substances when heated on a Nernst filament. 423 
positive electricity from sodium phosphate heated in ditferent 
gases*, it is pointed out that this curve is similar to those obtained 
when the negative emission is under test, which suggests that 
ionisation by collisions comes in at certain pressures. In the 
paper referred to this view is ruled out because, at that time, 
it was generally thought that much larger electric forces than 
those employed were necessary to obtain ionisation by collisions 
with positive ions. Some experiments recently made by Dr 
Pavloff in the Cavendish Laboratory have, however, shown that 
ionisation by collisions with positive ions occurs even with 
potential differences of about 10 volts. It is thus possible that 
the shape of the curve obtained with sodium phosphate is due to 
ionisation of the gaseous molecules present, by collisions with the 
positive ions from the glowing anode. 
The experiments made with uncoated filaments showed a 
pressure of maximum emission at about 30 or 40 mm. pressure, 
but near this maximum the thermionic current was very unsteady, 
varying capriciously over a range of as much as 50 per cent. in 
some cases. It was at first thought that this unsteadiness was 
due to the cooling effect of convection currents, but the emission 
at higher pressures was very much steadier, so it does not seem 
probable that convection is the cause. The points plotted in this 
part of the curve in fig. 2 represent the means of many deter- 
minations. In the case of the particular filament to which this 
curve refers, the pressure of maximum emission was at about 
30 mm. The effect of placing sodium phosphate upon the filament 
is to reduce this pressure of maximum emission very considerably, 
so that at pressures of a few millimetres there is a very much 
greater ionisation from the phosphate covered anode than from 
the filament alone. The ionisation at all pressures is increased 
by the presence of the sodium salt, but the increase is most 
marked at pressures of a few millimetres. 
_ The form of the current-pressure curve for a clean filament 
is similar to that given by a platinum strip which has been 
heated for a long time. In the latter case there are strong 
reasons for believing that the ionisation is due to gas evolved 
from the heated platinum, and it would appear probable that 
the ionisation from the Nernst filament is due to a similar cause. 
The question arises: can the form of the current-pressure curve 
be due to ionisation by collisions? The high pressure at which 
the maximum value of the current is obtained would seem to be 
against this view, but it must be remembered (a) that the electric 
force near the filament is very great; (b) that at the temperature 
employed, the mean free path in the surrounding gas or vapour 
* Proc. Camb, Phil. Soc. xvi. p. 89, 1911. 
