Guest Editorial: The dawn of gallium oxide microelectronics

Among semiconductors, Si is the foundational technology against which all others are compared. The bandgap is large enough to allow for the conductivity of the material to be controlled from semi-insulating to fully conductive but small enough such that both n-type and p-type impurities can be fully ionized. The level of integration achievable by Si is unmatched by any other semiconductor technology, and its application space is vast including digital logic, memory, RF, power switching, and even optoelectronics. Yet, as we approach the atomic limit of scaling for Si to reach the pinnacle of its performance, we see the fundamental limitations of its performance at the device level. There still remain applications and functions that are out of reach for this material. For power switching applications, the operating voltage is limited by the electric field strength at which breakdown occurs (E_br) and minimum achievable background doping in epitaxial drift layers while minimizing total energy loss is limited by resistive power dissipation during on-state current conduction and capacitive loss during dynamic switching. For RF applications, the maximum power achievable at higher frequency is limited by the power frequency product. These are fundamental performance limitations directly related to the E_br of the material which is strongly correlated with the bandgap.

In the quest for performance, various compound semiconductors with a larger bandgap have been developed for applications which are complementary to Si technology when Si cannot provide sufficient power density to achieve required metrics. Consequently, GaAs, SiC, and GaN have shown tremendous advances for both RF power amplification and power switching. However, materials with a bandgap greater than 3.3–3.4 eV, which is the value for SiC and GaN, come with many technical challenges. As the bandgap energy increases, the ionization efficiency of intentional dopants decreases, as does the likelihood of having both n-type and p-type conductivities. In addition, there exist no large-area, native substrates with a bandgap larger than 3.4 eV which can achieve a large range of conductivities. It is here that we turn to the transparent conductive oxide family of semiconductors, in which gallium oxide (Ga₂O₃) is unique and attractive.

Ga₂O₃ is by no means a new material. The properties of bulk single crystals were first studied in the 1950–1960s^1,2 with more recent melt and epitaxial growth technologies developed into the 1990s. However, in the last five years, Ga₂O₃ has attracted great attention due to a combination of unique material properties and substrate availability. The majority of the efforts to date have been devoted to monoclinic β-Ga₂O₃ due to its thermal stability, resulting in it being the most readily available. There are four other metastable polymorphs labeled α, γ, δ, and ε. Recently, these other polytypes have received increased attention due to unique properties not found in the β-phase. Corundum α-Ga₂O₃ is the second most studied polytype behind the β-phase because of its ease of heteroepitaxial growth on sapphire substrates. γ- and δ-Ga₂O₃ have defective spinel and cubic bixbyite-like structures, respectively. Hexagonal ε-Ga₂O₃ is of interest due to internal polarization that could lead to the formation of high-density two-dimensional electron gas (2DEG) at the (AlGa)₂O₃/Ga₂O₃ interface similar to AlGaN/GaN-based heterostructures. These metastable-phase Ga₂O₃ films are mostly obtained by low-temperature heteroepitaxial growth on foreign substrates and are easily converted into the most stable β-phase during high-temperature thermal annealing. Therefore, in this editorial, we will focus on β-Ga₂O₃.

The main motivation for research and development (R&D) on β-Ga₂O₃ is inspired by its incredibly large bandgap, which is estimated to be around 4.5–4.9 eV.^2–5 Due to the complex band-structure of β-Ga₂O₃, a wide range of high E_br from 5 to 9 MV/cm can be estimated from the bandgap using methods outlined by Hudgins et al.⁶ E_br is more than double the theoretical limits of SiC and GaN and translates to more than triple their power device performance predicted from the Baliga figure of merit, which is most commonly used to evaluate the suitability of a material for power switching devices.⁷ Additionally, a wide range of electron densities from ∼1 × 10¹⁵ cm⁻³ to >1 × 10²⁰ cm⁻³ has been demonstrated by intentional donor doping in epitaxial materials. The ability to reduce DC conduction losses in devices by minimizing on-resistance (R_on) and maximizing breakdown voltage (V_br) lends to high-power and high-voltage applications and, when combined with advanced scaling techniques, high-speed switches for low/medium-power applications. According to the Johnson figure of merit,^8,9 which is another benchmark for high-frequency power devices, the power frequency product for Ga₂O₃ is comparable to that for GaN because even though the saturation electron velocity in Ga₂O₃ is lower than that in GaN, the extremely large E_br of Ga₂O₃, which is more than double that of GaN, can compensate for the shortcoming. These properties combined with the availability of high-quality, large-area native substrates offer a complete platform for various applications such as high-performance power switching, RF amplifiers, and harsh-environment signal processing.

To date, Ga₂O₃ has achieved many key milestones for technology viability. The existence of high-quality Ga₂O₃ wafers manufactured from bulk single crystals that can be synthesized by several melt growth methods, such as float zone,^10–12 Czochralski,^13,14 vertical Bridgman,¹⁵ and edge-defined film-fed growth (EFG),^12,16 is one of the most important material features. Ga₂O₃ wafers with diameters as large as 2 in. for (001), 4 in. for (⁠$2¯$ 01), and 1 in. for (010) are presently available. The melt-grown bulk single crystals offer a significant advantage of Ga₂O₃ over SiC and GaN in terms of production cost because SiC and GaN bulk crystals require alternative synthesis techniques using higher pressure and higher temperature. From a physical perspective, high-quality Ga₂O₃ wafers also provide a unique opportunity to study epitaxial materials in which the lowest possible number of dislocations can be obtained.

Although polycrystalline Ga₂O₃ thin films have been used for decades as a transparent, charge dissipation layer for optical applications, epitaxial growth technology of single-crystal Ga₂O₃ thin films remains to be established. Several techniques such as molecular beam epitaxy (MBE),^17–23 pulsed laser deposition,^3,24,25 halide vapor phase epitaxy (HVPE),^26–28 metalorganic chemical vapor deposition,^29–32 mist chemical vapor deposition (mist-CVD),^33–36 and low pressure chemical vapor deposition³⁷ have been used for the epitaxial growth of Ga₂O₃-based materials on both native and non-native substrates. Each technique has some drawbacks and advantages. However, to date, MBE and HVPE have shown the best control over electron density.

With intentional shallow donors (Si, Ge, and Sn) and deep-level compensating acceptors (Mg and Fe), n-type conductivity is controllable over fifteen orders of magnitude, i.e., from highly conductive (∼10⁻³ Ω cm) to semi-insulating (∼10¹² Ω cm). However, as is typical with other oxide semiconductors, it is unlikely that p-type doping can be achieved with effective hole conduction in Ga₂O₃. Shallow acceptors are not predicted for Ga₂O₃, and hole transport is further restricted by its valence band structure. It is known from theoretical calculations that the valence band maximum of Ga₂O₃ has very small dispersion, leading to an extremely large hole effective mass.^4,38,39 Furthermore, a unique phenomenon that holes are inherently self-trapped in the Ga₂O₃ bulk has also been predicted.⁴⁰ Therefore, it is expected that unipolar Ga₂O₃ device applications will dominate, while bipolar devices will be difficult to realize.

β-Ga₂O₃ has multiple phonon modes in the low-energy range due to its band structure.^41,42 The low-energy polar optical phonons result in a dominant scattering factor that limits room-temperature electron mobility to 200–250 cm²/V s.⁴³ On the other hand, the saturation electron velocity is estimated to be around 2 × 10⁷ cm/s from theoretical calculations,⁴⁴ which is suitable for high-frequency operation. Although the mobility of Ga₂O₃ is lower than those of traditional compound semiconductors, the cubic dependence of DC conduction loss on E_br allows superior switching performance. Furthermore, for RF applications, the large E_br can support extreme scaling of the Ga₂O₃ field-effect transistors (FETs), allowing the electron velocity to reach saturation at a reasonable operating voltage.

Poor thermal conductivity is expected to be an engineering challenge for high-power Ga₂O₃ devices.^45–47 Packaging solutions for cooling devices will necessarily be a combination of topside thermal shunts and backside heat extraction through a thinned wafer mounted to a heatsink. Wafer bonding processes are particularly advantageous for overcoming poor heat dissipation caused by the low thermal conductivity of Ga₂O₃. Alternatively, structures consisting of thin Ga₂O₃ drift layers transferred to a supporting substrate with high thermal and electrical conductivities may be favorable for high-voltage and high-power vertical device applications.

Two fundamental device components for power conversion systems, Schottky barrier diodes (SBDs) and FETs are intensively being developed. Initial Ga₂O₃ SBDs made use of native n-type Ga₂O₃ substrates to evaluate the basic device performance.^48–50 However, Ga₂O₃ SBDs lagged behind FETs in development progress in the early years due mainly to a lack of suitable epitaxial growth techniques for thick n-Ga₂O₃ drift layers. The advancement of epitaxial growth technologies made it possible to obtain high-quality drift layers with a reasonably low electron density and led to acceleration of the SBD development.^51–53 One of the state-of-the-art Ga₂O₃ SBDs with an HVPE-grown drift layer demonstrated promising characteristics such as a reasonably low specific R_on, a reverse V_br over 1 kV, and an ideality factor close to unity.⁵⁴ Given more sophisticated edge termination structures used in Si and SiC power diodes, Ga₂O₃ SBDs should be able to be operated at much higher voltages. Another unique fabrication approach of free-standing α-Ga₂O₃ SBDs was also reported.⁵⁵ The α-Ga₂O₃ SBDs were fabricated through a lift-off of an epitaxial layer grown by mist-CVD on a sapphire substrate. The removal of the sapphire substrate and direct bonding of the device active area to a heat sink is effective in promoting heat dissipation.

The first single-crystal Ga₂O₃ transistors by means of metal-semiconductor FETs (MESFETs) were demonstrated in 2011.⁵⁶ The MESFETs developed into depletion-mode (D-mode) metal-oxide-semiconductor FETs (MOSFETs) in 2013,⁵⁷ which are the basic lateral Ga₂O₃ FET structures for current work. To date, D-mode Ga₂O₃ MOSFETs have demonstrated a large on/off ratio exceeding ten orders of magnitude,^57,58 stable operation up to 300 °C,^57,58 mean and peak electric fields in the Ga₂O₃ channel exceeding 3.8 and 5.3 MV/cm, respectively,⁵⁹ and pulsed drain current density exceeding 450 mA/mm.⁶⁰ Note that the peak value reported for E_br is much larger than the theoretical limits of SiC and GaN. Advanced D-mode Ga₂O₃ MOSFETs with a gate-connected field plate have achieved off-state V_br as high as 755 V.⁶¹ D-mode Ga₂O₃ substrate-gated, nanomembrane FETs with minimal self-heating effects demonstrated a record high drain current density of 1.5 A/mm,⁶² which is comparable to typical values achieved in GaAs- and GaN-based FETs. Enhancement-mode (E-mode) FETs enabling normally off operation are strongly desired for power switching for fail-safe functionality. Early E-mode results were obtained with wrap-gate devices achieving V_br > 600 V,⁶³ followed by lower voltage demonstrations using nanomembrane FETs⁶² and MOSFETs with an unintentionally doped channel layer.⁶⁴ Developments of vertical Ga₂O₃ FETs that are more suitable for high-voltage, high-power applications are starting to be reported.^65–67 

Device developments required for low series resistance and increased V_br in power switching FETs are mostly applicable to RF FETs. In addition, high-speed switching and high-frequency operation require short-gate device structures. The electron velocity in Ga₂O₃ is expected to saturate around 30 V/μm, which is almost equal to the E_br value of Si. Thus, highly scaled device topologies favoring high electric field profiles are required for high-performance operation of lateral RF FETs in Ga₂O₃. Highly scaled device structures with self-aligned electrodes fabricated by ion implantation are also considered to be viable and to be compatible with advanced scaling techniques used in industry. The first RF device characteristics of Ga₂O₃ MOSFETs with a gate length of 0.7 μm showed a maximum oscillation frequency of 13 GHz at a drain voltage of 40 V and a output power density >250 mW/mm at 800 MHz.⁶⁸ 

Very recently, significant progress in the development of (AlGa)₂O₃/Ga₂O₃ modulation-doped FETs (MODFETs), which are analogous to AlGaAs/GaAs MODFETs, has been demonstrated.^69–71 Unlike GaN-based heterostructures, there is no internal polarization in the monoclinic β-Ga₂O₃ structure. Therefore, formation of a high-density 2DEG induced by polarization doping is not expected at the (AlGa)₂O₃/Ga₂O₃ interface. However, the MODFET structure does provide an advantage for channel charge control by applied gate voltage, which is especially important for highly scaled devices.

Ga₂O₃ is a unique material with many attractive properties. Studies of material physics, defect characterization, dielectric interfaces, and development of device technologies on this new semiconductor have just begun and remain open-ended. Practical power switching and RF device applications require more advanced device architectures and thus demand further improvements in all aspects of melt/epitaxial growth, doping, and device processing and deeper understanding of fundamental material properties to be included in new modeling capabilities that are accurate under high-electric field operating conditions. R&D in the next decade is certainly crucial for the road to industrialization of Ga₂O₃ electronic devices. We hope that you will find this reference useful and be inspired to join in the development of Ga₂O₃ microelectronics.