CONDUCTIVITY-MODULATED SILICON RECTIFIER 999 



form flow through the junctions. This case is of particular significance 

 in high voltage rectifiers where small reverse currents result in relatively 

 large power. It has been pointed out in Sections 3.2 and 5.1 that body 

 avalanche breakdown is frequently not observed in these devices. 



A\'alanche breakdown current in silicon is carried by discrete pulses 

 of about 50 ^la at their onset and increasing with increasing current to 

 about 100 /ua. Approximate calculations"^ show that the ionizing regions 



o 



of these microplasmas are about 500 A in extent, have a current density 

 ?^ 2 X 10*' amp/cm^, and have a net space-charge density ?^ 10 /cm . 

 These pulses for junctions with ii'm.ix less than 500 kv/cm appear to be 

 independent of junction width and built-in space-charge. Rose considers 

 the statistical problem associated ^^^th a large number of pulses and pre- 

 sents a picture which is consistent with most of the experimental data. 

 He calculates the temperature rise, assuming the avalanche power is 

 1 X 10 " watts and is dissipated uniformly in a sphere. The maximum 

 temperature rise for a cluster of two or three pulses is in the order of 

 25° C. For the picture Rose presents, the temperature rise due to, the 

 microplasma should be relatively insensitive to the breakdown voltage. 

 Thermal collapse of rectification, i.e., increase of temperature until the 

 silicon is intrinsic, will probably not occur in the region of avalanche 

 multiplication. Two important conclusions can be obtained: 



1 . Avalanche breakdown should occur as a random process with a uni- 

 form probability over the junction. Large temperature rises due to a 

 breakdown of microplasma will probably not occur since the resulting 

 temperature rise would cause the breakdown voltage in that spot to 

 increase. The power is dissipated throughout the path of the current 

 pulse in the space-charge region. 



2. A thermal effect in silicon due to heating by the small plasma has a 

 very short time constant of the order of lO""^" seconds.'^ It is not possible 

 to separate a thermal effect of this type by reducing the pulse width. 

 The heating and cooling time is short compared to the pulse time in these 

 experiments. 



The pulse properties of a junction would be ciuite different if the break- 

 down occurred at one spot instead of many spots distributed over the 

 junction. Breakdown at a single spot on the surface has been observed. " 



6.2 Experimental Results 



Many rectifiers were given a voltage pulse which carried them into 

 breakdown. There was a wide distribution of V-I characteristics. Many 

 diodes did not show a negative resistance up to the maximum instan- 



