488 UNIVERSITY OF VIRGINIA PUBLICATIONS 



the liquid was removed after diffusion had proceeded so far as seemed 

 desirable. Experience showed tliat a period of from six to eight days was 

 required to give a satisfactory quantity of iodine in the upper layers. All 

 of these experiments were performed in a gas heated thermostat that kept 

 constant temperature to within ±0.1° if the gas pressure was constant. 



The first experiments made by the writer were made with the apparatus 

 just described; except that an electrically heated thermostat was sub- 

 stituted for the one heated by gas. This thermostat was very sensitive 

 and would hold the temperature constant to within about 0.2° for a week 

 if the temperature of the room did not vary more than 5 or 6° — and if 

 the current was not cut off in the interval. As we had no cellar at our 

 disposal, the temperature frequently did vary far more than this, and in 

 consequence we frequently lost set after set of experiments. Also the 

 current was frequently interrupted by various accidents to the city power 

 plant — repairing, rewiring, and so forth. 



New apparatus used. It was decided to use a cylinder much shorter 

 than before, and to have only two, instead of four, layers. That is, equal 

 quantities of iodine solution and of pure KI solution were used. The 

 cylinder was of the same general type as before (see fig. 1) but somewhat 

 smaller in diameter and very much shorter. The length of the new 

 cylinders actually used was about 7.5 cms., and the pipette used to measure 

 each layer held about 15 cc. The best results can be obtained with this 

 apparatus when the time is from three to four days. The advantages 

 of this shortening of the time is obvious and very great, but it is not at- 

 tained without some disadvantage. The four layer system gives theoreti- 

 cally three independent equations, from which to evaluate the diffusion 

 constant. It is true that in general some of the values are uncertain for 



mathematical reasons, i.e., the value of — corresponding to a given titra- 

 tion figure changes very little for quite large changes in the titration 

 figure. In spite of this the four layer system usually gives more than one 

 independent value for "k" in a single experiment. In the two layer 

 system, the onl.y check is to run separate experiments, either at different 

 times or with different cjdinders. This of course means more work, j^et 

 we found that the results were much more satisfactory than ■with the four 

 layer system. 



The method of procedure in this case was the same as for the four 

 layer system, the iodine solution forming the bottom layer. In general 

 four cyhnders were filled at the same time. After a period of from three 

 to four days the top laj^er was removed as in the four layer case. As 



