806 



THE BELL SYSTEM TECHNICAL JOURNAL, JULY 1950 



Table IX — Experimental Design of Randomized Block 



Experiment 



Note: Shelf represents storage for 24 hours. 



used is a 5 X 6 randomized block experiment with multiple subgroups.^* 

 Table IX shows the general plan of the experiment. The six columns 

 represent different etch treatments, and the five rows represent some pos- 

 sible variations in storage conditions. Each subgroup represents five 

 transistors, and the experiment represents a total of 150 transistors made 

 on germanium from the same zone-leveled slice, given 30 different treat- 

 ments. Although nine measurements were made for each transistor, the 

 figure of merit appeared to be most significantly dependent on the 

 treatments. 



As expected from the results already quoted, the major variability 

 was found in units formed on surfaces freshly treated with HF, with 

 considerable improvement in formability during storage. However, 

 the looked for influence of storage ambients does not appear when the 

 column F has been removed from consideration. One concludes that the 

 variation between treatments is small, and the effect of ambient is even 

 less than the effect of the treatments. Thus when surface treatment does 

 not vary to extremes, the effect of storage ambient is relatively minor. 

 Thus variations found in such experiments as exemplified on the manipu- 

 lator line in Fig. 13 must be attributed to a still unknown factor. 



4.2.5 Effect of Contamination Before Etching 



Since etching removes the damaged surface and is usually done with 

 highly corrosive materials, it seems unlikely that any contamination 



