260 



H. E. J. Q. du Bois on Kerr's 



§ 13. On account of the slight rotation no very sharp con- 

 clusions can be drawn from these numbers. The ratios of the 

 last column are certainly all of the same order of magnitude ; 

 the first two, corresponding to residual rotations, appear to 

 be equal and less than the third. This is easily explained by 

 the plausible assumption of a residual distribution differing 

 from the temporary one in giving comparatively less mag- 

 netization near the mirror. The result appears to me to lead 

 to the following interpretation : — 



For the law of proportionality it matters not whether the 

 magnetization be residual or kept up by external induction. 



This would tend to prove that the rotation is an immediate 

 effect of magnetization and depends upon it alone*. 



§ 14. Effect of temperature. — The ovoids could be fixed in 

 a brass tube closed at one end, and fitted at the other with a 

 number of diaphragms, through the central apertures of which 

 the mirror remained visible. This tube was heated by a 

 horizontal perforated burner ; the necessary draught being 

 obtained by two waterjet-exhausters, which thus acted like 

 chimneys. This arrangement could be introduced inside the 

 coil, separated from it by an asbestos jacket. With a current 

 of 30 amperes the coil and its contents could be electrically 

 heated up to 170° ; by means of the burner the ovoid could 

 be raised to higher temperatures still. Magnetic and optical 

 observations were then made simultaneously during cooling. 

 The rate of cooling could be modified at will by regulating 

 the enclosing coil's temperature by the current. The 

 temperature was given with sufficient approximation by a 

 thermometer having its bulb near the mirror and stem pro- 

 jecting from the apparatus. Of course the metals were heated 

 only for a short time and never beyond the temperature at 

 which they begin to show surface-colours ; as the least film 

 of oxide on the mirror completely modifies the phenomena |. 



§ 15. The results are given in Table I. The numbers for 

 cobalt, iron, and steel on the whole lead to the conclusion 

 that the value of e /3 / is but little variable with temperature. 

 At the same time 3' varies so little J with 6 that the distribu- 

 tion may be safely considered constant. Consequently e /3 

 ( = K) is proportional to e /3'j *"• e -> by the above, practically 

 constant. 



For nickel, however, 3' diminishes considerably while 6 



* du Bois, Wied. Ann. xxxi. p. 973 (1887). 

 t Kundt, Wied. Ann. xxvii. p. 199 (1886). 

 I See footnote t next page- 



