332 H. NAGAOKA. 



the periuînieiit twist. On the coiitrary, the luitwisthig for the twist 

 of 861° is greater than that for tlie twist of 1021°. The course of the 

 curve, after passing the ur.txiiuuni becomes steeper with tlie larger per- 

 inauent twist as the coui|)arison of (1) (2) with (o) (4) will show. 

 Thus, when the twist is large, and the magnetizing force sufticiently 

 great, the curve may he expected to cut the line of no twisting. 



Another difference in the curves of torsion ol)t: lined f )r different 

 permanent twists consists in the course of the ciu'\'e on the removal of 

 the m;ignetizing f^rce. In curve (2), we find tliat the "off"' returns 

 below the " on " curve, while in curve (3), it returns above it. In 

 the former there is hysteresis or lagging, in the latter priming or 

 negative hysteresis. This distinctive feature in the curves obtained 

 for different twists also varies with the thickness of the wire. 



It is unnecessary to give numerical details f )r the various experi- 

 ments made with different wires and with different twists. The 

 characteristics above descrilied are illustrated in the curves of Fig. 3, 

 which gives the results f)r nickel wires of diameters 0.5, 0.4, 0.7 mm. 

 For these also the untwisting re-iches a maximum f )r a comparatively 

 low field, and a hoisting liegins to set in, and continues to the highest 

 field used. 



The f )llowing experiment shows that this twisting may proceed 

 so fir as to result in a final condition of twistednrss relatively to the 

 original condition of the wire. The wire, 0.34 nriu. thick and 30 cms. 

 long was twisted through eight complete revolutions of the torsion 

 circle, and then released. It thus acquired a ])ermanent twist of 

 2548°. The magnetizing current was derived from a shunt dyn-niio. 

 The current strength was ndjusted l)y the li([uid slide bef )re described. 



