534 
MR. 0. W. RICHARDSON ON THE ELECTRICAL CONDUCTIVITY 
temperature was determined by means of a copper-nickel thermocouple, C^Di, attached 
to a hollow semicircular cylinder of brass, E, placed at the middle point of the line 
CD. The brass piece E was cut from a tube which before the operation fitted easily 
into ABCD, so that when the coj^per and nickel wires were tied round it, it fitted 
quite tigditly. The spirals FF^ were made of composition tulniig wound tight round 
ABDC; they served to keep the ends of the tube cool and thus j^revent the sealing- 
wax joints from softening. They were fed with cold water at and emptied at G. 
The side-tube H from DI)i led to the pump and McLeod gauge. 
The manner in which the wires A^B^ and C,I)^ were fixed in at the ends is shown 
more clearly in the enlarged diagram on the left of tlie figure. The shaded j^arts 
represent the distribution of the sealing-wax which wms used to make the joints. 
The ends f fj of the tube ABDC fitted into an annular depression on the brass plate 
aa. The platinum wire Aj^B^ was soldered at each end on to a stout copper wire, bb^, 
which fitted fairly tight in a glass tube passing through the brass tube cc^. The 
whole was fixed in air-tight by means of sealing-wax. One of the leads from the 
thermocouple was iixed in exactly the same way into the brass tube dd^. 
The sodium was originally placed in the form of small cubes on and around E, and 
it was considered that after heatino; for a short time in a vacuum a fairly uniform 
distribution of sodium over the central portions of the steel tube would be obtained. 
This was certainly what happened in the case of tlie glass apparatus which had been 
tried previously and in which the effect could be oljserved. The leak from the hot 
sodium to the platinum wire electrode A^B^ was then measured, according to its 
magnitude, either by an electrometer or by a galvanometer. At the lower 
temperatures where tlje quadrant electrometer was employed, one of the t|uadrants 
was connected to the case of the instrument which was insulated. The other 
quadrant was connected witli a standard condenser and, l^y means of a wire jDassing 
axially on sealing-wax supports along a brass tube, with the electrode A^B^. The 
outside of the brass shielding-cylinder was connected with the case of the electro¬ 
meter. In making an experiment the whole of the electrometer system was charged 
to a given potential, and the time required for the spot to move over a given number 
of scale divisions wns noted. This measured the current from the insulated electrode 
