DEVELOPMENT OF BURN-IN TECHNOLOGIES FOR UNPACKAGED INTEGRATED CIRCUITS (DICE)

THE ABILITY TO MANUFACTURE LARGE MULTICHIP MODULES WITH REASONABLE YIELDS AND COSTS IS HINDERED PRIMARILY BY THE PROBLEM OF BURN-IN LOSSES, WHICH CAN BE AS HIGH AS 5% PER CHIP. ONE SOLUTION WOULD BE TO PERFORM A "BURN'IN" OF THE BARE CHIPS (DICE) PRIOR TO THEIR ASSEMBLY INTO A MULTICHIP MODULE. NO PROVEN MANUFACTURING PROCESSES EXIST YET TO DO THIS. IT IS THE ULTIMATE OBJECTIVE OF THIS PROJECT TO DEVELOP SUCH A CAPABILITY. OUR GENERAL APPROACH IS TO USE A RELATIVELY LOW-COST HIGH-DENSITY INTERCONNECTION SUBSTRATE SUCH AS A "SILICON CIRCUIT BOARD" OR A FLEX-CIRCUIT TO DISTRIBUTE POWER, CLOCK, AND DATA LEVELS TO AN ARRAY OF DIE MOUNTING SITES ON ITS SURFACE. EACH DIE WOULD BE POSITIONED WITH ITS I/O BOND PADS FACING THE INTERCONNECTION SUBSTRATE, WITH MEANS FOR TEMPORARILY ELECTRICALLYCONNECTING THESE BOND PADS TO THE SUBSTRATE FOR THE DURATION OF THE BURN-IN. THE PHASE I EFFORT WILL FOCUS ON EVALUATING THE FEASIBILITY (BOTH ANALYTICALLY AND EXPERIMENTALLY) OF VARIOUS PROPOSED INTERCONNECTION METHODS, WITH EMPHASIS ON RELIABLE ATTACHMENT AND DETACHMENT TO ALUMINUM BOND PAD METALLURGIES, WHILE AVOIDING ANY DAMAGE OR CONTAMINATION TO THE DIE AND ITS BOND PADS.