Ultrawide-bandgap (UWBG) semiconductors, with bandgap energies much greater than the 3.4 eV of GaN or 3.2 eV of SiC, represent an emerging new area of intensive research covering a wide spectrum of materials, physics, devices, and applications [1]. As the critical electric field of avalanche breakdown increases super-linearly with increasing bandgap energy (detailed analysis discussed in a recent study led by Sandia and MIT Lincoln Lab [2]), UWBG semiconductors can tolerate high fields to enable high-power electronic devices for telecommunications, motor drives, power grid, electric vehicles, industrial and locomotive traction control, and various other applications. In addition, light emission from UWBG materials occurs in the deep ultraviolet (UV) part of the electromagnetic spectrum, which is attractive for extending the working wavelengths of photonic devices beyond the UV–visible (UV–vis) spectrum to enable potential applications in deep-UV optoelectronics, quantum information science, and bio-chemical sensing. This new class of semiconductors is also being explored for device applications in harsh environments by taking advantage of their thermal stability and radiation hardness. Compared to the development of GaN and SiC, all UWBG materials are relatively immature and still at a nascent stage. Most research efforts in UWBG focus on aluminum gallium nitride alloys (AlxGa1–xN), boron nitride (BN), diamond, and a large family of binary (typified by β-phase gallium oxide (β-Ga2O3)) and ternary oxide semiconductors. The extensive research activities on these materials are motivated by their reliable dopability and high carrier mobilities, the availability of substrates for thin-film growth, and successful demonstrations of devices. In this article, we provide an overview of the aforementioned different types of UWBG semiconductors, whose recent developments and state of the art in materials and applications are featured in this focus issue.

AlGaN

AlxGa1–xN is an alloyed UWBG semiconductor that typically possesses a hexagonal wurtzite structure. Its bandgap can be tuned from 3.4 eV (GaN at x = 0) to over 6 eV (AlN at x = 1) by varying the aluminum composition x in the material [1]. This tunability permits the ready formation of heterostructures [3], which allows great flexibility in the types of electronic devices that can be realized. Additionally, because it is a direct-gap semiconductor, it is suitable for the fabrication of UV emitters. Visible- and solar-blind photodetectors have also been demonstrated [4]. AlxGa1–xN can be doped n-type by incorporating Si, which is a shallow impurity up to a composition x of 80–85% [5]. Mg is used as a p-type dopant, although it is somewhat deep in the gap for GaN (~ 160 meV) and becomes deeper as the Al composition x is increased [6]. In addition to impurity doping, a great advantage of AlxGa1–xN is the presence of spontaneous and piezoelectric polarization [7], which aids in doping of the material through the formation of two-dimensional electron gases (2DEGs) as well as three-dimensional electron slabs [8]. Additionally, this approach of polarization-induced doping overcomes the large thermal activation energy of Mg and enables efficient p-type doping by field-ionizing the acceptors [9]. Growth of AlxGa1–xN is typically done by metal–organic chemical vapor deposition (MOCVD), but other growth techniques have also been employed, such as plasma-assisted molecular beam epitaxy (MBE) [10]. AlxGa1–xN may also be grown on a variety of substrates including sapphire, although growth on AlN substrates is becoming more common and has been instrumental for electronic devices [11] as well as UV optoelectronics. Growth on GaN [12] and even Si [13] substrates is also possible. While AlxGa1–xN is typically grown in the polar c-orientation, growth of semi-polar AlxGa1–xN has also been reported [14]. The physics of deep levels in AlxGa1–xN is likewise a very rich topic, with sensitivity to growth conditions and Al composition documented [15]. Deep levels are critical not only because they can compensate intentional impurity dopants [16], but also because they can impact electronic device performance due to carrier trapping and de-trapping and optical device performance due to absorption and non-radiative recombination. AlxGa1–xN may also potentially be used in photonic integrated circuits with applications in positioning, navigation, and timing [17].

Numerous electronic and optoelectronic devices have been demonstrated by groups around the world. The breakdown electric field is expected to increase approximately quadratically with the AlxGa1–xN bandgap [18], and this has enabled power switching devices based both on lateral-transport AlxGa1–xN/AlyGa1–yN heterostructures [19] and on vertical-transport devices with thick drift layers such as pn diodes [20] and Schottky diodes [21]. Moreover, the high breakdown electric field coupled with the favorable carrier velocity saturation properties of AlxGa1–xN [22] has permitted the demonstration of radio-frequency devices with encouraging performance [23]. One challenge with electronic devices is alloy scattering, which impacts not only the low-field mobility [24] but also the thermal conductivity. An additional major challenge is the formation of ohmic contacts, and various approaches not only to achieve linear contacts but also to reduce the specific contact resistivity have been demonstrated, including compositional grading [25]. Similarly, the integration of dielectrics on AlxGa1–xN is a topic of current research, which is especially challenging due to the large bandgap of the semiconductor and the resulting small conduction-band offset with the dielectric [26]. Regarding optoelectronics, the ultrawide direct bandgap of AlxGa1–xN has permitted the realization of UV emitters including light-emitting diodes [27] and lasers [28], with applications such as water purification in mind, and such devices are now commercially available. Several representative figures from a review paper on AlxGa1–xN-based UV optoelectronics by Kirste et al. [29] are reproduced below and are intended to convey key aspects of AlxGa1–xN, notably the high quality of epitaxy in Fig. 1 and exemplary device performance in Fig. 2.

Figure 1
figure 1

copyright 2021.

(Left) High-resolution transmission electron microscopy image of Al0.8Ga0.2N grown on AlN. (Right) 5 × 5 µm2 atomic force microscopy image of Al0.7Ga0.3N grown on AlN with root-mean-square roughness of less than 50 pm. Reprinted by permission from [29],

Full size image
Figure 2
figure 2

(Main Figure) Voltage (blue curve) and optical output power (red curve) plotted vs. absolute current (bottom axis) and current density (top axis) for AlxGa1−xN-based UV edge-emitting diode. (Inset) Electroluminescence spectrum showing emission at around 271 nm. Reproduced, with permission, from [28]. Copyright (2019), the Japan Society of Applied Physics.

Full size image

BN

BN is a compound isoelectronic with carbon. Like carbon, BN can possess sp2- and sp3-bonded phases [30], which are the analogs of graphite and diamond, respectively. The thermodynamically stable phase under standard temperature and pressure is sp2-bonded hexagonal BN (h-BN), whose wide bandgap of ~ 6 eV and ability to form single layers make it especially attractive for two-dimensional (opto)electronics or as an interlayer for heteroepitaxy [31,32,33,34,35]. The physics of high-temperature chemical vapor deposition (CVD) of h-BN using carbon-free precursors has been studied by Bansal et al. [36].

In its tetrahedrally coordinated sp3-bonded structure, BN can occur in multiple polymorphs. The wurtzite structure (w-BN) can potentially be alloyed with III-nitrides and grown lattice-matched on GaN template substrates to achieve wider bandgaps, which are useful for charge confinement in high-power electronic devices or quantum barriers in UV–vis optoelectronic devices [37, 38], However, BN is very dissimilar to the III-nitrides in terms of lattice parameters and stable crystalline phase; therefore, random alloys of B-III-N with more than a few percent of boron are difficult to attain [39,40,41,42]. In their review, Sarker and Mazumder summarize the developments of B-III-N alloys and provide insights into the microstructures of B-III-N films as demonstrated by local microstructural and atomic-scale chemical analyses [43].

Another sp3-bonded polymorph of BN is zinc-blende cubic (c-BN). This structure has a very large bandgap energy (~ 6.4 eV) [44], boasts an extremely high thermal conductivity second only to diamond [45], possesses arguably the highest breakdown field in the UWBG family, and can be doped either n-type or p-type [46,47,48,49,50,51], thereby making it a strong competitor for future high-power, high-frequency, and high-temperature electronics. To date, however, it has remained difficult to achieve device-quality c-BN using conventional MBE or MOCVD techniques since the cubic structure is metastable under ambient conditions. In general, it is believed that the nucleation of c-BN requires energetic bombardment of the growing surface with charged or neutral ions, regardless of the synthesis approach [52]. A multitude of factors, including degree of ion bombardment, strain, impurity concentration, and growth temperature, is important in determining the quality of the c-BN films [53,54,55,56,57].

Diamond

Recent progress on large wafers and device processing technologies has propelled diamond, which has a bandgap of 5.5 eV, onto the stage of high-power and high-frequency electronics [58]. The figures of merit of diamond devices are extremely high because of high carrier mobility (4500 cm2/V s for electrons, 3800 cm2/V s for holes) [59], large breakdown field (> 10 MV/cm), and high thermal conductivity (2200 W/m K) [60].

CVD is widely adopted for the growth of both substrates and epitaxial layers of diamond. Two promising techniques for realizing large wafers at low production cost are the direct wafer method (lift-off) and hetereoepitaxial growth. The former technique involves implanting carbon ions into the subsurface region of a diamond seed crystal to create a defective layer, followed by homoepitaxial growth on the seed crystal by CVD that simultaneously turns the defective layer into graphite. The grown film, a so-called clone-plate, is lifted off by etching the graphitized layer electrochemically [61]. Heteroepitaxial growth of diamond can be performed on Si, 3C-SiC, Pt, and Ir with lattice mismatches of 52%, 22%, 10%, and 7.6%, respectively [58]. Thanks to the small misfit presented by Ir, freestanding diamond wafers larger than 90 mm in diameter [62] and wafers with excellent structural quality [63] have been realized.

High-power capability of diamond devices was first confirmed on Schottky barrier diodes (SBDs) with boron-doped p-type layers [64]. The breakdown field of diamond SBDs reaches > 7 MV/cm without edge termination [65]. Diamond SBDs also show low leakage current and short turn-off transient with small reverse recovery time and charges even at elevated temperatures [58]. However, their specific on-resistance is high at room temperature and decreases only when the ambient/junction temperature exceeds 200 °C to increase the activation of the relatively deep boron acceptors (ionization energy ~ 370 meV [66]). To solve this problem, a new structure known as the Schottky pn-junction diode (SPND) has been proposed [67]. By precisely controlling the donor concentration in the SPND, holes injected from the impurity band of a p+ anode drift across a depleted n-type layer with high velocity even under forward-bias conditions to realize extremely low on-resistance with high forward current density > 20 kA/cm2 at room temperature. The SPND is expected to be useful for high-power radio-frequency applications because of its low capacitance, specific on-resistance, and forward voltage drop [68].

For switching devices, high-voltage or high-breakdown-field operation has been reported for metal–semiconductor field-effect transistors (MESFETs) [69], junction FETs (JFETs) [70], and deep-depletion diamond metal–oxide–semiconductor FETs (D3MOSFETs) [71]. Those FET devices use a bulk boron-doped p-type layer as the conducting channel, which leads to low current density at room temperature and increasing current density at elevated temperatures, similarly to the SBDs. Current controllability at room temperature can be dramatically enhanced using a two-dimensional hole gas (2DHG) channel [72,73,74], formed using surface-transfer doping when the diamond surface is terminated by hydrogen [75]. Hydrogen termination lowers the ionization energy of diamond, driving electron transfer from the valence band at the diamond surface into an acceptor layer consisting of molecular adsorbates or a transition metal oxide such as MoO3 [76]. This gives rise to p-type surface conductivity, with holes confined to a thin subsurface layer useful for reducing short-channel effects in high-frequency devices [73]. Output current densities of 2DHG channels reach 1.3 A/mm and 12 kA/cm2 for lateral [77] and vertical [78] configurations, respectively. Inversion p-type channel has also been confirmed on MOSFETs fabricated on phosphorous-doped n-type diamond [79], which realizes normally off operation and opens the door for practical applications. Reduction of interface-state density, which can be accomplished by using a 2D material free of dangling bonds as the gate insulator, is key to the improvement of channel mobility and current capability of the devices [80], as shown in Fig. 3. Recently, a hydrogen-terminated diamond FET with high channel mobility of 680 cm2/V s using a h-BN/diamond heterostructure was reported [81]. It is predicted that a strain-mediated rippled structure developed in the h-BN layer can enhance charge transfer across the h-BN/diamond interface [82].

Figure 3
figure 3

Correlation between field-effect mobility and interface-state density for inversion channel diamond and 4H-SiC MOSFETs. Adapted from [80], with the permission of AIP Publishing.

Full size image

Oxides

Oxides present a fascinating range of tunable physical properties, including conductivities ranging from insulating through semiconducting to superconducting, magnetism, and piezo-/ferro-/antiferro-electricity. This versatility makes oxides a materials class with high potential for new generations of electronic devices. Applications in power electronics and solar-blind UV detection can particularly benefit from UWBG semiconducting oxides. Compared to the widely explored III-nitrides, these oxides are still in their infancy, except for β-Ga2O3 which has been intensely studied over the last decade.

Like the quest to engineer GaN, harnessing oxides for (opto)electronic devices requires high-quality growth of bulk crystals and epitaxial layers with well-defined doping. A broad understanding of device-relevant physical properties is also essential. Bulk oxide crystals provide not only workhorse materials for the extraction of material properties, such as the mobility of electrons as discussed by Galazka et al. [83], but also substrates for homoepitaxy of thin films to achieve the highest structural quality. A versatile tool proven to pioneer novel oxide thin-film systems with highest quality is MBE, as discussed in a review by Nunn et al. [84].

In the following, we briefly review a selection of promising UWBG oxide materials ranging from binary oxides to ternary spinel and complex oxides, whose key materials properties are summarized in Table 1. A common feature of most UWBG oxides is that they can only be doped n-type with maximum Hall electron mobilities on the order of 100 to 300 cm2/V s [83], which limits their application space to unipolar devices unless hetero-pn-junctions are used. Often, the bandgaps of oxides were measured by the optical absorption of thin films. As a consequence, only direct transitions with large absorption coefficient or dipole momentum (e.g., from conduction band minimum to a lower-lying valence band at the Gamma point) were identified, resulting in a wider apparent “optical” bandgap than the fundamental indirect or direct but dipole-forbidden bandgap with weak optical absorption. Table 1 lists both types of bandgaps since the optical bandgap is relevant to optoelectronics applications (UV detectors, transparent conductors) and the fundamental bandgap determines the breakdown electric field for power electronics applications.

TABLE 1 Physical properties of selected UWBG oxides. “UID” stands for “unintentionally doped” due to point defects, hydrogen, or unspecified impurities.
Full size table

Binary oxides

β-Ga2O3

The most mature, benchmark binary UWBG oxide is β-Ga2O3, whose monoclinic β-gallia crystal structure is the most thermodynamically stable polymorph of Ga2O3 [85]. This material has attracted significant attention for power electronics applications because of a large 4.8-eV bandgap, controllable n-type doping with Si/Sn/Ge, and relatively high electron mobility of ~ 200 cm2/V s [86,87,88,89]. With a critical field strength approximately three times that of SiC and GaN, β-Ga2O3 offers greater intrinsic power conversion efficiencies and further expansion of the operating-voltage–switching-frequency power electronics application space. Melt-grown native substrates are available for β-Ga2O3, with 4-inch substrates already commercially available, indicating a path to commercially viable Ga2O3 devices [90]. Due to its wide bandgap, broadband transparency, low cost, and high thermal/chemical stability, β-Ga2O3 has also emerged as a new platform for UV–vis nonlinear optics and integrated photonics such as waveguides and solar-blind photodetectors [91, 92]. Successful development of processing techniques, such as ohmic contacts [93] and etching [94, 95] (Fig. 4), enables complex device structures to be fabricated. The monoclinic lattice of β-Ga2O3 leads to pronounced anisotropy in optical, dielectric, and thermal properties [96,97,98,99,100,101,102,103,104], yet surprisingly maintains near-isotropic electrical conductivity [105]. Lack of p-type doping [106,107,108] and low thermal conductivity [103, 104, 109] pose fundamental limitations to the design of β-Ga2O3 devices.

Figure 4
figure 4

Copyright 2019, American Chemical Society.

β-Ga2O3 fin arrays formed by metal-assisted chemical etching (MacEtch) with high aspect ratio and low-damage sidewall surfaces. Reprinted with permission from [94].

Full size image

Epitaxial growth techniques developed for β-Ga2O3 include MBE [110,111,112], halide vapor phase epitaxy (HVPE) [113], MOCVD [114], pulsed laser deposition (PLD) [115], and low-pressure chemical vapor deposition (LPCVD) [116]. In the early stages of development, epitaxy of β-Ga2O3 was mostly explored by MBE, with typical growth rates of 1–5 nm/min for oxygen-plasma-assisted growths and about 2× higher for ozone-assisted growths. While MBE is suitable for the epitaxy of lateral β-Ga2O3 devices, HVPE has been a popular technique for growing vertical devices owing to its capabilities of achieving much higher growth rates (> 10 μm/h) and low background electron densities (< 1013 cm–3) [117], which are desirable attributes for obtaining thick, lightly doped drift layers required for high breakdown. As a strong competitor of HVPE, MOCVD has also been adopted for growing homoepitaxial Ga2O3 films with smooth morphology, controllable n-type doping, and fast growth rates up to about 10 μm/h [118]. High-purity epitaxial films demonstrating superior electronic qualities, including a room-temperature carrier concentration of low 1014 cm–3 [119], a compensating acceptor concentration as low as 2 × 1013 cm–3 (< 0.1% donor compensation) [120], and a low-temperature electron mobility exceeding 104 cm2/V s, have been reported [121].

Alloys between Ga2O3 and Al2O3 present a rich material space with unique properties that make them attractive candidates as UWBG semiconductors. Modulation-doped field-effect transistors (MODFETs) that utilize β-(AlxGa1–x)2O3/Ga2O3 heterostructures can offer advantages of high sheet-charge density and excellent electron mobility from a 2DEG localized at the heterointerface [122,123,124,125,126,127]. However, alloying within the (AlxGa1–x)2O3 system is complicated by the variety of structures and local coordination environments that can be adopted by both parent compounds Ga2O3 and Al2O3 (Fig. 5) whose ground-state crystal structures are the monoclinic β phase and the corundum α phase, respectively. The Al composition of most epitaxially grown β-(AlxGa1–x)2O3 has been limited to about 30%, with higher incorporation tending to drive structural degradation, phase segregation, and the formation of γ-phase inclusions [128,129,130,131,132]. The highest Al content of 52% in β-(AlxGa1–x)2O3 is achieved by MOCVD in (100)-oriented films [133, 134]. The thermodynamics of Al incorporation in Ga2O3, and the resulting effects on crystal structure and the optical and electronic properties of (AlxGa1–x)2O3 alloys, are reviewed by Varley [135].

Figure 5
figure 5

copyright 2021.

Calculated phase diagram of (AlxGa1–x)2O3 alloys. Experimental points are included for monoclinic β-(AlxGa1–x)2O3 (triangles) and corundum α-(AlxGa1–x)2O3 (circles) grown by different techniques. Reprinted by permission from [135],

Full size image

α-Ga2O3

Lately, there is increasing interest in the metastable polymorphs of Ga2O3, among which α-Ga2O3 possesses the largest bandgap of about 5.3 eV [136, 137]. α-Ga2O3 can be grown heteroepitaxially with high quality on the c-, m-, a-, or r- planes of isostructural α-Al2O3 substrates (sapphire) by mist-CVD [136, 138,139,140], MBE [141,142,143], MOCVD [144, 145], or PLD [146, 147], enabling the full compositional range of α-(AlxGa1–x)2O3 from Ga2O3 to Al2O3 to be covered without miscibility gaps to allow bandgap engineering for α-Ga2O3-based heterostructures from 5.3 to 8.8 eV [138, 142, 145, 147]. Strain relaxation of α-(AlxGa1–x)2O3 on sapphire is anisotropic and its mechanisms have been investigated [148, 149].

Rutile GeO2

A series of recent publications predicts rutile GeO2 (r-GeO2) to be an UWBG semiconductor that can outperform Ga2O3 in terms of device efficiency [150,151,152,153]. r-GeO2 has a bandgap (~ 4.7 eV [150, 154]) similar to that of Ga2O3. Its dielectric constant, predicted electron and hole mobilities [152], and thermal conductivity [151] are, however, higher than those of Ga2O3, thus bringing beneficial consequences for power-device efficiency. In addition, the predicted prospects for p-type conductivity in GeO2 [150, 155] is extremely attractive as it would enable the realization of r-GeO2 pn-junctions, thereby dramatically widening the device application space to bipolar devices, including pn-junction field management for high-voltage devices. Sb and Al are theoretically predicted to be viable donor and acceptor dopants, respectively [150]. The feasibility of bulk growth by solution top-seeding [156] and from the flux [155], as well as of epitaxial growth by MBE [157] (yet at a low growth rate of 10 nm/h) and mist-CVD [158] (at much higher growth rates of up to 1.7 µm/h), has already been experimentally demonstrated. Definitive understandings of doping and charge carrier transport properties are yet to be developed.

Rutile SnO2

A relatively narrow bandgap of about 3.7 eV does not qualify the classical binary oxide rutile SnO2 as a true UWBG semiconductor. Nonetheless, this material’s combination of significantly higher thermal conductivity and higher electron mobility than most UWBG oxides are appealing for device applications, not to mention the potential for bandgap engineering when alloyed with r-GeO2.

Ternary oxides

Spinels

The ternary spinel oxides ZnGa2O4 and MgGa2O4 provide bandgaps similar to that of β-Ga2O3. Their lower electron mobilities [83] and structure-related propensity for antisite defects [84] that carry detrimental implications on the control of carrier concentrations may, however, explain the fact that these oxides are so far only investigated for applications in photodetectors [159] or as phosphors [160, 161] rather than for power electronic devices.

Complex oxides

Ternary complex oxides of general stoichiometry ABO3 (with cations A and B) provide, based on the choice of A and B as well as strain state, a wealth of (emergent) physical phenomena resulting in a fully tunable spectrum of conductivities as well as magnetic and dielectric properties [84]. At the same time, their common cubic (or pseudocubic orthorhombic) perovskite crystal structure provides the basis for monolithic integration of those properties in epitaxial, multifunctional heterostructures for novel devices. While the prototypical wide-bandgap semiconducting complex oxide SrTiO3 suffers from a low electron mobility (< 10 cm2/V s) at room temperature, the stannates have been demonstrated to alleviate this issue. To this end, BaSnO3 has recently been demonstrated to exhibit the highest room-temperature electron mobility (> 200 cm2/V s) among the complex oxides [83, 84], and SrSnO3 has already been made into a demonstrator MESFET device [162] since it offers reasonable electron mobilities in combination with a large bandgap.

p-type oxides

UWBG p-type oxides are rare, largely unexplored, and typically suffer from ultra-low hole mobilities (< 1 cm2/V s). A review by Zhang et al. [163] on p-type transparent conducting oxides indicates that p-type oxides with optical bandgaps above 4 eV, such as the delafossite CuBO2 (with exceptionally high hole mobility of 100 cm2/V s) or perovskite Sr-doped LaCrO3 (with low hole mobility of 0.03 cm2/V s), exhibit fundamental bandgaps (either dipole-forbidden direct or indirect) around 2 eV. A potential exception is the double perovskite oxide Ba2BiTaO6 [164] with an optical bandgap > 4.5 eV (fundamental bandgap yet to be experimentally explored) and hole mobility of 30 cm2/V s, albeit at low achievable hole concentrations on the order of 1014 cm–3. Kaneko and Fujita have demonstrated how alloying a p-type oxide α-Ir2O3 (bandgap 2.6 eV) with α-Ga2O3 results in a true p-type UWBG semiconducting α-(Ir,Ga)2O3 [165].

Summary

This focus issue provides an opportunity for the reader to get a glimpse of the recent advancements in UWBG materials, physics, and related technologies. Despite being in its early years, tremendous progress has been made in this research field in exploiting the fascinating properties of UWBG semiconductors. Fundamental materials-level work in AlxGa1–xN, diamond, β-Ga2O3, and other emerging UWBG materials has begun to produce device results commensurate with the fundamental advantages that these materials promise. Open questions remain in UWBG semiconductor research while new ones continuously evolve, to which first-principles computation techniques working in tandem with experimental studies have proven indispensable for improving device performance, discovering new materials with targeted functionalities, and stimulating new research directions [174].