Diversification lies at the heart of Kyma Technologies’ vision for its future. It first made a name for itself as a leading supplier of wide bandgap materials, but it is now expanding its offerings and has starting to provide plasma vapour deposition (PVD) equipment and photoconductive switches, explains the company chief executive officer, Keith Evans.

When you hear the name Kyma Technologies, I bet you think of bulk GaN. But that's not the only product developed by our team at Raleigh, NC. In addition to being a provider of high-quality HVPE-grown GaN materials, we offer a diverse and growing portfolio of products, including materials, equipment, and devices.

Three of our most important product offerings are discussed in this feature:  AlN templates for LEDs and power electronics; PVD crystal growth systems for fabricating AlN templates; and a novel, photoconductive GaN switch for optically isolated, rapid switching at high powers.

AlN templates

The first of these, our AlN templates, consist of a thin layer of this crystalline wide bandgap material, grown on either sapphire or silicon. AlN deposition is carried out with a patented, proprietary process called PVDNC - PVD of NanoColumns.  This creates AlN crystalline layers with a defined nanostructure, consisting of very dense array of AlN nanocolumns with a dislocation density that is very low – and possibly zero. According to atomic force microscopy and cross-sectional transmission electron microscopy measurements, the typical diameter for these nanocolumns range from 40 nm to 80 nm, and they have a very smooth surface with low root-mean-square roughness. Between the nanocolumns, which are oriented along the c+ axis, the AlN is of lower crystalline quality.

4-inch, 6-inch and 10-inch diameter AlN on sapphire templates produced by Kyma Technologies using their patented and proprietary PVDND technology.

We routinely deposit our AlN nanocolumns on both flat and patterned sapphire substrates. The AlN templates that result enable our customers to form GaN with higher quality earlier in buffer layer than they would get if they uses just sapphire. Realizing a better buffer earlier brings two major benefits: Producers of epiwafers and devices can thin their GaN buffer, saving time and ultimately bosting throughput and reducing cost; or engineers can maintain the thickness of the buffer, leading to a lower defect density in the buffer and device active regions. 

Our collaborators at John Muth’s group at North Carolina State University (NCSU) have documented the improvement in thermal conductivity in GaN as the defect density is reduced. That means that our PVD AlN template customers attain higher thermal conductivity GaN buffer layers faster during their buffer layer growth. Since the device active region for most devices replicates the structural quality of the top portion of the buffer layer, we can conclude that the entire device wafer and especially the device active region has lower thermal impedance, always a desirable attribute for a device epiwafer. . The higher thermal conductivity of the GaN buffers and device active regions resulting from our AlN templates have enabled our customers to report higher LED yield and better device performance.

It wasn’t obvious that our PVDNC process would work on patterned sapphire. However, it shouldn’t be that surprising that it does. After all, although today’s patterned sapphire substrate topologies contain micron or many-micron sized features, they are on a much larger length scale than our nanocolumns. The way we like to phrase this is that our PVDNC process puts a nanostructure onto the microstructure of a patterned sapphire substrate.

We have strong IP covering our growth technology, which is based on the formation of nanocolumnar III-N layers by PVD. When NCSU researchers Gerald Cuomo and Mark Williams founded our company, they brought their PVD technology with them. NCSU then licensed key patents to us, and we have strengthened our IP with additional patents – together they cover several important aspects of growth technology and equipment design relating to AlN PVD.

For many years, we kept our PVDNC AlN template growth process under wraps. During that time, we learnt and refined how to use it to improve our HVPE process for growing GaN. As with MOCVD growth of LEDs, defect density falls and film uniformity increases when HVPE-grown GaN layers are formed on top of our AlN template. What’s more, with this approach we don’t have to perform a two-step buffer layer growth, making our process faster while simultaneously providing added IP protection for our HVPE GaN growth processes.


Kyma’s PVD growth systems.

Marketing templates

When the market for sapphire substrates for LEDs took off around 2005, we decided to launch our PVD-grown AlN templates. This was our first addition to our bulk-GaN products, and it marked the beginning of our transition to a more diversified product portfolio.

At the outset, LED manufacturers were very reluctant to evaluate our material. A stigma surrounded PVD growth for epitaxial growth applications, and it did not help matters that our AlN layer characteristics were markedly different from most of the MOCVD-grown AlN-on-sapphire materials being used in many of the commercial LED recipes. But persistence on our part paid off, and a handful of companies in Taiwan and the US gave them a try.

Initially, these templates did not work well for the chipmakers, because the AlN layer was too thick. So we devoted time and effort to optimizing the thickness of AlN and its growth conditions, working in partnerships with a few companies. Although each firm settled on a different optimal thickness, their preferred values were quite similar. That didn’t surprise us, but we were caught off-guard when we found that these LED makers had their best results using buffer layers that were far thinner than the ones we used in-house for our own trials. We quickly migrated to these thinner AlN buffers, and very positive reports followed from our customers.

We have subsequently mulled over possible explanations behind the benefit of employing a thinner AlN buffer. We cannot divulge many details regarding our conclusions, but we can reveal that the benefits are not limited to LEDs, and extend to any application involving MOCVD growth of a GaN buffer layer on top of the template. 

A great strength of the PVDNC process is its scalability. Until about two years ago, we carried out our growth on small-scale PVDNC equipment built in-house a decade ago. However, as our sales started to climb, we needed more capacity and a more robust platform, so we talked to commercial PVD tool suppliers.

Our conversations with them were not that helpful. None had ever built a PVD tool operating in the growth window that our process required, and after attempts at forming creative partnerships, we concluded that it would be best to go it alone. This has yielded significant success: Our homebuilt, higher-capacity tool may not be fully automated, but it has produced thousands of templates, and it has also enabled us to demonstrate growth of AlN on 12-inch silicon and on 10-inch sapphire. We are now in the process of further automating this tool, as well as actively developing plans for a larger platform.

PVD tools

As we worked hard to increase our AlN template sales, we found that many customers wanted to make these engineered substrates in-house. Often they asked if we could supply them with PVD tools. Deciding if we would fulfil their wishes was very tricky, due to our heritage as a material company that employed proprietary growth tools and refined deposition technologies to give ourselves a competitive edge. Could we allow all this knowledge to enter the market place? Our view was that we could. We believe that it is important for us to listen to the views of our customers and act in their interests, and we fully recognize that expansion into the equipment market is an excellent commercial opportunity for us.

Process engineers that use our equipment discover that growth times are very short, thanks to a fairly fast growth process and the need for a relatively thin buffer of AlN to ensure optimal device performance. We see PVDNC growth cycle times shrinking to below 20 minutes as we mature the equipment platform. Thanks to these short deposition times, the addition of AlN costs much less than the purchase of sapphire.

Our first tool to market can be configured in many ways: 19 x 2-inch, 3 x 4-inch, or single wafer growth on substrates with diameters ranging from 6–inch to 12-inch. We are now working on a bigger tool, which we expect to support multiple 6-inch wafers and have the capacity to produce more than 40 x 2-inch wafers in a single growth run.

Both tools are highly automated, with cassettes and multi-wafer platters enabling up to 24 hours of growth on dozens of platters of wafers without operator intervention. Market response to these products is positive, and we are now evaluating the outsourcing of manufacturing, in order to support anticipated hike in shipments of dozens of PVDNC growth systems during the next few years.

Kyma major breakthroughs in materials, growth tools and products.

From materials to devices

Last September, we passed another important milestone – the launch of our first device. It’s a photoconductive semiconductor switch (PCSS), and its most important attribute is that it is the highest power semiconductor switch of its size with a sub-nanosecond response time. When naming this switch, we intentionally went for a double entendre: ‘Kyma Optical’ and ‘Knock Out’. The latter is deserved, because we have knocked out some of our own electronics when scrutinizing it in our high-speed switch testing facility!

This happens because when high electrical power is switched at such speed, electromagnetic radiation is generated with frequencies above 1 GHz. We are still learning about the details of the waveform but expect that a broadband pulse is generated with frequencies as high as 10GHz or higher. Such an electromagnetic pulse can induce currents in nearby electronics. If the induced currents are high enough, the equipment can fail irreversibly.  At lower currents or durations, the equipment may just need a reboot, which we have done many times.

Our motivation for developing the KO-Switch can be traced back as far as 2006, when we read a paper written by scientists at Lawrence Livermore National Laboratory: Wide bandgap extrinsic photoconductive switches. In this, they compared the performance of a PCSS made from GaN with one built from SiC. The impression we had after reading the paper was that GaN can be superior to SiC in this arena, because it enables the fabrication of devices that respond more quickly and have a lower on-state resistance.

We were very excited about the potential for GaN PCSS, and we explored many avenues that might allow us to supply materials or devices into a larger programme. Unfortunately, our initial efforts bore no fruit, and at times we felt that SiC had won the day – at least, for a while now. That’s not surprising, given that SiC wafers are larger, more plentiful and cheaper than our specially grown KO-GaN.

Despite this setback, we persisted, and we were rewarded for our efforts. We won a couple of modest US DoD contracts to build our own PCSS devices with support from the Air Force Research Laboratory, and these switches have produced very promising results. It took us very little time to beat the values for standoff voltage, on-resistance, and switching current reported by LNLL, and now, even though this work is still in its infancy, we can switch over 10 kV and 10 kA in less than 1 ns with very low on-resistance.

The testing of our switch has involved a high performance YAG laser equipped with an optical parametric oscillator and frequency doublers. This allows us to probe our device with a broad range of excitation wavelengths and pulse energies. We find that this switch has a very broad spectral response, which bodes extremely well for ultimate implementation using commercial, off-the-shelf fibre lasers and diode lasers. Applications for our switch include electric circuit fault protection, low jitter pulsed power, and support of several applications in the commercial, defence, and homeland security sectors. 

Our first KO-switches were built as part of a government-sponsored R&D programme, and we are still delivering devices under DoD support. Several devices have been sold this year, and we are now in discussions with several key corporate and government players. They are all interested in exploring this technology for US defence and homeland security applications.

What’s Next?

One of the biggest benefits of diversification, whether it is achieved organically or by acquisition, is that it leads to an expansion of our core competencies. In turn, this inevitably leads to new product opportunities, which sometimes draw on a portfolio of in-house expertise. In our case, this has happened with the KO-Switch.

We are committed to growing the business associated with our existing products, while continually seeking out new commercial opportunities. This year our goals are: To increase our PVD AlN template manufacturing capacity; get our equipment products into the market; and surpass certain performance milestones with our KO-Switch, which would make a compelling case for its insertion into several demanding applications. On top of this, we are aggressively pursuing the development of several new products in partnership with Duke University and others. Watch out for exciting new product announcements during the remainder of this year.

·         We thank the Air Force Research Laboratory for supporting our construction of a high speed switch testing facility.