62 PROCEEDINGS OF THE AMERICAN ACADEMY. 



from the laboratory stock, and the latter was made for me by Mr. J. B. 

 Churchill, whom I wish to thank for j^roviding me with a very consider- 

 able amount. Since the temperatures were read on a mercury ther- 

 mometer in the cell itself, no attempt was made to attain absolute iiurity 

 in the boiling substances. 



The most of the cells measured were liable to variations due to other 

 causes than change of temperature of about 0.001 volt. This was the 

 accuracy of measurement aimed at. As it was found to require a change 

 of about 5° ill temperature to cause a change of 0.001 volt, the tempera- 

 tuies were read on a thermometer graduated to degrees, and considered 

 constant as long as tiiey did not vary more than 1°. 



In order to standardize the thermometer used, it was tested with pure 

 naphthalene. According to Crafts,* uai)hthalene boils under 756 mm. 

 pressure at 21 7°. 9. The thermometer used for these experiments regis- 

 tered 218" as the boiling point under tlie same pressure. Tlie actual 

 readings of the thermometer were therefore taken as correct.f 



Change in Concentration. — In order to render the danger of change 

 ill concentration in the neighborhood of the electrodes as small as pos- 

 sible, the tubes A were given the shape shown in the figure. In the 

 preliminary measurements, however, there was still trouble on this ac- 

 count. This was due to the fact that the salts were put into the tubes in 

 the unmelted state. Air was unavoidably retained when the salt was 

 fused, which must necessarily be blown out after the tubes were lowered 

 under the level of the connecting solution, and the latter was thus 

 brought into the tubes. As the potentials were measured by means of a 

 capillary electrometer, even when the air filled the whole cross-section of 

 the capillary, the measurement could be made; the value was, however, 

 alwavs different from the true one. In order to avoid this trouble due 

 to air bubbles, the filling of the cell tubes with the pulverized salts was 



* Amer. Chem. Journ., V. 307, 1883-84. 



t Graebe (Annalen [Liebig's], CCXXXVIII. 362) finds tlio boiling point of 

 diphenylamine to be 302°. The earlier determinations of Hofmann, Girard and 

 Wilson, and Kreis, etc., are evidently incorrect, as Graebe standardized his ther- 

 mometers with benzophenone, according to the figures of Crafts. Whether the 

 difference between Graebe's 302° and the reading 298° of my thermometer in the 

 cell surrounded by the diphenylamine was due to impurities in the latter, to a 

 change in the value of the marked degree between 218° (the point found to be 

 correct), or to the fact that the thermometer in the cell was at a little lower tem- 

 perature than the surrounding vapor, I have not attempted to determine, since a 

 change of even 4^ in temperature would cause a change in voltage in the cells 

 measured of less than 0.001 volt. 



