522 Prof. H. L. Callendar on a 



and would permit the attainment of a much higher order of 

 comparative accuracy than the gas-thermometer. 



Selection of the Standard Platinum. 

 It is desirable to select a particular sample of platinum for the 

 standard, because experience has shown that the scale may be 

 most accurately reproduced in this manner. Thermometers con- 

 structed of the same sample of wire agree so closely throughout 

 the scale that their differences are almost, if not quite, beyond 

 the limits of observation. The standard ohm is similarly defined 

 as the resistance of a particular coil at a given temperature, 

 because the ohm can be more accurately reproduced by copying 

 such a standard, than by constructing a mercury resistance, 

 or referring to absolute measurements with a Lorenz apparatus. 

 It would be possible, however, to reproduce the platinum scale 

 with greater ease and accuracy than the standard ohm, even 

 if the standard thermometers and all their copies were lost. 

 It has been shown by many experiments with different quali- 

 ties of platinum — ( 1) That considerable variations in the purity 

 of the platinum (e. </., a variation of 20 per cent, in the coeffi- 

 cient) produce very slight variations (less than 1 per cent, at 

 1000° C.) in the value of pt. (2) That variations of purity 

 so small as to defy chemical analysis can be easily detected 

 by their effect on the temperature-coefficient of the wire. 

 (3) The identity of any new sample with the original standard 

 could be further tested by reference to well-determined fixed 

 points such as the B.P. of oxygen and sulphur, and the F.P. 

 of gold. It is conceivable, of course, that there may be 

 exceptions. It is not possible to prove a universal negative 

 by any number of experiments ; but our confidence in the 

 general truth of the experimental results (1) and (2) is 

 considerably strengthened by the knowledge that similar 

 results have been obtained in the case of nearly all other 

 metals. 



The choice of platinum as the standard metal is the outcome 

 of a very wide and general agreement among all the experi- 

 mentalists who have attacked this problem. Iridium, though 

 less fusible than platinum, is too rare and too difficult to work. 

 Similar objections apply to the other platinum metals. Alloys 

 are unsuitable for the purpose for two reasons. The change 

 of resistance of an alloy is much smaller than that of pure 

 metal, and owing to the difficulty of perfect mixing of the 

 constituents, the variations of different pieces of the same 

 wire are generally much greater than in the case of a pure 

 metal. Going outside the limits of the platinum group, the 

 metal gold has been proposed for lower temperatures on 



