Do the Significant Performance Benefits of SiC MOSFETs Warrant the Retirement of the IGBT?

Operating in ranges over 600 V and handling high currents (from 100 A to 3,600 A), silicon IGBTs (insulated gate bipolar transistors) have long been an essential component in high-power electronics.

IGBTs were developed specifically to address the limitations of silicon MOSFETs in these high-power devices. And since the device was first demonstrated in 1982, the IGBT has been used in a vast number of applications.

According to estimates (admittedly by the IGBT inventor, B Jayant Baliga) the technology has cut electricity usage during its operation by 75,000 TWh.

However, recent years have seen innovations in silicon carbide (SiC) MOSFETs and the technology offers compelling advantages in efficiency, thermal management, power density, and overall performance versus silicon IGBTs.

These benefits are due to the physical properties of SiC, with silicon carbide having a bandgap roughly three times that of silicon (3.26 eV versus 1.12 eV), with higher breakdown voltages and lower on-resistance. SiC has a higher electron saturation velocity and thinner epitaxial layers, leading to faster switching speed. It also has a higher thermal conductivity, making it more robust at high temperatures.

So, with these advances, and with the IGBT now in its fifth decade, is it time for the IGBT to be retired?

Switching losses

While efficient at low frequencies, IGBTs are bipolar devices and therefore suffer from slower switching and tail current conditions that result in higher turn-off losses compared to SiC MOSFETs.

IGBTs have achieved widescale adoption as a result of their lower on-resistance than silicon MOSFETs (note, silicon, not SiC), which is achieved through conductivity modulation from injecting holes (minority carriers) into the drift region. However, this injection of minority carriers does lead to tail currents during turn-off, and this creates significant switching losses.

In contrast, the drift layer impedance of SiC devices is much lower than that of Si devices. This allows SiC MOSFETs to achieve high breakdown voltages and low on-resistance without the need for conductivity modulation. Since MOSFETs do not generate tail currents in principle, replacing IGBTs with SiC MOSFETs can significantly reduce switching losses.

Fig 2: A comparison of switching losses of IGBTs and SiC MOSFETs during the turn-on and turn-off phases

From figure 2, we can see that the turn-on speed of SiC MOSFETs is comparable to that of silicon IGBTs: typically measured in tens of nanoseconds. And even though the forward voltage of the body diode in SiC MOSFETs is relatively high, it exhibits a fast recovery capability, which allows for a significant reduction in turn-on losses versus a silicon IGBT (3.3 mJ vs 10 mJ for the conditions stated in figure 2).

The same can also be seen when we look at the turn-off losses. Indeed, a key advantage of SiC MOSFETs lies in the turn off process and the lack of the tail current, which as stated is commonly observed in IGBTs, and increases with temperature.

This characteristic allows SiC MOSFETs to maintain their fast-switching capabilities even at breakdown voltages above 1200V and turn-off losses can be reduced by approximately 90% compared to IGBTs.

Conduction Losses

IGBTs have a threshold voltage that causes a constant voltage drop across the device during conduction. This leads to higher power dissipation, especially at lower currents.

With an insulating breakdown field strength that is roughly an order of magnitude higher than for silicon, MOSFETs based on SiC exhibit lower conduction losses compared to IGBTs as well as higher breakdown voltages with low impedance and a thinner drift layer.

For the same voltage rating, SiC devices can achieve a much lower specific on-resistance per unit area, which enables both a lower on-resistance in a compact package as well a reduced gate charge and junction capacitance.

For context, SiC devices can easily achieve breakdown voltages above 1700 V with very low on-resistance and this lower on-resistance scales more favorably with current, leading to lower conduction losses.

Thermal Management – and Power Density / Component Size

SiC has a thermal conductivity of 4.9 W/mK, which compares with 1.3 W/mK for silicon. This makes SiC MOSFETs significantly more robust at high temperatures.

IGBTs typically operate at junction temperatures up to 150°C and they also require extensive cooling solutions to maintain reliability and prevent thermal runaway. This need for robust thermal management often results in larger and more complex cooling systems, which can increase system size and limit the overall power density (see section 4).

And while it should be said that thermal runaway can also occur in SiC MOSFETS, this happens when the gate-source voltage is too low (less than 13 V) and the MOSFET is therefore at risk of not being fully turned on. To fully exploit the inherently lower on-resistance of SiC, the device’s gate-source voltage should be at least 18 V.

To compare, SiC MOSFETs can operate at junction temperatures in excess of 200°C. This higher thermal tolerance allows for higher power densities, with fewer and smaller heat sinks as well as reduced cooling requirements, advantage is particularly beneficial in applications such as data centers, aerospace and automotive, where space and/or weight are critical design constraints.

Switching Speed

As per the above, the turn-on and turn-off energies are significantly reduced for SiC MOSFETs, leading to lower switching losses/heat generation.

These characteristics enable the devices to operate effectively at higher switching frequencies, up to several hundreds of kHz, without significant heat buildup. In contrast, the heat emitted as a result of switching losses in IGBTs means they typically cannot operate efficiently at frequencies above 20kHz.

Reliability

On reliability, the devices can be said to be equivalent.

it should be noted that SiC’s gate oxide breakdown reliability is still a concern, however, SiC MOSFETs exhibit higher breakdown voltages and can withstand greater electric fields. Indeed, recent advances have reliability in SiC MOSFETs improve greatly and seen their intrinsic lifetime now matches that of the IGBT.

Cost

As we’ve seen, SiC MOSFETs are more efficient, have better thermal properties and thus systems using them require fewer, smaller and less costly heat sinks. The devices also enable higher switching speeds and (due to recent advances) now have a similar intrinsic life versus IGBTs.

Where IGBTs do outperform SiC MOSFETs is on cost. Being over 50, the IGBT is an exceptionally mature technology and (before heat cooling is considered) is less expensive per device. A 2021 forecast (to 2030) by the PGC Consultancy placed the die cost of SiC MOSFETs at roughly 3X versus the IGBT in 2021.

2022 $1,794M

2028 $8,906M

However, prices of SiC MOSFETs are falling. As the forecast states, economies of scale, generational advances, and (of particular note) the move from 150 mm to 200 mm wafers, will see the price of SiC MOSFETS fall significantly.

As the report stated: “The cost of a 1200 V/100 A MOSFET die made on a 200 mm substrate in 2030 could be 54% less than the cost in 2022, from a 150 mm substrate”.

Conclusion

We began this article asking if the significant performance benefits of SiC MOSFETs warrant the retirement of the IGBT. The answer is: not quite. But it won’t be long.

IGBTs may not yet be quite ready for retirement, but they have already been relegated to applications where cost outweighs the need for cutting-edge performance, for example home appliances, AC motor drive inverters and sub 20 kHz power supplies.

However, the list of applications is shrinking as the SiC generational improvements occur and the cost falls. As such, EV powertrains, solar inverters, data centers and power components in almost any aerospace application, where weight is vital, have already seen the IGBT swapped out.

So, is the IGBT ready for retirement? No. Perhaps the better question is: With the IGBT’s career lasting more than 40 years, retirement is nearing… will this happen before its 50^th anniversary in 2032?