SiC Device Development When Voids in SiC Superjunction Structures Are Acceptable

Related Vendor

FAULHABER Drive Systems

Voids formed during the epitaxial filling of SiC trenches reduce the breakdown voltage. However, eliminating them can increase the on-resistance. A simulation study examines how this trade-off can be resolved—and why the position of a defect is critical.

In SiC superjunction devices, narrow, nearly vertical columns offer a favorable compromise between breakdown voltage and specific on-resistance R_DS(on),sp, i.e., the resistance multiplied by the active area in the conducting state. However, their fabrication is challenging: When deep trenches are filled epitaxially, the opening can close before the underlying space is completely filled with semiconductor material. This leaves behind a void that alters the charge distribution and the electric field.

Wider trenches with steeper sidewalls make it easier to fill them without leaving voids. However, this geometry is not necessarily more advantageous from an electrical standpoint. It is precisely this relationship that Bailing Zhou’s team investigates in the study “Simulation study of void impact in trench-filling-epitaxy-based 4H-SiC semi-superjunction Schottky rectifiers”, published in Power Electronic Devices and Components. The authors combine device simulations with numerical optimization and compare two approaches: avoiding voids through geometry or specifically accounting for them during the design process.

Charge Compensation Meets Manufacturing Limits

Superjunction structures utilize adjacent p- and n-doped columns whose space charges largely compensate each other in reverse-bias operation. This allows the current-carrying n-region to be more heavily doped than would be possible with a conventional unipolar drift region having the same breakdown voltage. The electrical resistance decreases. In the Schottky diodes made of 4H silicon carbide examined in this study, the charge compensation structure extends only over a portion of the drift region. Below the columns, there is also a weakly n-doped layer. The study uses the term “semi-superjunction” (SSJ) to describe this configuration.

The study considers two methods for fabricating the columns. In epitaxial trench filling with p-doped SiC (P-type trench filling, PTF), trenches in an n-doped region are filled with p-doped SiC. In the case of n-doped SiC trench filling (N-type trench filling, NTF), the n-pillar is formed by filling a trench in a p-doped region. This difference determines whether a void is located in the charge-compensating p-column or directly in the current-carrying n-region.

The authors investigate trench widths of 1.5 and 3.0 µm, as well as trench sidewall inclinations relative to the vertical of 2°, 5°, and 8°, omitting any geometrically unsuitable combinations for NTF. The cavity shapes are derived from experimental cross-sectional images and incorporated into two-dimensional models for physical device simulation (Technology Computer-Aided Design, TCAD). The electrical parameters are derived from simulations using Synopsys Sentaurus. A mathematical Ersatz model (response surface model) based on these results describes how the electrical parameters depend on dopant concentrations and accelerates the optimization process. Selected designs are then re-evaluated using TCAD.

A Defect in the P-Column Can Be Partially Compensated for

If the void is located in the p-column, part of the semiconductor volume is missing there, and thus the charge of ionized acceptors available for compensation. With the doping remaining unchanged, the ratio between the p- and n-columns becomes unbalanced. The breakdown voltage decreases.

The narrow PTF structure with a 2° sidewall angle illustrates just how significant this effect is: Without a cavity, it reaches a breakdown voltage of 1,470 V in the simulation. With a cavity and unchanged doping, the breakdown voltage remains at 1,173 V. If the p-doping is adjusted to account for the reduced semiconductor volume, the value increases to 1,324 V. The specific on-resistance remains at around 0.22 mΩ·cm² in the study’s table.

The adjustment restores the charge balance integrated over the entire column region. However, it does not eliminate the local disturbance along the column. Therefore, a loss of 146 V remains compared to the cavity-free reference. For wide PTF trenches, this remaining loss is significantly smaller: The study cites 15 V at 2° and 12 V at 5°.

The Current Path is Further Restricted in the N-Pillar

In NTF structures, the void also affects the conductive cross-section of the n-column. In addition to the breakdown voltage, this causes the on-resistance to deteriorate. For the narrow 2° structure, the breakdown voltage drops from 1,510 to 1,306 V after the nominal charge balance is restored. At the same time, according to the authors, the specific on-resistance increases by about 25%.

For the NTF designs examined, this study therefore favors a wide, inclined structure filled without voids. For PTF, on the other hand, even a narrow, nearly vertical trench with a void remains a possible design option: its low resistance can offset some of the disadvantage in terms of breakdown voltage. Which variant is suitable thus also depends on the required blocking capability.

Smaller Cavities are not Necessarily Less Harmful

Another finding is relevant to process development, as the cavity area alone does not adequately describe the electrical effect. In the narrow PTF structures, it decreases as the sidewall angle increases. Nevertheless, after the global charge balance is restored, the 8° variant causes a greater loss in breakdown voltage than the 2° variant: 254 V instead of 146 V.

As the angle changes, the shape, position, and distance of the cavity from the column flank change simultaneously. These parameters influence local charge compensation and field distribution. A process that creates smaller cavities does not necessarily result in components with more favorable electrical properties. For the evaluation of cross-sectional images, this implies that defect geometry and component structure should be considered together.

A Broader Doping Window Requires a Clear Reference

The simulations also show lower sensitivity to certain doping variations. The doping window reported in the paper is, in some cases, significantly widened by voids. The authors define the window as the range in which the breakdown voltage reaches at least 80% of the value at nominal charge balance.

This relative reference is crucial for interpretation. If the breakdown voltage of the reference state decreases, the 80% threshold is also lower. A wider window therefore does not necessarily indicate a higher production yield if the absolute reverse voltage requirement remains unchanged. To determine this, the permissible doping range would need to be evaluated against this specific requirement.

The practical value of this work lies primarily in the joint optimization of the process and the component. It quantifies the electrical drawbacks caused by a void and the potential drawbacks that can arise if the trench geometry is designed solely to avoid it. From this, candidates for further manufacturing trials and electrical tests can be identified. 

Bailing Zhou: „Simulation study of void impact in trench-filling-epitaxy-based 4H-SiC semi-superjunction schottky rectifiers“, Power Electronic Devices and Components, 2026, Artikel 100153.

Power Measurement Technology

When Static Measurements Fail in SiC, GaN, and AI Infrastructures