· Lower on-resistances, which result in lower conductivity losses and higher overall efficiency
· Higher breakdown voltages:available Si Schottky diodes have breakdown voltage up to 300V, the first commercial SiC Schottky diodes are rated at 600V
· Higher thermal conductivity compared to Si: this leads to a lower junction-to-case thermal resistance, allowing more efficient heat transfer
· Higher temperature operation: SiC devices can operate up to 600°C, while Si devices can operate at a maximum junction temperature of only 150°C
· Forward and reverse characteristics that vary only slightly with temperature and time
· Lower reverse recovery current, reducing switching losses and electromagnetic interference (EMI)
· Operation at frequencies >20 kHz, which is not possible with Si-based devices at power levels of more than a few tens of kilowatts
· Higher voltage input/output ratios that allow a single stage for DC-DC conversion from 48V to 1V, compared with a silicon power MOSFET converter that would normally require two or more stagesInitially, GaN-based devices seemed a likely fit for applications where the high voltage and power handling capability, coupled with conversion efficiencies than were higher than the silicon equivalents, would create a high-value niche. The advantages of wide bandgap materials for power electronics applications have partly been borne out by the deployment of SiC devices in hybrid electric vehicles, but the penetration of GaN-based power devices may be more likely in lower-voltage applications, where SiC is proving to be too expensive. In the power management segment, power conversion applications offer the greatest potential for GaN. These applications typically require fast, efficient switches, where the higher carrier mobility inherent to GaN transistors is advantageous. While silicon devices are widely regarded to be at or approaching their performance limits, GaN offers opportunity for improvement, particularly in form factor and electrical efficiency. A key figure of merit used in power management applications is the combination of on-resistance and switching speed. There is a lot of development underway to improve the performance of GaN FET devices. Figure 2 shows the Ron performance for two GaN devices from EPC Corporation. It is evident from the curves that the GaN devices maintain lower on-resistance for a given breakdown voltage versus the silicon MOSFET devices listed. The chart also shows that the theoretical maximum performance for GaN devices is superior to SiC. Figure 2: Resistance versus Breakdown Voltage The advantage goes beyond the on-resistance, however. The gate charge required to switch a GaN FET is much lower than that of a silicon device. The product of on-resistance and gate charge is a useful figure of merit for power transistors. A lower figure indicates a low resistance to turn-on, combined with a fast switching speed. As shown in Figure 3, EPC Corporation claims their GaN devices have an RQ product more than 10 times lower than silicon alternatives. Figure 3: “RQ” Product Comparison: GaN vs. Silicon MOSFET Source: EPC Corporation EPC Corporation is one of the leading proponents of GaN for power management applications, and they make a compelling case for this technology. In Figure 4, they illustrate the theoretical performance comparison for GaN technology versus the incumbent silicon MOSFET technology and SiC technology. Figure 4: Theoretical “RQ” Performance Source: EPC Corporation In practice, a low RQ figure of merit means that circuit designers do not have to sacrifice low on-resistance for fast switching, as is normally the case. A side benefit of this performance is that GaN DC-DC converter devices will operate at higher bandwidths. Silicon power MOSFETs have difficulty with pulse widths below 100ns (corresponding to a 250 KHz bandwidth), while GaN FETs can be turned on or off in as little as 4ns. Figure 5 summarizes these metrics with the performance of an EPC GaN device versus some representative silicon MOSFET devices. Note that the “RQ” product of the GaN device is almost an order of magnitude lower than best MOSFET. Figure 5: “RQ” Product Comparison: GaN vs. Silicon MOSFET Source: EPC Corporation The implication is that a point-of-load (POL) converter using GaN can convert from 48V to 1V in a single stage, while an equivalent silicon converter would typically require one stage to convert to 12V and a second stage to convert to 1V. At higher drain-source voltages, the superiority of GaN becomes even more apparent, enabling entirely new architectures in power management, according to proponents such as EPC and International Rectifier. In the past several years, there has been a concerted effort to develop enhancement-mode (normally off) GaN transistors, especially for power management applications. These enhancement-mode devices are attractive for these applications because they operate in a similar fashion to the incumbent MOSFET technology, but have much better performance characteristics. A number of Japanese companies, including Sanken Electric, a collaboration between Fuji and Furukawa Electric, Panasonic and NEC, have all invested in the development of E-mode GaN HEMTs. In March 2010, Efficient Power Conversion (EPC) Corporation introduced a number of E-mode GaN-on-silicon power management devices, branded as eGaN, which they continue to refine and improve. These E-mode devices have different driver requirements than their depletion-mode GaN and silicon MOSFET counterparts. Companies like National Semiconductor and Texas Instruments have developed lines of compatible drivers for E-mode GaN devices. These driver efforts are likely to speed adoption of E-mode GaN devices by making the final package of driver and transistor as easy to use as MOSFET devices. For all the inherent material and performance advantages, developers of GaN transistor technology continue to address a number of challenges. The lack of suitable native GaN substrates complicates production of GaN devices because it is very difficult to grow lattice-matched, defect-free epilayers analogous to the processes used for GaAs or silicon transistor fabrication. There is substantial activity using a hydride vapor phase epitaxy (HVPE) method to produce thick layers that can serve as quasi-bulk substrates. Shortfalls with this technique have given rise to development of ammonothermal growth techniques. In August 2012, Soraa, a developer of GaN-on-GaN solid-state lighting technology, was selected by Advanced Research Projects Agency-Energy (ARPA-E) to lead a project on the development of bulk GaN substrates. The attraction of native GaN substrates lies in an anticipated performance improvement that may result in significant energy savings for LEDs. The LED market can provide the high-volume pull to develop a technology that can benefit other market segments with the availability of native GaN substrates. Even with a production ready process, GaN substrates are not expected to compete with silicon substrates purely on wafer cost. However, proponents of native GaN substrates point out that a simplified process will result in cost savings that may make the product cost more manageable. In the absence of production-scale, single-crystal GaN wafers, manufacturers must instead use foreign host materials such as sapphire, SiC or silicon. To date, semi-insulating SiC has been the material of choice for microelectronic components, thanks to its relatively close lattice match to GaN and its excellent thermal conductivity properties. However, the quality of epitaxial layers depends on both the lattice-match and the underlying substrate quality. Historically, suppliers of semi-insulating SiC have struggled to produce material with defect density levels comparable to substrates such as silicon or GaAs. More recently, improvements to semi-insulating SiC material quality, along with the availability of larger substrates, have made SiC a more viable economic choice for GaN growth. This has helped to improve the reliability of GaN microelectronic devices. However, SiC remains a difficult and expensive material to produce and it provides cost challenges for cost-sensitive, high-volume applications. In recent years, high-resistivity (HR) silicon has become a viable alternative to SiC in certain applications. Although it does not have as close a lattice match with GaN and possesses poorer thermal properties, silicon can offer a lower-cost path for some applications. GaN-on-Si also shows potential to transition to high-volume manufacturing processes because it is amenable to existing CMOS (Complementary Metal-Oxide-Semiconductor) semiconductor fabrication technology using commercially available, large diameter silicon wafers. There is a lot of development work in this area with manufacturers like AZZURRO Semiconductors developing 150mm GaN-on-silicon wafers with a roadmap to 300mm diameter wafers in the future. Future The size of the power electronics component market and the perceived advantages of wide bandgap technologies are leading to significant process and product development efforts. The US government realizes the importance of improvements in the efficiency and reliability of the electrical grid network and the US Department of Energy’s Office of Electricity Delivery and Energy Reliability (OE) has requested slightly more than $169 million in the FY2014 US budget to address issues and developments in this area. The budget request is up nearly 25% from 2012 spending as the OE recognizes the importance of the topic and realizes that industry will need some help to enable significant advances. In April 2011, the OE released their “Power Electronics Research and Development Program Plan” that details the vision, activities, challenges, needs and partnership strategies for this market segment. The activities focus clearly on wide bandgap materials, with a heavy emphasis on developing and refining GaN-on-silicon processes and devices. By 2016, the program hopes to have GaN-on-silicon devices operating at 5kV and 15A performance levels, with a goal of 20kV and 50A device performance by 2026. Market Estimates It appears clear that increasing demands for energy and an emphasis on alternative energy sources will lead to a very large market opportunity for power electronics components. The advantages of wide bandgap technologies appear very compelling, but there are challenges that must be addressed before these technologies will displace incumbent silicon technologies with any sort of production scale. The promising development for GaN usage in this market segment is that industry and government agencies are recognizing the importance of this technology and devoting significant resources to technology and process development. Considering all these factors, we estimate revenues for GaN devices used in power electronics applications will grow strongly, reaching slightly more than $73 million in 2017. While this revenue level represents only a very small portion of the total high power electronics market, the CAAGR (Compounded Average Annual Growth Rate) for GaN technology will be greater than 108%, indicating the revenue will roughly double every year, on average. GaN technology usage is in its infancy for high power electronics applications and we anticipate continued strong growth throughout the decade. Figure 6 shows this estimate. Figure 6: GaN Device Revenue in High Power Electronics Applications Conclusions Increasing energy consumption, the shift to a “digital economy” and the evolution to more efficient, more distributed sources of energy generation will all increase the available market for power electronics devices. The incumbent technology is silicon IGBTs and MOSFETS, but there are concerns that the limitations of these devices and technologies will hamper the growth and refinement of the electric grid. Wide bandgap materials, most notably SiC and GaN offer promising performance advantages and these technologies are generating a lot of interest. Both these technologies have challenges to overcome and while SiC devices have made early advances, we believe GaN-on-silicon products will grow quickly to reach slightly more than $73 million in revenue in 2017. Author : Eric Higham - Director - GaAs Strategic Technologies Practice References 1Jennifer Weeks and The Daily Climate, “U.S. Electrical Grid Undergoes Massive Transition to Connect to Renewables,” Scientific American, April 28, 2010, http://www.scientifi camerican.com/article.cfm?id=what-is-the-smart-grid&print=true.