1982 · Czochralski Silicon Ingot · Work 872 of 1,871
A materials store of a chip factory in Santa Clara, on a stainless steel worktable an open box of plain white foam, in it a silicon crystal lying lengthwise, a cylinder sixty centimetres long and ten and a half centimetres across, almost black grey with a mirror sheen, finely ringed, at the left end a short conical shoulder tapering into a thin seed rod with a small bulge, the right end sawn flat, beside it a pale plastic cassette of mirror-bright wafers ten centimetres across, two hands in white gloves lifting the front wafer, grey metal shelving with white foam boxes behind, fluorescent light from above.
What it was used for
In the Czochralski method a seed crystal is dipped into the silicon melt and slowly withdrawn; the single crystal from which wafers are later cut solidifies on it. The background section of a US patent on necking summarises the usual sequence: first a thin neck 2 to 4 millimetres in diameter, pulled at up to 6 millimetres per minute and up to 100 millimetres long, and only then the cone and the cylindrical body. Dipping produces dislocations by thermal shock which would otherwise propagate through the whole crystal.
What was new
Necking goes back to William Dash, "Growth of Silicon Crystals Free From Dislocations", Journal of Applied Physics, volume 30, issue 4, 1959, pages 459 to 474. That it is generally accepted in the industry that dislocations cannot reliably be removed from necks wider than 10 millimetres is stated in the same patent. The barb is structural: the entire crystal hangs from that thread. Crystals of 200 millimetres diameter with a Dash neck are therefore usually limited to 100 kilograms or less - if the neck fractures, the body drops into the crucible, molten silicon splashes, crucible, susceptor and heater can be destroyed, and the melt is lost.
What the sources leave open: The figures come from the background section of a 1990s patent and describe the state of the art with 200 millimetre crystals, not that of 1982. No source giving crystal weights and neck dimensions for 1982 was found, and Dash's 1959 paper is documented here only through the patent's citation, not opened in the original.