199 
less affected by those plieiioniena, liitlierto unregaixled, wliicli 
it is the object of this pa2)er to explain. 
AVlien, therefoi'e, the ionising agent fs feeble, the only- 
part of the curve which can be altered by varying the cur- 
rent is that where the potential gradient is small : the feebler 
the agent the smaller the gradient. lA-t us now consider 
whether our hypothesis makes it probable that we can alter 
the shape of the rest of the curve by any variation of the 
conditions of the experiment- 
Now, if initial recombination takers ])lace because the 
ejected electron does not get far enough away from its parent 
atom before it is stopped by encountering another atom, then 
diminution of jiressure ought to make it much easier to satu- 
rate. But this is a well-known fact (Rutherford, ''Philosophi- 
cal Magazine," vol. xlvii., p. 160). In order to obtain results 
comparable with those we had already obtained at ordinary 
pressures, we made several experiments in which all the con- 
ditions were the same, except that the pressure was less than 
that of the atmosphere. Curve C in Fig- 2 shows the results 
of such an experiment. If this is comj^ared with the other 
curves in the same figure it will be clear that alteration has 
taken place in the very portion of tlie curve where we should 
have expected it, and where change in the strength of the 
current has small influence, viz., all along the uppei part of 
the curve up to the high potential end. The saturation 
current per sq.cm. was about 10~^'' amp. In further sup- 
port of our hypothesis it may be pointed out that it gives a 
ready explanation of an experiment due to Rutherford, and 
described by him in the ''Philosophical Magazine,'' vol. xlvii-, 
p. 158. He found that the saturation value of the current 
through a gas could be obtained for a much lower potential 
gradient when the gas was drawn away from the uranium 
which ionised it, and treated in a separate vessel. This is 
to be expected whan it is considered that under tlie circum- 
stances of the experiment initial recombination was wholly 
absent. 
It is now convenient to consider these phenomena as they 
are manifested in other gases than air. It is well known that 
the relations between current and potential in carbon dioxide 
are in some way abnormal. But the peculiarities of this gas 
are even intensified in ethyl chloride (CoH-.Cl). The fact 
is that this effect, which makes it difficult to draw all the ions 
to the electrodes in the case of air, is far oreater in more 
complex gases, and thus it is extremely difficult to obtain the 
saturation current unless very high potentials are employed. 
We find it necessary to use a potential gradient of two to 
three thousand volts per cm. in the case of ethyl chloride at 
60 cm. pressure. In the investigations which were made by 
us i -philosophical Magazine," September. 1905). with regard 
