The Wide Bandgap Challenge

About a year ago, I had a great idea. Set up a blog to go with our website. Fast forward to now and we’ve managed to post just one article – and that was at the start. The last year has been a blur of fund-raising efforts and getting technology programmes up and running – both have been successful and we are full speed ahead.

During the fund-raising effort, a question that was asked time and time again was why do we think diamond can make an impact when so much investment has already been made in other materials, in particular silicon carbide (SiC) and gallium nitride (GaN)? The glib technical answer is just look at the respective physical properties of the materials and make your own conclusions. However the real answer is more complex and takes some time to explain.

For a start, the real competitor is silicon. It has become a commodity market. The base material is manufactured using a liquid process and wafers up to 450mm diameter can be processed – that’s a lot of devices on a single crystal piece of silicon. Also despite people attempting to write silicon off with regular monotony for the last 20 years, it has constantly found ways to push the material to its absolute limits. But now, finally, consensus is that 2021, or so, will see feature sizes finally hit the buffers of what is possible – however already thoughts are moving to things like vertical architectures, 3D architectures and hybrid approaches with 2D compounds and the like.

All this should be happy days for those looking beyond silicon at electronics based on materials that simply by replacing silicon with them could yield an order of magnitude improvement in performance. But here’s the rub. The fact is that the majority of the electronics market is a commodity market relying on the economies of scale. Anyone trying to compete with this has to either have something that is so compelling on performance that people will happily pay a premium or it can be made more cheaply than silicon is today (and ideally deliver the former as well).

Let’s look at the second point first. Just about all wide bandgap semiconductor materials are produced from the vapour or gas phase – this is significantly more energy intensive than is required to produce silicon and is at least an order of magnitude slower to grow. So the substrate materials are themselves expensive – although this is somewhat compensated by the fact that you generally need less wide bandgap material to make devices. You also then need to process that material to make devices. Here wafer size does matter, the cost of putting a 300mm or 450mm diameter wafer through a fabrication process is not significantly more than the cost of putting a 100mm or 200mm diameter wafer through a process (notwithstanding the eye watering costs of the infrastructure as you scale up).

Going back to the first point on performance, you also need to outperform silicon. That devices based on SiC and GaN can deliver this is not in dispute. However they have yet to yield the performance advantages that justify their current premiums, except in specialised applications where their use can yield overall savings in the bigger picture. Put simply, it is now possible to make a laptop supply that would be a quarter of the size of the one you currently use and it runs significantly cooler too, but would you pay a 3 to 4 times premium for it?

This is the problem that GaN and SiC now face. The devices that are being made are largely competing with ones where there is an equivalent in silicon that is 10%-20% of the cost to produce. Both materials have struggled to deliver a compelling case, and the outcome is huge investments in infrastructure to try to achieve the economies of scaled mass production in order to compensate for the fact that they can only deliver a good but not compelling case for replacing silicon. And their current roadmaps don’t look like they are going to deliver something that will generate clear blue water over silicon anytime soon either.

Already there are signs that both material technologies are consolidating. Take Infineon’s recent acquisition of Cree’s SiC power electronics business and International Rectifier for GaN devices (and complementary silicon devices), and Transphorm’s effective acquisition of Fujitsu’s GaN interests. Tacit admission that there is limited leeway for both these materials to make a dent into silicon?

When you crunch the numbers GaN, SiC and diamond at scaled production are, carat for carat, not that dissimilar in cost to produce. So the question boils down to how much starting material you need; how expensive is the manufacturing process; and, most importantly, is there a compelling performance advantage over silicon?

In the case of (single crystal) diamond, we need around a one fifth of the material that GaN and SiC need to achieve the same rating. The technology that Evince is developing has an inherent elegant simplicity that has less than half the manufacturing steps needed to make a conventional semiconductor device. Finally, where we are heading is towards a family of devices and applications that go where silicon, GaN and SiC presently just can’t go. This is where diamond will achieve early success and sustainability as it makes the journey towards commercial mass production.

Is the diamond electronics revolution is coming? You bet!