UNDER THE INFLUENCE OF CHEMICAL ACTION. 
37 
on the right-hand sides of curves I. and II., but the low-voltage parts of the curves 
are quite different. Apart from the difference in shape, the low-voltage parts of I. and 
II. are much broader than in the case of III. relative to the rest of the figure. How¬ 
ever, the breadth here is determined entirely by the assumed position of the zero. In 
curve II. this is at 1 volt in fig. 17 and corresponds to the zero in fig. 4. This zero has, 
however, been shown to be at least 0T2 volt outside the possible limits for the true 
zero as determined by the photo-electric data, and it may quite well be out by as much 
as half a volt. In that case curve I. would have to come up from the volt axis at about 
1 *5 volts instead of at 1 volt, which would have made the curve show a much stronger 
resemblance to curve III. The position of the true zero for the chlorine data plotted 
in curve II. is still more doubtful. These curves illustrate well the difficulties involved 
in the analysis of the energy distribution unless the position of the zero is correctly 
known. 
In fig. 18, which refers to the data for COCl 2 with the cylindrical electrode, the position 
of the actual zero is known to within 0-10 volt, and it is obvious that these experimental 
points (marked thus ©) show a much closer resemblance to the requirements of the 
Maxwell distribution. The crosses in this figure do, in fact, represent a Maxwell dis¬ 
tribution for the temperature T = 3600°K, and the same range dV = 0 -20 volt as the 
experimental points deduced from fig. 12, and the full curve is drawn to meet the theo¬ 
retical points. It will be seen that the points given by the experiments are very close 
to the theoretical curve, and, in fact, the agreement is as close as could be expected^ 
except in the immediate neighbourhood of the maximum. As has been explained 
already, there are disturbing causes which are likely to produce deviations in this region, 
and these deviations should be in the direction indicated by the discrepancy shown 
by the diagram. It should be pointed out that the only quantity which can be arbi¬ 
trarily varied in fitting the theoretical curve to the experimental points is the single 
parameter T. The only other variable entering is the stretch dV, which is already 
fixed by the way in which the experimental points are reduced from the observational 
data. Given T and dV ¥ the value of the ordinate for any particular value of V is deter¬ 
mined absolutely. 
I believe that the data exhibited in fig. 18 furnish strong evidence that the distribu¬ 
tion of energy among the chemically emitted electrons is of a very simple character 
and is identical with that which would be possessed by the molecules of a gas at a certain 
definite temperature. The fraction where kinetic energy lies between u and u -f- du 
appears, in fact, to be given by 
dn du u -A 
n JcT JcT 
For COCl 2 acting on NaK 2 the value of T appears to be close to 3600° K. The evidence 
would, of course, be stronger if it were unequivocally supported by the analysis of 
figs. 4 and 7. It may be that too much emphasis is being laid on the uncertain data 
and the disturbing causes which affect curves I. and II. of fig. 17. It may be that 
