UNDER THE INFLUENCE OF CHEMICAL ACTION. 
size this coating would fracture from time to time, exposing the bright alloy underneath 
the crack. Simultaneous with each fracture a sudden increase of the rate of electro¬ 
meter deflection could be observed, owing to the increased emission from the clean 
alloy. Needless to say, no quantitative data are obtainable from experiments under 
such conditions. In the experiments on which reliance is placed, the rate of the drops 
was so rapid and the pressure of the attacking gas so low that no visible change occurred 
in the appearance of the surface of the drops. 
In general it has not been found possible so to regulate the pressure of the attacking 
gas that the saturation current would stay constant. There is, however, no great diffi¬ 
culty in maintaining conditions so that the saturation current increases or diminishes 
in a regular manner with lapse of time. For example, a common method of procedure 
in dealing with COCl 2 has been the following :—Initially the COCl 2 would be strongly 
cooled by immersion of the generating tube in liquid air. The cock between the gene¬ 
rator and the testing apparatus would then be opened and all the gas pumped out of the 
whole apparatus to a pressure of 0 -001 mm. or less. The connection to the pumps 
would then be shut off and this vacuum would be maintained, provided the COCl 2 was 
well immersed in the liquid air. The alloy would then be allowed to drop, charged to 
a suitable negative potential, and the copper ball connected with the electrometer. If 
the electrical conditions were satisfactory there would be no deflection under these 
circumstances. The next step would be gradually to reduce the depth of immersion 
of the COCl 2 in the liquid air. This could be done by turning a screw table which 
supported the thermos bottle below the generator. Meanwhile continuous observation 
was kept on the electrometer spot, and at a certain stage a small movement would set. 
in, showing that some active gas was beginning to reach the apparatus. It was generally 
convenient to raise the thermos bottle a little at this stage, as owing to a lag in the 
thermal changes the lowering process was generally somewhat overdone before the 
effects were perceptible. At about this stage transient effects would frequently be 
observed. These may be attributable either to a more volatile active contaminant 
present in the chemical used, or to some small trace of the latter which had condensed 
on the upper walls of the generator. These walls would no doubt warm up quickly 
when the liquid air was reduced. These transient effects would disappear after a little 
while, and the electron currents under a given applied voltage would be found to increase 
very steadily and deliberately as the generating tube gradually warmed up. If the 
rate of increase became inconveniently large, it could be checked by raising the liquid 
air and repeating the processes substantially as described until a complete set of observa¬ 
tions had been obtained. 
In some cases it was found preferable to work in a reverse manner, by allowing an 
excess of the active gas to flow into the apparatus and to observe the phenomena under 
examination as this gas was consumed. The consumption takes place rather slowly, 
probably owing to adsorption of the gas by the walls of the apparatus. A typical 
example of the diminution of the saturation current with time after admitting COCl 2 , 
