OF ARTS AND SCIENCES. 231 



XVI. 



CONTRIBUTIONS FROM THE PHYSICAL LABORATORY OF THE 

 MASSACHUSETTS INSTITUTE OF TECHNOLOGY. 



XXX. — BOILING POINTS OF NAPHTHALINE, BENZO- 

 PHENONE, AND BENZOL UNDER CONTROLLED 

 PRESSURES, WITH SPECIAL REFERENCE TO 

 THERMOMETRY. 



By S. W. Holman and W. H. Gleason. 



Presented January 11, 1888. 



The employment of the melting and boiling of various substances as 

 a means of testing or of graduating thermometers at temperatures 

 above 100° C. is a practice of long stauding, especially among chemists. 

 But it will be readily conceded that the results have been in general 

 but roughly approximate, owing to several causes of error, e. g. im- 

 perfect purification of substances, faulty apparatus (permitting under 

 or over heating), incomplete systems of thermometry, and errors in 

 the values assumed as the meltiug and boiling temperatures, these 

 arising, in turn, from causes similar to those just mentioned. 



The value of steam as a means of fixing one point on the thermo- 

 metric scale comes in part from the facts that water does not change com- 

 position on boiling at ordinary pressures; that it can be readily obtained 

 in a state of sufficient purity, so that the temperature of its vapor, or 

 rather of a clean thermometer placed in its vapor, can be relied upon 

 as sensibly reproducible under a given pressure ; and that this tem- 

 perature under more than the ordinary range of atmospheric pressure 

 has been measured (by Regnault and Magnus) with sufficient accuracy 

 for thermometric uses. The primary measurement of temperatures 

 above 10L)°C. is a process of extrapolation from the ice and steam 

 points, and thus possesses the liability to error common to all extra- 

 polation, the magnitude of the error depending upon the method, in- 

 struments, and skill employed. It is obvious, therefore, that, whatever 

 be the system of thermometry, a decided gain in accuracy and con- 

 venience would accrue to the art of temperature measurement if by 



