THE CONSTITUTION OF THE ELECTKIC S|-.\|;K. 211 



as deduced from joints which are not in close proximity to the poles, would not be 

 large. Some nf niir photographs give evidence of being affected by this error. 



Other inaccuracies which cannot readily be evaluated are introduced by the curva- 

 ture of the spark. 1C \vr imagine the molecular streams to follow the course of the 

 spark, the appearan<v "ii our photographs might be considerably modified. In par- 

 ticular, if the curvature lies in a plane passing through the slit, the molecules near 

 the pole will move towards the slit or away from it, and the c<>m]M>nents <>f the 

 velocity resolved along its direction will have a smaller value than near the centre of 

 the spark, where its direction is nearly always parallel to the slit. The error intro- 

 duced is in a direction opposite to that found in the previous case, the molecular 

 velocities near the poles now apj)euring too small. 



One further jxjint remains to be noted. Photographs like those illustrated by 

 fig. 29 and fig. 30 cannot be easily explained, unless we take it that the metallic 

 molecules actually carry the electric current. If that is the case, and if the process 

 of discharge is similar to that advocated by one of us in 1882* and now generally 

 accepted, the molecular stream will carry with it the ionic charge, at any rate 

 in the positive part of the discharge. Now, considering successive oscillations, 

 electric forces must continue to act on the metallic molecules, first accelerating, then 

 retarding, and then again alternately accelerating and retarding. Under the condi- 

 tions of our experiments, and with the velocities measured by us, the molecules would 

 only have got to a distance of about 1 millim. from the pole before the second accelera- 

 tion takes place. If that is a correct view, we should expect a sinuous curve for the 

 molecular path, and there seems indeed a tendency towards such a form in some 

 photographs, as for instance that given in fig. 22. The only way to overcome this 

 complication will be to increase the period, so that we may be able to measure the 

 velocity of the stream separately for each oscillation. The effect of magnetic forces 

 on the velocity of the stream may also supply useful information. 



9. Effect of Self -Induction in tlw Spark Circuit. 



In the course of our investigation we were led to insert a coil of win- into the spark 

 circuit in order to separate the oscillatory discharges. We discovered in tliis way a 

 curious effect on the appearance of the spectrum, the air lines completely disappearing 

 wl len the coil was inserted, provided the self-induction was sufficient. The most plausible 

 explanation seems to be that the air lines are produced entirely by the first initial 

 discharge, when the spark gap contains no metallic vapour. The subsequent oscillations 

 pass, on the contrary, through the metal vapour, which in the meantime has had time 

 to diffuse away from the electrodes. By inserting the coil, the initial discharge takes 

 place more slowly and apparently does not heat up the air sufficiently to yield the line 

 spectrum. The whole duration of the spark is considerably lengthened, and the 

 * ' Bakerian Lectures,' 1884 and 1890 ; Boy. Soc. Proc.,' vol. 37, p. 317, 1884, and vol. 47, p. 526, 1890. 



2 2 



