Substrates and epitaxial growth dominate ECSCRM discussions

The key talk of the conference was given by Daisuke Nakamura, from the Toyota Central R&D labs in Japan, who outlined research that he claimed could "open the world to SiC device technology". Nakamura explained that he and colleagues at Toyota and DENSO Corporation have rendered SiC substrates that are "almost dislocation free". "It will be possible in the near future to eliminate dislocations perfectly, and to enlarge the [substrate] diameter to several inches," he said.

Until now, SiC wafers have been plagued with high concentrations of defects. These limit commercial applications of the material to just a few devices, such as Schottky diodes. Bipolar SiC devices have tended to suffer from a degradation of the material s electrical properties, which seems to originate from in-plane dislocations.

Nakamura s team made the breakthrough by employing "repeated a-face growth" (RAF). The team started with a single SiC crystal grown on an a-face, which consequently had a high dislocation density, and took a section of this crystal along its a-axis (see figure 1). The researchers then allowed the crystal to develop on its other a-face, before continuing with conventional c-face growth. Repeating the a-face step appears to eliminate stacking faults and suppress dislocations.

In a 20 mm-diameter substrate grown with RAF, the average etch-pit density (EPD) was 75 cm-1, which is one-thousandth that of conventional SiC substrates. The result for micropipes was even better, because these were completely eliminated in the RAF-grown wafers. The team also made some pin devices using the technique. According to Nakamura, these devices were much more reliable than is normally the case.

Nakamura explained that to make the technology suitable for commercial use the team tried to increase the crystal size, and fabricated a 3 inch substrate with an average EPD that was about one-hundredth of that found in conventionally grown material.

The availability of affordable, high-quality, large-diameter substrates was discussed by Ilya Zwieback, from US-based II-VI. Zwieback focused on semi-insulating substrates for GaN-based RF and microwave devices. He explained that if devices are to be reliable at elevated temperatures then semi-insulating substrates with resistivities as high as 1011 Ωcm are required.

To manufacture crystals, II-VI uses an advanced physical vapor transport process (VPT) that is "principally the synthesis of SiC from separate sources of silicon and carbon". The company produces semi-insulating 6H-SiC substrates in 2 inch, 3 inch and 100 mm diameters, as well as 2 and 3 inch n+4H-SiC substrates, which are required for SiC-based power/switching devices.

Zwieback believes that to produce high-quality boules it is important to prevent crystal defects by controlling precisely the temperature gradient. II-VI uses growth temperatures of 2000-2300 °C and pressures of 5-50 mm Hg, backed up with computer-aided thermal simulation. The company s semi-insulating 6H substrates are fabricated through compensation - the addition of precisely controlled amounts of vanadium - and assessed by a room-temperature non-contact resistivity mapping system (see figure 2).

Zwieback explained that material with high background contamination, usually resulting from shallow impurities such as boron and nitrogen, requires a higher concentration of vanadium doping for full compensation, but this is not always desirable. II-VI s response was the recent introduction of a new process with reduced background contamination, and a small and precisely controlled vanadium flux. "We [now] produce crystals with incredibly high resistance, greater than we can measure," said Zwieback. The "low-vanadium-doped" substrates often exceed the upper measurement limit of 1012Ωcm. II-VI is now working on its next goal, to grow undoped 2 and 3 inch 6H substrates. Resistivities as high as 1010Ωcm have already been reported.

Greater control of crystal growth through the addition of a pipe to the growth chamber was described by Peter Wellman from the University of Erlangen, Germany. Wellman explained that SiC crystals are traditionally grown by PVT in a closed graphite crucible at temperatures above 2000 °C. This approach prevents adjustments of gas flows, such as the carbon:silicon ratio and dopant concentration, and is susceptible to changes in the temperature distribution of the crucible, which can lead to growth instability.

A process called modified PVT (MPVT), in which a pipe is introduced to the standard set-up to enable a small amount of gas to be added, offers several advantages. These include controlling polytypism (the crystalline form of SiC), and reducing defect density through a silicon-rich gas phase by adjusting the carbon:silicon ratio (see figure 3).

Wellman s co-worker, Ralf Mueller, described how MPVT is used to produce highly aluminum-doped SiC substrates for high-power bipolar device applications. Aluminum is rarely the first choice of dopant for SiC - when added as a powder it initially causes high defect generation before depleting and causing a large variation in the doping concentration across the boule. Boron, today s preferred choice, provides greater doping uniformity, but the penalty is a deeper doping level and consequently lower device efficiencies. With MVPT, the depletion of aluminum is avoided. This allows aluminum-doped wafers with resistivities of 0.09 Ωcm for 6H-SiC and 0.2 Ωcm for 4H-SiC. Both of these values, which are suitable for high-power device applications, are far lower than the best published value of 0.45 Ωcm, said Mueller.

Exploiting different crystal orientations to improve boule quality by PVT was discussed by Christoph Seitz, from the University of Erlangen-Nuernberg, Germany. He told the audience that growth in the conventional direction, [0001], is straightforward, but disadvantages include polytype switching and micropipe formation. After outlining the benefits and pitfalls of other directions, such as the [10-10] direction, he posed the question: "What about the in-between directions? Do we add up the positive aspects, or the negative ones?".