34 HEAT. 



ences they have introduced are very slight. If we do not require very 

 great accuracy, we may take 



a = -0001800 

 /3= -00000002. 



Callendar * has modified the second method used by Regnault by 

 employing six pairs of hot and cold columns placed in series as repre- 

 sented diagrammatically in Fig. 25A, the successive columns being alter- 

 nately cold and hot, as marked by C and H. If the mercury when in 

 equilibrium stands at a in the gauge-tube connected tfl the first cold 

 column, and at z in the gauge-tube connected to the last hot column, the 

 difference of level to be measured, represented by a', z, will be six times 

 that due to a single pair of hot and cold columns. As the columns were 



f 



c d g l h. 



I o 



FIG. 25A. 



FIG. 25B. 



nearly 2 m. long in place of 1*5 m., the length of Regnault's columns, 

 the expansion was nearly eight times as great. In the actual apparatus 

 the cross tube ef was doubled back, so that fg lay behind be, and ih 

 behind ed, and so on, as shown in plan in Fig. 25B. All the cold columns 

 were placed in one tube, and all the hot columns in another, the tubes 

 containing oil rapidly stirred. One was surrounded by ice and the 

 other was electrically heated. For further details we refer the reader 

 to the original paper. 



Callendar found that if O a t is the mean co-efficient of expansion 

 between C. and t C., then 



10 10 x O a t = 1805553 + 12444^/100 + 2539^/10000. 

 For approximate work we may put 



10 8 x O a,= 18006 + 2*. 



* Phil. Trans., A. 211, p. 1, 1911. 



