Magnetisation of Liquids. 187 



mutual induction of the primary on the secondary of a solenoid 

 arising from the presence of a liquid in the solenoid against known 

 small inductances. Thus, if the sum of the inductances be reduced 

 to zero, as shown by the galvanometer in the secondary giving no 

 deflection, the balance will be disturbed to the extent 47T&M, due to 

 the insertion of a liquid into the solenoid whose coefficient of mag- 

 netisation is &, and the galvanometer in the secondary circuit will 

 give a deflection when the commutator revolves. An adjustable 

 inductance is then reduced by a known amount, m, till the deflection 

 disappears ; so that we get 



47T&M = in .'. k = m/4 7r\f, 



where m and M are quantities easily calculated. 



Since the formula does not contain either the rate of the rotation 

 of the commutator nor the value of the primary current, no particu- 

 lar precautions are necessary to keep these quantities constant. 



In all the determinations the magnetising force was varied from 1 

 to 9 centigram units, and in no case was there any variation in k. The 

 densities of the salts in solution were also varied over large ranges, 

 and showed that the coefficient of magnetisation for ferric salts in 

 solution depended only on the quantity of iron per c.c. that was 

 present, giving the formula 



10 7 fc = 2660 W 7*7 



for ferric salts, where W is the weight of iron per c.c., the quantity 

 7'7 arising from the diamagnetism of the water of solution. 



A similar result was obtained for ferrous salts, the corresponding 

 formula being 



10 7 & = 2060 W 77, 



the temperature being 10 C. 



The following table shows the coefficient of magnetisation for the 

 different salts examined, w being the weight of the salt per c.c. of the 

 solution : 



10? &. 



Fe 2 Cl 6 916w-7-7 



Fe a (S0 4 ) 3 .... 745^-7-7 



Fe,(NO,) 6 ..... 615 w- 7-7 



FeCl 2 908w-7'7 



FeSO 4 749*0-7-7 



The effect of temperature was also estimated, the results of the 

 experiments being shown by means of a curve (fig. 1), the x ordinates 

 of which denote the temperature, and the y ordinates are proportional 

 to the coefficient of magnetisation, a length corresponding to 50 

 being subtracted from each for convenience of representation. 



VOL. LX, p 



