MAGNETISM AND TWIST IN IRON AND NICKEL. 507 



approximately proportional to the current density, that is, within the limits of the 

 experiments, to the mean magnetising force acting on the strip. 



14. General Summary. — For convenience, we shall, before taking up the second 

 section of the present paper, summarise the preceding results under four headings. 



A. — When an iron, nickel, or cobalt wire or rod, magnetised longitudinally, is 

 subjected to a cyclic variation between equal positive and negative values of a current 

 passing along it, or when the same, carrying a steady current, is subjected to a similar 

 cyclic variation of a longitudinal magnetic field, the amounts of twist are, for the same 

 combination of current and field, in general different. In the case of iron and nickel, the 

 current-reversal twist is less than the field-reversal twist in high fields (see Plates I. and 

 II.). In very low fields, however, the current-reversal twist is the greater. To each 

 line current there corresponds a particular field, for which the two modes of reversal give 

 the same twist. This point of equal twists occurs in higher fields as the current is taken 

 stronger, and in lower fields as the longitudinal tension is taken greater. The stronger the 

 line current, other things being equal, the less is the difference between the two twists. 

 In the few observations made on cobalt, the current-reversal twist was always the smaller. 



B. — Detailed study of the various stages of the cyclic twisting in iron and nickel 

 shows that for every same value of the changing current or field there are in general two 

 values of twist, so that the cyclic graphs (see Plate III.) form closed areas. When the 

 cycle is established by variations of the longitudinal field, and when the field is taken 

 between high enough limits, the cyclic area has three loops, the ascending and descending 

 branches intersecting each other near the points of maximum twist. Otherwise the 

 cyclic graphs for the twist are very similar to the cyclic graphs for magnetic induction, 

 studied by Warburg and Ewing. 



C. — The direction of the twist is in all three metals exactly what would result if we 

 try to explain it in terms of the simpler magnetic strains of elongation and contraction as 

 studied by Joule, Barrett, Bidwell, and others. In moderate fields, iron twists right- 

 handedly when the line current and longitudinal field are co-directional ; nickel and cobalt, 

 on the other hand, twist left-handedly. 



D. — The comparison of the various phenomena as they exist in iron and nickel gives 

 an ever-strengthening argument of a cumulative character in favour of this explanation, 

 first suggested by Maxwell, that the Wiedemann effect is essentially determined by the 

 Joule effect. In Part I. I have given a formula applicable to thin tubes, which connects 

 the magnetic elongations with the magnetic twists. To apply this formula to the case 

 of wire cannot be expected to lead to accurate numerical results ; and yet a direct 

 calculation by means of this formula, from observed twists in given circumstances of 

 current and field, of the ratio of the elongations of iron and nickel, was wonderfully 

 concordant with Mr Bidwell's direct observations of these elongations. The continuously 

 diminishing effect of increasing tension upon the twist in iron, and the somewhat more 

 complex effect of increasing tension upon the twist in nickel, correspond accurately with 



