Accurate Modeling of Gate Leakage Currents in SiC Power MOSFETs

TL;DR

This work tackles gate leakage in SiC MOS devices by developing a self-consistent physics-based framework that couples electrostatics, quantum tunneling, 1D Boltzmann transport, impact ionization, and nonradiative multiphonon hole trapping within the Comphy platform. The approach yields quantitative agreement with gate-leakage $J_g$-$V_g$ measurements across 80–573 K without empirical fitting, revealing that electron-initiated impact ionization in the a-SiO$_2$ oxide generates electron–hole pairs and that captured holes amplify leakage, shifting the flat-band condition. A key finding is that the conduction-band offset $\Phi_c(T)$ acts as the tunneling barrier rather than the combined barrier $\Phi_{\text{eff}}(T)$, highlighting the role of Fermi-level offsets in thick oxides. The framework provides a predictive tool for oxide reliability and lifetime projections in wide-bandgap devices and is applicable to advanced Si CMOS processes that share the same gate dielectric stack.

Abstract

Silicon carbide (SiC) metal-oxide-semiconductor field-effect-transistors (MOSFETs) enable high-voltage and high-temperature power conversion. Compared to Si devices, they suffer from pronounced gate leakage due to the reduced electron tunneling barrier at the interface between SiC and amorphous silicon dioxide (a-SiO$_2$). We develop a self-consistent, physics-based simulation framework that couples electrostatics, quantum tunneling, carrier transport, impact ionization, and charge trapping for both electrons and holes. The model quantitatively reproduces measured gate-current-voltage characteristics of SiC MOS capacitors over a wide temperature (80-573 K) range and a wide bias range without empirical fitting. Simulations reveal that conduction electrons in a-SiO$_2$ can trigger impact ionization, which generates electron-hole pairs, and leads to capture of holes in the oxide bulk, thereby enhancing gate leakage current. The framework captures these coupled processes across multiple orders of magnitude in time and field, providing predictive capability for oxide reliability. Although demonstrated for SiC devices, the methodology also applies to Si technologies that uses the same gate dielectric.

Figures (20)

• Figure 1: Device-physics framework and experimental validation. (a) Self-consistent one-dimensional MOSC (1D MOSC) model: electrostatics set the potential and band profiles across SiC, a-SiO$_2$, and n$^+$-Si; quantum tunneling injects electrons into the oxide; drift transport carries hot electrons, which undergo impact ionization to generate holes; these holes are captured by defect states in the oxide. Conduction-band edge $E_\text{c}(z)$, valence-band edge $E_\text{v}(z)$, and Fermi levels $E_\textsc{f}$ are indicated in each region. (b) Gate leakage current density $J_\text{g}$ versus gate bias $V_\text{g}$ at 80, 298, and 448 K for the MOSC (oxide thickness 54 nm). Experimental data (points) are compared with simulations that include only tunneling (dashed) or both transport and charge capture (solid).
• Figure 2: The 1D MOSC model for gate leakage current simulation. Top: External circuit used for analysis, with voltage and conventional current (arrow) defined as positive. Middle: Illustrative space-charge profile $\rho(z)$ showing oxide charge trapping and possible interface charges at the SiC/a‑SiO$_2$ and n$^+$-Si/a‑SiO$_2$ interfaces, respectively. Bottom: Illustrative current density profile $J(z)$ across the MOSC, indicating Fowler-Nordheim tunneling near $z_\text{tm}$ and amplification of electron and hole currents via impact ionization.
• Figure 3: Band structure of 4H-SiC. (a) Crystal structure representation: (i) (0001) plane showing lattice points in direct space (open circles) and reciprocal space (filled circles), with first Brillouin zone boundary (orange hexagon) and Cartesian coordinate axes; (ii) three-dimensional view of the first Brillouin zone highlighting high-symmetry points ($\Gamma$, M, K, L, A, H) and crystallographic direction $k_z = [001]$. (b) Energy dispersion along high-symmetry directions: (top) conduction band modeled by two Kane-type non-parabolic bands ($\Delta E = 0.115$ eV) with effective masses and non-parabolicity parameters annotated, showing excellent agreement with DFT calculations (symbols) massDFTplusexp_KaczerPRB1998; (bottom) valence band structure comprising heavy-hole , light-hole, and split-off hole bands, illustrating spin-orbit coupling [dashed (slod) lines include (exclude) spin-orbit coupling] as reported in Ref. VBofSiC_PersonLindefeltJAP1997. Note $k_\parallel = [001]$, while $k_\perp$ is in the $\Gamma$MK plane.
• Figure 4: Nonradiative multiphonon–mediated charge transitions in a-SiO$_2$. (a) Configuration–coordinate diagram for a two-state defect: parabola 1 represents the potential-energy surface when the electron is bound in the trap, parabola 2 when the electron occupies the conduction band. Their intersection defines the capture barrier $\varepsilon_{21}$, the emission barrier $\varepsilon_{12}$, and the relaxation energy $E_\textsc{r}$. (b) Band-diagram view of a trap-band exchange in the oxide. A defect at energy $E_\text{t}$ and depth $z_\text{t}$ may undergo a capture/emission event to the oxide conduction/valence band locally at $z_\text{t}$ (multiple-trapping), and/or non-locally at the Fowler-Nordheim turning point $z_\textsc{fn}$ (phonon-assisted tunneling).
• Figure 5: Methodology for modeling tunneling from the semiconductor space-charge region into the insulator. (a) Device electrostatics at $V_\text{g} = 40$ V, T = 298 K: carrier and ionized dopant profiles (top) under band bending (bottom), showing local carrier contributions $\Delta Q_n(z_k)$ to the surface charge. (b) Implementation of a weighting scheme for tunneling: (i) weight function $W_q(z)$ derived from surface charge contributions, (ii) energy-resolved electron distributions $n(E_\parallel,z)$ at selected depths, (iii) total weighted distribution for tunneling calculation.