independent of Changes of Temperature. 



219 



The coefficient changes from positive to negative between the 

 lengths 6 and 9 cm. And hence if the change between these points 

 is nearly linear, a length of abont 8 cm. should have a zero coeffi- 

 cient, and it might also be calculated that the permanent loss 

 would be 0262. A fresh length of exactly 8 cm. was cut from the 

 same coil of wire and was found to have a coefficient of — 0*000015, 

 and a permanent loss of 0*281 . A piece of this wire, a very little 

 less than 8 cm. long, would without doubt, have a strictly zero 

 coefficient. 



There are thus two practicable ways of obtaining zero tempera- 

 ture coefficients, either (1) by altering the hardness, or (2) by 

 altering the dimension ratio ; and the latter may be effected by 

 varying the diameter for a constant length, or the length for a con- 

 stant diameter as may be the more convenient. In addition, the 

 material of which the magnet is made must have certain chemical 

 and physical properties, not yet determined, of which, as far as some 

 experiments I have made can decide, the physical rather than the 

 chemical properties are the more important. 



Some of the results in Tables IY, Y, and YI are here plotted as 

 curves and exhibit interesting features. 



The curve of the relation of coefficient to dimension ratio (diameter 

 ■constant) from the data of Table YI, Diagram I, curve (1), has a 

 double inflexion between which it crosses the axis of abscissae and 



■at either end apparently approaches to horizontal asymptotes. This 

 curve is probably typical of the behaviour of music wired. 



Carve (2) on this diagram traces the series of experiments on 

 No. 33 wire. The two first points on the left correspond to the 



B 2 



