The next thing to decide on was, of course, the underlying principle 

 to make use of. The only one that appeared to warant development 

 was that of using- a reagent, expelling the moisture with heat, condens- 

 ing the vapor atid collecting it in a vessel graduated to read directly in 

 percentages. 



The next points to settle were the best form of apparatus and the 

 best inexpensive reagent to use. 



At the outset we determined to use metal in the apparatus, so far 

 as possible, rather than glass, owing to the danger of breaking the latter, 

 and to use no rubber connections. We thought to make the condenser 

 and heating flask of aluminum, if possible, owing to its many desirable 

 qualities. As no satisfactory solder for aluminum has, as yet, appeared 

 we could not attach an evaporating cup to the condenser, unless we used 

 some form of shoulder-and-nut connection, which did not seem altogether 

 desirable, nor could we attach an inner condenser tube to an outer water 

 jacket unless we used rubber stoppers at the ends. Further, upon 

 experimenting to determine the effect, if any, of reagents upon sheet 

 aluminum, we found that the aluminum became roughened too much to 

 warrant its use as an inner condenser tube. 



We then tested the effect of reagents on other metals, finding that 

 block tin showed no ill effect whatever. We decided then to use a block 

 tin tube for the inner condenser tube. This was sufficiently cheap, and 

 could be easily soldered to the outer water jacket and to the evaporating 

 cup. We had the outer water jacket made of thin sheet brass. The 

 best size of tin tubing to use had to be determined by experiment and also 

 the size of the water jacket. 



W T e decided that to insure complete condensation of the water, with- 

 out making the condenser too cumbersome, the vapor should be driven 

 downward instead of upward. This would thus insure the lower part of the 

 condenser being cold, since the warmed water would remain at the top. 

 We wished further to have a condenser large enoupm to hold sufficient 

 water not to need refilling under average conditions in a factory. 



The object of using a reagent was to have something to keep the 

 butter from foaming too much during the heating process, to help carry 

 the condensed moisture out of the inner condenser tube, and to wash the 

 tube out thoroughly afterward. We tested the use of amyl butyrate at 

 this stage, since it was almost insoluble in water and was sufficiently 

 volatile for part of it to be driven off with the moisture, and at the same 

 time had a boiling point sufficientlv high for part of it to remain after 

 all the moisture was expelled from the sample, and pass over later to 

 serve as a wash for the inner condenser tube. We were unable to pro- 

 cure a suitable metal evaporating cup at first, and in the meantime used 

 a glass flask, connecting it with the condenser by means of a perforated 

 rubber stopper. We used the condenser in a slanting position at first, 

 but found that in this position the inner tube did not become thoroughly 



