Prof. H. L. Oallendar on Platinum Thermometry. 217 



hvdrogen and platinum thermometers, adopting my method 

 of enclosing the spiral inside the bulb of the air-thermometer. 

 The majority of their observations were taken while the tem- 

 perature of the instrument was slowly rising. This method 

 of procedure is very simple, but it is open to the objection 

 that the mean temperature of the spiral is not necessarily 

 the same as that of the gas enclosed, especially when, as in 

 their apparatus, the spiral is asymmetrically situated in an 

 asymmetrical bulb. If we take their observations in melting- 

 ice, in solid C0 2 , and in liquid air, which are probably in this 

 respect the most reliable, and calculate a difference-formula 

 in terms of pt, we shall find c' = *003621, d , = l'69. Calcu- 

 lating the values of t' by this formula, we find that all the 

 rest of their observations make the temperature of the plati- 

 num spiral on the average 1° higher than that of the gas. 

 This might be expected, as the temperature was not steady, 

 and the warmer gas would settle at the top of the bulb, the 

 spiral itself being also a source of heat. 



If we take their own formula, and calculate the equivalent 

 difference-formula, we find c' = '003610, d'= 1*79. This agrees 

 fairly well with the values found above, as they appear also 

 to have attached greater weight to the observations in C0 2 

 and liquid air. But, if we take the formula calculated by 

 Dickson (Phil. Mag. Dec. 1897), who attaches equal weight 

 to all their observations, we find c' = '003527, d'=2'43. The 

 excessive difference in the values of the coefficients deduced 

 by this assumption is an index of the inconsistency of the 

 observations themselves*. 



Behaviour of Pure Wire at Low Temperatures, — In the 

 case of ordinary platinum wire, with a coefficient c = '0035 or 

 less, the effect of the curvature at low temperatures of the 

 t, R, curve, as represented by the positive value of the dif- 

 ference-coefficient d, is to make the resistance diminish more 

 rapidly as the temperature falls, and tend to vanish at a point 

 nearer to the absolute zero than the fundamental zero of 

 the wire itself. When, however, the value of pt° is numeri- 

 cally less than 273°, the effect of this curvature would be to 

 make the resistance vanish at some temperature higher than 

 the absolute zero. If, therefore, we may assume that the 

 resistance ought not to vanish before the absolute zero, we 

 should expect to find a singular point, or a change in sign of 

 the difference-coefficient, at low temperatures. If this were 

 the case, it would seriously invalidate the difference-formula 

 method of reduction, at least at low temperatures, and as 



* Contrast the close agreement of Dickson's reduction in the case of 

 Fleming's observations. 



