G58 THE BELL SYSTEM TECHNICAL JOURNAL, MAY 1956 



Table I — Average Resistivity Variations 

 (A) Along length axis. Grand Length Average ± 10%. 



Growth Rate 

 Mils per Second 



0.9 

 0.8 

 0.09 



n-Type 



±% 



9.9 

 7.6 

 9.0 



No. of Ingots 



27 



12 



108 



p-Type 



±% 



10.9 



17.4 



9.3 



No. of Ingots 



33 



16 



137 



Average 



± /o 



10.4 



13.2 



9.2 



generally coarse — • changing over distances 2 to 5 times larger in dimen- 

 sion than the 50 mil dimension in question. Thus the p data summarized 

 here should give a reasonably valid representation of the true p variations 

 in the ingots measured. 



Table I summarizes the resistivity variations recorded as percentages 

 of the mean resistivity of each ingot. These variations are separated into 

 those observed (a) along the length axis and (b) over the cross-section, 

 for the different growth conditions and resistivity types. 



It is readily seen that the average variation along the length, about 

 ±10 per cent, is larger than the average cross-sectional variation. The 

 variations are not systematic along the length of the ingot and are 

 chiefly due to fluctuation in the length of the liquid zone. An appreciable 

 part of this variation is due to the effect, mentioned earlier, of discon- 

 tinuities in the unmelted charges between 1 inch lengths of crystals 

 that were being releveled. A smaller length variation of p, about ±7 

 per cent, was observed in those ingots grown from continuous charges. 



Part B of the table shows that the variation of p over the cross-section 

 is sensitive to the growth rate in the range covered. For slow growth, it 

 is small, and one would reasonably expect that if further improvement 

 in p variation were required, it should first be sought by improving the 

 control of the zone length. 



(2) Macro Perfection 



Macro perfection of the pilot production product is extremely high. 

 There were essentially no cases of polycrystallinity, or twinning, except 



