ELECTRICAL NOISE I\ SKMICONDUOTORS 0")") 



led to a hypothesis regarcHiis the noise meclianism, which is by no means 

 a compk^tc^ exphination, hut which nuiy h(^ a useful step in that dii-ec- 

 tioii. This hypothesis resulted largely from tlie experimental woik, hut 

 it seems worth while to desci'ihe it first to h(>lp appreciate the siji;nifi- 

 cance of some of the experimental results. 



It has been observed that in many semiconductor structures the 

 noise \'oltag"e is approximately proportional to the dc bias current. 

 This relation sufi;jj;ests that tiie noise is tiu^ result of fluctuations of the 

 conduct i\ity of tlie materiah wiiich modulate the bias current and j)ro- 

 duce a tiuctuatino; voltage across tlie specimen. Such fluctuations in 

 conducti\-ity could result from variations in concenti'ation of the mi- 

 nority carrier (holes in /;-type material, electrons in p-type). The mag- 

 nitude of the ol^served noise and the type of spectrum seem to demand 

 that the fluctuation be coarse-grained in time to a much greater extent 

 than could i)e accounted for by random statistical fluctuations of carrier 

 density. Experiments of Haynes^ on lifetime and transit of injected car- 

 riers in rods of germanium have occasionally indicated finite sources of 

 minority carriers in the material. Our hypothesis is that such sources 

 of carriers are rather generally distributed over the material (although 

 mostly too small to be noticed in experiments of the Haynes type), and 

 that their acti^•ity is being modified at a slow rate by some unspecified 

 local influence in a suitable way to agree with the observed noise spec- 

 trum. 



The experiments described below involving noise correlation phenom- 

 ena and the effect of a magnetic field on noise point strongly to an im- 

 portant role for the minority carrier in the noise mechanism, and hence 

 strongly suggest some such hypothesis as that just described. 



III. NOISE IN SINGLE CRYSTAL FILAMENTS 



It was found se^'eral years ago that a filament cut from single crystal 

 germanium of high purity exhibits noise well above Johnson noise when 

 a dc current is flowing in it. It is not clear whether this noise arises in 

 the body of the material or on the surface, but to date no method of 

 preparing the sample has eliminated this noise, and it is a prominent 

 feature even at bias fields as low as 10 volts per centimeter. This noise 

 seems to have most of the characteristics of the noise in diodes and 

 transistors: it has the l/f spectiiim, is current dependent, and is quite 

 stable with time. It has been the subject of considerable study in the 

 hope that a better understanding of it would illuminate the whole field 

 of semiconductor noise. 



