Prof. H. L. Callendar on Platinum Thermometry. 211 



attainable accuracy. In the case of the platinum thermo- 

 meter these effects are relatively much smaller, owing to the 

 large change of resistance with temperature, and can be 

 completely eliminated in a very simple manner. 



Ag. F.P. Method of Reduction for Impure Wires. — The 

 simplest method of reduction for such wires at high tempera- 

 tures, would be to take the Ag. F.P. as a secondary fixed 

 point instead of the S.B.P. for the determination of the 

 difference-coefficient d. This would in general lead to a very 

 close agreement at temperatures between 800° and 1200°C, 

 but would leave residual errors of 3° or 4° at temperatures in 

 the neighbourhood of the S.B.P. To obtain a continuous 

 formula giving results consistent to within less than 1° 

 throughout the range, it would be necessary to adopt the 

 method which I suggested in my last communication (Phil. 

 Mag., Feb. 1892), assuming d to be a linear function of the 

 temperature of the form a-\-bt, and calculating the values of 

 a and b to make the instrument agree with the pure wire at 

 both the S.B.P. and the Ag. F.P., taking the latter as 960°'7. 

 We should find for the wire (c = "00320) above quoted, d= 

 1-580 at the Ag. F.P. If we apply this value at the Au. F.P., 

 we should find £ = 1063 o, 0. But if we employ the second 

 method, and calculate a linear formula for d to make the 

 results agree throughout the scale, taking d= 1*751 at the 

 S.B.P., we obtain d=a+ fa = T898 — 000331*. Hence 

 the appropriate value of d to use at the Au. F.P. would be 

 d= 1;547, giving for the Au. F.P. £ = 1060°-0, which is in 

 closer agreement with the value 1060 o, 7 given by the pure 

 wire. This method has also the advantage that it gives 

 practically perfect agreement at the S.B.P., and at all points 

 between 0° and 1000°. In the case of the mercury thermo- 

 meter,, or the thermocouple, a similar cubic formula is required 

 to give an equally good agreement between 0° and 200° 0. 



In the original paper in which the suggestion was made, I 



couple very extensively in their investigations. They inform me that 

 they were compelled to abandon the method shortly before my appoint- 

 ment, because in spite of every precaution which their experience could 

 suggest they found it impossible, owing to these residual thermal effects, 

 to effect a sufficiently accurate calibration of the Pt-PtRh thermo- 

 couple at temperatures between 0° and 100° C. The substitution of 

 baser metals such as iron and german-silver at low temperatures would 

 no doubt partly meet this difficulty, but would involve the abandonment 

 of the wide range and constancy and uniformity of scale characteristic of 

 the platinum metals, which are qualifications so essential for a standard. 

 We conclude on these grounds that the application of this thermo- 

 couple is limited to high temperatures, and tnat the contention that it is 

 preferable to the platinum thermometer as a secondary standard cannot 

 be maintained. 



