CHEMICAL INTEKACTIONS AMONG DEFECTS IN Ge AND Si 595 



represents a formidable task. Although work along these lines is being 

 done we shall content ourselves, in this article, with a less quantitative 

 approach. The following plan has been followed. 



A rectangular wafer of semiconductor uniformly doped with ac- 

 ceptor to the level, Na , is uniformly saturated with lithium to a level, 

 Nd , slightly less than Na • Thus, the resulting specimen is well compen- 

 sated but not converted to n-type. Lithium is then allowed to diffuse 

 out of the specimen, and because of the thinness of the wafer, this 

 process may be regarded as plane-parallel diffusion normal to its large 

 surfaces. Low resistivity p-type layers therefore develop near the sur- 

 faces. If the thin ends of the wafer are put in contact with a source of 

 current, current will flow parallel to its axis, so that the equipotential 

 surfaces will be planes normal to this axis. The flow of current will be 

 one dimensional because the inhomogeneity in lithium distribution oc- 

 curs in the direction normal to its flow (see Fig. 21). 



If two probe points are placed at a fixed distance apart on the broad 

 surface of the wafer (see Fig. 21), then the conductance measured be- 

 tween them is a reflection of the total number of carriers in the low 

 resistivity layers, i.e., a measure of the total amount of lithium which 

 has diffused out. A more detailed connection between this conductance 

 and diffusivity is derived in Appendix E. For the moment, however, 

 attention will be confined to the description of the general plan of ex- 

 periment. 



According to the formulas derived in the early parts of this section, 

 and also to (B14) and (B15), the diffusivity is something like Do/2 in the 



CURRENT 



CURRENT, 

 (I) 



Fig. 21 — Diagram illustrating measurement of dependence of diffusivity on 

 ion pairing (see Section XI). 



