MR. T. ROYDS ON TIIE CONSTITUTION OF THE ELECTRIC SPARK. 
343 
first have to be destroyed. No such change in the direction of motion of the vapour 
has been found in the streamers. In some cases, however, the envelope is not straight, 
but sinuous; where one streamer overtakes the previous one and falls into the 
envelope it indicates a higher velocity than the end of the previous streamer, and 
afterwards has a smaller velocity. This sinuosity is noticed in several photographs, 
but in the majority the envelope is straight. The acceleration and retardation of the 
vapour is coincident with the alternations in the field, but the retardation is not 
sufficient to reverse the direction of motion. The photographs show that the charge 
of the particles is negative. The successive alternations in velocity are 550, 130, 670, 
190, 460 met./sec., the variation becoming smaller as the oscillations are dying out 
and as the vapour approaches the centre of the gap. 
Unless discretion is used in selecting the point whose co-ordinates are measured, 
such variation in the velocity might be introduced where none exists. If a streamer 
does not actually run into the one following, there might be a tendency to take a 
reading on the steep part of the latter, which does not indicate the velocity of the 
vapour. 
This sinuosity apparently would indicate that sometimes the metallic vapour is 
ionised. However, the action of the electrical field on charged particles cannot be the 
sole cause of the motion of the metallic vapour, for, owing to the reversal of the 
electrical field, the mean velocity during several periods would be small. 
And, lastly, there is no evidence to support the hypothesis that the vapour is 
projected with the required velocity from the electrode. 
It is, perhaps, fortunate that in any spectrum only two velocities are obtained. 
More than two would be difficult to explain. According to the simplest conception, 
viz., that we are dealing with a case of diffusion, the velocity would be inversely 
proportional to the square root of the vapour density, and it would seem simplest to 
suppose that the vapour which gives rise to the second class of line is more highly 
dissociated than that emitting the first class. If this is the case we must have, that 
if the vapour emitting the short lines of the bismuth spectrum is monatomic, that 
emitting the long lines contains five atoms to the molecule, which is very improbable. 
The impossible values obtained for the temperature of the vapour show that the 
velocity of the second class of line is not due to diffusion. 
Another explanation might be that the short lines are emitted by charged molecules 
which are impelled with high velocity by the electric field, whilst the long lines are 
due to non-ionised vapour. 
(b) Propagation of Luminosity through the Vapour .—Since we are most probably 
dealing with the diffusion of vapour from the electrodes, there will be vapour present 
in the space between the electrode and as far as the vapour starting at the first 
oscillation has reached. The electric current which is re-established in each oscillation 
can raise this vapour again to incandescence, and a streamer thus be produced. 
Where this streamer will start, whether at the electrode where there is always most 
