812 



THE BELL SYSTEM TECHNICAL JOURNAL, JULY 1952 



designated B and C, the former being from the same soiu'ce as "sihcon 

 B" referred to in the paper by Scaff et al.^ A typical analysis is given 

 therein. The C silicon was from another source and a spectroscopic analy- 

 sis indicated it was somewhat more impure than B thus agreeing with 

 observed differences in its electrical and optical characteristics. An opti- 

 cal variation of considerable interest is shown in Fig. 10 where the 

 spectral transmittance of the two grades of silicon is compared in the 

 infra-red for polished plates each 0.0195'' thick. 



The transmittance of B decreases a little with increasing wavelength 

 but C goes down much more. Both however, start to get transparent at 

 about the same point, 1.1 ^ and also show corresponding absorption 

 bands superimposed on the main curve. Briggs* has compared the trans- 



0.2 



6 



16 



18 



20 



« 10 12 14 



WAVELENGTH IN MICRONS 



Fig. 10 — Spectral transmittance of B and C grades of silicon, polished 

 0.0195" thick. 



mittance from 2 ^ to 12 /x of the A and B silicons in Scaff 's paper where 

 the former was much more impure than the C grade. The absorption 

 of the A sihcon increased so rapidly out in the infra-red that a much 

 thinner sample was used for the measurements than for the B material. 

 If this difference in thickness is allowed for, the effect of impurity is 

 very striking. 



The spectral response of large area cells made of the B and C materials 

 and bombarded with 1000- volt helium positive ions is shown in Fig. 11. 

 The two curves are similar in shape except the one for C silicon is some- 

 what narrower and in addition is shifted toward the blue. Both have 

 some of the secondary humps noted previously. 



All the cells shown in this paper have indicated a long wave limit of 

 about 1.2 jLt. Actually some response can usually be detected out to about 

 1.3 \i. Measurements made some years ago on the internal barrier units 

 also gave a limit around 1.3 /x but relatively more response at 1.2 /i with 

 peaks close to 1.10 /i- This difference is reasonable because light was 



