1903.] of the Diffusion of Salts in Solution. 215 



Ci, the top stratum passes down through the tube D and is reserved 

 for weighing. In the same manner all the remaining strata are 

 drawn off. 



The specific gravity of each stratum is ascertained. 



These specific gravities when plotted give a curve from which * can 

 be calculated. 



Fig. 2 shows how this was done with a solution of chloride of 

 sodium of density 1026-892. The contents of the 25 c.c. sp.g. bottle 

 weighed 25 '6723 grammes. I assumed (which is not quite correct) 

 that the weight of salt was 0-6723 gramme, and plotted this as 

 300 mm. 



In the same way the weights of salt in 25 c.c. of all the other strata 

 were plotted. The small circles show the curve so obtained. And 

 from this curve K can be calculated by means of Fourier's equation 



u = 





. ( 4 J)' + i . l( * Y- etc. ). 



*Wlrf/ -V2v/K// / 



In this particular case diffusion went on for 11 1J hours and the 

 average value of nt for 111J hours was 3'95, therefore K (for one hour) 

 = 3-95/1 llj = 0-0355. 



This equation gives a curve which is symmetrical with respect to 

 the line AB. The small crosses were obtained by marking the small 

 circles on tracing paper and turning the tracing paper through 180 

 round the point O. The near agreement of the two curves is a good 

 test of the accuracy of the observations. 



The fluids and the room where the experiment was made were at 

 about 14 C, but it was not possible to keep the temperature of the 

 room constant. The vessel A was very completely enclosed in 

 flannel. 



The labour of calculating */ from any observed curve is obviated by 

 means of the curves shown on fig. 3. Here the values of u at the 

 distances 1'25, 2*5, 3'75, and 5 cm. were calculated from the above 

 equation, for successive values of */ from 1 to 10, and plotted to the 

 scale of 300 mm. = unity. For example, the figure 98 -45 (the average 

 of]98-2 and 98 '7 see fig. 2) where x = 1'25 cm. is seen by the curve to 

 give the value 3'93 for Kt. Inasmuch as these calculations of u are 

 very laborious, I have given the values in the following table. Each 

 must be multiplied by 300 if 300 mm. is taken as unity. With such 

 curves the value of Kt is obtained by inspection. 



