Statistical Analysis of Micropipe Defect Distributions in Silicon Carbide Crystals

Abstract

The density and distribution of micropipes (MPD's), a screw dislocation with a hollow core, are significant criteria for SiC wafer selection in device fabrication. For this reason the measurement of micropipes in SiC is an important problem. The micropipe distributions of 2’’ diameter 6H SiC wafers from three different PVT furnace designs in three different laboratories were statistically characterized. The micropipe distributions were measured in two ways: by KOH etching the wafer and then counting the hexagonal pits formed on the wafer surface using a microscope and imaging software; and by manually counting with an optical transmission microscope the micropipes on a un-etched wafer within a grid of 0.5mm squares superimposed over the wafer. None of the wafers had a random distribution of micropipes i.e. they did not follow a Poisson distribution. The micropipes were found to associate with each other i.e. there is a higher probability of there being other micropipes near a location where a micropipe has been found. In the eight samples characterized, 1.9% of the wafer area contained 32% of the total micropipes. We have analyzed the repeatability of the optical transmission microscopy method of micropipe characterization and found that it is dependent on the micropipe density; the standard deviation of the measurement increases with increasing MPD. The number of micropipes counted increases by about 10% with increasing magnification and 15% when the wafer is double side polished.