CHEMICAL INTERACTIONS AMONG DEFECTS IN Ge AND Si 599 



and the measured diffusivity will appear larger than the real diffusivity. 

 This condition can be partially corrected by etching the surface chemi- 

 cally until it is fairly smooth. 



When dealing with such thin layers, the no-space-charge assumption 

 becomes invalid and the diffusion potential ought really to be considered. 

 Considering all the difficulties, i.e., concentration dependence of diffusion 

 coefficient, possible existence of space charge, and roughness of surface, 

 it is apparent that only qualitative effects are to be looked for in the dif- 

 fusivities which have been measured. 



The most that can be predicted is that for specimens containing a 

 given amount of acceptor, the measured D (some average quantity) 

 should be less than Do , the disparity increasing with decreasing tem- 

 perature. At high temperatures D should converge on Do . Furthermore, 

 at a given temperature D should decrease with an increase in concen- 

 tration of acceptor. These tendencies are in line with the idea that reduc- 

 tion of temperature or increase of doping leads to an increase in pairing. 



Runs Avere carried out on specimens etched with Superoxol^^ at the 

 temperatures 25°, 100°, and 200°C. In the 25°C run the wafer was allowed 

 to remain in the measuring apparatus under the two probe points in air, 

 and S was measured from time to time. At 100°C the specimen was 

 immersed in glycerine containing a few drops of HCl, the temperature 

 of the bath being controlled. Periodic removal from the bath facilitated 

 the measurement of 2. At 200°C glycerine was again used as a sink for 

 lithium, the sample being removed periodically for measurement. 



Fig. 23 illustrates some typical plots of 2/So versus \/t. They are 

 all satisfactorily straight. Fig. 24 shows a plot of log Do against \/T, 

 extrapolated from the data of Fuller and Severiens.^^ In this illustration, 

 Aalues of log D (obtained from the above measurements by determining 

 the slopes S and employing (11.12)) are also plotted at the temperatures 

 of diffusion. For ■& the case of complete pairing was assumed. 



The first thing to note is that the points for log D all lie below log Do 

 except at 200°C and satisfy the qualitative requirement outlined above.* 

 Moreover they drop further below log Do as the temperature is reduced, 

 ^\■hile at 200°C they have almost converged on log Do . 



The results for zinc are particularly interesting. Zinc is supposed to 

 have a double negative charge in germanium. Hence we would expect 

 very intense pairing to occur. This is indicated in the difi'usion data 

 where the sample containing zinc at the rather low level, A^^ = 2.7 



* The long range nature of the interaction forces becomes evident when one 

 considers that the diffusivities are being altered by impurity (acceptor) concen- 

 trations of the order of 1 part per million. 



