528 
PROFESSOR J. A. EWING ON EXPERIMENTAL 
by amounts which vary from point to point along the axis of the bar, while after the 
magnetising current is reduced to zero there remains a demagnetising field, also 
unequal along the bar, due to the magnet itself. The magnetism which remains is 
therefore only that part of the true residual magnetism which is not removed by this 
self-demagnetising force: by “ true residual magnetism ’ is meant the magnetism 
which would be found if after being magnetised the metal were left entirely free 
from all magnetising force. The influence of the ends has been recognised by many 
writers. To secure uniform magnetisation Yon Quintus Icilius experimented with 
ellipsoids of revolution in place of cylindrical bars. Stoletow* and Rdwland,! by 
using endless magnets, have given their determinations of magnetic susceptibility and 
the maximum of magnetisation a value incomparably greater than that possessed by 
previous observations made on short bars. It is clear that the question which my 
experiments were, in the first instance, specially directed to examine, namely, whether 
hysteresis occurs in the relation of magnetism to field when the field is varied ivithout 
change of sign (§ 2), could not be answered by any experiments with short bars, since 
in these the ends produce a magnetising force of reversed sign when the external force 
due to the solenoid is withdrawn. To answer this question we must use methods of 
experiment of the same rigorous character as are necessary in examining magnetic 
permeability or the coefficient of induced magnetisation. 
To eliminate the action of the ends, and so secure the conditions suitable for exact 
experiment in the relation of ^ to 3, Stoletow (on the suggestion of Kirchoff) and 
Howland, used magnets whose form was that of a closed ring, and Rowland also 
used rods so long that the influence of the ends became negligible. With closed 
rings, as he has remarked, we can do no more than observe sudden changes in 
magnetism by using an induction coil, wound on the ring, in circuit with a ballistic 
galvanometer. With rods, on the other hand, we can examine the actual magnetic 
state at any instant, either ballistically, by drawing off a short induction coil from the 
middle portion of the length, or by direct magnetometric measurement. My own obser¬ 
vations show that it is only when the length of the rod (if of iron) is about 300 or 400 
times its diameter that the effect of length becomes insensible. The principal draw¬ 
back to the use of very long rods is the difficulty of forming them (when the section 
is at all considerable) so as to be of sufficient length and at the same time homogeneous 
and uniform in section. By using a ballistic galvanometer whose needle, although 
sufficiently heavy to act ballistically, was comparatively light, I have been able to 
obtain good results with rods of very small section —wires in fact—which by the 
operation of drawing are scarcely less uniform in section than turned rods, and are in 
* Phil. Mag., xlv. (1873), p. 40. 
t Phil. Mag., xlvi. (1873), p. 140, and xlviii. (1874), p. 321. An admirable summary of the work of 
Stoletow and Rowland, as well as the earlier researches of Wiedemann, Von Quintus Icilius, Thalen, 
and others, and the later contributions of many other observers, has been lately published by Professor 
Chrystal in the article “ Magnetism,” Enc. Brit., ninth edition, vol. xv. 
