device technologies. This gap in the literature presents a significant hurdle toward commercial and strategic deployment of real-world systems that utilize UWBG deep-UV sensors for imaging.
In this Perspective, we look at the status of epitaxially realized deep-UV detectors based on two most widely studied UWBG semiconductors, i.e., AlxGa1−xN and Ga2O3, and reflect on the challenges and the road ahead. Although there are impressive device results based on nanostructures (e.g., nanowires and nanobelts)12–15 of UWBG materials and amorphous/polycrystalline Ga2O3,16,17 yet large-scale manufacturability and uniformity of devices on such platforms are extremely challenging to achieve as of date. This is not to conclude that such developments are impossible: nanowire LEDs18,19 are there in the market and quantum dot-based displays are ubiquitous. However, our discussion will be confined to epitaxial UWBG devices only; we shall seek to understand what it would take for the materials and device communities to translate UWBG detectors into a mature and robust solar-blind imaging technology such that market adoption of these sensors becomes a reality. Sections II and III deal with AlGaN- and Ga2O3-based deep-UV detectors, respectively, while in Sec. IV, we summarize our concluding thoughts and remarks on this topic.
II. PERSPECTIVE ON AlGaN DEEP-UV DETECTORS
A. General discussion: Material quality, substrate, doping, and dark current
There are several excellent review articles20–22 and book chapters23 published on the status of AlGaN deep-UV detectors, including those that include reviews of Ga2O3-based counterparts as well.24 Many of these reviews have provided extensive and exhaustive coverage of all aspects of AlGaN UV detectors including the status on material growth25 and doping, structural qualities, device development efforts, and the exploration of UV FPAs based on the same. The status of AlGaN UV detectors in terms of the state-of-art device performance parameters such as responsivity, dark current, transient response, UV-to-visible rejection ratio, and photo-to-dark current is often captured in such reviews, either graphically or in tabular formats. Thus, we shall eschew the detailed discussion of the status of AlGaN photodetectors and their material development vis-à-vis their performance as reported by various groups. Instead, our focus will be on the pragmatic assessment of the broad area of UWBG detectors.
In the sections ahead, we may invoke the device status of the commercially successful InGaAs detectors for a baseline comparison of AlGaN deep-UV detectors to understand whether basic shortcomings in the device technology have led to a gap between academic research and market viability for the latter. It must be mentioned that AlGaN and InGaAs are fundamentally different material systems, and deep-UV and IR detectors have very different application spaces, requirements, and expectations. Yet, from a material-agnostic and application-agnostic point of view, a qualitative and broad comparison among different photodetector technologies could help us calibrate the status of UWBG detectors. This could aid us in evaluating what has prevented AlGaN detectors from competing with CCD/CMOS-based UV sensors in the market. This is an important aspect to ponder because AlGaN/GaN high electron mobility transistors (HEMTs) and AlGaN-based deep-UV photodetectors emerged almost around the same time, which is mid to late 1990s. Yet, while AlGaN/GaN HEMTs are fast entering the market26 and the strategic sector,27 AlGaN deep-UV detectors are still confined within the realm of academic research. It is true that HEMTs and UV detectors have widely different applications, market spaces and competition; however, it is imperative to pragmatically assess the road ahead for AlGaN deep-UV detectors in terms of their real-world usage.
Since the Al-content in AlGaN can be tuned to tailor the bandgap and, hence, the cut-off wavelength between 200 and 365 nm, there are several studies on the demonstration of AlGaN-based UVB detectors as well.28–30 However, for solar-blind detection (<290 nm), the Al-content in AlGaN needs to be at least 40%, and this poses a few challenges from the point of view of materials development. The epitaxy of AlGaN with high Al-content (>40%) is itself non-trivial when it comes to controlling the background impurity concentration, doping,31 stress,32 dislocation densities,33,34 and other defects like V-pits28,35 and spiral hillocks,36 all of which substantially effect the device performance. On top of that, p-type doping is a major difficulty, given that Mg ionization energy is quite high for AlGaN.37,38 In fact, even the study and development of a highly conductive n-type AlGaN39,40 with a high Al-content is still an important topic for research.
Epitaxial growth of AlGaN, particularly by metal organic chemical vapor deposition (MOCVD), and subsequent design, fabrication, and characterization of deep-UV photodetectors have been studied since 1990s. Since native AlGaN substrates do not exist, these epi-layers are grown hetero-epitaxially, usually on a sapphire substrate although growth on silicon,41,42 AlN bulk crystal,43 and SiC44,45 have also been reported. To alleviate material issues such as V-pits and threading dislocations (TDs) arising due to heteroepitaxial growth, various growth techniques like pulsed atomic layer epitaxy (PALE),46 lateral epitaxial overgrowth (LEO),47,48 and migration enhanced epitaxy49 have been reported toward the growth of superior quality AlGaN epi-layers. In the last 25 years of research, thus, the MOCVD growth of AlGaN (>40% Al-content) on sapphire has matured significantly, leading to the demonstration of excellent quality epi-layers even without the use of techniques such as PALE and LEO. For instance, use of a low-temperature (LT) AlN followed by a high-temperature (HT) AlN nucleation layer significantly helps reduce coalescence boundaries, thereby enabling the growth of low dislocation density layers with sub-nm surface roughness and facilitating stress engendering for subsequent AlGaN layers.50 TD densities (TDDs) in the low- to mid-108 cm−2 are now commonplace for AlGaN on sapphire although PALE and LEO can reduce this by another order of magnitude albeit at the price of higher process complexities and costs. In fact, the combination of LT and HT AlN nucleation layers has led to the demonstration of superior quality AlGaN (Al > 40%) epi-layers even on silicon (111) substrates.51
For solar-blind UV detectors, sapphire is more commonly used than other substrates due to the fact that sapphire is transparent to deep-UV (>190 nm), allowing for back-illumination of the devices, which is required for FPAs in flip-chip-bonded configuration.52 Lower lattice mismatch with AlGaN, lower cost (than SiC and bulk AlN), good growth control, and large-scale manufacturability also favor the choice of sapphire. The epitaxy of AlGaN on silicon is much more difficult with a larger thermal coefficient mismatch, higher dislocation densities, and a significant wafer bow that could adversely affect large-area device fabrication steps. Yet, due to cost and economy of scale, AlGaN on silicon is a promising route to UWBG deep-UV detectors,53,54 and more so, because the existing CMOS fab-lines can be used for mass production of such epi-wafers on silicon. Monolithic integration of AlGaN-on-silicon-based FPA imaging arrays with existing CMOS backend electronics is, however, still not likely since the former is on silicon (111) while CMOS is based on silicon (100). However, a silicon platform allows for an easy removal of the complete substrate and subsequent heterogeneous integration of such devices on literally any other platform, which could hold promise for the emerging area of “system scaling.”55 Also, several space-related deep-UV applications require imaging at still shorter wavelengths, going all the way down to 10 nm,56 which forms the region of vacuum UV. A sapphire substrate is a poor choice for back-illumination in such a case as it is not transparent below 190 nm. On the other hand, selective or even complete etching or removal of silicon makes AlGaN-on-silicon a more feasible option for vacuum UV detection under back-illumination. However, as mentioned already, there are only a handful of reports on AlGaN UV detectors on silicon,53,54,57–60 with high Al-content. This is not withstanding the fact that epitaxy of AlGaN (with high Al-content) on silicon is widely studied due to its importance as transition/buffer layers61,62 in GaN HEMTs-on-silicon, which are of great commercial significance. AlGaN growth on silicon for GaN HEMT, however, does not mimic the requirement of AlGaN on silicon for deep-UV detectors. The latter may require doping (both n- and p-types), which is not a requirement for GaN HEMT; second, non-radiative recombination centers, minority carrier lifetime, and diffusion length that are important to consider for assessing the optical quality of a material are not critical for GaN HEMTs where C-doping is usually implemented to lower the leakage.
The epi-layer growth and development of the material platform must be done in conjunction with the device feedback; thus, a correlation among the growth conditions, structural qualities, and device performance is critical to understanding, evaluating, and improving the status of UWBG deep-UV technology. The photodetector device parameter that can be easily measured is the dark current or leakage current, the value of which can be used for a quick assessment of the overall material quality of the epi-layer(s). For an undoped sample, the AlGaN epi-layer(s) is (are) expected to be highly insulating, and the most common type of photodetector to be realized on such platforms is of MSM or metal–semiconductor–metal geometry that employs two back-to-back Schottky contacts. The dark current in MSM detectors, which flows laterally, would, however, depend on the overall device dimensions, including the width and the number of the inter-digitated fingers and on the spacing between adjacent fingers. These device dimensions do vary among the various reports published in the literature but usually not by a significant margin. The fact that ultralow dark currents in the range of femtoamperes have been quite often reported for MOCVD-grown AlGaN MSM devices on sapphire63–66 testifies to the maturity of AlGaN epitaxy. The role of screw and edge-type dislocations in mediating dark current, and the transport mechanism behind such leakage have also been quite widely studied67–69 as a part of correlating device performance to the structural quality of AlGaN on sapphire.
MSM detectors with their inter-digitated fingers and large peripheries in a lateral geometry, though the simplest to realize in terms of epitaxy and device processing, do not inspire the development of FPAs where indium bumping and flip-chip bonding are essential for integration to RoIC. Photodetectors with vertical and quasi-vertical architectures with smaller periphery and top/bottom contacts such as p–i–n and Schottky diodes are favorable for implementing FPAs. The literature abounds in reports of AlGaN p–i–n70–80 and Schottky81–87 photodetectors exhibiting excellent performance in the solar-blind regime. There have been quite a few reports71,88–94 on arrays and FPAs based on such vertical and quasi-vertical photodetectors including the demonstration of deep-UV imaging as well. The development of Schottky and p–i–n photodetectors on sapphire, however, necessitates doping and more process complexities such as mesa isolation and passivation, making these more resource-intensive and expensive to realize. Particularly, doping is a non-trivial exercise in AlGaN with high Al-content. Although reasonably good conductivity in n-doped AlGaN (Al-content > 40%) is achievable,31,39,95,96 p-type doping is extremely challenging due to large dopant ionization energies97,98 and incomplete solubility of the dopant (Mg) atoms in AlGaN.99,100 This makes highly conductive p-AlGaN (Al-content > 40%) extremely challenging to realize. The conductivity of the doped AlGaN layer depends on the dislocation density101 and, hence, on the underlying buffer layers also. Edge dislocations are known to trap carriers and also act as Coulombic scatterers as the presence of negatively charged acceptor-like charged states is often observed along their lines.102 Controlling the crystal quality, by using a superlattice buffer,103,104 for instance, is, therefore, of paramount importance in realizing highly conducting doped AlGaN epi-layers.105 Figure 2 shows some of the state-of-art conductivity values for p-doped74,99,106–114 and n-doped AlGaN39,115–124 as reported by various groups in the literature as a function of Al-composition. As expected, as the Al-content increases, it becomes more difficult to achieve high doping levels or conductivity values in AlGaN. For Mg-doped AlGaN, the use of reverse polarization grading, i.e., grading the composition of AlGaN from a high to a low Al-content along the growth direction, leads to polarization-enhanced p-doping and is a possible route toward realizing AlGaN with relatively higher p-conductivity as seen in Fig. 2. More recently, highly conductive p-doped AlN has been reported125 employing metal modulated epitaxy (MME) using a beryllium source as a p-dopant. Thus, despite the challenges in achieving high conductivity in AlGaN—especially, p-AlGaN, there has been reasonable progress on this front in the recent years, and this presents an important area of research in the broad area of UWBG opto-electronics.
Besides the challenges in doping, leakage currents in p–i–n and Schottky diodes are extremely sensitive to screw dislocations acting as shunt paths, making material growth, the most crucial aspect of enabling this technology. While p–i–n and Schottky diodes with ultralow dark currents76,81,86,87 have been reported on sapphire substrates by careful engineering of the buffer layers, epi-layers on silicon exhibit higher dislocation densities and, hence, pose a more serious problem. Shunt leakage paths arising out of high dislocation density are serious issues that have deleterious effects on device reliability and long-term operations. Growing AlGaN on bulk single crystal AlN substrates126,127 is another route toward reducing dislocation density and is being widely reported for the demonstration of high-performance deep-UV LEDs; however, scalability, the development of large-area AlN substrates, and the associated costs make it less attractive for possible commercial applications. That being said, performance rather than cost is the major driver in strategic and space-based applications. Thus, if AlGaN on bulk AlN substrates can enable high-performance solar-blind detectors that are highly reliable and robust, these could be potentially attractive for military and space-borne applications. Given that there are few studies128 on a solar-blind AlGaN photodetector on the bulk AlN substrate, this presents a fertile and interesting area of research for UWBG detectors.
One of the key areas in AlGaN solar-blind photodetectors where significant research efforts are still required is avalanche photo diodes (APDs) with an ultrahigh gain and superior noise performance. In this context, it must be mentioned that APDs in the AlGaN system are often demonstrated primarily using a p–i–n geometry and seldom in a Schottky configuration. However, the difficulty in effective p-type doping AlGaN with high Al-content and the difficulty in estimating the impact ionization coefficients for AlGaN with different Al-compositions have slowed down the progress of AlGaN APDs for solar-blind operation. There have been some studies on estimating the impact ionization coefficients of AlGaN,129–132 and this is expected to aid in the further development of high-performance AlGaN APDs. Yet, there are reports of AlGaN APDs exhibiting gain (G) > 104 for epi-layers realized on sapphire substrates133,134 and even reaching 105 for devices on single crystal AlN substrates,128 making them attractive for space-borne deep-UV imaging. Thus, there is sufficient scope for the development and optimization of APDs based on AlGaN with high Al-content on both sapphire and on AlN substrates, especially in terms of controlling the structural quality via dislocation reduction, understanding leakage mechanisms vis-a-vis breakdown, and precisely evaluating the impact ionization coefficients for AlGaN with different Al-contents or for compositionally graded AlGaN.
Are challenges in p-doping, high dislocation density in AlGaN epi-layers, and consequent reliability concerns in the devices the primary reasons that have held back UWBG detector technology from being used in real-world sensing or imaging despite 25 years of research? In absolute terms, dark currents at the limit of instrument capability (fA) have been achieved; so, dark current cannot be a reason for inhibiting real-world deployment of AlGaN UV detectors. Cost, large-area growth, and large-scale manufacturability cannot be the primary bottlenecks as well, since the multibillion-dollar white LED industry thrives on epitaxy and device processing on GaN/InGaN epi-layers grown on large-area (8- to 12-in.) sapphire wafers. Controlling the uniformity of AlGaN (high Al-content) in terms of thickness and composition over a 6- or 8-in. sapphire is something that must be carefully studied and optimized for this. However, this cannot be expected to be a showstopper.
To understand whether limitations and challenges in device development have held back successful commercial deployment and market penetration of AlGaN detectors, we discuss about the performance of AlGaN solar-blind detectors vis-à-vis their figures of merit.
B. Device performance and a new figure of merit
Dark current alone cannot capture the entire essence of the structural quality of the material vis-à-vis the detector performance. Responsivity, external quantum efficiency (EQE), photo-to-dark current ratio, UV-to-visible rejection ratio, noise equivalent power (NEP), specific detectivity, dynamic linearity, and 3-dB bandwidth (or transient response) are important device parameters to consider for solar-blind photodetectors. The relative importance of each of these figures of merit depends on the application. For instance, in deep-UV imaging of astronomical objects, often, the speed or bandwidth may not be quite important as the UV detector may have to wait for a long time to even detect a single photon. In certain astronomy-related applications, fast response even for deep-UV photodetection maybe highly desired. In any case, these devices need to be extremely sensitive with excellent visible rejection and noise performance. On the other hand, in missile plume detection in the UV, the transient response (speed) of the detector is super important besides its sensitivity.
There have been several reports on AlGaN solar-blind detectors with excellent performance in terms of many of these device parameters. Most of these reports have been about AlGaN detectors with Al-content between 40% and 70% (250 < λ < 290 nm), grown usually on sapphire. There are studies on devices based on AlN135,136 as the active absorber layer also, but such reports are quite few. In general, for most of the AlGaN detectors including those that are on the silicon substrate as well, the device parameters that are usually reported include responsivity (and gain), EQE, photo-to-dark current, UV-to-visible rejection, and transient response. For a conventional detector, gain (G) is defined as the ratio between the measured responsivity to the theoretical responsivity (for EQE = 100%) at a particular wavelength. However, for avalanche photodiodes (APDs), gain is defined as the ratio of the difference in the photo and dark current near the onset of avalanche to that at near zero bias. Usually, the responsivity and, hence, the gain depend strongly on the bias applied in all kinds of AlGaN photodetectors. For instance, high responsivity values of >200 A/W (G ∼ 1000) had been reported for MSM68-type AlGaN photodetectors at peak wavelengths of 265 nm operating in a non-avalanche mode. Such a high gain, indicative of an excellent photosensitivity, also suggests the presence of traps that induce barrier lowering and subsequent enhancement of the photocurrent.69 Photodetectors exhibiting high gain on account of traps, however, are slow to respond. The rise time in a transient response is loosely proportional to the inverse of the bandwidth and is quite extensively reported by various groups to be in the range of a few milliseconds to a few seconds.
In this context, reference must be made to the persistent photoconductivity (PPC) effect that had been widely studied137 and reported for photodetectors based on AlGaN and on AlGaN/GaN HEMTs.138 PPC is the observed phenomenon of photocurrent sustaining for a long time even after the source of light is turned off and leads to a slow transient. For photodetectors based on AlGaN/GaN HEMTs,139 PPC has been attributed to deep impurity levels or DX centers in the AlGaN barrier140,141 and the energy barrier for electron capture corresponding to these DX centers have been reported to be about 0.3 eV. Similar observations have been made for AlGaN MSM detectors also where trapping at deep level centers, microscopic defects like dislocations and vacancies,142 as well as macroscopic defects like cracks. Even for an n-doped AlGaN, PPC is attributed to deep level states rather than energy states corresponding to silicon donors.143,144 Usually, such PPC effects can be fitted with a stretched exponential, and with higher temperature or with higher bias, persistent photoconductivity can be mitigated to a significant extent.145 It must also be noted that PPC is oftentimes reported for sub-bandgap illumination, and this makes PPC an important phenomenon to understand and suppress for UWBG UV detectors because it could lead to spurious detection at longer wavelengths making them lose their solar-blind property.
It is noteworthy that the frequency spectra of AlGaN detectors or the measurement of 3-dB bandwidth are seldom reported. Among the few studies that report this, we must mention the demonstration of bandwidths of 5.4, 4.1, and 1.65 GHz, respectively, for AlGaN MSM (λ = 222 nm),146 Schottky (266 nm),86 and p–i–n (267 nm)147 photodetectors, corresponding to transient response periods of sub-100 ps. These numbers indicate that AlGaN solar-blind detectors can be realized with ultrafast response times, rivalling the more mature and widely commercialized InGaAs IR detectors in terms of response speeds. More recently, response time < 25 ps has been reported for AlGaN MSM devices.148 Thus, we can conclude that the transient performance of AlGaN solar-blind UV detectors is not a reason that has held back the real-world adoption of UWBG UV technology.
It must be noted here that there is a fundamental trade-off between gain (or sensitivity) and the 3-dB bandwidth (or the speed). Thus, detectors with high gain usually exhibit longer rise times and vice-versa. The physics of such trade-offs and limits to the performance of photodetectors are extensively discussed in the literature.149,150 The gain-bandwidth product (GBP) is, therefore, a useful figure of merit (FoM) that captures both aspects (i.e., sensitivity and speed) of a photodetector and indicates how good it is on a relative scale. However, this FoM may mask the assessment of a detector for a particular application: for example, in UV astronomy, the sensitivity needs to be super high while the speed is usually of no consequence; but in terms of this gain-bandwidth product FoM, an excellent detector for UV astronomy may fare poor due to its lower speed.
Given that very few reports exist on the measurement of 3-dB bandwidth of AlGaN deep-UV detectors, G/τav could be used as a figure of merit commensurate with GBP, where G is the gain and τav is the harmonic mean of the rise time and the fall time as reported in transient measurements. It is noteworthy that if the responsivity is less than the theoretical value given by qλ/hc, then G < 1 and is ideally not termed as a “gain.” In such cases, the EQE is more appropriate to use although we shall refer to G as the ratio between the measured to the theoretical responsivity value at a particular (peak) λ for non-avalanche mode photodetectors. For avalanche detectors, G will be used in the conventional sense. This also enables us to compare the performance of AlGaN deep-UV detectors with their InGaAs counterparts as far as figure-of-merit is concerned without worrying about absolute responsivity values in A/W which, theoretically speaking, would depend on the wavelength of detection. However, the parameter G/τav would not capture the noise performance of a photodetector and, therefore, cannot be a comprehensive figure-of-merit that may be used to evaluate detectors across different geometries and material systems, even for deep-UV applications.
The noise performance including NEP and specific detectivity (D*) of AlGaN deep-UV detectors are seldom measured81,83,151–153 and reported; however, it is a common practice to estimate or calculate D* based on thermal and dark current limited noise, applying standard equations.137 In fact, most of such estimates for AlGaN detectors, for instance, ignore the dark current component and calculate the thermal noise limited D* only or vice-versa. Thus, the values of calculated D* do not actually capture the true behavior of a photodetector as far as noise measurements are concerned. Since only a few reports exist on the noise measurement of AlGaN deep-UV detectors, the calculated values of maybe used to capture the essence of D* instead, where Rλ is the responsivity (at the bias considered) and ID is the corresponding dark current. The expression for D* indicates that it is roughly proportional to even if both dark current (shot noise) and thermal resistance factors are accounted for. For Schottky and p–i–n diodes, ID would depend on the area, and hence, the normalized values (A/cm2) should be used. For MSM detectors, to a first order, we may ignore the variation of device dimensions as mentioned above and, hence, rule out normalization. However, it would be more useful if the figure of merit is dimensionless so that detectors of different geometries operating across different wavelength ranges may be compared. may be divided by qλ/hc (i.e., the theoretical responsivity at EQE = 1) as discussed in the preceding para, and this may be called “G” independent of whether it represents an actual gain or not, which is representative of the EQE. For MSM devices, ID maybe normalized to Ilimit = 1 fA, which is usually the instrument limit for most measurement tools. Besides, to the best of the authors' knowledge, there has been no report of dark current in MSM detectors measured or reported below 1 fA for AlGaN deep-UV detectors. Thus, is a dimensionless FoM that captures the noise performance of detectors.
However, to benchmark photodetectors by accounting for both noise performance and their transient speed, a more comprehensive FoM is required, which encompasses all key parameters. Given that G/τav captures the GBP trade-off in detectors, a more holistic FoM may be introduced by combining it with the noise FoM, and labeling it as the Comprehensive FoM (CFOM) for a photodetector,
Here, τDR = 1 ps is the lowest possible transient response time in an AlGaN-based detector and is invoked to make CFOM a dimensionless quantity. This number was arrived at by considering the fact that at least 100 nm of the absorber layer is required for any AlGaN photodetector to absorb an appreciable fraction of the incident UV light; with a saturation velocity of ∼107 cm/s, the transit time turns out to be ∼1 ps. Even if RC delay is neglected, it is almost improbable to get a sub-picosecond transient response in AlGaN Schottky and p–i–n deep-UV detectors. The constant γ = 1 ps (1 fA)1/2, which is used to make CFOM a dimensionless quantity, may be ignored in plotting the data for comparison.
The CFOM captures the role of noise and that of bandwidth in assessing the overall performance of a photodetector. A higher value of the CFOM implies a superior photodetector, broadly speaking. Usually, detectors (excluding APDs) exhibit high gain due to traps that inherently slow down the device and usually result in a high dark current. This CFOM would, therefore, be useful in comparing detectors operating in the non-avalanche mode but with high gain due to defects. On the other hand, APDs usually have slower responses, and so, it would be unfair to compare APDs with non-avalanche mode photodetectors with respect to CFOM. In Fig. 3, we plot the CFOM for p–i–n and Schottky AlGaN-based solar-blind deep-UV74,83,86,154–159 and commercially available InGaAs-based160–165 near-IR photodetectors. In many cases, especially with commercial detectors, only the rise time is provided; hence, we have taken rise time for all cases. The data points in Fig. 3 correspond to reverse biases in the range of 5–20 V; this is because published reports in the literature do not mention or plot the various values of responsivity, transient response, and dark current at the same bias. However, the plot would only marginally change due to this fact.
Although InGaAs detectors have ultrafast transients, it is seen that AlGaN p–i–n and Schottky detectors have also been reported with similar speed (tens of picoseconds) but with superior dark current due to the ultrawide bandgap nature of AlGaN. This results in both p–i–n and Schottky detectors in AlGaN exhibiting CFOM at par or even better than that of their InGaAs counterparts, spanning 250–290 nm of operation. However, it must be noted that these data points for the deep-UV regime correspond to AlGaN detectors realized in academic and other research labs, while for near-IR, the InGaAs detectors are commercially available, implying those are industry-qualified and reliability-tested, rugged, and robust devices.
From Fig. 3, it is abundantly clear that purely from the device performance parameters’ point of views, state-of-art AlGaN deep-UV photodetectors are easily at par with the highly mature and successful InGaAs near-ID detectors. That is, there is no fundamental or even practical limit to the performance of AlGaN photodetectors (in p–i–n and Schottky geometries at least), which prevent them from being useful for deep-UV detection or imaging applications requiring high-speed and excellent noise performance. Although reliability and degradation mechanisms of AlGaN detectors require further studies, we can conclude that the basic parameters of device performance (bandwidth, EQE, and noise) are not the reasons that have prevented noticeable real-world adoption of these devices.
As mentioned earlier, in space-borne deep-UV astronomical imaging, photon counters are often required as it may take several hours for just one deep-UV photon to arrive. These detectors, thus, require super high gain, which is why MCPs and PMTs are commonly employed to boost the gain in conventional CCD-based deep-UV imaging solutions. In this context, AlGaN APDs are, therefore, highly promising, particularly if they can rival the gain of MCPs and PMTs. However, APDs are usually not fast in their responses. A much lower rise time for APDs would make their CFOM values much lower than those corresponding to non-APDs, which is why the proposed CFOM may not be the best figure of merit to compare or benchmark AlGaN APDs. For APDs, usually gain and noise are the two most critical device parameters to consider for any benchmarking. However, there is barely any report on the measurement or study of noise performance of AlGaN APDs. So, we refrain from making any comparison of AlGaN APDs with noise as a parameter. Instead, another parameter that is quite important from the point of view of practical applications of AlGaN APDs in a space-borne platform is their ability to reject visible and IR parts of the incident radiation. This is equivalent to asking: how excellent is the UV-to-visible rejection ratio (UVRR) of a solar-blind photodetector? Presently, optical filters and Cs-decorated cathodes are commonly employed in space-borne platforms to reject longer wavelengths in CMOS/CCD-based deep-UV imaging. AlGaN APDs can easily get rid of this complexity and unnecessary weight by virtue of being intrinsically solar-blind. This is a critical advantage offered by UWBG devices. Various studies published in the literature usually report of UVRR being in the range of 103–106 for AlGaN, defined as the ratio of the responsivity values at the peak wavelength to that at 400 nm.
In real-world space-borne platforms, for solar-blind imaging, photocathodes made of or decorated with Cs–Te or similar compounds are used for rejecting the visible and IR radiation while PMTs provide the necessary gain. Usually, Cs–Te exhibits UVRR exceeding 104 while PMTs166 exhibit gain exceeding 105 at nearly a kV of bias although it can exceed 107 at a bias of a couple of kV. Thus, to be a viable competitor to replace the Cs–Te photocathode and PMT combination, AlGaN photodetectors need to exhibit both gain and UVRR values that are at par or better than the above ones. In this context, the product of UVRR and gain (UVRR*G) would be a fair yardstick to benchmark the state-of-art in UWBG UV detectors with respect to the solar-blind PMTs in vogue. In Fig. 4, we scatter plot some data points for this product with respect to the peak wavelength of detection for AlGaN solar-blind photodetectors in p–i–n, MSM, and Schottky geometries as reported in the literature.69,70,74,76,83,86,130,134,144,145,147,167–175,230,240,245,246 A few data points for AlGaN APDs are also reported along with a lone point for an AlGaN detector that exploits a two-dimensional electron gas (2DEG) at the interface of two AlGaN layers with different Al-contents. At 280 nm of peak wavelength, this 2DEG-based AlGaN photodetector has the highest product of UVRR and gain, outperforming APDs and even commercial-grade solar-blind PMT154 as is evident from Fig. 4, and, thus, testifies to the immense promise that UWBG detectors hold for lightweight, compact, and efficient solar-blind UV imaging. The PMT, however, has a peak at 240 nm, and around such wavelength ranges, AlGaN detectors are yet to be demonstrated with superior gain and UVRR.
Next, we look at FPAs and device reliability of AlGaN deep-UV detectors vis-à-vis their real-world applications.
C. FPAs, integration, and device reliability
While discrete detectors are useful for sensing applications, an FPA is what is required for imaging. One of the most important and widely used areas for deep-UV detection is UV astronomy. AlGaN FPAs are quite attractive for this, given their intrinsic solar-blind nature. Usually, compound semiconductor-based photodetectors need to be integrated in a hybrid-configuration with a CMOS ROIC using indium bump for flip-chip bonding. This is because III–V or III-nitride devices cannot be monolithically realized on a single CMOS silicon (100) substrate. Flip-chip-bonded FPAs require back-illumination that is why sapphire as a substrate is appropriate for the growth of the epi-stack if the target imaging spans λ ≥ 200 nm. However, deep-UV imaging for astronomy, for instance, requires detecting signals at much shorter wavelengths such as for λ < 100 nm. Sapphire is opaque to such wavelengths and, thus, prevents AlGaN-based deep-UV FPAs from being useful for vacuum-UV imaging in astronomy. The silicon substrate, on the other hand, can be completely removed with either grinding or deep reactive ion etching (DRIE); the resultant “membrane” of the III-nitride epi-layers may be transferred and flip-chip bonded to ROIC, thereby enabling imaging down to a sub-100 nm wavelength. However, maintaining the membrane integrity and mechanical stability would be extremely challenging if the silicon substrate is removed completely from the entire wafer. Therefore, silicon is etched away selectively to yield a honeycomb pattern.176 This maintains the membrane integrity at the cost of reduced fill factor and efficiency. There are only a few studies on deep-UV imaging based on FPAs of AlGaN photodetectors on sapphire52,73,90,177–181 and much fewer on silicon.59,164,182 Most of these are p–i–n based although a few reports exist on FPAs realized using Schottky detectors. These FPAs are usually 2D in nature, and typical AlGaN FPAs have been reported with 320 × 256 or 256 × 256 pixels with a pitch of 30 × 30 or 25 × 25 μm although more recently, FPA with 640 × 512 pixels is reported. FPAs with good pixel uniformity and the state-of-the-art discrete detector performance in terms of responsivity and dark current have been reported. Most of the AlGaN FPAs studied so far have demonstrated imaging at 260–290 nm wavelengths although imaging at shorter wavelengths is also reported.56 The imaging demonstrated is almost always based on an artificial object illuminated by a broadband UV source or of beam shaping in a synchrotron line; solar-blind imaging of natural sources such as stellar objects in UV astronomy (using UV detectors on board spaceships) using AlGaN-based photodetectors is yet to be reported to the best of the authors' knowledge. Thus, there exists an enormous scope to study, design, and improve AlGaN-based FPAs on both sapphire and silicon substrates toward realizing space-based imaging of astronomical bodies. This is especially true for AlGaN-based APDs, given that there has been no study so far on the demonstration of FPAs based on the same for solar-blind applications. This could possibly be due to (a) the overall challenges in realizing AlGaN APD as mentioned earlier and (b) the stringent requirements on the device-to-device uniformity in terms of gain, across the wafer.
Pixel uniformity, i.e., device-to-device variability and readout noise, is extremely important to consider while realizing FPAs. This necessitates excellent control on the epi-layer growth in terms of uniformity in the thickness, doping, and composition of AlGaN layers across the wafer. This also requires a stringent process control in each of the unit processes of device fabrication to ensure that variations in device parameters such as contact resistance, etch depth, sidewall leakage (suppressed by passivation), etc., remain within tolerance with as low an error margin as practicable across the wafer. For large-scale manufacturability, AlGaN FPAs would have to be realized on epi-layers grown on 6- or 8-in. sapphire or silicon wafers, and this would involve addressing a different set of challenges including the yield, wafer bow (for growth on silicon), compositional uniformity across such large wafer dimensions, and so on. However, before that, AlGaN FPAs on a 2-in. sapphire and on silicon need more extensive studies, including detailed reliability tests for discrete devices and arrays, and correlation between material growth, structural properties such as dislocation density and the detector reliability outcomes. Particularly, for AlGaN epi-layer growth on silicon, a major challenge is to keep the layers crack-free post-growth. Even if there are a few cracks or micro-cracks, the bigger challenge is to keep these cracks from multiplying further or propagating during reliability tests. Epitaxial Lateral Overgrowth (ELO) and Reduced Area Epitaxy (RAE) are two approaches adopted for reducing the strain in the layers as well to obtain smoother surfaces with relatively lower dislocation density. Crack-free detector arrays on silicon have exhibited excellent performance and uniformity,56 and the feasibility of adopting such techniques for 6- or 8-in. wafers on a commercial scale needs to be evaluated.
For use in real-world applications, AlGaN-based deep-UV detectors need to be reliable. It is indeed difficult to find in the literature the industry standard for qualifying deep-UV detectors although reliability tests to qualify InGaAs IR photodetectors for space-based applications are reported.183 There are only very few reports studying the reliability of AlGaN deep-UV detectors. In fact, there is just one report on studying the degradation mechanism of AlGaN MSM detectors on sapphire184 under extended hours of operation under UV illumination. It was reported that unpassivated AlGaN MSM detectors with Al-content of 50% exhibited an appreciable increase in dark current under continuous front illumination (λ = 250 nm) for 25–60 h during which the device was continuously operating. Oxidation of ambient moisture near the top metal fingers was found to be the reason for this. When passivated with SiN, this moisture-induced degradation was mitigated.
Not surprisingly, there is no such report until the date on the investigation of degradation in AlGaN Schottky or p–i–n detectors on sapphire under continuous UV illumination. Thus, we sought to check whether similar moisture-induced degradation was observable for Schottky detectors too. For this, we had grown a stack as shown in Fig. 5(a) using MOCVD. A variation in the crystalline quality of the AlN buffer grown using the two-temperature step growth method, and consequently, the absorbing AlGaN epi-layers were achieved by varying the inter-nucleus distance or the density of nuclei on the growth surface by changing the thickness of the low-temperature nucleation layer. Two samples were, thus, grown with different low-temperature AlN nucleation layer thicknesses to result in different threading dislocation densities. Sample A had no low-temperature (LT) AlN layer, while for sample B, the LT AlN layer was grown for 2 min. Both samples A and B had the same stack as shown in Fig. 5(a). The V/III ratio, growth temperature, and growth pressure for the AlGaN layers were 850 °C, 1100 °C, and 40 mbar, respectively, for the two samples. The Si-doping level in the n-AlGaN layer was 1 × 1019 cm−3. Schottky diodes were then fabricated using standard photolithography with Ni/Au as the top Schottky contact and Ti/Al/Ni/Au as the Ohmic contact after a dry etching of the top UID layer. The active area of the devices was 0.16 mm2.
From the FWHM values of rocking curve XRD scans along various off-axis crystal planes, the screw and edge-type dislocation densities for samples A were estimated to be around ∼109 cm−2 and mid-1010 cm−2, respectively, while the corresponding numbers were <106 cm−2 and <7 × 109 cm−2 for sample B. As expected, sample B exhibited dark current, which was several orders of magnitude lower than that of sample A. While the details of the investigation with respect to the correlation between electrical, optical, and structural properties of the samples would be published in a future study, here, we seek to highlight the degradation of AlGaN Schottky photodetectors with continuous UV exposure vis-à-vis the dislocation densities. Both the samples were exposed to continuous high intensity (10 mW/cm2) UV exposure from the front, and dark current, photo current, and responsivity values were measured after 10, 100, 1000, 2000, 5000, 8000, and 10 000 sec.
Sample A exhibited severe degradation in terms of both its dark current and zero-bias responsivity as is obvious from Fig. 5(b): the dark current increased by a few orders of magnitude from ∼10 μA to approximately a few mA while the already-low zero-bias responsivity of 0.6 mA/W plummeted down further with a continuous UV exposure and reached almost 0.1 mA/W after 10 000 s of the same. We choose to consider responsivity at zero bias and not a finite reverse bias so as to prevent any trap-mediated gain mechanism to kick in at higher fields that could mask the sub-optimal performance of devices on sample A. On the other hand, even after 10 000 sec of UV exposure, devices on sample B exhibited negligible change in dark current (<1 pA) and zero-bias responsivity (109–106 mA/W). Interestingly, under backside illumination, sample A exhibited less than 2% change in dark current and responsivity values even after 10 000 s of continuous UV stress (not shown here). Sample B exhibited negligible change even under backside illumination.
These experimental observations suggest a dislocation-dependent and Schottky-electrode-mediated degradation mechanism operating close to the UID AlGaN surface, analogous to what had been reported for AlGaN MSM detectors on sapphire by Brendel et al.173 Detailed studies on the moisture-induced changes in the AlGaN surface near the top electrode for these Schottky detectors would be published in a future work.
However, such moisture-induced degradation can be avoided by back-illumination, which is how usually flip-chip-bonded FPAs are operated, or by passivation with SiN, if it requires front illumination. The more important reliability tests include those that could qualify AlGaN detectors for space applications; however, there is barely any report on the same, for any kind of AlGaN detector configuration or FPAs. For instance, commercial InGaAs-on-InP photodetectors operating at 1060–1600 nm are subjected to proton irradiation of about 30–35 MeV with a total dosage of 15–50 krad, gamma irradiation of about 1 MeV, and mechanical shock and vibration tests corresponding to mil-std tests such as mil-std-883 Method 2007 and Method 2002. For tests involving mechanical shock and vibration, the detectors need to be attached to the driver, properly packaged and mounted, restrained firmly to the vibration or the shock platform, and the leads need to be adequately secured. After a detector qualifies all such tests, it still needs to be launched into space in the payload so that it stays in the space for an extended duration, say a year or two, to subject it to real space conditions. The detectors get exposed to outer space radiation such as proton radiation from the inner Van Allen belt and galactic cosmic rays besides being subjected to vibration and shock during the launch. Also, the detector module gets subjected to extreme temperature cycles (−157 to 121 °C), several times a day as it orbits the earth in the space station or a satellite. Its behavior and performance degradation, if any, would then be evaluated after it is brought back to the ground. InGaAs photodetectors185 have been found to exhibit no change in their performance such as dark current and responsivity after undergoing spaceflights for more than a year172 besides being subjected to lab-based radiation and mil-std tests. This testifies to the mature status of InGaAs detectors that are used for space-based imaging and in ultrasensitive gravitational wave detectors.186
While there has been no report of such tests on AlGaN detectors or FPAs either in the lab or in a space mission, there are a couple of reports studying climatic and ageing tests for FPAs based on Schottky Al0.40Ga0.60N on silicon.164 The silicon substrate was thinned down and etched in a honeycomb pattern to enable back-illumination. The detectors exhibited a slow decrease (∼10%) in responsivity when irradiated with a fluence of 1016 photon s−1 cm−2 at Lyman α. For climatic and ageing tests, the detectors were subjected to three cycles of 400 h at 60, 80, and 100 °C each while keeping them operational. Besides, the FPAs were also subjected to thermal cycles of −110 to 80 °C, which was not, however, as extreme as temperature cycles for space-bound InGaAs detectors. No change in dark current, transient response, and responsivity was found after these tests although the evolution of cracks in the layers was studied, indicating that the AlGaN Schottky detectors were robust and had good material qualities.
There is nonetheless a great need for performing such reliability tests for AlGaN MSM and p–i–n photodetectors also, including those that are realized on sapphire and on silicon substrates. More studies are certainly needed even for Schottky detectors with different Al-contents used in the active layer (for cut-off at different wavelengths). The performance change in the UV-to-visible rejection ratio, noise power density, and detectivity of AlGaN photodetectors after extreme thermal cycling and proton and gamma radiation need to be comprehensively investigated. Moreover, the operation of such devices at elevated and at cryogenic temperatures for extended periods of times need to be studied to understand how deep acceptor states, traps, and dislocation-mediated shunt paths affect leakage and photo-response over a wide range of temperatures. For mil-std tests, packaging and mounting are quintessential; however, there is barely any report of the study of packaging187 of solar-blind AlGaN detectors. Figure 6 shows the sequence of steps followed in the authors' lab for realizing a wire-bonded 1 × 5 linear array of MOCVD-grown AlGaN-on-silicon MSM detectors. These are quasi-vertical MSM detectors where the bottom contact is formed by etching 100–200 nm of the top AlGaN absorber layer. These detectors in a linear array were mounted on a printed circuit board (PCB) and wire-bonded. Some of the details of this study correlating the defects in AlGaN epi-layers including dislocations and the evolution of cracks with possible performance degradation in a linear array of MSM devices were reported earlier. Particularly, reliability tests for such arrays could be of immense importance to the community in terms of making this a viable technology. Thus, these are the areas of research in AlGaN UWBG deep-UV detectors that we believe to require sufficient thrust to establish a mature technology.
D. What is preventing market penetration of AlGaN UV detectors?
Schottky AlGaN solar-blind photodetectors are commercially available188 although MSM and p–i–n detectors are not (to the best of the authors' knowledge). Figure 7 shows an image of an AlGaN solar-blind Schottky photodetector available commercially from Genicom. It is housed in a TO 39 package and gives a peak responsivity (254 nm) of 0.06 A/W at zero bias. However, it is going to be quite difficult to find out to what extent (if any) such commercially available AlGaN deep-UV detectors are used in industrial, space-borne, strategic, or any other real-world applications. An important area where solar-blind photodetectors are of immense importance is UV astronomy as has been mentioned earlier. As discussed in Sec. II C, detailed reliability tests and space qualification of AlGaN-based photodetectors are yet to be reported. There is no public information on whether AlGaN-based detectors have been used by any space agency such as NASA, ESA, JAXA, etc., for deep-UV imaging of the sun or any other stellar source. However, there have been studies on such devices reported from NASA189,190 although it is not mentioned whether any AlGaN deep-UV detector has ever been flown in a spaceship or a satellite. Similarly, solar-blind detectors are useful for early warning in war such as in missile plume detection, but given its highly strategic nature, such information is almost impossible to obtain.
While it is abundantly clear that the field of AlGaN deep-UV detectors is still lacking in comprehensive reliability tests for qualifying the discrete devices and/or the FPAs, it is not necessarily the only reason why this technology has not yet entered the market. In the areas of compound semiconductor technologies, the material scientists or the epitaxy experts cannot work in isolation to grow hetero-epitaxial layers; they must work with the device engineers to understand the material requirements from device perspective. Similarly, the device experts need to work with the systems and circuit engineers to have the devices put in real-world systems. Thus, to bring a semiconductor device technology to fruition, material scientists, device physicists/engineers, and system experts must work in tandem. In addition, such endeavors are extremely resource-intensive, which is why a strong pull from the industry and/or the strategic sector is a must to fund and propel the growth of any device technology. In the context of GaN, the blue LED was disruptive, and so, there was an increasing pull from the LED industry that accelerated and sustained the development of blue and later green LEDs and laser diodes based on GaN since the late 90s. Similarly, GaN HEMT on SiC received enormous pull from the defence sector for applications in microwave power amplifiers that eventually led to a rapid expansion of this technology, including GaN-on-silicon HEMTs for power-switching applications. In addition, system engineers have worked with device and material scientists to ensure that GaN HEMTs can now be used in DC–DC converters and monolithic microwave integrated circuits (MMICs). Such collaboration between material/device engineers and system experts for AlGaN UV detectors is rarely observed or reported. In this case, the people in an imaging team who work with the front-end electronics and integration of imagers would constitute “systems experts,” and there is a real need for device engineers to work closely with the former to assess and propel the growth of UWBG detector technology.
Unfortunately, for AlGaN-based solar-blind detectors, we believe a strong pull from the industry and/or the strategic sector is missing for the most part. What could be a possible reason?
A general rule of the thumb is that it is extremely difficult or nearly improbable for compound semiconductors to compete with the incumbent silicon technology unless one of the following two conditions is met:
Either silicon cannot enable a particular functionality such as emitting light or detecting in mid-IR, and so, compound semiconductor technologies become indispensable (e.g., LEDs, laser diodes, and mid-IR detectors); or
The emerging technology offers vastly superior performance and added advantages (such as compactness, radiation-resistance, high temperature operation, etc.) for the similar cost* as silicon technology. For instance, GaAs- or InP-based HEMTs and HBTs are used in cell phones for RF transmission/receivers, or GaN HEMT on silicon is a strong contender for power switching at 600 V node although it is still quite difficult and unlikely to totally replace existing silicon MOS technology. (*“Cost” factor usually does not apply if it is for the defence/space sectors).
Condition (i) is certainly not met for AlGaN deep-UV detectors because silicon-based device technologies are being routinely used for almost all applications that require solar-blind imaging including UV astronomy, flame monitoring, or other industrial uses. AlGaN deep-UV detectors could fit into condition (ii); however, their advantages over silicon-based detector technologies probably have not outweighed the sheer advantages of the latter.
While it is true that radiation hardness and rejecting the visible and IR wavelengths require more work (added complexities) for silicon CMOS- and CCD-based sensors imaging in deep-UV, yet, their cost, maturity, scalability, and, most importantly, their ease of integration with the front and back-end electronics make these the dominant players in the market, including space-based imaging. Oftentimes, gain is a very important parameter for a detector to capture space-based images in deep-UV as the number and frequency of arrival of such short wavelength photons is extremely low. Thus, photo multiplier tubes (PMTs) or micro channel plates (MCPs) and/or their hybrids (MCP-PMTs) are used in conjunction with CCD- or CMOS-based sensors in space-based telescopes to provide large gains (∼106),191 while optical filters are used to reject visible and IR wavelengths. These certainly add to extra weight and so does the battery used for biasing the CCD/CMOS sensors and the cooling arrangements in some cases. Even then, their super high gain and excellent noise performance offered by PMT/MCP/MCP-PMT in conjunction with the ease of integration make silicon-based sensors the de facto choice for solar-blind imaging. To add to the woes of UWBG detectors, the photocathode in a PMT or an MCP may be made of a solar-blind material such as Cs–I or Cs–Te. In fact, Ce-decorated GaN may be used as photocathode192 whereby the role of GaN is not to utilize its bandgap but its low electron affinity when decorated with Cs that leads to the generation of electron-charge by a photo-electric effect.
For industrial and civilian uses, cost matters significantly, and for applications such as flame monitoring or solar UVB monitoring (in healthcare), silicon sensors thump over UWBG detectors any day, given their economy of scale and ease of availability.
Figure 8 shows a schematic of solar-blind imaging using CCD/CMOS sensors, while Fig. 9 shows the same for a possible AlGaN-based imaging setup. Three important factors working in favor of AlGaN-based detectors are as follows:
Their intrinsically solar-blind nature, requiring no optical filters, which is expected to cut down on weight, a critical parameter to factor in for space-borne imagers;
Their radiation-hardness; and
Their battery-free operation, especially for Schottky and p–i–n detectors, which exhibit impressive responsivity even at zero-bias. This further reduces the weight of the module.
However, advantage (iii) would not apply in the case of FPAs based on AlGaN APDs, which would require high bias to achieve high gain.
Of late, however, there has been report193 of solar-blind detection using silicon photodetectors without the need for optical filters to reject the visible. This is achieved by carefully tailoring the depletion width in an asymmetric junction and taking advantage of the varying absorption depths in silicon for short wavelength photons. This, however, requires external bias and so adds to the total weight. However, this has the potential of posing severe threat to UWBG photodetectors as far as intrinsic solar-blind nature is concerned although the latter can naturally reject visible light to a much better extent and that too without any applied bias.
In any case, the AlGaN photodetector community must exploit the advantages of this material system and further improve two key performance parameters, i.e., gain and UV-to-visible rejection ratio while operating at as low bias as possible. In that case, these devices could perhaps compete with CMOS/CCD sensors, at least for space-borne deep-UV imaging. Enabling FPAs based on AlGaN APDs (Al-content > 40%) with gain >106 and achieving more than seven orders of magnitude of visible rejection, for instance, would require significant improvement in the structural quality of the epi-stack such as further reducing the dislocation density and defects and ensuring compositional and thickness uniformity over large-area (6 in. or more) growths. An additional requirement would be to develop these on silicon as opposed to sapphire for reasons stated earlier including the possibility of imaging at sub-200 nm wavelengths.
If we draw an analogy with GaN HEMT technology, GaN microwave transistors on SiC could offer 10× higher output power than rival technologies based on GaAs or silicon LDMOS while being more efficient, more compact, and lighter. This is why these got rapid traction for military applications such as in radars without much competition from the rival technologies. However, GaN-on-silicon HEMTs for power switching (e.g., 600 V node) are certainly penetrating the booming power electronics market but with serious competition from silicon MOS and SiC FETs, and a clear winner cannot be easily predicted. This is because the advantages offered by GaN-on-silicon 600 V HEMTs—superior breakdown field, higher current densities, higher switching efficiency, and frequency at more compact sizes—are offset by their not-yet-mature reliability, relatively high cost, and accelerated ageing performance where silicon MOS leads by an undisputed edge despite a lower switching efficiency.
While there have been more studies and certainly more industry-pull for GaN HEMTs-on-silicon than for AlGaN-based solar-blind detectors, yet AlGaN UV sensors have the potential to be an important player, given the predicted multi-billion-dollar market size for UV sensors. However, for this, an assessment of the usefulness and advantages of AlGaN detectors for visible-blind UVB and UVA sensing also needs to be done vis-a-vis the cost and performance of their silicon-based counterparts. In any case, to transition the UWBG solar-blind detectors from the realm of academic research to being viable technologies in the market, the following points maybe considered as concluding thoughts:
Material quality requires further improvement in terms of lower defects, precise and uniform control of doping and Al-content across large wafers, especially for enabling FPAs based on AlGaN APDs operating at the solar-blind regime. And such APDs with ultrahigh gain need to be developed with a robust and uniform performance across the wafer.
Ultrahigh UV-to-visible rejection ratio needs to be demonstrated, either by device engineering or by disruptive improvements in the material quality to enable super sensitive or single photon detection in the space even without the use of PMT/MCT.
More comprehensive reliability tests for space qualification including std-mil tests need to be carried out for all kinds of AlGaN deep-UV detector topologies, including those on discrete detectors and FPAs on-wafer as well as on flip-chip-bonded modules with substrate thinning.
Finally, material scientists and device engineers need to work more closely with imaging teams and system developers to realize deep-UV imaging modules with AlGaN devices, including testing, performance evaluation, and long-term operationability, to understand and exploit the advantages these devices offer over silicon CMOS and CCDs.
III. PERSPECTIVE ON GALLIUM OXIDE UV DETECTORS
In the area of Ga2O3 deep-UV photodetectors, there has been an explosive growth in terms of journal publications and research articles as captured in Fig. 10. Many research groups have been working on all aspects of this material growth and device development, including the growth and/or the synthesis of Ga2O3 bulk crystals,194–196 thin films,197,198 and nanostructures,199 including nanowires,200 nanobelts201 and exfoliated flakes,202,203 and demonstrating solar-blind photodetectors on such platforms. In the category of thin films, there are several reports on deep-UV detectors based on polycrystalline204,205 and amorphous Ga2O3206,207 also. Since there are comprehensive review articles208–210 on the status of and challenges for Ga2O3 deep-UV photodetectors, in this section, we shall touch upon the overall status briefly and, instead, discuss more about the challenges ahead and the lessons learnt from their AlGaN counterparts.
The fact that there are five polymorphs of Ga2O3 (α, β, ɛ or κ, γ, δ) makes this a rapidly expanding area of research for both material scientists and device engineers. However, studies on γ- and δ-Ga2O3 are barely reported211,212 due to their unstable nature. β-Ga2O3 is by far the most widely and commonly studied polymorph as it is the most stable, and most of the photodetector studies are also based on this polymorph. It has a bandgap of 4.6–4.9 eV, corresponding to a cut-off in the range of 250–270 nm in its photo response. This makes it an attractive candidate for solar-blind UV detection, just like AlGaN. The major disadvantage compared to AlGaN, however, is that its bandgap is fixed as it is a binary compound; so, tailoring the cut-off wavelength by a reasonable margin is possible only if β-Ga2O3 is alloyed213 with Al2O3 and/or In2O3. The study of the growth and optical characterization of such alloys214–219 is an emerging but nascent area of research, which is why our discussion shall be primarily focused on Ga2O3. On the other hand, a major advantage of Ga2O3 over AlGaN is that single crystal, bulk substrates can be grown from the melt, at least for β-Ga2O3, which makes large-area manufacturability and scalability quite feasible besides potentially lowering the cost. The availability of bulk, single crystal substrates of β-Ga2O3 eliminates the laborious task of engineering and optimizing nucleation and transition layers during the growth to reduce dislocations and other defects like V-pits, making homoepitaxial β-Ga2O3 a superior quality material compared to say, AlGaN, which is usually grown on sapphire or silicon. This is immensely promising for enabling devices that are reliable, robust, and uniform as opposed to those that are realized on epi-layers with a high threading dislocation density. However, bulk β-Ga2O3 wafers may pose a challenge in back-illuminated, flip-chip-bonded detectors of FPAs, which will be discussed toward the end of this section.
As has been widely reported and discussed in the literature, the absence of p-type doping or rather the high improbability of achieving p-type conductivity is the Achilles' heel for Ga2O3. This is due to the very low formation energy of self-trapped holes (STHs), the large formation energy of the native acceptors, and the absence of dispersion of the energy levels near the valence band maxima.220 Both cation and anion substitutional dopants for possible p-doping in Ga2O3 are reported221 to exhibit acceptor transitions with high (>1 eV) energy levels with respect to the valence band, making it quite unlikely to achieve hole conductivity in this emerging UWBG material. Even at high doping concentration, hole conductivity is unlikely due to the high ionization energies of the dopants. It is undisputed a fact that reliable, controllable, and stable p-doping is quintessential to expanding the application space and usefulness of this emerging UWBG material. The realization of p–n junctions in Ga2O3, for instance, could lead to p–i–n photodiodes with superior performance, deep-UV light emitting diodes, and a host of power electronic devices such as Ga2O3 thyristors and insulated gate bipolar transistors (IGBTs), which can potentially create disruptive changes in the technology landscape today. In the context of deep-UV detectors, p-type Ga2O3 could enable APDs and vertical p–i–n type photodiodes with ultralow dark currents that are difficult to realize using Schottky diodes. On an optimistic note, more recently, there has been a demonstration222 of p-type conductivity in pulsed laser deposition (PLD)-grown Mg-doped β-Ga2O3−δ where the transfer characteristics of a top-gated field effect transistor (FET) are shown to provide evidence of hole transport in gallium oxide. Further work and investigation in this direction could be promising for bipolar opto-electronics.
A. State-of-the-art in β-Ga2O3 detector performance
For an emerging UWBG semiconductor such as Ga2O3, it is largely about understanding and, thus, improving the material growth and the structural quality of the crystal; therefore, the device performance is quite closely tied with the progress in demonstrating superior material quality. With improvements in the bulk crystal growth of β-Ga2O3, MSM and Schottky photodetectors with good figures of merit have been reported on crystals grown by float zone and edge-defined film fed (EDFF) methods.223–226 More recently, self-powered responsivity as high as 9.78 A/W227 for Schottky detectors on an EDFF-grown bulk crystal was demonstrated although no explanation was offered on the origin of such an ultrahigh gain at zero bias. These Schottky devices exhibited transient response in the range of a few μs while demonstrating dark current in the range of ∼10−8 A/cm2. Responsivity as high as 1000 A/W at 240 nm with a visible rejection ratio of 106 had also been demonstrated228 for Schottky diodes realized on the bulk β-Ga2O3 wafer grown by the FZ method, indicating that highly sensitive photodetectors with excellent visible rejection are achievable on bulk crystals of gallium oxide. Several groups have also realized both MSM and Schottky photodetectors on β-Ga2O3 grown homoepitaxially on β-Ga2O3 bulk substrates.229–231 These include Si-doped232 and Sn-doped233 β-Ga2O3 epi-layers grown by epitaxial growth techniques such as MBE, PECVD,234 MOCVD, and PLD. Most of these detectors exhibit good responsivity with dark currents in the range of nA or even lower. Zero-bias EQE exceeding 87% has also been achieved in homoepitaxial Si-doped β-Ga2O3 on bulk β-Ga2O3.235
On the other hand, hetero-epitaxially grown β-Ga2O3 on sapphire has been quite extensively investigated236–241 for MSM detectors, especially in terms of device responsivity, dark current, visible rejection, and transient response. Heteroepitaxy, despite its cons of dislocations and other defects arising out of lattice mismatch and strain, offers tremendous flexibility in band engineering and in growing epi-layers on different types of substrates. Hence, the epitaxial growth of β-Ga2O3 on substrates such as sapphire is being widely explored and reported. This has the potential to lower the cost further since sapphire substrates are almost as cheap as silicon. Typical growth approaches include molecular beam epitaxy (MBE), MOCVD, atomic layer deposition (ALD), pulsed laser deposition (PLD) or laser MBE (LMBE), and mist-CVD. For instance, MBE-grown β-Ga2O3 exhibiting a responsivity in excess of 200 A/W and dark current of a few tens of pA have been demonstrated.242 Similarly, PLD-grown β-Ga2O3 detectors with responsivity exceeding 18 A/W and dark current in the range of ∼10 pA have been reported.219 These suggest that despite challenges associated with epi-layers grown on sapphire, photodetectors with high gain and ultralow dark currents are achievable for β-Ga2O3.
However, slow transient response is a problem inherent to almost all types of photodetectors on gallium oxide, including those realized on bulk crystal and homo or hetero-epitaxially grown β-Ga2O3 epi-layers. Typically, Ga2O3 photodetectors exhibit high gain221 due to hole trapping and subsequent photo-induced barrier lowering. This leads to high responsivity values albeit at the cost of slower response (∼ms). Most of the studies on photodetectors realized on either bulk or hetero-epitaxially grown β-Ga2O3—including both MSM and Schottky type devices—report high responsivity in the range of a few A/W to a few hundred A/W in conjunction with transient responses that do not go below a few milliseconds, indicating that this gain-bandwidth trade-off is quite severe in the case of gallium oxide. Oxygen vacancies243 are attributed to be a major reason for causing hole traps, which is why the device performance has been shown to improve after an anneal in oxygen ambient.244 Nonetheless, for applications where fast responses are required, epitaxial as well as bulk β-Ga2O3 detectors have a long way to go in terms of improving their bandwidth. This will require materials’ engineering such as alloying, doping, tailoring the growth conditions, stress, and dislocation engineering (for hetero-epitaxial growths) to suppress oxygen vacancies, complexes, and other point defects that potentially slow the device by trapping holes. It must be noted, however, that since p-type doping of β-Ga2O3 is highly improbable, the transport of holes (such as those generated by optical illumination) in gallium oxide is itself a highly interesting topic of research. It is reported that such holes usually form polarons245,246 at lattice distortions such as vacancies and get localized, which prevents p-type conduction. Thus, polarons in β-Ga2O3 need more studies in conjunction with a careful material and device engineering to exploit superior transient responses in this material system by suppressing hole capture while not sacrificing much on the gain. On the other hand, in applications where speed is not of primary importance, such as in deep-UV imaging of astronomical sources, β-Ga2O3 detectors do hold promise, given that these have been shown to exhibit high responsivity values (>100 A/W), excellent visible rejection (>106), and ultralow dark currents (∼pA).
Since β-Ga2O3 is a binary compound with a constant bandgap, instead of plotting CFOM against the peak wavelength of photoresponse, we plot vs corresponding to some of the state-of-art MSM detectors as reported in the literature.65,69,134,145,193,221,247–259 This is shown in Fig. 11 where a few data points from AlGaN MSM photodetectors are also scatter plotted. The peak wavelength of photoresponse for AlGaN detectors varies between 250 and 290 nm for this plot. For any data point, the gain and the dark current values correspond to the same applied bias value; however, this bias may not be the same for different data points. For instance, in some, data points are taken at −5 V, while others are taken at −20 V. However, this variation in bias (−5 to −20 V) causes less than an order of magnitude change in dark current or gain. Hence, Fig. 11 does give a reasonably accurate feel for comparing MSM detectors in terms of CFOM to a first order. The top left corner in Fig. 11 is the desirable area for the detector performance as it represents devices with ultralow dark current, high gain, and faster responses. What is remarkable is that despite being relatively less mature than AlGaN, β-Ga2O3 detectors perform quite well when all the three key parameters (gain, speed, and dark current) are considered together. In fact, the highest value of corresponds to a β-Ga2O3 device, albeit at a relatively high dark current. However, at ultralow dark current levels, AlGaN still holds the upper edge in terms of values.
Two other attractive routes to exploiting superior device performance in Ga2O3-based photodetectors are the use of exfoliated flakes260 and nanostructures. A single crystal of Ga2O3 cleaves easily along the (100) and (010) planes; thus, using simple mechanical exfoliation, flakes of the same can be placed on different substrates, leading to ultra-thin layers of high-quality Ga2O3. Using such flakes, graphene-gated phototransistors have been demonstrated261 with a responsivity of >2000 A/W and even exceeding 105 A/W262 and a photo-to-dark current ratio exceeding 108 while extracting a specific detectivity of ∼1014 Jones. However, it is noteworthy that such graphene-gated FETs with ultrahigh responsivity and PDCR do not turn off under illumination even when the gate is biased below the pinch-off voltage; besides, the transient response is usually slow in such devices with ultrahigh gain, with decay times running into a few seconds. Similarly, nanostructures of Ga2O3 such as nanowires and their arrays (including core–shell nanowires), nanobelts, nanosheets, nanodots, nanorods, and their arrays have been explored for realizing deep-UV photodetectors with promising performance. Exhaustive reviews on the same are available.195,263 Some of these studies have reported responsivity values in excess of a few hundred A/W, indicating their highly sensitive nature to deep-UV radiation, with dark currents as low as <1 pA, suggestive of their potential excellent noise performance. However, typically, even for detectors based on exfoliated flakes and on nanostructures, the transient response values have been in the range of a few milliseconds to a few seconds; and for devices whose rise and fall times are in the range of μs, the responsivity values are usually quite low. This indicates that achieving a high bandwidth could be a challenge even for such non-epitaxial and exfoliated Ga2O3 detectors. Notwithstanding the fact that large-scale manufacturability, uniformity, and yield of such devices are quite challenging, nanostructures and exfoliated photodetectors are at an emerging stage of research and need further improvement in their material quality to lower the rise/fall times while not trading off on their responsivity values. These platforms also require scalability of the growth or synthesis processes to offer a viable economy of scale.
It is noteworthy that the UV-to-visible rejection ratio for most of the photodetectors is usually between 103 and 105, irrespective of their type (Schottky, MSM, or photoconductive), irrespective of the material (bulk, homoepitaxial, and hetero-epitaxially grown), and even for exfoliated and nanostructure-based Ga2O3. While there are a few reports with a higher visible rejection ratio going all the way up to 6 × 107, particularly using phototransistors,264 there needs to be further studies on ways to improve this critical device parameter, especially in conventional two-terminal photodetector configurations such as MSM and Schottky diodes. This will primarily involve suppressing sub-bandgap absorption by reducing defects such as complexes and vacancies, which give rise to defect levels within the bandgap.
Another attractive aspect of β-Ga2O3 photodetectors is their integration with a wide range of dissimilar materials to realize heterojunctions, both in planar/epitaxial geometry and using nanostructures or flakes. This opens the possibility of a nearly endless combination of β-Ga2O3 with other materials to exploit band alignments and energy band engineering for superior device performance. For instance, due to the absence of p-type doping of Ga2O3, several groups have demonstrated p–n and p–i–n junction photodetectors based on n-Ga2O3 and a p-type semiconductor such as p-silicon,265 p-GaN,266,267 p-NiO,268 p-CuO,269,270 and p-ZnO271 with impressive figures of merit. In addition to these, inorganic copper-based hole transport materials including Cu2O,272 CuI,273 CuSCN,274 CuMO2,275 and CuGaO2276 have been recently reported to form p–n heterojunctions with Ga2O3. Junction with the electrolyte and the α/β vertical junction nanorod array have also been reported for deep-UV detection.277,278 However, the trade-off between gain and transient response persists even for heterojunctions as well. For heterojunctions that exhibit rise times in the range of a few tens of μs, the responsivity is usually quite low such as in the case of monolayer graphene/cleaved Ga2O3 (100);279 and heterojunctions with high responsivity (>28 A/W at zero bias) usually exhibit rise times barely less than a second such as p-GaN/Ga2O3 junction.280
In terms of arrays, both linear and two-dimensional arrays of β-Ga2O3 photodetectors have been reported,208,281–283 including the study of pixel-to-pixel variation in responsivity and dark current values. Using an active pixel sensor consisting of a discrete Ga2O3 MSM photodetector monolithically integrated with CMOS amplifier, flame detection in the deep-UV has also been recently demonstrated.284 Although conventional FPAs have not been demonstrated using β-Ga2O3 photodetectors, basic imaging of simple patterns based on 4 × 4 MSM arrays has been reported.285 Similar pattern imaging using 32 × 32 arrays has also been shown for MSM detectors based on amorphous Ga2O3.260
From the preceding discussion, we may conclude that although significant work has been accomplished in the area of β-Ga2O3 detectors, yet, it is still at a research phase and has to evolve some more before it reaches a level of desired performance, which could make it suitable for real-world applications. β-Ga2O3 is relatively more recent compared to AlGaN although it has now been around for more than a decade. Given that single crystal wafers of β-Ga2O3 are available commercially both in semi-insulating and in n-doped configurations, it is high time now for the community to give it a directed thrust to elevate its deep-UV detector research into the next level. On the other hand, as of now, it would be premature to discuss about the promise of real-world applicability of heterojunction or exfoliated flake-based or nanostructure-based photodetectors realized using β-Ga2O3.
B. Photodetectors based on α- and ɛ-Ga2O3
Although much less explored compared to β-Ga2O3, there has been an increasing focus on both α- and ɛ-Ga2O3 in terms of their material growth and, hence subsequently, on deep-UV photodetectors realized on these platforms. α-Ga2O3 has a corundum structure analogous to that of sapphire and has the largest bandgap (5.5 eV) among all polymorphs of gallium oxide. This makes it an attractive candidate to detect shorter UV wavelengths as compared to β-Ga2O3 polymorph, and besides, a high-quality material can be realized on sapphire, which is quite cheap. On the other hand, ɛ (or κ)-Ga2O3 has an orthorhombic structure, and its growth has been demonstrated on WBG materials such as III-nitrides286 including AlN and GaN, SiC,287 MgO,288 and sapphire,289 making it an attractive candidate for designing of deep-UV opto-electronic devices via heteroepitaxial integration.
The growth of α-Ga2O3 and of ɛ-Ga2O3 has been reported using techniques such as MOCVD, mist CVD, ALD, and HVPE with corresponding photodetector studies. These phases are thermodynamically semi-stable and, thus, present material as well as device challenges in terms of realizing high-performance solar-blind detectors. As of now, the materials’ development aspect of these phases in terms of understanding and improving their structural quality, controlled doping, and hetero-epitaxy needs to be more actively pursued, which would then enable the development of photodetectors with various configurations. However, these phases do hold tremendous promise even beyond conventional deep-UV photodetectors such as in easy integration with sapphire or with other WBG substrates, exploring ferroelectricity and polarization for novel, multi-functional devices.
For the growth of α-Ga2O3, mist-CVD290 has proven to be the technique of choice as there have been several reports covering different aspects of growth. For instance, n-type doping of α-Ga2O3 using Sn and Si (separately) as substitutional dopants has been reported291,292 using mist-CVD besides demonstrating Schottky diodes and FETs based on the same.293–295 In fact, deep traps in mist-CVD grown α-Ga2O3 has also been studied using the photo-capacitance method.296 Mist-CVD growth of α-Ga2O3 has been achieved on a-, c-, m-, and r-planes of sapphire;297,298 extensive structural studies such as LEO,299 dislocation generation and annihilation,300 as well as misfit relaxation301 have also been investigated. Optical properties such as excitonic binding energy for such epi-layers are also reported.302 Given such an increasing focus on the mist-CVD growth of α-Ga2O3, it is expected that there will be more studies on deep-UV detectors and their arrays based on Schottky diodes and MSM detectors of the same. This could help understand the correlation between dislocation density with the dark current and gain and establish a robust platform for achieving high peak responsivity in sub-240 nm wavelength regime with excellent transient response. Due to the nascent stage of solar-blind photodetectors based on α-Ga2O3, the study of heterojunctions of this material with other semiconductors such as III-nitrides, layered 2D materials, and metal-oxides for dual-band, multi-spectral and broadband UV sensing presents a fertile area of research.
The growth of ɛ-Ga2O3, on the other hand, has been reported using HVPE and MOCVD265 on sapphire and on a few other substrates, with subsequent demonstration of solar-blind photodetectors. However, even for ɛ-Ga2O3, mist-CVD is increasingly becoming a more widely used technique toward developing and understanding the growth, given its low cost and simple setup. Not only has ɛ-Ga2O3 been grown using mist-CVD on different substrates such as MgO, yttria-stabilized zirconia,264 AlN,303 NiO/sapphire,304 and SnO2/sapphire,305 but the structural aspects of the same have also been studied including the stoichiometry, rotational domains, and the microstructure. The stabilization of phase pure ɛ-Ga2O3 is, however, non-trivial, and there needs to be a careful and detailed understanding of the role of carrier gas flow rate, for instance, in tailoring the phase content (α or ɛ) of Ga2O3.
Among other factors, the growth chemistry has a pivotal role in controlling the metastable phase of the gallium sesquioxide. In one of our ongoing studies, we observe that the presence of Cl− facilitates the ɛ-phase over the α-phase. Figure 12 shows the custom-built mist-CVD setup at the authors' lab using which growths were performed. Gallium acetyl-acetate (0.02M) was used as the Ga-source, while the growth temperature was fixed at 450 °C with 1 atm pressure. Figure 13(a) shows the θ–2θ XRD scan of pure α-Ga2O3 phase deposited using CH3COOH as the dissolution agent, and Fig. 13(b) shows the corresponding scan for mixed phase of ɛ-Ga2O3 and α-Ga2O3 grown with a dissolution agent such as HCl. The presence of stable Al–Cl bonds at the c-plane sapphire may change the surface energy of sapphire and the interfacial energy between the oxides, thereby promoting growth of metastable ɛ-Ga2O3. Similar observations are also reported in previous studies.306 There are many factors that combinedly dictate the growth mechanism of these phases as already covered in recent review by Bossi et al.307 Figure 14 shows a picture of a 1 × 1 cm2 ɛ-Ga2O3 sample grown on sapphire using mist-CVD in the authors' lab (details to be published later).
In this context, we must mention about the ambiguity in the nomenclature of κ- vs ɛ-phase of gallium oxide. The crystal structure of ɛ-Ga2O3 was reported to be hexagonal with P63mc space group symmetry.308 The nomenclature for this phase was debated in past years for reasons pertaining to the crystal symmetry in which it crystallizes. Later it was characterized using TEM measurements and found to be composed of orthorhombic (Pna21) domains at 120° with each other giving it a “pseudo hexagonal” symmetry at the macroscopic scale.309 It has now been unanimously accepted as orthorhombic κ phase.
Hitherto vast majority of Ga2O3 literature report this phase as ɛ primarily due to the legacy reasons; in this article, we also adhere to the same nomenclature of the orthorhombic phase.
Nonetheless, ɛ-Ga2O3 is at an even more embryonic stage of research than α-Ga2O3; even substitutional n-type doping, electron mobility and conductivity measurements, nature of deep traps, and the generation of dislocations are some of the fundamental aspects of this material, which are yet to be studied. Though there are a few studies on MSM photodetectors based on ɛ-Ga2O3, there is no report yet on the demonstration of deep-UV detectors based on Schottky diodes of the same for which controlled n-type is a pre-requisite. This suggests that there exist abundant opportunities for the gallium oxide device and materials’ community to explore these emerging polymorphs in terms of understanding and improving the performance of deep-UV detectors.
C. Lessons from AlGaN and the road ahead for Ga2O3 photodetectors
In Sec. III B, we had made a critical assessment of the development of AlGaN-based deep-UV detectors and what it would require to likely push these devices into real-world applications such as in space-borne telescopes for UV astronomy. Hetero-epitaxy of AlGaN has shown outstanding progress leading to excellent detector performance, and yet, there are milestones to reach and challenges to address. These lessons learnt from the AlGaN deep-UV detector development effort can be utilized for their Ga2O3 counterparts too, so that research into the fundamentals of material growth and device engineering can be complemented by a proportionate effort toward understanding and overcoming bottlenecks, which otherwise prevent UWBG solar-blind detectors from entering the market.
In this context, we believe the following points are worth considering, especially for β-Ga2O3 photodetectors:
A concerted effort is needed between materials and device engineers to address the universal problem of slow transient in epitaxial β-Ga2O3 detectors. Understanding the formation of polarons and hole trapping is critical in this direction so that transients in the range of μs to ns are routinely achievable without compromising too much on the gain. In general, there should be comprehensive studies on the correlation between the structural quality of the epi-layers such as point defects, dislocations, and other material aspects, vis-a-vis the performance parameters of the photodetectors such as responsivity, dark current, and transient speed. Such studies have been reported for AlGaN detectors, for instance,178 and it is high time now for studies to be conducted on Ga2O3 detectors also.
There have been a few reports of Ga2O3 detectors with fast response time (∼μs), especially for nanostructures and flake-based devices; however, there is no report so far on the measurement of the 3 dB bandwidth of such detectors. Similarly, noise measurements on Ga2O3 detectors are rare to be found except for a few reports.260,261,310,311 We believe that the noise performance of both MSM and Schottky detectors needs to be investigated more actively, and the correlation of NEP with the growth conditions and material quality needs to be assessed toward making gallium oxide a mature and well-rounded detector technology.
How robust and reliable the detectors are is a very crucial aspect to consider while discussing the future roadmap of UWBG devices. Competing with the dominant silicon CMOS/CCD technology for deep-UV sensing is extremely challenging; thus, reliability and qualification of Ga2O3 detectors are of paramount importance. In this context, extreme temperature performance, radiation resistance, and the effect of continuous UV stress need to be investigated more extensively for Ga2O3 detectors. This will help understand the failure mechanisms and safe operating area for these devices toward assessing their reliability. Given that this is an emerging area of research, there are very few reports on the effect of radiation,312,313 thermal cycling,314 and on the high temperature operation315 of β-Ga2O3 detectors, and none on the UV stress induced degradation studies of the same. In addition to these, the effect of proton316 and gamma317,318 radiation also needs to be studied for both homoepitaxial and heteroepitaxial β-Ga2O3 detectors.
Arrays of MSM detectors based on β-Ga2O3 have been demonstrated toward imaging simple patterns. The next natural progression could, therefore, be to develop FPAs, to flip-chip bond with RoIC, and to demonstrate deep-UV imaging via back illumination. For this, integrated arrays of Schottky detectors need to be more extensively studied, especially in terms of their dark current (shunt paths), gain and transient response, and the improvements in their performance with field plates and passivation, all of which present fertile areas of research for the device community. Though Schottky detectors are being reported for bulk and homoepitaxial β-Ga2O3 films, yet the vast majority of photodetector studies on hetero-epitaxial films are of MSM geometry, given their ease of growth and fabrication.
In the context of back-illumination for flip-chip-bonded β-Ga2O3 FPAs, absorption by the substrate will become a serious bottleneck for both homoepitaxial and heteroepitaxial layers. While the Ga2O3 substrate will absorb the radiation at band edge, sapphire will absorb deep-UV with wavelength shorter than 190–200 nm, thereby preventing the detection at vacuum-UV. One approach could be substrate thinning, but it would require elaborate studies to understand the mechanical stability of the membrane and cracks that may propagate. For Ga2O3 substrates, there could be selective thinning or etching of the substrate using a honeycomb pattern though such approaches would lead to severe reduction in the EQE. Putting the Schottky contact at the mesa etched undoped or mildly n-doped Ga2O3 and the Ohmic contact on the top, i.e., on the n-doped Ga2O3, would improve the device performance under back-illumination. For sapphire substrates, the removal or thinning of the substrate would be much more difficult. Another interesting but challenging approach would be to grow Ga2O3 on silicon and then to completely remove the silicon and transfer the Ga2O3 membrane onto an RoIC or other front-end electronics without flip-chip bonding for front illumination, i.e., adopt heterogeneous integration. Or the silicon substrate may be selectively etched or thinned in a honeycomb pattern as well to enable back-illumination when bonded to an RoIC. Either way, the growth of high-quality Ga2O3 on silicon is going to be quite challenging although there have been a few reports on the same.319,320
The absence of p-doping makes demonstration of all-Ga2O3 APDs a remote possibility although there have been some reports on APDs based on Ga2O3-based heterojunctions and nanostructures. APDs in Ga2O3 are realized usually with other semiconductors such as SnO2 (Ref. 321) or in a core–shell microwire322 configuration. The other semiconductor(s) with which such heterojunctions are made usually do not have the same bandgap as or higher bandgap than Ga2O3, which is what required for efficient carrier absorption and multiplication processes. One interesting idea in this direction could be to explore heterojunctions between Ga2O3 and III-nitrides. While photodetectors based on heterostructures of GaN and Ga2O3 are demonstrated,323,324 what would be interesting is a heterojunction with AlGaN with Al-content >60%, which has a bandgap wider than that of Ga2O3.
Finally, there must be a synergistic effort between the material/device scientists and systems’ development experts to help test, integrate, and validate these detectors in real-world applications. β-Ga2O3 detectors must exhibit super high visible rejection ratio while providing impressive gain and ultralow dark currents to eliminate the need of PMT/MCP and to compete with incumbent CCD/CMOS sensors for deep-UV imaging.
While the foregoing pointers have been in the context of β-Ga2O3, it must be reiterated that detectors based on the other polymorphs of gallium oxide, primarily ɛ and α phases, are at an embryonic stage of research; thus, further efforts are required to establish a mature growth platform for these phases in terms of understanding the growth kinetics, the evolution of defects, hetero-epitaxy on different substrates, controlled doping, and so on. Subsequently, understanding and improving unit processes for device development such as reducing the contact resistance, minimizing the leakage, optimizing passivation, controlled anisotropic etching, etc., need to be undertaken before any discussion on possible real-world application of detectors based on ɛ and α phases of gallium oxide becomes pertinent. A major milestone to achieve for these phases of Ga2O3 is the bulk crystal growth of large-area wafers, for, without that, the economy of scale will never favor market adoption.
Solar-blind deep-UV photodetectors based on UWBG semiconductors such as AlGaN and Ga2O3 do hold tremendous promise, especially in terms of reducing the system complexity and weight while improving the EQE and robustness. Although the AlGaN system is more mature and better studied than the Ga2O3 platform, yet the former still requires further improvements in terms of large-area epitaxy (6 in. or more), reduced dislocation density, and the development of APDs with ultrahigh gain and visible rejection to make it a serious contender in the area of space-borne deep-UV imaging. Ga2O3, on the other hand, despite its fast progress, still has a long way to go in terms of improving the basic device properties such as rise/fall times, which would necessitate more extensive studies on point defects vis-à-vis the growth conditions and the substrate. Besides, the cost of a single crystal bulk Ga2O3 wafer is still quite high.
We believe any shortcoming in device performance and material quality cannot be the primary reason or at best cannot be the only reason, why these (especially AlGaN deep-UV detectors) are not adopted in real-world applications yet. We believe the excellent performance from low-cost and highly mature CMOS/CCD platforms with high integrability is the reason UWBG devices are not able to enter the fort yet. With MCP-PMT and solar-blind photocathodes such as Ce–Te, CMOS and CCD platforms achieve ultrahigh gain and the ability to image in vacuum UV (sub-200 nm) with solar-blind capability. To compete against this incumbent imaging technology for space-borne applications, UWBG photodetectors need to outperform the combined benefits of MCP-PMT and Cs–Te photocathode, i.e., provide ultrahigh gain (∼106) with ultrahigh visible rejection (>107). Only then, UWBG detectors would perhaps poise themselves as a truly attractive alternative to CCD/PMT-based imaging platforms. On the other hand, the ability of image in vacuum-UV is in fact a critical aspect for space-borne telescopes because a lot of astrophysical activities can be detected at sub-200 nm wavelengths. In addition, this presents a gap in the UWBG photodetector literature because there are very few studies that report responsivity, EQE, bandwidth, and other aspects of a photodetector for sub-200 nm radiation. For instance, in most of the UWBG detector studies, the photoresponse starts to fall off, albeit gradually, at wavelengths shorter than the band edge. This is most likely due to the high surface recombination velocity arising out of surface states, given that most of the shorter wavelength photons would be absorbed near the surface. Thus, surface passivation is a critical component of device development toward enabling UWBG detectors to image in vacuum UV. Also, it is not trivial to measure photoresponse in vacuum as most academic institutes do not have a QE tool housed inside of a vacuum chamber.
Apart from space-borne, deep-UV imaging of astrophysical processes, another important application for solar-blind detectors is detecting the UV emission from the plume of a missile or a rocket, and this is a highly strategic area. Thus, it is quite difficult to find reports and studies325 in the public domain which discuss about solar-blind imaging of missile plumes. Given that both the target aircraft and the missile would move in any direction, with any speed, at a widely varying distance, and over a wide range of altitudes, detecting the plume faithfully (i.e., ensuring the UV signature is from a missile and not something else) and with a fast response time to enable locating the time-varying co-ordinates of the missile is an extremely challenging task. This must be done fast enough to activate the defensive mechanism of the target aircraft such as decoys or flares. Also, the strength of the UV signature that arrives at the target aircraft would be extremely weak, necessitating a high gain in the detectors. UV FPAs need to be on-board the aircraft and must provide a wide field of view, sensitivity, and speed. Usually, a global positioning of UV detectors at six locations on the aircraft would be needed to provide an overlapping field of view to ensure coverage across 360°. Thus, UWBG photodetectors for this application would require to be fast (∼μs of rise/fall times) in addition to exhibiting high gain and low dark current.
In the area of industrial applications, deep-UV imaging is particularly useful for detecting alcohol and hydrogen flames, which are “blue” and/or invisible to the human eye. There is an entity called Ofil Ltd.,326 which, for instance, claims to use a proprietary optical filter technique to make CCD-based pure solar-blind deep-UV detectors. Their proprietary DayCor@II camera that employs this is claimed to be used in about 120 utilities (as per their website). Unlike UV astronomy or missile plume detection, applications such as flame monitoring or combustion detection in industrial settings do not need ultrafast detectors with ultrahigh gain. We believe this is an area where UWBG photodetectors, with their advantage of being lightweight without optical filters, can start off immediately. In fact, given that UWBG detectors have been shown to exhibit high EQE even at zero-bias,74 battery-free flame monitors based on AlGaN or Ga2O3 can be mounted on industrial settings.
As a concluding snapshot, Fig. 15 provides a qualitative representation of the status of deep-UV photodetectors based on AlGaN and Ga2O3 of β, α, and ɛ (or κ) phases. The approximate correspondences to various TRLs (technology readiness levels) are also mapped against the status. It is time for the materials/device community now to go beyond reporting just responsivity or the variation of photocurrent with incident power and to actually realize FPAs that are then tested for imaging. In parallel, there needs to be a continual effort toward improving material properties in conjunction with smart device engineering to push the performance envelope of UWBG detectors. Only then, we believe that UWBG solar-blind photodetectors stand a chance to compete with the incumbent silicon-based detector technologies.
See the supplementary material for the values of the parameters (such as rise time, responsivity, dark current, etc.) taken from the literature and commercially available devices to compute and plot Figs. 3, 4, and 11 that are provided in different sheets in an MS Excel file.
This work was funded in part by SCL/ISRO and by U.S. AOARD (Program Manager: Ali Sayir). We also acknowledge funding support from MHRD through the NIEIN Project, from MeitY and DST Nano Mission through NNetRA for supporting the facilities at CeNSE. We thank Dr. Matteo Bosi from IMEM CNR, Parma, Italy for fruitful discussions on kappa vs epsilon phase of gallium oxide.
Conflict of Interest
The authors have no conflicts to disclose.
The data that support the findings of this study are available from the corresponding author upon reasonable request.