ELECTRICAL NOISE IN SEMICONDUCTORS 



9()7 



of the noise will bo <>;eiienited in tlu> inlei'ior, hut a hole }»;enerate(l in the 

 interior will be much more likely to recombine on the sui'tace. The 

 (lotted (•ui'\'e of Fiii;. 10 has been calculated assumiii<>; a uniform dis- 

 tribution of noise sources throujihout the interior of the filament and 

 ('([ual and very larj>;e i'ecoml)ination constants for the two surfaces. It is 

 S(HMi that for this rase the reduction of lifetime i)i'(Ml()miiiates, ;ind there 

 is a monotonic decrease in noise witii increasint>; magnetic held. 



Experimental work has given ivsults which in most cases are in fair 

 (lualitative agreement with the calculated relations. Measurements for 

 three filaments, each of which had one high recombination and one low 

 recombination surface, are shown in Fig. 12. The recombination param- 

 eters, as shown on tlu^ curves, were of the order of ^ = 10 for one sur- 

 face, and \f/ = 0.5 for the other. The general shape of the curves is cjuite 

 similar to the calculated curves of Fig. 11. The maxima are of the right 

 order of magnitude, and occur at reasonable values of the field param- 

 eter $. The lack of detailed agreement between the measured and cal- 

 culated curves is not surprising, because the experimental conditions 

 did not fulfill the assumptions made for the calculations in several re- 

 spects. The filaments were neither wide enough nor long enough so that 

 edge and end effects could be overlooked. The recombination properties 

 of the surfaces could not be measured directly, but had to be estimated 

 from other filaments which had been similarly treated. One experi- 

 mental ciu've shows a secondary maximum on the opposite side of the 

 origin. This might indicate a defective portion of one surface having an 

 anomalous recombination constant. 



Experimental results are shown in Fig. 13 for four filaments, each of 

 which had nominally equal recombination (constants for the two sur- 

 faces. These may be compared with the calculated curves of Fig. 10. It 

 will be noted that the experimental curves are not symmetrical about 

 $ = 0. This lack of symmetry is probably due to dissymmetry in the 



zy 



-12 



12 



-8-4 4 8 



MAGNETIC FIELD PARAMETER, <|) 



Fig. 13 — Experimental magnetic etTect for similar surfaces. 



