CONTEMPORARY ADVANCES IN PHYSICS 93 



If the voltage between the electrodes is augmented rapidly instead 

 of slowly, the breakdown-potential may be greater; it is as though the 

 discharge were delayed for an appreciable time after the proper 

 critical P.D. was reached, during which time the voltage is over- 

 shooting the mark and giving rise to error. I mention this because it 

 has bearing on what follows. 



If the voltage is supplied from a "source of high frequency" of 

 one of the types customary before the development of the vacuum- 

 tube oscillator — for instance, an induction-coil or an interrupter — it 

 arrives as a sequence of highly-damped high-frequency wavetrains with 

 longish intervals between. At the end of each interval, the P.D. 

 between the plates rises suddenly and rapidly, and if it rises far enough, 

 breakdown takes place. The difference between the rise which (were 

 it not interrupted by breakdown) would end in the attainment of a 

 thenceforward constant voltage, and the rise which (were it not 

 interrupted by breakdown) would be followed by successive falls and 

 smaller rises and alternations of direction, is practically small. True, 

 breakdown might occur, in the latter case, during the second rise when 

 it had missed the first; or after the completion of one damped wave- 

 train, the gas might be left in an abnormal state lasting until the 

 coming of the next and facilitating breakdown by the next. But this 

 does not seem to happen in practice, and if it did, there would be 

 obvious advantages in studying it with trains of undamped waves such 

 as nowadays can be produced. For successions of damped wave- 

 trains, therefore, I will merely quote the general result applicable to 

 air at atmospheric pressure: the voltage producing sparkover, between 

 definite electrodes at a definite distance, is almost if not quite in- 

 dependent of frequency up to such high values as a million cycles — 

 what changes have been observed are generally increases and may be 

 ascribed to the fact just mentioned, that when the voltage is increasing 

 very rapidly it may overshoot the minimum value sufficient for spark- 

 over before the spark gets started. 



Turning now to sinusoidal wavetrains such as modern technique 

 makes available: if such a one be applied while its amplitude is yet too 

 small to cause a breakdown, and then the amplitude is gradually 

 increased (or alternatively, the distance between the electrodes is 

 diminished) it will gradually modify the gas by reproducing ionization 

 in ever-increasing amount — the "self-amplifying" effect of the ioniza- 

 tion, which I mentioned above; and this will eventually bring about 

 breakdown. We know a great deal about this preliminary process for 

 steady voltages, but as yet we can only infer it for alternating voltages. 

 Thus for steady voltages, there must be at least two modes of ioniza- 



