Semiconductor Technology from A to Z

Silicon has dominated power electronics for decades, but it runs into physical limits: at high blocking voltages (in high-voltage transistors), a Si device needs a thick, lightly doped drift region to safely support the electric field – which costs on-resistance and therefore losses. This relationship follows directly from the Poisson equation: the lower a material's critical breakdown field strength, the thicker the drift region must be designed for a given blocking voltage, and the on-resistance grows disproportionately as a result – roughly with the second to third power of the blocking voltage. In silicon this effect already becomes noticeable above a few hundred volts, and at several kilovolts the drift-region resistance dominates the total device resistance almost completely.

Compound semiconductors such as silicon carbide (SiC) and gallium nitride (GaN) have a significantly higher breakdown field strength than silicon – a direct consequence of their larger bandgap, which binds electrons more strongly and thus raises the threshold for impact ionization. This allows thinner, more highly doped drift regions at the same blocking voltage – and thus noticeably lower conduction losses combined with a smaller chip area. In practice this means: at identical blocking voltage and identical on-resistance, a SiC or GaN device can be realized on a fraction of the chip area of a Si device, or conversely deliver a markedly lower on-resistance at the same area. This material advantage runs as a common thread through the entire development of power electronics over the last twenty years and is the real driver behind the commercial breakthrough of both material systems.

A direct comparison shows clear differences: silicon has a bandgap of 1.12 eV, 4H silicon carbide reaches 3.26 eV, and gallium nitride 3.4 eV – both compound semiconductors are clearly wider-gap materials and belong to the class of so-called wide-bandgap (WBG) semiconductors. This larger bandgap not only affects the breakdown field strength but also allows higher junction temperatures: while silicon devices are typically operated at junction temperatures around 150–175 °C, 200 °C and above are common practice for SiC devices, since the intrinsic carrier density at a given temperature remains many orders of magnitude lower.

The critical breakdown field strength rises from about 0.3 MV/cm for silicon to roughly 2.8 MV/cm for SiC and 3.3 MV/cm for GaN. SiC additionally stands out with high thermal conductivity (about 4.9 W/cmK versus 1.5 W/cmK for silicon), which makes heat dissipation at high power levels considerably easier and allows more compact cooling concepts – an advantage not available in the same way for GaN, due to its lower thermal conductivity (roughly 1.3 W/cmK) and the typical heteroepitaxy on foreign substrates. GaN, in turn, reaches a very high electron saturation velocity, enabling especially high switching frequencies; combined with the high mobility in the 2DEG channel (see the GaN chapter), this results in switching frequencies in the MHz range that are practically unattainable with Si or SiC devices. Taken together, this yields a clear division of labor: SiC where high blocking voltage, high power and good heat removal are required, GaN where switching frequency and form factor take priority.

To objectively compare semiconductor materials for power switches, B. Jayant Baliga defined a figure of merit in 1989: BFOM = εr · µ · Ec³, formed from permittivity, carrier mobility and the critical field strength raised to the third power. It indicates how small the specific on-resistance can theoretically become at a given blocking voltage, and it builds on the dependence of drift-region resistance on breakdown field strength already discussed above: since Ec enters to the third power while mobility enters only linearly, the field-strength advantage of WBG materials dominates the result almost entirely.

Because the critical field strength enters to the third power, the material advantage of SiC and GaN has an especially strong effect: both exceed silicon by more than two orders of magnitude in the Baliga figure of merit. In practice this means that, at identical blocking voltage, a SiC or GaN device could theoretically be designed with a 100- to 300-fold lower specific on-resistance than a Si device. In reality this theoretical potential is not fully exploited – interface defects, immature fabrication processes and packaging limitations (see the following chapters) noticeably reduce the achievable real-world advantage. Nevertheless, the Baliga figure of merit remains to this day the central benchmark against which new semiconductor materials for power electronics are measured, and it explains why the industry is consistently switching to SiC and GaN despite higher substrate costs.

SiC plays to its strengths mainly at high voltages and power levels – for example in traction inverters for e-mobility or in industrial applications such as solar inverters, UPS systems and railway power supplies, where blocking voltages from 650 V to over 10 kV and high currents are required. SiC's high thermal conductivity also allows more compact heat sinks and higher power densities, which is an important selection criterion especially in applications with limited installation space, such as in vehicles.

GaN, on the other hand, excels at high switching frequencies and medium voltages, for example in chargers and fast-charging infrastructure, where compact form factor and low switching losses matter most. Because GaN devices are predominantly lateral in structure (see the GaN chapter) and commercially available devices have so far mostly reached up to the 650 V to 900 V range, their field of use typically remains limited to medium voltage classes – a range in which the high switching frequency delivers the greatest economic leverage, since it allows transformers, inductors and capacitors to be shrunk dramatically. This complementary positioning of SiC and GaN is no coincidence; it follows directly from the different material properties and fabrication routes of the two systems. The following chapters first cover crystal growth and substrates, then the devices of both material systems in detail.