What Is Silicon Carbide (SiC)? Why EV Inverters Use It

In everyday life: from abrasives to electric cars

SiC is not actually a new substance. In 1891, E. G. Acheson in the US found it by chance during experiments with an electric furnace. It is very hard, so it spread under the name "Carborundum." It became an abrasive, a material for grinding and polishing things (The Electrochemical Society).

SiC became widely known as a semiconductor thanks to electric cars. A teardown reported in 2018 found 24 SiC parts in the motor drive unit of Tesla's Model 3. They were made by STMicroelectronics. This unit is an inverter. It turns the battery's direct current (DC) into alternating current (AC) for the motor.

As of 2026, SiC is mainly used in these places.

• Electric car inverters: The benefit is largest in cars with high-voltage batteries. In 2021, Infineon released SiC parts for such cars. Hyundai of South Korea, which adopted them, said the driving range grew by more than 5% compared with silicon
• On-board chargers and fast chargers: SiC is used in the charger inside the car, which turns AC from a home outlet into DC for the battery. It is also used for the conversion inside charging stations (ROHM's materials)
• Solar power: SiC is used in the device that turns DC from the panels into AC for homes and the power grid. When Cree released its SiC MOSFET in 2011, this was one of the main uses it listed
• Railways: In Japan, for example, Mitsubishi Electric announced in 2014 an order for inverters made only with SiC, for trains of Odakyu Electric Railway. The example of Japan's N700S bullet train is in "What Is a Power Semiconductor?"

What kind of material is SiC?

SiC is a crystal of silicon (Si) and carbon (C) atoms in equal numbers, bonded in turn. The bond between silicon and carbon is very strong, so the crystal is very hard. On the Mohs scale, a common measure of hardness, it rates about 9.5. That is close to 10 for diamond, the hardest material.

SiC comes in many crystal forms with the same atoms but different arrangements. They differ in how each layer of atoms is shifted as the layers stack up. These "sibling" crystals are called polytypes. More than 200 are known for SiC.

Power semiconductors use 4H-SiC, which repeats the same stacking every 4 layers. The "4" stands for 4 layers, and the "H" for hexagonal (six-sided). A different arrangement also changes how easily electrons move. 4H-SiC is chosen because electrons move fairly easily in it, and the direction makes little difference.

How it works ①: Three numbers compared with silicon

Three numbers capture a material's character. "What Is a Power Semiconductor?" explains what each one means. Here we just line them up (from the materials of Toshiba Electronic Devices & Storage).

eV (electron volt) is a unit for the height of the step. MV/cm means "how many million volts per centimeter of thickness it can withstand." W/(cm·K) shows how easily heat passes through. Compared with silicon, SiC's step is about 3 times as high. It withstands nearly 10 times the electric push, and it carries heat about 3 times as well. The second number, the electric push, matters most. As the next section shows, it greatly changes the thickness of a key layer inside the part.

How it works ②: Holding back high voltage with a thin layer

Inside a power semiconductor is a thick layer that conducts electricity poorly on purpose. When the part is off, this layer holds back the high voltage. When it is on, current passes straight down through the layer. This layer is called the "drift layer."

The voltage it can hold is roughly the electric push it can withstand times the layer's thickness (this is simplified). With about 10 times the strength, the same voltage fits in about one-tenth the thickness. A simple calculation for 1,200 volts gives about 80 micrometers for silicon and about 9 micrometers for SiC. A micrometer is one-thousandth of a millimeter.

A thinner layer is not the only gain. A small amount of another element, an additive, is mixed into the layer to supply electrons that carry current. More additive makes the layer weaker against the electric push. But SiC is strong to begin with, so it can take much more. With a thin layer and many electrons, the resistance when the part is on drops sharply. This is the on-resistance. For layers that withstand the same voltage, SiC's is in theory several hundred times lower than silicon's (ROHM's explanation).

Lower resistance means less heat when the same current flows. With less heat, the cooling equipment can be smaller, so the whole device gets smaller and lighter. But the real upper temperature limit also depends on other materials, such as the package around the chip.

How it works ③: SiC MOSFET or silicon IGBT?

For the high voltage and large current that drive a motor, silicon has mainly used a transistor called the IGBT. A silicon MOSFET (a transistor switched on and off by voltage) would have too much resistance in its thick drift layer. An IGBT lowers the resistance by flooding the layer with two kinds of particles. They are electrons and holes (gaps left where electrons are missing).

The price is paid at turn-off. When an IGBT turns off, the current trails on a little until the leftover particles leave the layer. This is called tail current, and it causes a loss at every switch. So IGBTs are usually used for switching up to tens of thousands of times a second (tens of kilohertz).

With SiC, the drift layer is thin, so even a 1,200-volt-class MOSFET can have low resistance. A MOSFET carries current with electrons only, so it has no tail current. In January 2011, Cree of the US (now Wolfspeed) announced a 1,200-volt SiC MOSFET. The company calls it the "first commercial" one.

The table is a rough guide. In practice, engineers weigh the price of the parts against the money saved by lower losses. A SiC part may cost more. But if the cooling system or the battery can be smaller, the car as a whole can come out ahead. For tables of the differences, see "Si vs SiC" and "IGBT vs Power MOSFET."

Why SiC is hard to make and expensive

The crystal is grown from a gas

A silicon crystal is grown by pulling it up from melted silicon ("What Is a Silicon Wafer?"). SiC does not become a liquid at normal pressure. Instead, it turns directly from a solid into a gas. A solid turning into a gas without first becoming a liquid is called "sublimation."

So SiC is grown using sublimation. SiC powder, the raw material, goes in the bottom of a graphite container. A seed crystal is attached to the underside of the lid. The container is heated to 2,100–2,500 °C, and the powder turns into gas. On the slightly cooler seed, the gas turns back into a solid. It piles up in the same pattern as the seed. This is called the "sublimation method" (according to a 2024 paper).

This method is based on the seeded sublimation method shown in 1978 by Tairov and Tsvetkov of the Soviet Union. The crystal grows about 0.3–0.8 mm per hour. Pushing for faster growth tends to bring in defects and crystals with other arrangements (2024 paper). At this speed, growing just a few centimeters takes several days (our own calculation).

Flaws inside the crystal

A crystal grown from a gas easily picks up flaws in its arrangement of atoms. A big problem for a long time was a thin, tube-like hole running straight through the crystal. This is called a "micropipe." Parts built on top of one fail easily. Thanks to steady improvements, high-quality crystals now have fewer than 1 per square centimeter.

A flaw that still needs care is a misalignment running sideways along the layers of atoms. This is called a "basal plane dislocation." A SiC MOSFET contains a part that works as a diode. When current flows through it, electrons and holes meet and cancel out. The energy released can make this flaw spread.

The spread flaw becomes a plane where the stacking order of the layers has gone wrong. This is called a stacking fault. Stacking faults resist current, so the voltage across the part when it is on rises with use. One countermeasure is a layer that limits the harm of these flaws. It is placed between the base crystal and the layers of the part.

Very hard to cut and polish

The grown crystal is sliced thin and polished to a mirror finish to make wafers. SiC is nearly as hard as diamond. So even wires and grinding wheels coated with diamond grains take a long time, and the tools wear down. SiC is hard but also brittle. Its edges chip easily, and hidden damage forms easily below the surface.

On the polished wafer, the part's layers, such as the drift layer, are grown from raw materials in gas form at over 1,500 °C. This technique grows a thin crystal layer that follows the arrangement of the base crystal. It is called epitaxial growth, and here too, care is needed to reduce flaws.

Larger wafers are hard to make

For a long time, most SiC wafers were 150 mm across. Silicon's largest wafers are 300 mm, so that is one-quarter of the area. In July 2021, STMicroelectronics announced it had made 200 mm wafers for prototypes. A 200 mm wafer has almost twice the usable area of a 150 mm one. ST says it yields 1.8 to 1.9 times as many good chips.

Factories are also moving to 200 mm. Wolfspeed opened a 200 mm factory in New York State in April 2022. Infineon opened the first phase of a 200 mm factory in Malaysia in August 2024. In May 2024, ST announced plans to build a 200 mm factory in Italy (production is planned to start in 2026). In January 2026, Wolfspeed announced it had made a 300 mm single-crystal wafer, but volume production is still to come.

The boundary with the gate film

In a MOSFET, a thin, glass-like film sits between an electrode called the gate and the path for electrons ("What Is a Transistor?"). SiC can grow the same oxide film as silicon by oxidizing its surface. But at the boundary between SiC and the film, many "traps" that catch electrons tend to remain. These traps are called interface states.

With many traps, electrons in the path move less easily. Often they move one-tenth as easily as inside the crystal, or less. As the traps catch and release electrons, the voltage at which the switch starts to turn on also drifts little by little. This is the threshold voltage (2021 paper). Makers refine how they form the film and run test after test to confirm the parts last.

Industry trends: from crystal to finished part in-house

In SiC, the quality and supply of wafers decide the performance and price of the parts. So major makers have tried to do everything within their own groups, from growing crystals to making parts. ROHM bought SiCrystal of Germany in 2009, and ST bought Norstel of Sweden in 2019. In November 2021, onsemi also bought GT Advanced Technologies, a US company that grows crystals.

A major maker of SiC wafers and parts is Wolfspeed of the US. Cree, which also made LEDs and other products, narrowed its business to SiC and changed its name in October 2021. On June 30, 2025, it filed for Chapter 11 of the US Bankruptcy Code to reduce its debt. This process lets a company keep running while it sorts out its debts. It finished the process that September and announced it had cut its debt by about 70% (company announcement).

Infineon, Mitsubishi Electric, Toshiba, Denso and others also make SiC parts. Plans by ROHM and a Toshiba group company to expand production are covered in "What Is a Power Semiconductor?"

Common myths

By the numbers

SiC's strengths and weaknesses grow from the same root. Its atoms bond strongly, so it stands up to both a strong electric push and heat. That same strength makes it hard to melt and very hard. So it is difficult both to grow and to process.

Learn more

• Compare in a table: Si vs SiC, SiC vs GaN, IGBT vs Power MOSFET
• Terms for the next step: SiC (silicon carbide) (detailed explanation), 4H-SiC, step-flow growth, wide-bandgap semiconductor
• Try the numbers: the die per wafer and yield calculator lets you compare how many chips you get from 150 mm and 200 mm wafers
• Related articles: "What Is a Power Semiconductor?", "What Is GaN (Gallium Nitride)?", "What Is a Silicon Wafer?"