Lnigfusion of Hydrogen through Hot Platinum. 5) 
hydrogen at each end. The heating would then promote 
circulation of the gas inside the tube; in fact, we should 
expect a constant stream of hot gas up the hot tube, equalized 
by a current down the arm MH. These convection currents 
would effectually prevent any accumulation of impurities 
within the tube itself. 
The objection to this scheme caused by the putative 
rigidity of the glass framework was avoided by means of the 
loose joint E. The upper tube was blown out a little about 
1 em. and slightly constricted about 3 mms. from the end, 
the edges being pressed outwards with a file. The lower 
tube was drawn down a little at its end so as to project into 
the bulbous inflation of the upper tube, and still have plenty 
ot play in the shoulder formed by the constriction lower 
down. The tubes were then joined by means of thick-walled 
rubber tubing outside and the joint drowned internally with 
mercury between the two tubes. This both prevented the 
hydrogen from coming in contact with the rubber and also 
made the joint absolutely air-tight. With this arrangement, 
the upper tube being clamped at L, it was found that the 
joint allowed several centimetres’ play at FE’ before the 
platmum tube was sealed in. It is evident therefore that 
there could be no appreciable strain due to the much smaller 
expansion caused by heating. 
The temperature of the tube was determined by means of 
a thermocouple formed of very fine platinum and platinum- 
rhodium wire. The two wires were welded very carefully 
by pulling them against the hot tube so that they both made 
contact at exactly the same point. In this way the galvano- 
meter deflexion in the thermocouple circuit was found to be 
almost independent of the direction of the heating current, 
showing that only a very small fraction of the potential-fall 
along the tube was shunted through the galvanometer. 
A small correction was made for this. The thermocouple 
wires H H' ran through glass tubes to a point about 1 metre 
distant, where their ends were soldered to copper wires 
leading to the galvanometer, both the junctions being placed 
in test-tubes in a vessel containing cold water, whose 
temperature could be read. As trouble was at one time 
experienced owing to the sensitiveness of the D’Arsonval 
galvanometer changing during the experiments, an arrange- 
ment was set up by which the thermocouple reading could 
be compared with that produced by a constant small fraction 
of the H.M.F. of a storage-cell by simply reversing a 
commutating switch. The E.M.F. of the cell was checked 
by means of a sensitive voltmeter from time to time. — 
