GaN: Gaining Traction, But Still Short of Fulfilling Its Promise in Power Electronics
By Keith Dawson, Mouser Electronics
If vertical GaN is to gain market acceptance in a big way, improvements will be needed in a number of areas to scale up production and bring down prices.
Wide band gap semiconductors have long been boosted as the next-generation technology that will make inroads into silicon's half-century reign atop the world of semiconductors, and one day come to dominate in applications from RF to power to transportation.
Figure 1: Properties of WBG vs. Silicon Carbide (SiC) vs. Silicon demonstrating the high mobility of GaN
Wide band gap (WBG) semiconductors promise improvements in nearly all performance dimensions over conventional silicon: they are more efficient, switch faster, tolerate higher operating temperatures, feature higher breakdown voltages, and can handle higher currents. They are also smaller and lighter. These promises come as silicon technology begins to run out of steam, in terms of sustaining the rates of improvement whose macro effect is Moore's Law. Silicon IGBTs, MOSFETs, and thyristors are closing in on their physical limits in terms of power density, breakdown voltage, and operating frequency.
The leading contenders today for WBG semiconductor applications are silicon carbide (SiC) and gallium nitride (GaN); the latter is offered in three forms. "Wide band gap" refers to the large energy gap -- compared with silicon -- between the top of the valence band and the bottom of the conduction band in solid materials. The term is mostly a way of distinguishing other semiconductors from silicon; the size of the band gap (measured in electron volts, eV) does not alone determine the qualities of a semiconductor material, but rather is one factor in determining the material's quantum behavior.
SiC has a band gap of 3.2 eV, and GaN 3.4 eV, compared to silicon's 1.1 eV. Other WBG materials with possible future promise in electronics include gallium oxide (Ga2O3, 4.8 eV), diamond (5.5 eV), and aluminum nitride (AlN, >6 eV).
Below is a comparison of technologies for wide band gap power electronics.
- SiC — Targets high performance diodes; 1200V JFETs and MOSFETs
- GaN on Si — For high power low cost diodes and transistors
- GaN on SiC —High performance, high power, high cost for RF applications
- GaN on GaN — Still in development
- Gallium oxide — Very early development, for transistors and solar cells
- GaN on Diamond — Future development. Diamond has the highest thermal conductivity of any material
In power applications the leveling off of progress in silicon components is perhaps most evident. Renewable energy calls for inverters operating in harsh situations where silicon faces serious challenges. Automotive applications, especially hybrid / plugin technologies, demand more stringent temperature and power profiles than silicon can easily meet.
The high voltages of the electrical power industry offer opportunities for more efficient WBG power semiconductors for control and conversion. Throughout the power grid, conversion losses are estimated at 10%. With power semis employed everywhere, including power grids, switching power supplies, adaptors, cars, and appliances, every percentage point in efficiency gain matters. Power conversion systems using WBG materials add about 5% to the efficiency of SiC power conversion systems. This is mainly due to the lower resistance of WBG materials, which lowers power losses from heat and increases switching efficiency.
Grid power usage is expected to grow worldwide by a significant amount in the coming years. How large the predicted growth will be depends on which study is consulted. The US Energy Information Administration in 2013 projected 56% growth to the year 2040. The International Energy Agency in 2014 estimates a median growth rate of 37% to 2040. So a few percent improvement in the efficiency of power components will be magnified over time.
Lead Time
SiC has been studied and developed for semiconductor applications for more than 20 years (see Timeline), whereas GaN, a much more difficult material to fabricate and to purify, is about a dozen years behind, and also more expensive along all the points in the supply chain. The difficulty of growing high-quality GaN explains much of the lead that SiC now enjoys.
Silicon devices are grown homoepitaxially: that is, the active layers are deposited on a base or substrate of the same material (in some form). Homoepitaxy provides a number of advantages, chief among them the avoidance of a mismatch at the boundary between substrate and active material. When a crystalline material is grown on a foreign substrate, lattice mismatches are inevitable, causing strain and dislocations. The boundary also impedes heat flow to the extent the thermal coefficients of the two materials differ.
Figure 2: Brief History of Wide Bandgap © 2016 Mouser and Mouser Electronics. All rights reserved.
SiC devices are now almost all homoepitaxial. But due to the historical difficulty of obtaining bulk GaN material of sufficient purity, GaN semiconductors are mostly heteroepitaxial, grown on either SiC or Si. (Sapphire, which is the dominant substrate in GaN LEDs, is also sometimes employed, but its poor thermal conductivity limits its usefulness.)
In such devices current flows laterally, along the semiconductor's surface. This limits both the maximum current and the breakdown voltage compared with homoepitaxial devices in which the current moves vertically.
Homoepitaxial, vertical GaN devices are only now reaching commercial availability. Published Figure-Of-Merit (FOM) results for vertical GaN transistors have been well past the SiC limit while approaching the theoretical limit for GaN, resulting in dramatically improved switching performance.
Another factor boosting the cost of vertical GaN devices is the smaller size of substrate wafers available in the less mature technology, compared with SiC or Si. The few companies investigating vertical GaN are using using 100mm (4") GaN substrate wafers. GaN-on-SiC is now transitioning to 6-inch wafers and 8-inch are on the horizon. GaN-on-Si is produced on 6-inch, 8-inch, and 12-inch wafers. Larger wafers result in greater economies of scale.
Current Market and Prospects
Almost all of the GaN semiconductor devices on the market today are heteroepitaxially grown, lateral GaN. Most are based on SiC or cheaper Si substrates. Lux Research has predicted a 24%
compound annual growth rate for GaN-based power electronics, to $1 billion in 2024. By that time SiC will account for $1.9 billion. The total of WBG semiconductors will be only 13% of the power electronics market worldwide in 2024; 87%
will be silicon.
Lux predicts that in the GaN segment of the power market at 2024, GaN-on-Si will dominate with 90%; GaN-on-SiC will be 9%; and vertical GaN, 1%
.
After completing a degree in physics and English from Carnegie Mellon University, Keith Dawson has enjoyed a long career as tech writer, software developer, manager, tech marketer, Internet consultant, unelected pundit, and community builder — not necessarily in that order. As a technology journalist and editor he wrote for Media Grok, Media Unspun, Slashdot, The CMO Site, Business Agility, and served as Editor-in-Chief for Develop in the Cloud and All LED Lighting. A seminal blogger, Dawson has been advocating and practicing social media since before the birth of the web.
