an Explanation of Hall's Phenomenon. 261 



correspondence between the direction of the thermoelectric 

 current and the sign of HalFs effect. With regard to the 

 aluminium, a piece of the foil was mounted upon glass, and 

 Hall's experiment performed with it. As was anticipated, the 

 sign of the " rotational coefficient " was found to be + like 

 that of iron, zinc, and cobalt. It is probable, therefore, that 

 the specimen of aluminium with which Mr. Hall worked 

 differed in some respect from that used by myself. The 

 anomalous specimen of gold being in the form of wire, could 

 not be submitted to the same test : it probably contains some 

 disturbing impurity. 



With regard to cobalt, since it could not be obtained in the 

 form either of wire or of sheet, it was necessary for the obser- 

 vation of the thermoelectric effect to treat it in a somewhat 

 different manner from that previously described. A rod of 

 the metal was held firmly by one end in a vice, and both ends 

 were connected by wires with the terminals of the galvano- 

 meter. The middle of the rod was gripped with gas-tongs 

 which had been made hot in a Bunsen flame, and stress was 

 produced by twisting. It was found that, while the stress 

 continued, a current passed from the strained to the free 

 portion of the rod through the hot part. The direction of the 

 current was the same whether the twist was clockwise or 

 counterclockwise, and it ceased the moment the stress was 

 discontinued. For the sake of comparison, a copper rod was 

 treated in a similar manner, and the resulting current was in 

 the reverse direction. For the purpose of this experiment, 

 torsion is clearly equivalent to traction, both tending to draw 

 asunder the molecules of the metal. 



Some months ago, when experimenting with pieces of iron 

 cut into different siivpes, I tried the effect of making a longi- 

 tudinal slit along the middle of the plate. The result was, 

 that the Hall effect was considerably weakened even though 

 a stronger current was used than before the slit was cut. In 

 the light of subsequently acquired knowledge, I have varied 

 the experiment thus : — A plate of iron was prepared as usual, 

 but the galvanometer-wires, instead of being attached in the 

 ordinary manner, were connected to points nearer to the 

 middle of the plate than to the edges. When the magnet was 

 excited, the ordinary Hall effect was found to occur, but it 

 was somewhat weakened. Two longitudinal slits were then 

 cut along the middle of the plate, a connecting piece about 

 4 raillim. wide being left at the centre. A current from two 

 bichromate cells was passed through the plate, and the posi- 

 tions of the galvanometer connexions were carefully adjusted, 

 until, on depressing the key K (see fig. 5), not the smallest 



