On the fine scale revealed by TEM, the material is characterized by predominant dendritic 3C crystals. All these dendritic crystals have equivalent orientations and are not polycrystals like those found in dendritic growth of metals. The branching-off characteristic of 3C dendritic growth is thus due to symmetry of its having three equivalent close-packed directions. Since both trunks and branches have exactly the same orientation and since the direction of the spines is parallel to the < 1 1 1 > directions, S i c crystals seem to grow in one of the < 1 1 1 > directions under the present conditions. Some X crystals were also found to grow in a columnar fashion in one of the < I 11> directions, much as in S i c whiskers. These columnar crystals exhibit twin bands perpendicular to the growth direction.Both the 2H and the one-dimensionally-disordered crystals are uniaxial and grow in acolumnarfashion in thec direction. SinceSiC does not usually form regular crystalline boundaries between crystals of different orientation, as observed in common metals, presumably because of the high energy involved in dislocation formation, these crystals do not form dendritic structures with branching-off as do the 3C crystals. However, several onedimensionally-disordered crystals were found to grow together with a columnar habit, making small angles with one another. Precise details of the structure of such small-angle grain boundaries are not known .The one-dimensionally-disordered crystals are characterized by the fact that the stacking order of the crystal is almost completely random, whereas the structure is perfect in the other two dimensions. The existence of the one-dimensionally-disordered structure reflects the general characteristic of polytypism of S i c , i.e. that the energy of the structure does not depend on its stacking order and that any polytype is equally pr~bable.~.'" The situation is also reflected by the extremely low stacking-fault energy.6 The stacking order of S i c crystals thus seems to be affected by local growth conditions,'" as demonstrated by the distribution of disordered crystals in the present case. One-dimensionally-disordered crystals, however, should not be considered simply as heavily faulted 3C crystals. ,A study of such one-dimensionally-disordered crystals by highresolution electron microscopyz1 revealed that the structure con-sisted of random layers of thin a-and P-phase lamellas of various thicknesses.2t Acknowledgments: The writers are indebted for valuable discussions to J. E.