Perspective on breakdown in Ga₂O₃ vertical rectifiers

There is currently a debate on the optimum choice among ultrawide bandgap (UWBG) semiconductor materials for high-power electronic systems.^1–3 The candidates include Ga₂O₃, diamond, and AlN. Each has its advantage and drawbacks, as listed below. Ga₂O₃ is currently the most advanced in terms of reports in the literature on power rectifiers and transistors,^1–4 but enthusiasm for its development has waned in the US, where the focus is more on AlN and diamond. A part of the issues related to this is the large number of reports where materials, processing, or testing limitations lead to leaky devices with poorly controlled reproducibility. In this Perspective, we suggest that this is a normal part of the maturity cycle and more work is needed to understand even the experimental breakdown field.

Below, we present a short summary of the shortcomings of the candidate ultrawide bandgap semiconductors.

The most mature of the polymorphs for this material is β-Ga₂O₃, with monoclinic symmetry.¹ This leads to anisotropic electrical and thermal properties. The main drawbacks include the following:

Low thermal conductivity in all directions:

1. The absence of shallow p-type dopants, which has been partially addressed using p-type oxides such as NiO to create heterojunctions. A p-NiO/n-Ga₂O₃ heterojunction diode (HJD) forms an abrupt p + −n junction, with the depletion region predominantly in Ga₂O₃ under reverse bias.

2. Commercial availability of device-quality epi and substrates is essentially limited to one supplier, Novel Crystal Technology, in Japan.

3. The process technology is still immature, which leads to variations in device performance reported in the literature from nominally similar starting rectifier structures.

With arguably the best combination of materials properties, the main drawbacks include the following:

1. Diamond substrates are expensive, and supply of electronic grade structures is also limited to a few vendors. Single crystal diamond substrates are extremely difficult to upscale in size (current limit ∼1 in.). Diamond can be doped as a p-type material in a wide range of hole concentrations; however, n-type conductivity is much less attainable, often limited to surface conductivity.

2. There are difficulties in achieving consistent and high-quality diamond films for device fabrication. These include challenges in controlling impurities and defects.

3. Developing reliable and efficient diamond-based rectifiers requires advancements in device design and fabrication techniques.

4. While diamond has excellent thermal conductivity, efficient heat dissipation in high-power applications can still be challenging, especially for large-scale devices.

5. The long-term reliability and stability of diamond-based rectifiers in harsh operating environments need to be thoroughly evaluated and proven.

With a very wide bandgap and excellent thermal conductivity, there is promise of a very high-power operation. However, current limitations include the following:

1. AlN substrates remain expensive, although the recent scaling to 4-in. substrates is an advancement.

2. There are difficulties in achieving consistent and high-quality doping for device fabrication. The ionization energy of Si increases with increasing Al content in AlGaN and undergoes a DX transition in Al-rich AlGaN, acting as compensating acceptors.² In other words, Si DX centers act as acceptor-type compensating point defects, which lowers free electron concentration.² For acceptors, the theory shows that small energy differences between delocalized and trapped holes, except Be, strongly trap holes, forming localized states. Mg has the lowest ionization energy among acceptors but limits the maximum p-type carrier concentration to <10¹⁷ cm⁻³. A major concern is that if doping is done only through hopping conductivity from relatively deep acceptors/donors, then is it possible to capitalize on such conductivity in devices using carriers in hopping conduction bands to inject holes or electrons, obtaining conductivity modulation in p-i-n structures, or avalanche amplification in avalanche photodiodes?

3. Contact resistances remain high due to the limited doping achievable and facile formation of Al₂O₃ by the oxidation of dangling Al bonds.

In these equations, E_CRIT is the critical field of the semiconductor, e is the electronic charge, N_D is the doping in the drift region, ɛ is the permittivity of the semiconductor, and W_PT is the depletion depth at punch-through. Reported literature values for the theoretical value of E_CRIT for β-Ga₂O₃ include 6.7,⁵ ∼8,^6,7 10.3,⁸ and 15 MV cm⁻¹ (Ref. 9). Thus, one of the first tasks is to establish the correct value for Ga₂O₃. Currently, the value is generally assumed to be ∼8 MV  cm⁻¹, with little critical analysis for its basis. It can be up to 13.2 MV  cm⁻¹ in Zn-doped Ga₂O_3.^10,11

The experimentally measured peak electric field has been reported to be in the range of 6–7.4 MV  cm⁻¹ in the best cases of high V_B/W_PT, namely, 8.3 kV/13 μm¹² and 13.5 kV/18.4 μm.¹³ These are both from small-area devices, with an anode radius of 75–100 μm. There is still a gap between the performance of large-area devices and small-area devices due to nonuniformity in material properties such as doping, thickness, and extended defect density. In our case, mm² devices fabricated on the same area as small devices achieving 13.5 kV breakdown typically exhibit V_B in the range of 3.9–7.2 kV, corresponding to peak fields of <4.8 MV  cm⁻¹. At these high fields, there is an absence of hole self-trapping effects and also evidence of conductivity modulation in p-NiO.¹⁰ 

Basically, all Ga₂O₃ rectifiers reported in the literature operate in a punch-through mode, shown schematically in Fig. 1(a). In Fig. 1(b), we show a compilation of V_B values in our punch-through rectifiers for different drift layer thicknesses and doping levels. In some cases, the experimental breakdown voltages are now reaching theoretical values, indicating that in some areas of some current wafers, the control of defect density and doping is sufficiently good to enable achievement of the full potential of Ga₂O_3.

What is the basis for the widely quoted 8 MV cm⁻¹ critical field for Ga₂O₃? The initial reference comes from Higashiwaki et al.,⁶ estimated from the treatment of Hudgins et al.⁷ The latter assumes a power law dependence of E_C on E_G that is not empirically accurate and also assumes that impact ionization is the breakdown mechanism, which has not yet been experimentally seen in Ga₂O₃.

As pointed out by Goodnick et al.,¹⁴ there is no such thing as a single “critical field” for a semiconductor material since the breakdown field depends on doping and temperature and also depends on drift region geometry, i.e., punch-through vs non-punch-through. Theoretical treatments are inherently one-dimensional, while impact ionization is anisotropic. Such approaches misrepresent real devices where 2D and 3D effects typically dominate (e.g., edge terminations) and transport within these simulations is oversimplified. The simulators also generally assume complete dopant ionization, which is universally untrue for ultrawide bandgap (UWBG) materials and assumes only simple low-field drift transport.¹⁴ As an example from the previous generation of power semiconductor materials, a literature review of reported impact ionization coefficients in GaN (α₀ and ɛ₀) yields a vast range of values for a critical electric field and a varying dependence on doping.¹⁴ This work demonstrates the impact of ionization coefficients on breakdown and the discrepancy between one-dimensional and two-dimensional models in predicting breakdown voltage.

In conventional semiconductors, avalanche breakdown occurs when field-accelerated electrons ionize an atom, promoting excess electrons to the conduction band, eventually leading to an electron avalanche. This process can be assisted by defect-mediated trap states that lower the barrier, as well as direct electron-tunneling effects.^15,16 Impact ionization is triggered when the kinetic energy gain by the electron from the field is larger than what can be relaxed via electron–phonon and electron–electron scattering processes. This is the von Hippel criteria,^17–19 which state that impact ionization occurs when the average energy gained by electrons from an electric field exceeds the average energy lost to phonons, leading to a runaway process of carrier multiplication and breakdown.

It is unclear whether avalanche breakdown occurs even in UWBG materials. For instance, the ionization coefficient of electrons in some wide-gap semiconductors has been shown be dramatically lower than that of holes,^17,18 because the width of the lowest-lying conduction band is narrower than E_g. The combination of a narrow conduction band and a large bandgap in electronic oxides makes it less probable for carriers to gain the critical energy required for impact ionization. This property contributes to the excellent dielectric properties and reliability of these materials in electronic devices. Furthermore, for some oxide-based semiconductors with a significant degree of ionic bonding, electron affinity is much larger than the bandgap, which is the opposite in many UWBG semiconductor materials. These drastic differences in electronic structures could lead to other dielectric breakdown processes involving bond breakage and eventual material degradation and/or electron emission that are unaccounted for by the von Hippel criteria. In UWBG materials, where bandgaps are larger than their electron affinity and conduction bandwidths, new breakdown processes may arise, such as direct bond breaking. It is important to note that while impact ionization may be less likely in wide bandgap electronic oxides, it is not entirely ruled out. Further research is needed to fully understand the breakdown mechanisms in these materials and to develop strategies to mitigate their impact on device performance.

To summarize, the basic parameters that define applicability to power electronics, such as the critical field, the presence of avalanche breakdown, temperature dependence of these parameters, and breakdown mechanisms are still not firmly established in Ga₂O_3.^20–29 

In terms of immediate issues to be addressed for the commercialization of Ga₂O₃ power rectifiers, the major materials challenges for Ga₂O₃ power devices to be commercialized based on the review by Kohei Sasaki¹ are the following:

1. Cost of epi wafers: Growing high-quality Ga₂O₃ crystals for device fabrication is expensive, especially for large-diameter wafers needed for power electronics.

2. Killer defects: Crystal defects like (100) plane twins and nanovoids can significantly impact device performance and yield.

3. Purity of the epitaxial layer: Achieving consistent and high purity in the epitaxial layer is crucial for device reliability.

4. Nondestructive inspection methods: Nondestructive characterization methods to measure defects in UWBG semiconductors include photoluminescence (PL) spectroscopy, which is limited by sensitivity and potential interference from other optical processes, Raman spectroscopy, which is limited by sample preparation requirements and potential interference from surface scattering, deep-level transient spectroscopy, which is limited by temperature range and sensitivity, and positron annihilation spectroscopy, which is limited by the need for a positron source and data analysis complexity. These methods provide valuable insights into the defect structure of UWBG semiconductors, but they have limitations in terms of sensitivity, specificity, and sample preparation requirements. Transmission electron microscopy images have been instructive, but this is a destructive technique and limited in analysis area.^30–32 

These challenges hinder Ga₂O₃ from being a currently viable alternative to existing power electronics materials such as SiC and GaN.

Additional areas with ongoing research that could address these limitations are as follows:

1. Developing capping layers to suppress the formation of high-resistivity layers during oxidation.^33,34

2. Investigating alternative p-type doping techniques.

3. Developing nondestructive inspection methods for Ga₂O₃ wafers.

What are some of the additional future device research issues with Ga₂O₃ rectifiers? One is the value of breakdown voltages obtained in small-area devices on “sweet spots” on the currently available commercial wafer structures, which typically involve areas of thicker drift regions and lower doping levels. We suggest that these values are important since they show what will be possible on a routine basis as growth and field management techniques are optimized. Similarly, differences in breakdown measured under static versus dynamic switching conditions are typical in immature semiconductor materials systems and will narrow as that technology matures.^20,21 Focusing only on large-area devices with Ampere range currents is also misleading at this stage, since nonuniformity is an issue on currently available wafers and differences would be expected between the performance of small- vs large-area devices. The breakdown voltage depends on diode size and gets smaller with larger devices, an indication that defects in the material still limit diode performance. No recently observed temperature dependence of breakdown voltage or measured ionization coefficients support the possibility of the observation of avalanche breakdown in β-Ga₂O_3.

In summary, Ga₂O₃-based rectifiers do have the potential to revolutionize power electronic applications and enable significant advancements in energy efficiency and renewable energy technologies. Moving forward, additional work is needed on the following basic science, fabrication, and testing areas:

1. Establishing the critical breakdown field of both β-Ga₂O₃ and its wider bandgap (5.2–5.3 eV), rhombohedral symmetry polymorph, α-Ga₂O₃.

2. Better understanding of the leakage current mechanisms at the interface of HJDs, including the role of interface states and tunneling of carriers from NiO to Ga₂O₃^27,28 and of the role of defects.^30–32 

3. More avalanche and surge current tests are needed, since from an application perspective, they are important for establishing the robustness of power devices. The observation of positive temperature coefficients for breakdown would establish the presence of impact ionization in reverse breakdown.

4. Wide bandgap electronic oxides may be resistant to impact ionization due to their narrow conduction bands, large bandgaps, and ionic bonding, which can lead to alternative breakdown mechanisms. The presence of specific mechanisms under different conditions needs to be established.

5. Ensure that the breakdown under dynamic switching conditions is the same as the static breakdown voltage. In GaN rectifiers, the difference between the two can be significant and on the order of a factor of two.²⁹ 

6. Optimization of the trade-off of on-resistance and on-voltage versus breakdown voltage for specific switching applications. Reducing switching losses is essential for improving the efficiency of Ga₂O₃-based rectifiers.

7. Greater realization among the general device community of the importance of avoiding damage to the Ga₂O₃ surface during patterning and deposition processes.^35–37 

8. Improved understanding of the figures of merit for specific power applications. Recent research has derived a power loss density limit for wide bandgap (WBG) and UWBG devices, which can guide thermal management design.^3,28 It finds that UWBG devices offer a lower total loss limit than WBG devices, especially at high voltages. The drift region design for switching performance optimization differs from that for static resistance optimization.

9. Improved understanding of the role of cables, probe tips, and probe station components in contributing leakage current to the I-V characteristics for very high voltage conditions. Our experience has shown that such contributions can be higher than diode leakage and give rise to an ohmiclike shape to the I-V’s at biases below breakdown. Many research groups even lack the capability of touching voltages above 3 kV.

10. Better design and control of packaging and thermal management of high-power devices. Currently, even the cleaving technology for Ga₂O₃ is inadequate.¹ 

The work at UF was performed as part of Interaction of Ionizing Radiation with Matter University Research Alliance (IIRM-URA), sponsored by the Department of the Defense, Defense Threat Reduction Agency, under Award No. HDTRA1-20-2-0002. The content of the information does not necessarily reflect the position or the policy of the federal government, and no official endorsement should be inferred. The author at the National Yang Ming Chiao Tung University would like to acknowledge the National Science and Technology Council, Taiwan, for their financial support under Grant No. NSTC 112-2628-E-A49-015.

The authors have no conflicts to disclose.

Jian-Sian Li: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). Chao-Ching Chiang: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). Hsiao-Hsuan Wan: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). Fan Ren: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). Yu-Te Liao: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). Stephen J. Pearton: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Project administration (equal); Writing – original draft (equal); Writing – review & editing (equal).

The data that support the findings of this study are available within the article.