598 THE BELL SYSTEM TECHNICAL JOURNAL, MAY 1956 



of D. The only exception to this statement occurs in connection with high 

 temperature experiments (above 200°C.) where both pairing and the 

 diffusion potential are of little consequence. The mere fact that 2/2o 

 plots as a straight line against -s/t is not evidence for the constancy of 

 D. In Appendix E it is shown that a straight line will result, even when 

 ion pairing is important, provided that the diffusion potential is based 

 on the no-space- charge condition, i.e. provided that D varies only 

 through its dependence on Nd • 



On the other hand, the last statement implies that the existence of a 

 straight line relationship is evidence that the diffusion potential has at 

 least been minimized. 



The most careful experiments were performed in germanium doped to 

 various levels with gallium, indium, and zinc as acceptors. The ger- 

 manium specimens were cut in the form of rectangular wafers of ap- 

 proximate dimensions (1.25 cm X 0.40 cm X 0.15 cm). Fresh lithium 

 filings, were evenly and densely spread on one surface of the wafer, and 

 alloyed to the germanium by heating for 30 seconds at 530°C in an at- 

 mosphere of dry flowing helium. Then the other surface was subjected 

 to similar treatment. 



After this the specimen was sealed in an evacuated pyrex tube and 

 heated at a predetermined temperature for a predetermined period of 

 time. The temperature was chosen, according to Fig. 5, so that the 

 saturated specimen would still be p-type and just barely short of being 

 fully compensated. Also attention was paid to the problem of avoiding 

 precipitation on cooling. The time of saturation was determined from an 

 extrapolation of the known lithium diffusion data, in germanium, of 

 Fuller and Severiens^^ which is plotted in Figure 19 for the range ex- 

 tending from about 0° to 300°C. 



After saturation the sealed tube was dropped into water and cooled, t 

 It was opened and the wafer ground on both sides, first with No. 600 ' 

 Aloxite paper, and then with M 303)^ American Optical corundum 

 abrasive paper. The final thicknesses of the specimens ranged from 0.025 ' 

 to 0.075 cm, the thinnest samples being used for the runs at the lowest 

 temperature. 



If the specimen is quite thin and highly compensated it is possible in 

 principle to measure very small diffusivities (as low as 10~ cm /sec) i 

 within a period of several hours. This is so because the low resistivity 

 layer formed near the surface, although thin, will carry a finite share of 

 the current in thin compensated specimens. On the other hand, additional 



o 



difficulties arise. Diffusion layers as small as 100 A may be involved. If 

 the surface is microscopically rough, diffusion will not be plane-parallel; 



