871 
In this way curves, as given in fig. 1, were obtained (2 times 
enlargement of the original negative) 
with the oscillograph, or as in fig. 7 
with the stringgalvanometer. 
We can divide one complete period 
of the interrupter into 4 nearly equal 
parts. The two first quarter periods 
represent the time during which the 
circuit is closed, the two last ones 
the break period. During the 2rd and 
3'¢ quarter period the armature mo- 
ves towards the cores; during the 
1st and 4 quarter period in an 
opposite direction. We know that 
the number of lines of force passing 
through the armature determines the 
force with which it is attracted by 
the electromagnet. We may even 
say that this attraction is very nearly 
proportional to the square of that 
number of lines of force. | 
Our curves show that the attraction during the second quarter 
period is very much greater than during the first. This fact was 
pointed out by Lord Rayiuten and has practically formed the basis 
of all later communications on this subject. But at the same time 
we see that during the 3'¢ quarter period, the current being broken, 
a strong attractive force still exists, which is notably stronger than 
the attraction which during the 4!" quarter period works against the 
movement of the armature. Even when the interrupter works under 
very different conditions as to frequency, current-strength ete. this 
fact remains unchanged. We may say that the attraction during any 
part of the movement of the armature towards the pole pieces, 
greatly exceeds the attractive force in any point during its course 
away from the electromagnet. Consequently there is no need for 
any retarding device for making the current with respect to the 
movement of the armature — as suggested by Lord RayLriga — in 
order to improve the working of the interrupter. Probably such a 
device would not only be inconvenient, but would hamper the 
working of the apparatus. 
Can we explain the curve for the attraction? For the ascending 
part this is certainly possible. We can even calculate it approximately. 
We first suppose the selfinduction to be constant during the make 
