28 
PHYSICS: D. L. WEBSTER 
Proc. N. a. S. 
the spectrum flash out stronger when the images coincided. This method 
worked especially well when they were still wet. The parts where the 
general radiation was not too strong were also examined by holding the 
film against a white paper in a strong light, and a negative lens was also 
used. Since these methods were needed it is obviously impossible to show 
the faint lines in a reproduction even when there was no doubt of their 
existence in the original. Therefore, no reproduction is attempted here. 
I wish here to express my thanks to Mr. D. S. Piston, who took most of 
the photographs in this investigation. 
The potential was furnished by a modified form of Hull's^ apparatus 
shown in figure 2. The 500-cycle current came from a Crocker- Wheeler 
motor generator loaned by the Submarine Signal Co., through the gener- 
osity of Mr. H. W. J. Fay. The transformers and inductances were made 
mostly from parts of induction coils contributed by the Boston City 
Hospital, through the kindness of Dr. F. H. Williams. The 60-cycle one 
had the small secondaries so placed that the heating current in either 
kenotron was the same whether the other was on or not. The condensers 
were of glass plates and sheet iron, immersed in oil and protected as in 
Hull's apparatus, by corona gaps. 
With connection A, shown by the short-dash line, only one kenotron 
is used, and the potential on it rises to more than twice that of the D. C. line, 
while the frequency of the fluctuation on that line is 500^, the same as 
that of the transformer. .But with connection B, using two kenotrons, 
the potential on each is that of the D. C. line itself, that of the transformer 
is half that amount, and the fluctuation frequency on the D. C. line is 
doubled. It should be noticed that to produce this double frequency 
it is most essential that no part of the transformer secondary or first 
or second condensers should be grounded. 
The constants of the apparatus were approximately as follows: Ci = 
3 X 10-3 jnf(j^ = Cs = 1.5 X 10 mfd, Li (sum of the two sides plus 
twice the mutual inductance) = 20,000 henries, L2 = 15,000 henries. 
Such large inductances are not strictly necessary, and were used only 
because they were at hand. They were measured by using 90 kv. (measured 
by spark) at 500^ with an A. C. voltmeter (current type) in series in 
the middle of the line. As the voltage fluctuation is approximately 
^ ^ ^ -r — ^, it is extremely small, probably not over a few volts under 
CiC2C3LiL2CO^ 
the worst conditions. The resistance of Li and L2 together, measured by 
D. C, was about 0.3 megohm. When running, a current type voltmeter 
in series with C2 and C3 showed no measurable current, though of course 
it showed easily the current in Ci. 
Incidentally, it may be noted that the above expression for the fluctua- 
tion should contain also the factor (1 — C2Moo^), where M is the mutual 
inductance between Li and L2. This factor could readily be used to neu- 
