190 PROCEEDINGS OF THE AMERICAN ACADEMY. 



obtained by the weight of solution employed and by the specific gravity 

 of the cold solution referred to water at the same temperature, the 

 volume corresponding to the observed cell-coustant ratio is obtained. 



YII. Preparation of the Scbstaxces and Solutions. 



The sodium chloride used was purified by precipitation with hydro- 

 chloric acid gas. It was then washed with hydrochloric acid, dried, and 

 finally ignited until decrepitation ceased. 



The potassium chloride was precipitated twice with hydrochloric acid 

 gas, crystallized from hot water, dried, and finally ignited. 



Solutions were made up, by weighing out the salts, so as to be almost 

 exactly 0.1 and 0.01 normal at 4°. The solutions of smaller concen- 

 tration were made by diluting the 0.01 normal one with the help of two 

 graduated flasks. 



The equivalent weights used are as follows: K = 39.14, Na = 23.05, 

 CI = 35.46. All weighings were reduced to a vacuum. 



The water used throughout this investigation was prepared by redis- 

 tilling ordinary distilled water, to which alkaline permanganate solution 

 was added, from a steam-jacketed copper still with a tin condenser. 

 The first quarter of the distillate was rejected, and the following portions 

 were condensed hot (between GO'' and 90°). The water had a specific 

 conductivity of (0.7 to 1.0) X 10~^ reciprocal ohms. 



VITI. Discussion of the Systematic Errors and their 



Correction. 



Errors Affecting the Specific- Volume Values. — 1 . In calculating the 

 specific-volume, the volume of the bomb was directly determined at about 

 135°, as described in Section VI, and the expansion of the metal from 

 this point to the temperatures of the experiments was corrected for. An- 

 drews,* working with "soft" cast steel, which corresponds to the mate- 

 rial from which the bomb was constructed, found the mean coefficient of 

 cubical expansion between 100° and 300° to be 0.0000450; and this 

 value was adopted for the corrections. The difference between his steel 

 and that used in the bomb can hardly be great enough to cause an ap- 

 preciable difference in the coefficient of expansion, since his values for two 

 steels as different as Bessemer steel with 0.15 per cent combined carbon 

 and cast steel with 0.45 per cent differ by only 6 per cent ; and an 

 error of even 6 per cent in the coefficient of expansion would produce 



* Proc. Roy. Soc, 43, 299 (1887). 



