Key Takeaways

  • Gallium nitride (GaN) and silicon carbide (SiC) will displace silicon (Si) in power conversion due to higher performance and lower cost.
  • GaN and SiC will service different segments of the market. GaN will take over consumer, telecom, and computer applications, while SiC will be most prominent in industrial applications that require higher voltages and current.
  • Manufacturing costs of GaN and SiC will come down in the next three years. The cost of GaN devices is dominated by growing a thin layer of the crystal on top of a standard silicon wafer. The cost of SiC devices is dominated by growing the SiC crystal in a bulk ingot.

Technical Characteristics of
GaN and SiC

GaN and SiC both start with a large fundamental advantage over silicon in the power conversion market. The devices can be made much, much smaller in size for the same relative voltage and current handling capability. This reduction in size is due to the higher bonding energy of the atoms in the crystal – thus they are called wide band gap semiconductors. In the case of GaN, the crystals also have a higher mobility of electrons. The relative theoretical performance of these three crystals can be seen in figure 1, where the vertical axis shows the relative size of the device and the horizontal axis shows the device’s ability to block voltage. As the voltage increases, so does the relative die size for Si, GaN or SiC. But, at all voltages SiC and GaN devices can be several orders of magnitude smaller than silicon!

Relative theoretical performance of GaN, Si, & SiC transistorsOn the market today there are GaN and SiC transistors that are 5-10 times superior to the theoretical limit of silicon, and with the anticipated technological progress, following a “Moore’s Law” rate, much higher performance is expected from GaN and SiC over the next few years.

With superior performance, the remaining questions are: (a) what are the relative advantages and disadvantages of GaN and SiC, and (b) why haven’t GaN and SiC devices already replaced Si?

Relative Advantages: GaN vs. SiC

The initial successes of GaN and SiC over Si have been at opposite ends of the voltage spectrum. GaN has made inroads in applications requiring 600 V and below, while SiC has made inroads in applications requiring 1200 V and above. The reason for this split is two-fold: speed and cost.

GaN devices are not only smaller than Si power MOSFETs (the dominant type of silicon transistor at voltages below 600 V), but also much faster. The speed of GaN stems from (a) the size advantage – electrons do not have to travel as far in a smaller device, and (b) the higher mobility of electrons in a GaN crystal – electrons can move more quickly. Power conversion applications at voltages of 600 V and below become much more efficient when devices can switch faster. For this reason there is a premium market that forms the pool of “early adopters” for GaN transistors.

SiC transistors are finding success at higher voltages primarily due to their size advantage. Today, a very high voltage SiC device (1200 V or higher) can be an order of magnitude smaller than a Si IGBT, currently the dominant type of silicon transistor at these voltages. This size advantage makes a significant difference in industrial applications such as UPS systems, motor drives, and high voltage DC-DC transmission.

Relative theoretical performance of GaN, Si, & SiC transistors

The advantage of GaN over SiC at lower voltages comes from the differences in the respective manufacturing technologies used today. The most cost effective way to manufacture a GaN transistor is to grow a thin epitaxial layer of the crystal on top of a silicon wafer. This wafer can then be processed into active transistors in a standard silicon CMOS foundry at a very low cost. Furthermore, low voltage GaN transistors do not require the expensive and bulky, and performance degrading packages that are required for Si MOSFETs and SiC transistors. As a result, GaN transistors are extremely cost effective at lower voltages compared to both Si and SiC.

The advantage of SiC over GaN at higher voltages comes from the structure of the device. At high voltages, the transistor is more efficient when the current is conducted through the wafer. This type of transistor is called a vertical device as compared with the low voltage GaN transistor, which is called a lateral device. The vertical structure necessitates that the entire SiC wafer be made of a single crystal type. SiC crystals are much lower cost than GaN crystals today and thus enjoy an early adoption lead in this market.

Elements of Manufacturing Cost

A technologically superior device that costs more to manufacture will enjoy success in niche markets that require the enhanced performance. In order to completely displace a technology a product needs both higher performance AND lower cost. The key elements of cost in manufacturing a transistor are (1) the starting material, (2) growing the epitaxial crystal (low voltage GaN and Si only), (3) wafer fabrication, and (4) the cost of packaging. Table1 compares GaN costs with Si power MOSFETs on the left, and SiC costs with IGBTs on the right. The comparison looks three years ahead to illustrate the evolution of cost components expected over that time.

Table 1: Comparison of relative manufacturing costs: (a) GaN FETs vs. Si power MOSFETs and (b) SiC devices vs. Si IGBTs
GaN (600 V and lower)
  2013 2016
Starting Material lower lower
Epi Growth higher same
Wafer Fab lower lower
Assembly lower lower
Overall higher lower
SiC (1200 V and higher)
  2013 2016
Starting Material higher same
Epi Growth n/a n/a
Wafer Fab higher same
Assembly same lower
Overall higher lower

Both GaN and SiC enjoy the advantage of smaller device size compared with Si. In the case of GaN, this translates into an immediate advantage in the cost of starting materials, since the GaN transistor is grown on a standard silicon wafer. SiC, however, requires a relatively expensive substrate, even taking its size advantage into account. Over the next few years the cost improvements in SiC crystal growth are expected to neutralize the advantages of Si IGBTs giving SiC a relative cost advantage at 1200 V and higher.

Today, when comparing GaN transistors with MOSFETs, the epitaxial growth of GaN on the silicon wafer is the sole cost disadvantage. With advances in epitaxial growth equipment, however, this disadvantage is expected to equalize in the next few years, giving GaN a clear cost advantage over the aging MOSFET.

Displacement of Silicon in Power Conversion

With both performance and cost advantages over their silicon counterparts, GaN and SiC will displace Si in their respective segments of the $12B power transistor market in the not-too-distant future.