EXPERIMENTAL STUDY OF HEAT LEAKAGE. 705 



at the cumol point, where it has always been hard to secure conditions 

 sufficiently steady to warrant confidence in the results. In fact, 

 cumol was finally abandoned altogether because of this trouble. 



The difference in magnitude between the coefficients in the equation 

 of calibration of the differential thermometer 67 and the correspond- 

 ing coefficients in the other two equations is due to circumstances 

 which are worth a brief remark. The two thermometers 6 and 7 

 were made of consecutive pieces of platinum wire taken from a spool 

 marked 'Heraeus' purest,' and no pains were spared to make these 

 thermometers exactly alike. The two pieces of platinum were 

 annealed at a red heat, and their resistances were adjusted to equality 

 within less than 0.0005 ohm at room temperature. After winding, 

 the platinum was once more annealed and the net resistances of the 

 two thermometers were again equalized by making slight adjustments 

 in the compensating circuits. The resistance of each was also now 

 very nearly that of No. 4. The two thermometers were then inserted 

 in their respective tubes and the caps which close the bottoms of these 

 tubes were brazed on, This operation nullified all the efforts at equal- 

 ity of resistance which had been made. The resistances at zero dif- 

 fered by nearly 100 times as much as before brazing, and each exceeded 

 that of 4. Moreover, as the equation of calibration shows, the normal 

 resistance of 67 decreases slightly with rising temperature, although 

 it is so nearly constant over a considerable range of working tempera- 

 ture that the actual temperature need not be known accurately so far 

 as elimination of the normal resistance is concerned ; this is a conven- 

 ience rather than otherwise. 



Thermometer No. 4 was made of Baker platinum, and it is therefore 

 not to be expected that it would combine with either 6 or 7 to form a 

 differential thermometer as satisfactory as 67 itself; in fact, both R-n 

 and i?64 vary much more rapidly with the temperature than R^i, as is 

 indicated by the equations of calibration. The dissimilarity of 4 to 6 

 and 7 is further shown by a comparison of the Callendar constants for 

 the three thermometers. The constant 8 has the values 1.54, 1.65, 1.58, 



= ~i7v?rD — ) the values 0.003795, 0.003859, 



0.003882, for 4, 6 and 7 respectively. The value of 8 is ordinarily 

 taken as a criterion of the purity of the platinum: if 8 exceeds about 

 1.50, the platinum is regarded as impure. On the other hand, values 

 of A' as large as those given for 6 and 7 indicate a high degree of purity 

 in the platinum. There thus seems to be some doubt whether a large 

 value of 8 is, by itself, a reliable indication of impure platinum. The 



