Hybrid microcircuits (HMC's) designed for the Energy Research and Development Administration (ERDA) by Sandia Laboratory in Albuquerque, NM, are manufactured by the Kansas City, MO, Division of the Bendix Corporation employing beam lead devices (BLD's) as active components. The HMC's consist of sputtered tantalum-nitride (Ta2N) resistors and evaporated chromium-gold (Cr-Au) conductors on alumina substrates with applique components and thermocompression bonded ribbons, lead frames, fine wires, and beam lead devices. During recent testing of HMC's for a new program, failures of the BLD's resulted in unacceptable yields and increased rework. These BLD failures showed up primarily as increased leakage failures, intermittent failures that repeated on a random basis, and in worst cases, as direct device shorts. An investigation of the problem showed that cracks often appear on the edge of the silicon and on the silicon nitride lip that overhangs the edge of the beam. These cracks appear at various phases of the processing, but primarily during BLD bonding. High temperature reverse bias (HTRB) tests of diodes and transistors combined with analysis of the devices that failed during these tests indicated that gold in the cracks was a primary cause of failures. The gold in the cracks is the result of gold "relocation" or migration from the gold beams on the semiconductor. The gold was detected by energy dispersive X-ray analysis (EDAX) and electron-beam induced-current methods. Gold was observed on the silicon cracks above the beams and in silicon nitride cracks on either side of the beams. Gold resistive leakage paths of this type have been observed only in devices such as p-n-p transistors and p+ silicon substrate zener diodes, where one of the gold beams is more than 1.42 V positive with respect to the body of the silicon chip. This indicates that the gold leakage paths are formed by electromigration of gold from the beam to the silicon chip. Results of HTRB tests on n-p diffused zener diodes,1 p-n diffused zener diodes, and n-p-n transistors are consistent with the gold electromigration theory. Several conditions are theoretically required for gold e!ectromigration of this type: the gold beam must be positive with respect to the silicon chip by more- than some minimum voltage (the gold-gold electrolytic potential is 1.42 V); there must be a crack in the silicon nitride, or at the intersection of the beam and a chip, or a crack in the silicon above the beam; and an electrolyte such as moisture-containing halogen ions must be present in the crack. Vacuum baking of HMC's after cleaning and just before electrical testing, the elimination of water as a cleaning agent, and the elimination of trichloroethylene at final assembly cleaning have solved the leakage problem caused by gold electromigrafion in the devices.