356 Mr. Carl Barus on the 



38. Corrections are now to be applied for the thermal and 

 elastic volume-changes of glass. Direct measurements of the 

 elastic constants of my glass tubes, throughout an interval of 

 300° and 2000 atmospheres, are at present out of the question. 

 Hence I selected liquids of large compressibility. Fortu- 

 nately the correction is differential, and it will vanish in 

 proportion as the isometrics of glass approach those of the 

 above liquids (§ 43). 



For the expansion of glass throughout long intervals (0° to 

 300°), results of Dulong and Petit* and of Recknagelf are 

 available. The coefficients of the former are the smaller. I 

 will therefore accept them, seeing that inasmuch as the 

 thermal variations of the compressibility of glass are not 

 known, the smaller coefficients of expansion ($) are prefer- 

 able. Dulong and Petit found : — 



0° to 100°, /3 = 26/10 6 , 



0° to 200°, /3 = 28/10 6 , 



0° to 300°, /3 = 30/10 6 . 



From these the volume of the constant-volume tube in terms 

 of its capacity at 20° and 70°, respectively, are to be com- 

 puted. 



39. The compressibility of glass taken from Everett's tables 

 (/. c.) agrees fairly well with the values due to Regnault, and 

 more recently to AmagatJ ; but apart from the fact that com- 

 pressibility increases in marked degree with temperature, little 

 further can be conjectured. Hence, for want of data, the 

 volume of the glass tube under any hydrostatic pressure p 

 (atm.) must be taken as 1— pxl0 6 jk, where & = 4xlO n , 

 irrespective of temperature. 



The present and preceding paragraphs suffice for the com- 

 putation of a table of double entry (the arguments being p 

 and 0), from which the volume-changes Bv of the glass con- 

 stant-volume tubes, corresponding to any of the pairs of 

 values of Ap and A0 in Table VIII., may be at once desig- 

 nated. A list of the values Bv is given in Table IX., corre- 

 sponding in arrangement with Table VIII. The curious 

 feature of these data is that Bv, as here computed, is some- 

 times positive and sometimes negative. 



40. Having given these data, it is finally necessary to 

 find what correction Bp is to be added to Ap in order to com- 



* Dulong and Petit, Ann. de Ch. et Phijs. [2] vii. p. 113 (1817). 

 t Kecknagel, Ber. k. Bayrischen Acad. [2 J p. 327 (1866). 

 \ Amagat, Comptes Rendus, cvii. p. 618 (1888) ; cviii. p. 1199 (1889) ; 

 he finds l/& = 2405/10 9 . 



