A silicon crystal implanted with gold was heated to successively higher temperatures, reaching a maximum of 1250°C, and its surface examined by Auger emission spectroscopy (AES) and LEED at each stage. No recrystallization was observed until the crystal had been heated to 740°C, at which point the AES analysis could be interpreted in terms of the appearance of gold atoms at the surface. At 810°C the Auger peaks from gold were considerably larger than those from silicon, but decreased progressively thereafter as the temperature was raised until, at over 1000°C, the differential distribution was indistinguishable from that of clean silicon. The first LEED pattern observed, at 740°C, was not the Si(111)-7 pattern, but another familiar one, the Si(111)-(radical3 × radical3)-R30° pattern. With increasing temperature, the fraction one-third-order pattern spread over the whole surface and, above 900°C, was joined by another, a fraction one-fifth-order pattern, probably based on a domain structure. At the highest temperature no fractional orders were visible, only a Si (111)-(1 × 1) pattern, with good spot-to-background contrast. The normal 7-pattern could only be achieved by ion bombardment and annealing, not by heat treatment alone, suggesting that there was still a very high density of dislocations remaining from the implantation.