954 THE BELL SYSTEM TECHNICAL JOURNAL, SEPTEMBER 1952 



ture on noise behavior. Such experiments have been rather unsatisfac- 

 tory because the changes in impedance and gain characteristics as a 

 function of frequency are of the same order as the changes in noise prop- 

 erties. This makes the interpretation ambiguous. B}^ and large, such 

 experiments suggest that changes in noise with temperature are rather 

 small, perhaps of the order of the change in absolute temperature, and 

 not at all like the exponential changes associated with a diffusion process. 

 This observation does not necessarily rule out a diffusion-like noise 

 process; it might indicate merely that we are not looking at the right 

 part of the spectrum to observe exponential changes with temperature. 



II. A HYPOTHESIS REGARDING THE NOISE MECHANISM 



Considerable work has been done on the theory of current-dependent 

 noise having a 1/f spectrum. Among the earliest was that of Schottky^ 

 in connection with flicker noise in vacuum tubes. He considered the 

 arrival of foreign atoms on the emitting surface of the cathode as a 

 random series of events governed by a diffusion law with a charac- 

 teristic time constant, and arrived at a 1/f rather than a 1/f spectrum, 

 and a highly temperature sensitive process. Surdin pointed out that by 

 postulating a series of decay processes with suitably distributed time 

 constants a 1/f spectrum could be achieved. From physical arguments 

 regarding the emission process from cathodes, INIacfarlane' obtained a 

 range of relaxation times and a 1/f spectrum, in a process which was 

 highly temperature dependent. Richardson gave a very general anal- 

 ysis of the noise properties of sj^stems in which the conductivity was 

 governed by a diffusion process. One conclusion was that a geometrically 

 simple diffusion process in one, two, or three dimensions could not lead 

 to 1/f spectrum, although by some highly specialized assumptions about 

 the geometry of a contact surface he was able to obtain such a spectrum. 

 DuPre, in considering a hj^pothesis somewhat resembling that of Sur- 

 din, showed that the required range of activation energies was phys- 

 ically reasonable, and that the assumptions could be set up in such a 

 way as to make the process relatively temperature independent. Several 

 of the abo^•e authors and Van der Ziel discuss the physical basis for 

 applying flicker noise theory to the noise in semiconductors. Although 

 this theoretical work has contributed a great deal to distinguishing be- 

 tween suitable and unsuitable mechanisms, there is still no specific 

 physical theory of noise in semiconductors which can be tied in a (juan- 

 titative manner to experimental results. 



The experimental work described in the remainder of this paper has 



