Experiments on second breakdown in silicon-on-sapphire p
+-n-n
+diodes show that second breakdown involves filament nucleation, filament growth, and a melt transition. Experimental results will be reviewed relating the spatial and temporal development of current filaments to voltage waveforms during constant current pulse testing. Theoretical models will be presented which show: (1) The condition for a filament to nucleate within the junction is for the thermally generated saturation current to be equal to the total current at a particular site. When this happens, the junction voltage at this site decreases sharply toward zero. (2) The multiplicity of filaments is due to competitive current paths; junction channels with their series spreading resistances versus the normally avalanching junction regions. Spreading resistance is also the reason for for the quenching of avalanche adjacent to a junction channel and the negligible voltage drop seen externally when nucleation occurs. (3) The melt formation is accompanied by an abrupt voltage drop (the melt does not follow the voltage transition). The voltage waveforms and final melt radii, for constant current pulses of different amplitudes, computed are in good agreement with experiment. (4) The threshold current for second breakdown is proportional to

, where ρ is the resistivity of the n-region.