Silicon carbide is a highly desirable material for high-power electronic devices — more desirable even than silicon. And now the problem of producing large, pure wafers of the carbide could be solved.
After a period of intense scientific and industrial development, the semiconductor silicon carbide (SiC) is at last proving capable of outperforming silicon in electronic devices for high-power, high-frequency and high-temperature applications. Its potential as a replacement for silicon has been known since the 1950s, but its late emergence is principally due to the difficulty of growing large SiC crystals of sufficient quality. That difficulty is set to evaporate, thanks to the crystal-growing process introduced by Nakamura et al.1 on page 1009 of this issue.
A major problem encountered in growing crystals of SiC is that the material doesn't have a liquid form. This means that the traditional methods of crystal growth developed for silicon and other semiconductors, based on controlled solidification from the liquid phase, cannot be used for SiC. However, the invention in 1978 of the ‘modified Lely method’2 (or physical vapour transport, PVT) opened the way to the production of large-area SiC wafers. This is a seeded sublimation technique: supersaturated SiC vapour condenses onto a single crystal seed inside a graphite crucible. Impressive progress has been made, but the quality of the SiC grown in this way is still too poor and remains an obstacle to be overcome if SiC technology is to make a commercial breakthrough. Improving the structural properties and at the same time increasing the size of SiC wafers that can be grown are key areas of research in this field.
As well as the classic defects, such as different kinds of dislocation, that are common to most materials, SiC wafers produced by PVT tend to have some peculiar defects of their own. Known as ‘micropipes’, these defects are a consequence of the crystal structure of SiC and profoundly affect the reliability of electronic devices based on this material. The basic unit of the crystal structure is a silicon–carbon bilayer; within each bilayer, carbon atoms are arranged in a two-dimensional hexagonal pattern, with silicon atoms directly on top of them. Three-dimensional lattices are then built by stacking the bilayers along the normal direction, called the c-axis, of the hexagonal structure, and a surface polarity effect results, with one face of the crystal formed of silicon atoms and the other of carbon atoms (Fig. 1). In fact, several crystal structures, or polytypes, are possible — all described in terms of stacked hexagonal layers, but with different rotational offsets between the layers. Micropipes are hexagonal tube-like cavities, with diameters in the sub-micrometre to few-micrometre range, that develop parallel to the c-axis of the hexagonal structure (which is the common direction of crystal growth). They are the most harmful defects in SiC.
Intensive work on various aspects of the PVT growth technique — reactor design, growth conditions, seed orientation and surface preparation — has led to a considerable reduction in the number of these defects in SiC ingots and wafers. The best commercially available 3-inch-diameter wafers (of the 4H polytype) have micropipe densities of 10–100 cm−2 and dislocation densities of 103–104 cm−2. This quality of crystal is good enough for use in some commercial SiC devices, such as Schottky diodes3, but not for others. For instance, in high-power bipolar devices, there is a degradation in the material's electrical properties that seems to be related to the development of extended stacking faults, originating from in-plane dislocations in the SiC (ref. 4).
So reducing the density of dislocations in an SiC wafer, and in the epitaxial layer on top of it that forms the active part of the device, is an absolute necessity for the development of high-power SiC technology. Nakamura et al.1 have succeeded: their new process produces SiC crystals with a dramatically lower dislocation density — although, perversely, the first step of the process actually creates a high density of stacking faults.
Most SiC crystals are grown on the c-faces (perpendicular to the c-axis) of other crystals. In fact, it has already been shown that growing SiC crystals on surfaces with a different crystal orientation — the a-faces — fully suppresses the formation of micropipes. However, crystals made in this way are affected by basal-plane stacking faults. Nevertheless, Nakamura et al. started with a SiC single crystal that had been grown on an a-face, inheriting a high density of dislocations from its seed crystal. Then, taking a section of the crystal along the a-axis of that face, they allowed the crystal to develop on the other a-face (Fig. 2), afterwards continuing with classic growth on the c-face. Nakamura et al. call this the repeated a-face (RAF) growth process. It is the repetition of the a-face step that ensures that stacking faults are elimi-nated and dislocations suppressed. The ingots of SiC produced are, say the authors, “virtually dislocation-free”.
These results are spectacular: the RAF process is a major innovation in materials science. Silicon carbide has become, at last, a contender for silicon's crown.
About this article
Scientific Reports (2017)