200 Prof. H. L. Callendar on Platinum Thermometry. 



each of which is used for a comparatively limited number of 

 determinations. In such a case the trouble of drawing the 

 separate curves, with sufficient care to be of use, would more 

 than counterbalance the advantage to be gained by the 

 method. 



Hey cock and Neville's Method. — In order to avoid this 

 difficulty Messrs. Heycock and Neville, in their classical 

 researches at high temperatures * devised an ingenious modi- 

 fication of procedure, which has given very good results 

 in their hands, but is not quite identical with the simple 

 difference-formula. They described a difference-curve in the 

 usual manner, giving the value of the difference in terms of 

 ft as abscissa for a standard value rf = l*50 of the difference- 

 coefficient. The appropriate values of d were determined in 

 the case of each pyrometer by the S.B.P. method. In re- 

 ducing the observations for any given values of pt and <7, the 

 value of the difference corresponding to pt was taken from 

 the curve for J = l*50, and was then multiplied by the factor 

 d/1'50 and added to pt. This method is very expeditious 

 and convenient, and gives results which are in practical 

 agreement with the pure difference-formula, provided that, 

 as was almost invariably the case in their observations, the 

 values of d do not differ materially from the average 1*50. 

 If, however, the pure difference-formula is correct, the method 

 could not be applied in the case of values of d differing con- 

 siderably from the average. The difference between the 

 methods cannot be simply expressed in terms of either pt or t 

 for considerable variations in the value of d. But for a small 

 variation Sd in the value of d in the vicinity of the normal 

 value, it is easy to show that the difference St between 

 the true value of t as given by the difference-formula 

 t—pt=.dp(t), and the value found by the method of Heycock 

 and Neville, is approximately 



8t=8d(dt/dpt-l)p(t)t. 



Neglecting the variation of d entirely, the error would be 

 B r t=Sd(dt/dpt)p{t). 



For example, at £ = 1000°, p(t)=90, (d£/dpf) = l-40, we 

 should find for a variation of d from 1*50 to I'60, the values 

 & = 3°-8 (H. & N.), and 8't =12°'8 (variation neglected). 



This is an extreme case. In the observations of Heycock 

 and Neville, the values found for the coefficient d seldom 



* Trans. Chem. Soc. Feb. 1895, p. 162. 



t The value of dt/dpt at any point is readily found by differentiating 1 

 the difference-formula (2), dpt/dt=l-(t/50- ljd/100. 



