10 
PROF. 0. W. RICHARDSON ON THE EMISSION OF ELECTRONS 
voltmeter connected, in effect, across the gap between the drop of alloy and the copper 
sphere. The actual difference of potential across this gap is not equal to V, but is equal 
to V + Iv where K is the contact potential difference between the surfaces of the alloy 
and the copper sphere. If now the value of K is changing between one experiment 
and another, the observed results are completely accounted for. It will be necessary 
that K should exhibit an extreme variation of as much as 2 -5 volts, but this does not 
seem a surprising requirement when it is remembered that the contact electromotive 
force between the alkali metals and copper is comparable with this amount. The 
changes in the contact electromotive force which occur are to be attributed to changes 
in the surfaces of the alloy and of the copper due to the gases and vapours to which 
they are exposed, and to changes in the copper surface due to splashing with the alloy 
and possibly to absorption of the vapours of the alkali metals by the copper. 
To test this hypothesis, all the sets of observations with C0C1 2 which were sufficiently 
complete to form a reliable guide were collected together and the relative currents in 
terms of the standard tabulated. In cases such as curve IY. in fig. 3, where the current 
was not saturated at — 3 volts, the standard voltage was chosen about 3 volts negative 
to a voltage at which the current was about 50 per cent, of the final saturation value. 
The precise value of this voltage does not matter much, as the current in this region 
is not varying with the voltage to an extent ascertainable by these experiments. What 
is important is that the standard current should be saturated, and it was convenient 
to employ the least voltage that would make sure of this. It was also felt that until 
more information was available about the phenomenon, it was desirable to employ as 
the voltage for the determination of the standard current a voltage which would occupy 
the same position on each characteristic. For example, it was not, and is still not, 
known with certainty whether there is a small variation of current with voltage on the 
flat parts of the curves. If there is any such change, the values of the relative currents 
would be affected by the value of the voltage at which it was decided to measure the 
standard saturation current. The method adopted ensures that there are no errors 
arising from considerations of this character, which would be appreciable in comparison 
with the unavoidable experimental error. In all the curves but two the original three- 
volt standard could be retained. In fact, it approximately satisfied the condition just 
described, the current at zero volts in these cases being on the rapidly rising part of the 
characteristic. 
The relative currents thus obtained were then plotted for each series against the 
actual volts given by the voltmeter and the amount of displacement along the voltage 
axis was judged, which would be necessary to bring all the curves as nearly as possible 
to coincidence. This amount is, of course, in general different for the different curves, 
but is the same for every experimental point belonging to any one curve. The dis¬ 
placements are not applied to the curves, which constitute a secondary inference from 
the experimental data, but to the primary source of evidence, the experimental points 
themselves. The absolute position of the composite curve in relation to the scale of 
