Gallium Oxide Vertical Power Devices - Modeling, Material Growth, and Hetero-epitaxial Integration with Silicon

Abstract

𝛽-Ga2O3 is a candidate for a new generation of power semiconductors that can significantly lower the conduction losses and increase power density in the power converters. PFOM or power figure of merit (V2Br/Ron) is a ratio of the breakdown voltage squared and the area-specific on-resistance of a power device. PFOM is a measure of how low the conduction loss (through Ron) is for a given breakdown voltage (VBr). 𝛽-Ga2O3 has a material potential for PFOM < 34 GW/cm2, significantly better than GaN (8.6 GW/cm2), SiC (3.3 GW/cm2) or Si (0.01 GW/cm2). In the absence of p-type conductivity in 𝛽-Ga2O3, we propose heterojunction bipolar (vertical) devices with p-WBG semiconductors to increase the PFOM. 𝛽-Ga2O3 also suffers from very poor thermal conductivity. In this thesis, we motivate heteroepitaxial integration of 𝛽-Ga2O3 on the silicon substrate to alleviate this problem and significantly reduce the substrate cost. We (re)develop and characterize the unit processes for the optimal growth of UID (unintentionally doped) and doped Ga2O3, NiO, and Cu2O semiconductor films that we use in the devices described further.

We examine the possibility of 𝛽-Ga2O3 HBT with n-p-n configuration using first principles and TCAD simulations. Cu2O is a p-type semiconductor with a high minority diffusion length and a low conduction band offset with 𝛽-Ga2O3, making it suitable for the heterojunction base of HBT. Motivated by a report of higher than (empirical) theoretically expected VBr in the Cu2O- Ga2O3 diode, we investigate its origin using TCAD simulations. Two potential explanations are interface charges (by shielding the field in low Ec Cu2O) and heavily doped base (by increased avalanche breakdown field). In our attempts to repeat the Cu2O-Ga2O3 diode with the same stack, we don’t observe any improvement in the VBr with increasing Cu2O doping. In Cu2O TFTs (thin film transistors) that we make with different dielectrics, we find considerable field effect mobility (𝜇𝐹𝐸) degradation due to interface defects (from interfacial CuO). The values (of interface charges or heavy doping in the base) that explain the high VBr are detrimental to the on-state of the HBT. For the simple vertical architecture, an ideal Cu2O-Ga2O3 HBT cannot deliver higher PFOM than unipolar devices like MOSFETs, MESFETs, FinFETs, or HJFETs. To achieve higher than state-of-the-art PFOM, we identify alternative p-type semiconductors with wider bandgap (WBG). We estimate the PFOM bounds for p-WBG base HBTs and prescribe the design rules for p-type WBG candidates.

We develop a platform to integrate Ga2O3 power electronics on a Si substrate. As Ga2O3 is thermodynamically a less stable oxide than SiO2, we need a stable epitaxial buffer to prevent Ga2O3 reduction to GaOx. TiN is a conductive epitaxial buffer that has the potential to host vertical power devices without significant resistance losses. However, the fact that Ga2O3 is less stable than TiO2 raises a question of its suitability. MgO/TiN (with ultra-thin MgO) conductive epitaxial buffer allows a relatively abrupt interface with epitaxial Ga2O3 but drastically increases vertical resistance. Removing the MgO interlayer causes interfacial GaTiOx that dominates the resistance, although not as severely. To prevent the interfacial GaTiOx, we ramp up the TiN/Si substrate temperature in forming gas. We get a sharp interface with epitaxial Ga2O3. We achieve vertical power diodes with on-state resistance (5 mΩ cm2) comparable to state-of-the-art and with a negligible contribution (<1 mΩ cm2) from the Ga2O3-TiN interface. TMAH treatment for roughness reduction is counter-productive, leading to large etch-pits and through-Ga2O3 pipe holes around extended defects like dislocations or grain boundaries. We identify these defects as the primary source of off-state leakage. We must develop large lateral crystallite size Ga2O3 films to achieve the possible PFOM potential.