MBE: A game changer for GaN-on-silicon RF

On the back of remarkable results at 40 GHz, GaN-on-silicon HEMTs have a tremendous opportunity to serve in mobile applications requiring high-power millimetre waves.

BY ANDRE BONNARDOT, ELODIE CARNEIRO AND STEPHANIE RENNESSON FROM EASYGAN AND FABRICE SEMOND FROM UNIVERSITÉ CÔTE D’AZUR, CNRS-CRHEA

Do you think that GaN-on-silicon will be adopted for RF applications? If you asked that question a couple years ago, the answer could well have been along the lines of a cautious yes, tempered by the point that it’s hard to know when GaN-on-silicon will have an impact, and to what extent. And there was very good reason for this caution: it is challenging to produce GaN-based heterostructures on silicon with sufficient material quality in a reproductible, reliable manner.

Helping spur the development of GaN-on-silicon RF devices are decisions related to the future of the wireless industry. In March 2022, the 3GPP (3rd Generation Partnership Project) standardisation initiative propelled the 5G millimetre wave spectrum limit to 71 GHz, exacerbating the technical challenge for device manufacturers. The reality is that while it is not going to be easy, deploying millimetre-wave bands is inexorable – that’s because higher frequencies are essential for supporting the exponential increase in data traffic created by future 5G advanced networks.

The inevitable march of millimetre-wave standardisation will be challenging for pico-cell and handset vendors. They will be constrained when designing and integrating RF front-ends operating at those frequencies. In turn, this will create a new, high volume, cost sensitive RF device market opportunity for semiconductor manufacturers.

A great candidate for providing the high-power densities at high operating frequencies that’s demanded by this application is the GaN HEMT. The most suitable form of this transistor is the GaN-on-silicon HEMT: it is the technology of choice for GaN electronics, and it is already gaining momentum in power electronics applications.

Today, however, the GaN-on-SiC form of the HEMT is the mainstay in RF applications. Compared with the silicon substrate, SiC provides a superior thermal conductivity as well as more compatible lattice and thermal expansion mismatches (see Table 1). These assets make it easier to produce a device delivering high power at high frequencies. But there are drawbacks, in terms of cost, CMOS compatibility, supply availability and large diameter feasibility. Due to this, there is no question that the GaN-on-silicon HEMT will rise to the fore for high-volume deployment, once technological barriers have been removed.

At EasyGaN, a French start-up founded in 2017 by researchers of CRHEA, a CNRS laboratory that pioneered the field of GaN hetero-epitaxy on the silicon substrate, we are helping to accelerate this switch to GaN-on-silicon. Read on to discover how we are using MBE to transform the capability of this class of GaN HEMT in the millimetre-wave domain.