104 Professor II. L. Callendar [March 10, 



method, wc are met by the difficulty that the platinum begins to 

 soften and stretch at a temperature considerably below its melting 

 point. Owing to the smallness of the expansion, a very slight viscous 

 extension produces a relatively large error. In the resistance method 

 it is not necessary to subject the wire to tension, and a small strain 

 would in any case produce an inappreciable error on account of the 

 very large increase of resistance with temperature. To obtain an 

 equal degree of accuracy by the calorimetric method (2), or the 

 thermo-electric method (8), it is necessary to use a furnace in which 

 relatively large quantities of platinum can be melted. This has been 

 done by Violle for method (2), and by Barus and Holborn and Wien 

 for method (3). The latter used a linear formula for extrapolation, 

 although their gas-thermometer experiments appeared to indicate a 

 cubic formula for temperatures below 1200° C. 



The temperature of the melting point of platinum on the platinum 

 scale by the resistance method (4) is approximately ]pt = 1350°, and 

 varies but slightly for different specimens of platinum. The result, 

 when reduced to the scale of the gas-thermometer by assuming that 

 the rate of increase of resistance diminishes uniformly with rise of 

 temperature (according to the usual formula of platinum thermometry, 

 which has been verified with great care at moderate temperatures), 

 gives a temperature of 1820° C. on the scale of the gas-thermometer. 

 It is not improbable that platinum may deviate slightly from this 

 formula at the extreme limit of the scale in the close neighbourhood 

 of its melting point, but the evidence for this result is at least as 

 good as that obtainable by any of the other methods. The observa- 

 tions are very easy and accurate as compared with the calorimetric 

 method, and it is not necessary to make any arbitrary assumptions 

 with regard to the formula of reduction, as in the case of the thermo- 

 electric method. 



As the accuracy of this formula has recently been called in ques- 

 tion, on what appears to be insufficient grounds, by certain German 

 and French observers, it is the more interesting at the present time 

 to show that it leads to a result which cannot be regarded as 

 improbable at the extreme limit of the scale. A different formula 

 has recently been employed by Holborn and Wien, and supported by 

 Dickson (Phil. Mag., December 1897). The writer has already given 

 reasons (Phil. Mag., February 1899) for regarding this formula as 

 inferior to the original, of which, however, it is a very close imitation. 

 The above observations on the melting point of platinum, if reduced 

 by Dickson's formula, would give a result 2 = 1636° C, which appears 

 to be undoubtedly too low as compared with the results of other 

 methods, however great the margin of uncertainty we are prepared 

 to admit in these difficult and debatable regions of temperature 

 measurement. 



It should be observed that the results of Violle by method (2) 

 are consistently lower than those given by the resistance method in 

 the case of silver, gold and copper. We should, therefore, expect a 



