GaN on SiC: The Substrate Challenge

Wireless communication may have had its beginnings in the late nineteenth century, but it only really started a connectivity revolution in the 1990s, when commercial MOSFETS and RF Laterally Diffused Metal Oxide (LDMOS) circuits delivered the much-needed power, efficiency, and price point.

That revolution connected us globally through 2G, 3G and, more recently, 4G cellular links, carrying not just voice but large amounts of information that has already resulted in 1.5 zettabytes (ZB) of IP data being transferred annually among various types of devices (according to Cisco’s Visual Networking Index).

As all manner of devices begin to network, the need for speed is pushing the industry toward 5G – the next-generation telecom standard that promises greater bandwidth and reduced latency. Many 5G bands reside in frequency bands that are too challenging for traditional semiconductor devices based on silicon (Si).

GaN for high frequency

What RF LDMOS was to early cellular networks, gallium nitride (GaN) is to modern-day and high-frequency applications. Compared with gallium arsenide (GaAs), and Si LDMOS, GaN has long been known to offer advantages that are hard to beat:

1. Electron mobility is significantly higher than LDMOS. Increased electron mobility provides more gain at higher frequencies and better efficiency.
2. Significantly higher breakdown voltage compared with GaAs and LDMOS. The critical electric field for breakdown is estimated at over 4 MV/cm for GaN, 0.2 MV/cm for Si and 0.4 MV/cm for GaAs. GaN offers high reliability and ruggedness at supply voltages from 28V to 50V.
3. Reduced die size due to better thermal conduction.
4. High activation energy allows GaN devices to operate with less cooling and without costly and complicated structures for heat extraction.
5. High power density results in lower capacitances and higher impedances that are easier to match.
6. GaN HEMT characteristics lend themselves well to linearization techniques like predistortion.

The substrate challenge

The conversion to GaN is well underway, but there are technological challenges, including the difficulty in growing GaN epitaxial films. This is due to the difficulty in manufacturing native GaN substrates in a size and purity to be competitive for use in homoepitaxy.. So, another substrate – for heteroepitaxial growth – is needed.

The choice of that substrate material is critical. This material must not only have a high thermal conductivity but also low lattice mismatch with GaN.

While companies have manufactured GaN devices on Si, Silicon Carbide (SiC) and diamond substrates, only SiC currently best meets all requirements.

GaN on SiC or Si?

SiC’s closely matching lattice structure means that GaN epitaxy can be grown on it with lower dislocation density then other materials. This reduces leakage and improves reliability.

Si, on the other hand, matches neither GaN’s lattice structure nor thermal properties. This causes higher defect densities as well as reliability and manufacturability issues due to warping. To overcome these problems, companies have to use complex buffer layers and an overall more complicated epitaxial structure.

Since SiC, like GaN, has much better thermal conductivity than Si, it allows high power densities to be efficiently dissipated for realistic drain efficiencies, preventing the extreme channel temperatures that would result due to self-heating. The difference between the thermal conductivities of SiC and Si is enormous: 430 W/mK for 4H-semi-insulating (SI) SiC polytype against 146 W/mK for Si. This means that an equivalently powered GaN on silicon device has to be about 20% larger in die size versus the GaN on SiC counterpart in order to spread out the thermal load, even when the silicon substrate is heavily thinned.

Moreover, GaN on Si is susceptible to degraded performance as temperature increases compared with GaN on SiC. When the Si substrate heats up, its resistivity reduces significantly. A substrate with low resistivity equates to higher RF loss, negatively impacting RF performance.

Si substrates also have a higher parasitic capacitance than SiC substrates, which results in Si substrates having a more limited operating bandwidth.

The impact of GaN on SiC devices on the total costs, factoring in the higher efficiency, better high-frequency operation, smaller die size, and the savings in space and weight, make it a compelling solution over GaN on Si.

Wolfspeed meets the challenge

GaN on SiC has not only helped overcome the challenges to GaN manufacture but has also dominated the GaN market with significant penetration in 4G infrastructure due to its superior value proposition.

As the industry looks for 5G solutions, one vendor stands out for their extensive experience and market dominance in both SiC and RF GaN: Wolfspeed.

Wolfspeed ranks at the top in terms of the number of RF GaN related patents, with 118 patent families as of 2019, according to Yole Développement. This is nearly 23 percent more patent families than the second-place patent owner.

Find out more about GaN on SiC from Wolfspeed.