Power semiconductors: Si, SiC and GaN

Why EVs moved to SiC and chargers to GaN

Every inverter, charger and DC-DC converter chops current with switches that are either fully on or fully off, thousands to millions of times per second. A good power switch blocks high voltage when off, carries high current with little voltage drop when on, and moves between the two quickly. Its headline datasheet numbers are the blocking voltage (650-750 V parts for 400 V EV batteries, 1,200 V parts for 800 V batteries, leaving room for overshoot), the on-resistance Rds(on) or on-voltage Vce(sat), and the switching energies Eon and Eoff.

Two kinds of loss

Conduction loss is heat while the switch is on: P = I_rms²·Rds(on) for a MOSFET. Switching loss comes from each transition, when voltage and current overlap for a few nanoseconds to a microsecond. Each event costs roughly ½·V·I·t_transition, so P_sw = (Eon + Eoff)·f. Conduction loss grows with current and temperature; switching loss grows in proportion to frequency. Designers want high frequency because inductors, transformers and capacitors shrink as frequency rises, so the switch that wastes less energy per transition wins on size and cost.

Why silicon runs out of breath

To block voltage, a power device has a thick, lightly doped drift region that holds the electric field when off. Silicon breaks down at about 0.3 MV/cm, so a 1,200 V device needs roughly 100 µm of drift doped near 10¹⁴ cm⁻³, and that layer is resistive when the device is on. For unipolar devices, specific on-resistance rises with about the 2.5 power of blocking voltage (Baliga’s silicon limit). Superjunction MOSFETs such as Infineon’s CoolMOS bend that curve up to 600-900 V, but above that silicon needs a different trick.

IGBT versus MOSFET

The IGBT adds a p⁺ collector under a MOSFET structure. When on, it injects holes that flood the drift region with carriers (conductivity modulation), so a 1,200 V IGBT carries hundreds of amps at only 1.5-2 V. The price is paid at turn-off: the stored carriers must recombine, so current tails off for about a microsecond while the voltage is already high. That tail makes Eoff large and keeps IGBTs at roughly 2-20 kHz; IGBT traction inverters typically switch near 10 kHz. IGBTs also have a knee of about 0.7 V before current flows, which wastes power at light load, and their separate anti-parallel diode adds reverse-recovery loss. A MOSFET has no knee, no tail and a built-in body diode, but in silicon its drift resistance limits it to lower voltages.

Blocking 1,200 V: silicon needs about 100 µm of lightly doped drift region; 4H-SiC, with a 10× higher critical field, needs about 10 µm doped 100× more heavily, so its ideal drift resistance is hundreds of times lower. Substrates are drawn short.

Silicon carbide

4H-SiC has a band gap of 3.26 eV, about ten times silicon’s critical field and more than twice its thermal conductivity. For the same 1,200 V rating, the drift region can be about 10 µm thick instead of 100 µm and doped about 100× more heavily, so the ideal drift resistance drops by several hundred times. That makes 1,200 V SiC MOSFETs practical: unipolar, fast, no tail current, and rated for 175-200 °C junctions. Tesla’s Model 3, launched in 2017, was the first high-volume car with a SiC main inverter, using 24 SiC MOSFET modules from STMicroelectronics. With no knee voltage, SiC gains most at the partial loads where cars spend most of their time, commonly credited with several percent more range. 800 V platforms such as the Hyundai Ioniq 5 and Lucid Air need 1,200 V-class switches, and SiC has taken most new 800 V inverter designs. Its weaknesses are cost and maturity: substrates are hard to grow, wafers are moving from 150 mm to 200 mm (Wolfspeed opened the first 200 mm SiC fab in 2022), and gate-oxide reliability needs careful qualification.

Gallium nitride

GaN (band gap 3.4 eV) is usually built as a lateral HEMT on an ordinary silicon wafer. A thin AlGaN layer on GaN creates a two-dimensional electron gas with mobility near 2,000 cm²/V·s, without any doping. GaN HEMTs have tiny gate charge and no body diode, so no reverse recovery; they switch in a few nanoseconds and run converters at hundreds of kHz to a few MHz. That is why GaN took over compact USB-C chargers after Anker’s 2018 Atom PD 1 used a Navitas chip: higher frequency shrinks the transformer. Most GaN power parts are 650 V or 100-200 V because the lateral structure makes very high voltage harder.

Faster edges, louder EMI

Speed has a cost. SiC edges reach 20-50 V/ns and GaN about 100 V/ns, versus 3-10 V/ns for IGBTs. Every stray capacitance turns dv/dt into current (i = C·dv/dt): 100 pF from a motor winding to its frame at 50 V/ns carries 5 A spikes through bearings, shields and signal grounds. Loop inductance rings with device capacitance, and 10 nH at 5 A/ns gives 50 V of overshoot (V = L·di/dt). Designers trade speed against EMI with gate resistors, Kelvin-source connections, laminated busbars, snubbers and common-mode chokes.

Cost and suppliers

Silicon IGBTs remain the cheapest per amp; SiC dies cost roughly two to three times as much for a similar rating, but can save on cooling, battery and magnetics. SiC comes from STMicroelectronics, Infineon, Wolfspeed, onsemi and Rohm; GaN from Navitas, Innoscience, EPC, Infineon (which bought GaN Systems in 2023) and Texas Instruments.

Interactive: watch the losses move

Switching losses: Si IGBT vs SiC MOSFET

Amber is voltage, cyan is current, red is their product v·i, whose area is the energy lost in each switching event.

At 10 kHz, 25 °C, 100 A	Si IGBT	SiC MOSFET
On-state	Vce(sat) 1.50 V	Rds(on) 10.0 mΩ
Eon + Eoff per cycle	5.00 mJ	1.00 mJ
Conduction loss	150 W	100 W
Switching loss	50 W	10 W
Efficiency (of 20 kW)	99.01%	99.45%

Switching frequency10 kHzJunction temperature25 °CLoad current100 A

Time axis

Model: one 400 V half-bridge leg at 50% duty delivering 200 V × I. Each device carries I while on; the IGBT leg's diode drop is taken equal to its Vce(sat). Representative values for 650-750 V automotive parts, not a specific datasheet.

In practice

Practice

Losses in one inverter switch

A SiC MOSFET carries 100 A RMS and switches at 10 kHz. At 25 °C its Rds(on) is 10 mΩ and its Eon + Eoff is 1 mJ at this current. (1) Compute conduction and switching loss. (2) Repeat with the 150 °C Rds(on) of 14 mΩ. (3) Compare an IGBT with Vce(sat) = 1.5 V and Esw = 5 mJ. (4) At what frequency do the IGBT’s switching losses equal its conduction losses?

text

(1) P_cond = I²·Rds(on) = 100² × 0.010 = 100 W
    P_sw   = Esw·f      = 1 mJ × 10,000 /s = 10 W      → total 110 W

(2) Hot: P_cond = 100² × 0.014 = 140 W                → total 150 W

(3) IGBT: P_cond ≈ Vce(sat)·I = 1.5 × 100 = 150 W
          P_sw   = 5 mJ × 10,000 /s = 50 W            → total 200 W

(4) 5 mJ × f = 150 W  →  f = 30 kHz
    Above that, the IGBT spends more energy switching than conducting.

Check part (1) in the visualization: set 100 A, 10 kHz and 25 °C, and the SiC bars read exactly 100 W and 10 W. Then slide the temperature up and watch the conduction bar grow.

Real-world scenario

SiC inverter fails EMI after a drop-in device swap

A team moves a 400 V, 150 kW traction inverter from silicon IGBTs to SiC MOSFETs, keeping the existing power board, busbar and gate-driver design. The new inverter cuts losses roughly in half on the dynamometer, but the first CISPR 25 emissions pre-scan fails badly and the motor-position sensor misbehaves at high load.

Symptoms

Phase-node edges measure about 50 V/ns, versus about 5 V/ns with the IGBTs, with 150 V of overshoot ringing at about 60 MHz.
Conducted and radiated emissions exceed the CISPR 25 limit by 15-20 dB between 30 and 100 MHz, with peaks at the ringing frequency.
Resolver faults and CAN error frames appear only above about 200 A of phase current.

Interview questions

screenWhen would you choose an IGBT over a MOSFET?

IGBTs make sense at high voltage and high current with modest switching frequency, such as 1,200 or 1,700 V industrial drives switching at 2-10 kHz. Their bipolar conduction gives an almost flat on-voltage of about 1.5-2 V even at hundreds of amps, and silicon IGBTs are cheap and rugged. MOSFETs, especially SiC, win when you need higher frequency or better light-load efficiency, because they have no knee voltage and no tail current. In a 400 V EV inverter both work, which is why the choice there often comes down to cost.

screenCompute the losses: 100 A, 10 kHz, Rds(on) = 10 mΩ, Esw = 1 mJ.

Conduction loss is I²·R = 100² × 0.010 = 100 W. Switching loss is Esw·f = 1 mJ × 10,000 per second = 10 W, so 110 W in total. If the die runs at 150 °C, Rds(on) can be about 40% higher, which makes conduction 140 W. That is why I always read Rds(on) and Esw at the real junction temperature, not the 25 °C headline value.

screenWhy did Tesla move to SiC in the Model 3?

The Model 3, launched in 2017, was the first high-volume car with SiC MOSFETs in its main traction inverter, using 24 SiC modules from STMicroelectronics. SiC has no knee voltage, so at the light loads where a car spends most of its time it loses far less than an IGBT, which is commonly credited with several percent more range and less cooling. It also tolerates higher junction temperatures. The trade-offs were a more expensive device and more care with gate drive and EMI.

deepWhy do wide-bandgap materials allow thinner drift regions, and what does that buy you?

The critical electric field rises steeply with band gap: about 0.3 MV/cm for silicon and about 2.5-3 MV/cm for 4H-SiC. The drift region needs a width of roughly 2V/Ec to block a voltage, so SiC needs about a tenth of the thickness, and it can be doped about 100× more heavily. Drift resistance goes as W/(qμN), so the ideal specific on-resistance at 1,200 V is several hundred times lower than silicon's. That is why a 1,200 V SiC device can be a fast, unipolar MOSFET, while silicon needs an IGBT with conductivity modulation and a tail current.

deepYou moved a 1 MHz charger to GaN and see ringing and false turn-on. What is going on?

GaN switches in a few nanoseconds, so a few nanohenries in the power loop or gate loop cause big overshoot and ringing. GaN's threshold is low, around 1-1.7 V, so Miller current from a fast dv/dt through Cgd can lift the gate of the off device and cause shoot-through. I would shrink the power loop with tight bypass capacitors, use a Kelvin-source gate return and separate turn-on and turn-off resistors, and hold the off-state gate firmly at or below zero. Then I would remeasure with a high-bandwidth probe and a short ground spring, because a long ground lead creates ringing that is not really there.

Quiz

References and further reading

Where the ideas in this lesson come from, and what to read next. Books are the standard texts; papers are the original sources; standards are what employers work to.

Book
B. Jayant Baliga (2008). Fundamentals of Power Semiconductor Devices. Springer.
Read: The chapters on breakdown voltage and drift-region design.
Paper
B. Jayant Baliga (1982). Semiconductors for High-Voltage, Vertical Channel Field-Effect Transistors. Journal of Applied Physics 53(3), 1759-1764.doi:10.1063/1.331646
Read: The figure of merit that explains the SiC and GaN advantage in one equation.
Book
Ned Mohan, Tore M. Undeland and William P. Robbins (2003). Power Electronics: Converters, Applications, and Design. Wiley, 3rd edition.
Read: The chapter on power switches and their losses.