The IDTechEx report "Power Electronics for Electric Vehicles 2025-2035: Technologies, Markets, and Forecasts" assesses the advantages and potential challenges of using engineered substrates for electric vehicle power electronics, such as scalability, production volumes, and performance.
Engineered substrates could reduce the overall cost of SiC MOSFETs by decreasing the amount of monocrystalline SiC per device, and can increase the peak voltage of GaN HEMTs, offering advantages for use in electric vehicle traction. Furthermore, widespread availability will increase production volumes and reduce lead times, as epitaxial growth is slow and engineered substrates can be mass-produced. For GaN, new vertical devices will have higher power density, and for SiC, die areas will be reduced, also increasing power density. It should be noted that this is a general overview of what could happen based on analysis by IDTechEx. Source: IDTechEx
Traditionally, substrates are thin discs of single-crystal semiconductor material on which devices such as MOSFETs and IGBTs are grown. Since the epitaxial growth of the semiconductor is built upon this foundation, ideally the substrate should be the same crystal as the epitaxy (homoepitaxy) – for example, a silicon substrate for silicon devices, and so on. For silicon MOSFETs and IGBTs used in the power electronics of automotive modules, such as on-board chargers, traction inverters, and DC-DC converters, silicon substrates are the obvious choice. Global suppliers include SUMCO, Shin-Etsu, and GlobalWafers, and the bottom line is that this is a mature, reliable, and cost-effective technology.
Increased volumes of SiC MOSFETs with engineered substrates
The SiC wafer market has recently begun to globalize, and approximately 75% of SiC wafers will be produced in the United States starting in 2023. With companies like Wolfspeed and STMicroelectronics expanding in Europe and China, the emergence of other SiC wafer companies, and the transition from 150 mm to 200 mm substrates, the market is expected to become more competitive, which will reduce costs.
Alongside these advances, companies are creating engineered substrates for SiC power devices. These generally leverage silicon-on-insulator (SOI) technologies and consist of a layer of polycrystalline SiC bonded to a thin layer of monocrystalline SiC (a SiC wafer is typically made entirely of monocrystalline SiC). Companies utilizing some variant of this technology include Soitec, which has already announced collaborations with X-FAB and STMicroelectronics, and Sicoxs (a subsidiary of Sumitomo Metal Mining), which currently markets 6-inch engineered SiC substrates and anticipates mass production by fiscal year 2025.
With the globalization of the SiC substrate market and an increasing number of players, automotive original equipment manufacturers (OEMs) are seeking large volumes of SiC MOSFETs to create efficient powertrains, which can increase a vehicle's range by up to 7% compared to Si IGBTs. In terms of volume and cost, engineered substrates offer advantages and can lead to production volumes an order of magnitude greater than without engineered substrates using the same amount of monocrystalline SiC. To create engineered substrates, a mechanically strong polycrystalline SiC wafer is prepared, with steps including grinding and polishing to achieve a flat surface. A very thin layer of monocrystalline SiC substrate is then bonded tightly to this polycrystalline wafer. After processing, activation, and inspection, the engineered substrate is ready, and the remaining monocrystalline wafer is reused to manufacture more engineered substrates.
The result is a reduction in costs, since SiC substrates account for up to 50% of the total cost of a device, and an increase in volume, as a single SiC monocrystalline wafer can now produce between 10 and 50 substrates. Engineered substrates have also been shown to have lower "connection resistance" across the entire substrate surface, resulting in less heat generation per unit area and the ability to produce a greater number of chips per substrate.
Although engineered SiC substrates have potential, they are still a couple of years away from mass production, and questions remain about their primary commercialization strategy. For example, companies could license the bonding process, partner with foundries or device manufacturers to produce substrates at scale, or lend the necessary equipment. Design cycles and lead times in the automotive industry are notoriously long, spanning several years. Furthermore, for these processes to be suitable for manufacturing devices for electric vehicles, very stringent requirements and extensive testing are needed.
Substrates Designed to Make GaN Viable in Traction Inverters
: With SiC already used in electric vehicle power electronics, the mass commercialization of engineered substrates would help accelerate that development. In the case of GaN, several companies, such as Vitesco and VisIC Technologies, want to launch high-voltage GaN products for automotive applications, but the obstacles are slightly different. IDTechEx is aware of only one company, Odyssey Semiconductor, that produces bulk GaN substrates for GaN devices. In May 2024, it was acquired by Power Integrations, but it remains to be seen whether GaN substrates can be manufactured reliably and cost-effectively.
Historically, silicon substrates have been used to grow GaN devices (heteroepitaxy), and almost all devices on the market use silicon substrates. Since silicon substrates are already mass-produced for silicon-based devices, their use for GaN transistors seems obvious. Buffer layers manage distortion and lattice mismatch between the epitaxial GaN and the Si substrate, improving robustness. However, GaN can only be grown in thin layers on Si substrates, resulting in higher defectiveness than in homoepitaxial GaN and voltage-limited lateral GaN devices.
For GaN to be used in traction inverters, devices must be able to maintain high voltages and power levels for extended periods, which is where current GaN technologies fall short. Alternatives include GaN on sapphire and GaN on SiC, which can produce higher-voltage devices but are incompatible with CMOS processes, are expensive, and present other handling and thermal performance issues. Engineered GaN substrates mitigate these problems by creating a substrate that minimizes grid mismatch and aligns the coefficient of thermal expansion with that of GaN to produce thicker epitaxial GaN layers for vertical devices exceeding 1200 V. This is the approach taken by the factoryless company Qromis, with its QST (Qromis Substrate Technology). By tailoring the substrate's physical properties to GaN, the devices will have low defect rates and minimize breakage. Furthermore, while current GaN devices are grown on 150 mm Si substrates, the engineered substrates are scalable, allowing GaN to be grown on 200 mm substrates and, eventually, 300 mm substrates. Over time, this will increase production volumes and reduce the cost per device.
Given that GaN substrates are currently prohibitively expensive, the best approach is to manufacture something as close to GaN as possible, which is the goal of engineered substrates. Si substrates are cost-effective if large volumes are guaranteed, but they struggle to meet the demands of high-voltage power electronics with current technology. Alternative substrates, such as SiC and sapphire, have other drawbacks, so engineered substrates could be a key technology for enabling GaN to enter the electric vehicle power electronics market. However, IDTechEx is unaware of any adoption plans by automotive original equipment manufacturers or Tier 1 suppliers.
Author: John Li, technology analyst at IDTechEx
