Magnetization and Magnetic Change of Length. 643 



distance from the upper end of the ovoid, the rest being well 

 insulated with asbestos paper. The interspace between the 

 leading wires and the wall of the porcelain tube was tightly filled 

 with asbestos fibres, and thus protected as much as possible 

 from convection currents. The other junction was arranged 

 as in the second experiment. The thermoelectric current 

 was measured with a d'Arsonval galvanometer from Keiser 

 and Schmidt ^ the reading of which was corrected by the 

 authors with a mercury thermometer containing nitrogen 

 below 550° C, and by Professor JSTagaoka and Mr. S. 

 Kusakabe with the melting-point of sodium chloride. A low- 

 resistance galvanometer was, at the same time, employed to 

 measure the temperatures lower than 200° C. A simple 

 connexion permitted us to pass the thermoelectric current 

 through the d'Arsonval or the low-resistance galvanometer 

 as the case might be. 



The experiment was conducted in the following order : — 

 The adjustments of the magnetometer and the coils, as 

 described in the first experiment, were effected ; the heating- 

 coil with the specimen was then placed in the right position. 

 The magnetization at the temperature of the room was first 

 determined ; then a current from a dynamo was passed 

 through the heating-coil, till the temperature of the specimen 

 became constant. The direct effect due to the current in the 

 heating-coil was tested by breaking or reversing the current. 

 The small deflexion of the magnetometer, when there was 

 any, was completely eliminated by altering the form of the 

 leading wires. The demagnetization by reversals, while the 

 heating-current was passing, showed no trace of residual 

 magnetism, which indicates that the magnetization due to 

 the heating-current was insensibly small. When the tem- 

 perature became constant, the magnetizations at gradually 

 increasing fields were measured. Another stronger current 

 was next sent through the heating-coil and the same processes 

 were repeated as before. In this way, we measured the 

 magnetization in the stage of ascending temperature and then 

 that in the descending stage. During each set of obser- 

 vations the temperature was fairly constant, and even in very 

 unfavourable cases the variation did not exceed 2 degrees. 

 The temperature was always noted both before and after each 

 experiment and the mean was taken. When a series of 

 experiments was finished, the specimen was taken out of the 

 coil and the compensation tested. Excepting in a few cases, 

 we found the compensation undisturbed ; when, however, 

 the disturbance was such as to require a correction, it was 

 uniformly distributed. 



