Strong Fields at liferent Temperatures. 301 



optical measurements with the only approximately homo- 

 geneous " red " light less accurate than for cobalt and nickel 

 (loc. eit. § 21) ; another reason for attaching most importance 

 to these latters' curves. 



§ 13. Nearly all optical observations were made, with steam 

 on, at 100° ; cooling to 0° would not do on account of the 

 dew condensing on the mirror. A few measurements were, 

 however, made with the strongest fields at ordinary tempera- 

 ture ; the rotations proved greater than at 100° in every 

 case, and their values for 0° were obtained by simple extra- 

 polation ; they are given at the bottom of Table I. The 

 values of 3 marked x were calculated on the assumption of 

 absolute invariability of Kerr's constant with temperature, 

 whereas I have only shown the variations to be less than the 

 limit of experimental errors (a few per cent, per 100°, loc. cit. 

 § 15). The corresponding values of © are marked x 0° in 

 the diagrams of fig. 2 for steel, cobalt, and nickel. From a 

 comparison with the short temperature-curves of fig. 1 it 

 would appear (always on the above assumption) as if the 

 decrease of magnetization from 0° to 100° were much more 

 marked for fields of thousands of units t^han for lower ones. 

 The conclusion of § 7 is based on this inference. 



§ 14. Magnetite (see loc. cit. § 17). — Curve I. of fig. 4 

 represents 6 = funct. (^B'), experimentally determined for a 

 small crystal of magnetite by purely optical observations (at 

 ordinary temperature) ; its ordinates must be proportional to 

 those of 3 = runet. (9?')- Now the latter function is unknown 

 for this material, but it may be asserted to start from the 

 origin very nearly as a straight line 3=3?747r*. On in- 

 spection of curve I. the two observed points © nearest the 

 origin are accordingly seen to lie almost in line with it. 

 Now multiplication of these two values of e into the corre- 

 sponding ones of 47T/Q}' is easily seen to give Kerr's constant 

 as +*0122 and +'0115 for the two points respectively : the 

 mean value +*012 will be sufficiently near the truth for our 

 present purpose. 



We consequently divide the ordinates of e — funct. (33') by 

 •012 and thus obtain 3 = funct. (2$') in absolute measure. 

 Thence 3 = funct. (•§') is easily calculated and plotted as 

 curve II. (fig. 4). The magnetization evidently tends 

 towards a limit not much above 350 O.G.S., somewhat less 

 than that for nickel. Our magneto-optic method has thus 

 given an absolute magnetic curve, at least approximately, 



* du Bois, Wied. Ann. xxxi. p. 1)52 (1887); Phil. Mao-. Nov. 1887. 



