ON THE MEASUREMENT OF MAGNETIC HYSTERESIS. 
69 
of semi-cycles. We have not taken any special steps to secure this agreement 
because (1) we are satisfied that the sum of the throws for a pair of semi-cycles gives 
a correct measure of the hysteresis, and (2) it was more convenient to take a pair of 
semi-cycles than a complete cycle, since with the semi-cycles we could measure and 
W simultaneously. 
We found that the throw for a complete cycle depended a good deal upon the Avay 
in which the key was manipulated. On the other hand, the throws for each of a pair 
of semi-cycles were generally very regular. 
§ 44. In § 6 we show that, when, on account of the previous magnetic history of the 
specimen, the two throws of the dynamometer for a pair of semi-cycles are unequal, 
the sum of the throws still furnishes a correct measure of the energy dissipated in the 
complete cycle, if U = W. This inequality in the two throws gave us much trouble. 
It persisted in some cases in spite of very many reversals of the magnetic force. As 
tested by the ballistic galvanometer, the iron had reached a steady state, l)ut the gal¬ 
vanometer only shows the change of induction on reversing Hq, and not the actual values 
of B corresponding to -f- Hq or to — Hq. In the belief that after many reversals the 
cyclic B—H curve would lie symmetrically about the axes of B and H, we naturally 
did not look to the iron for the cause of the inequality in the throws. We spent 
some weeks in making changes in the dynamometer and other parts of the apparatus, 
but of course without result. At times the inequality would almost disappear 
(probably on account of a large number of reversals), only to reappear for some 
apparently slight cause, such as changing the range of the magnetic force, or using 
another specimen, and this uncertainty made the matter very perplexing. But 
though we felt that we had failed to discover the cause of the inequality, yet the 
agreement between the value of U, found by the dynamometer, with that of W, 
deduced from the cyclic B—H curve, made us certain that the method was giving at 
least approximate results, and we began to make experiments on the effects of tension 
and torsion upon hysteresis. During these experiments we found that tlie inequality 
varied as the stress was applied, and we were thus led to see that the origin of the 
inequality lay in the iron. When once we suspected the cause, it was easy to make 
experiments to satisfy ourselves that our suspicion was true. We were then able to 
employ the method without hesitation. 
§ 45. To study the inequality in the two throws for a pair of semi-cycles in a 
definite manner we made the following experiments :—A freshly annealed iron wire, 
which had not been magnetised since the annealing, was placed in the solenoid when 
no current was flowing, and the magnetic force Hq was then applied for the first time. 
We found that the throw of the dynamometer for the first reversal from Hq to — Hq 
was greater than for any subsequent reversal in either direction, and that up to 100 
reversals the throw is greater when the magnetic force changes from Hq to — Hq than 
when it changes from —Hq to Hq. In. these experiments the value of Hq was 5'77, 
and the value of Bq, after 100 reversals, 9990 ; the section of the wire was ’0265 
