Semiconductor Technology from A to Z

The traction inverter is the power electronics component that most directly determines an electric vehicle's range and efficiency, since it converts the entire drive current from the battery to the electric motor and must operate with as few losses as possible across all driving states – from low-speed starts to high-speed cruising. SiC MOSFETs have become the standard here over Si IGBTs, because their lower switching losses allow higher switching frequencies – this shrinks passive components (inductors, capacitors) and typically raises efficiency by 3 to 5 percentage points over the drive cycle.

This seemingly small efficiency gain has a substantial practical impact on range, since it applies across the entire drive cycle, especially in part-load-heavy city driving, where inverter losses account for a disproportionately large share of total energy consumption. In 800 V system architectures, which are increasingly becoming standard, the SiC advantage grows even further, since the higher breakdown field strength matters most at higher voltage classes: an 800 V system requires markedly thicker, more lossy blocking layers with Si IGBTs, while SiC MOSFETs handle the voltage step with comparatively little extra chip area. In addition, SiC's higher junction temperature allows for a more compact, lighter cooling system, which in turn indirectly improves overall vehicle efficiency – an effect easily overlooked when looking only at inverter efficiency in isolation.

For applications with medium voltages (48 V to a few hundred volts) and a need for a compact form factor, GaN plays to its strengths: chargers for laptops and smartphones use the high switching frequency of GaN HEMTs to drastically shrink transformers and filter components – a 65 W USB-C charger today fits into a housing that, just a few years ago, was needed for a third of the power. The size of a switch-mode power supply scales roughly inversely with switching frequency, since the transformer core and storage inductor need fewer turns and a smaller core cross-section at higher frequency; GaN unlocks frequency ranges in the high hundreds of kHz to low MHz that would not be economically achievable with Si devices.

DC fast chargers for electric vehicles are also increasingly adopting GaN in their front-end AC/DC stages, where the compact design helps make charging stations smaller and lighter without sacrificing the required efficiency in the range of 95 to 98%. GaN is also gaining traction in data centers, for example in highly efficient server power supplies, where every percentage-point improvement in efficiency translates directly into substantial energy and cooling cost savings given the enormous total power involved. This range of applications – from the small USB-C charger to the megawatt charging station – shows how universally GaN's frequency advantage is being leveraged across modern power electronics.

Roughly speaking: Si IGBTs remain the most cost-effective option today for low switching frequencies (below 20 kHz) at very high power levels, for example in industrial drives or megawatt-class railway inverters, where material cost per chip area is a decisive economic factor and the low switching frequency makes the IGBT's higher switching losses uncritical. SiC MOSFETs dominate the medium-to-high power range with frequencies reaching into the hundreds of kHz, for example in solar inverters, traction drives and industrial power supplies, where both high blocking voltage and moderately high switching frequency are required.

GaN HEMTs unlock MHz frequencies at lower to medium power levels, where form factor is the dominant design driver – for example in consumer chargers, telecom power supplies and, increasingly, in the drivetrains of smaller electric vehicles and e-bikes. These three technologies do not overlap sharply but rather form a continuum, within which the choice of the optimal device depends on the specific combination of target voltage, target power, desired switching frequency and cost constraints; in many modern systems, several technologies are even combined, for example a SiC main switch paired with a GaN auxiliary stage for particularly fast control loops.

Despite their clear technical advantages, SiC and GaN devices remain more expensive than comparable Si solutions – driven mainly by substrate costs and lower manufacturing volumes. The elaborate crystal growth processes described in the previous chapters (PVT for SiC, heteroepitaxy for GaN) are inherently slower and more expensive than the established Czochralski process for silicon, and the smaller wafer sizes available so far prevent the economies of scale that Si manufacturing has benefited from for decades.

That price gap, however, is shrinking steadily: 200 mm SiC substrates and larger GaN-on-Si wafers lower per-unit costs, while rising volumes in e-mobility and consumer electronics create economies of scale that affect both substrate production and downstream device processing in existing semiconductor fabs. For many applications, the break-even point over the system's lifetime (lower cooling costs, smaller form factor, higher efficiency) has already been reached today, even though the raw component price is still higher – a total-cost-of-ownership comparison therefore already favors SiC or GaN in many cases today, even where the isolated component price would not suggest that at first glance. As manufacturing volumes continue to grow, wafer sizes increase and processes mature, component prices are expected to converge further toward Si levels in the medium term, though a complete cost parity is unlikely to be reached even in the long run, given the inherently more elaborate crystal growth involved.