960 THE BELL SYSTEM TECHNICAL JOURNAL, SEPTEMBER 1952 



14 



12 



10 



8 



6 



4 



2 







-2 



-4 



-6 



-8 



-10 



•12 



-8-4 4 8 12 16 20 



MAGNETIC FIELD PARAMETER, $ 



Fig. 12 — Experimental magnetic effect for dissimilar surfaces. 



case where the two recombination parameters are 0.1 and 10, and for 

 a second case where the parameters are 0.5 and 50. In this figure, sepa- 

 rate curves have been shown for noise due to holes generated on each 

 of the two surfaces. The total noise would be gotten by adding the noise 

 powers represented by the two curves after appropriate weighting for 

 the contributions of the two surfaces. At present we do not see any way 

 of determining the weighting factor. 



A third case is that in which it is assumed that the noisy generation 

 of holes is uniform throughout the body of the filament, but that re- 

 combination takes place on the surfaces only. These assumptions seem 

 at first sight to be in contradiction to the statistical mechanical prin- 

 ciple of detailed balancing, which states that under equilibrium condi- 

 tions all processes occur with equal frequency in the forward and reverse 

 directions. Thus it would seem that if holes are generated in the interior, 

 we must consider recombination in the interior also. Actually this is 

 not necessary under the non-equilibrium conditions which prevail 

 during noise measurements. There is no necessity for the noise generated 

 by a source and a sink for holes to be simply related to the strength of 

 this source. Thus we may suppose there are relatively weak sources and 

 sinks for holes in the interior, but that the hole absorption and genera- 

 tion of the sources is very noisy compared to the recombination and 

 generation processes on the surfaces. If this is the state of affairs, most 



