186 PROCEEDINGS OF THE AMERICAN ACADEMY. 



range from to its maximum value, increasing continuously, and the 

 iron being initially unmagnetized. The curve obtained by back- 

 shearing is called the " normal " curve of magnetization for the kind 

 of iron used. As the applied field H' is now the same as the resulting 

 field H, the demagnetizing field having disappeared, this normal curve 

 gives us the true permeability /x and susceptibility k for every H, and 

 is therefore the characteristic curve of the iron which we must use 

 in order to get correct values for the physical quantities mentioned. 

 Ewing and other investigators have made much use of this back- 

 shearing process in working out hysteresis curves obtained for long 

 iron wires, it being assumed, while experimental determinations were 

 still lacking, that cylindrical iron wires could be regarded as behaving 

 magnetically like ellipsoids of the same length and cross-section, pro- 

 vided the ratio of length to diameter was not too small. 



The first attempt to find numerical values for the demagnetizing 

 eftect in cylindrical iron rods was made in 1894 by Du Bois^ in dis- 

 cussing the only magnetization curves with varying length of rods 

 which had up to that time been published: six by Ewing, obtained 

 ballistically,^ and a few by Tanakadatd, taken by a magnetometric 

 method.* From these results Du Bois constructed a table of values 

 for N for values of ill ranging from 10 to 1000, where in = ratio oi 

 length L to the diameter D, of the rod. He evidently considered that 

 N remains practically constant for the whole range of magnetic in- 

 tensity. Du Bois's values of N for cylinders are from 10 per cent to 

 20 per cent smaller than for the corresponding ellipsoids, that is ellip- 

 soids having the same ratio of length to maximum cross-section. 



In 1895 C. B. Mann published^ an extended series of results, 

 obtained magnetometrically, for the demagnetizing factors of iron 

 cylinders. The leading points brought out by this investigator, for 

 the rods experimented on, most of which were of small diameter, are: 

 (1) The .^'s for cylinders are very nearly constant for all intensities 

 of magnetization below /= 800 ; after this point they increase rapidly 

 as / increases. (2) For the range in which the iV's are practically 

 constant, they vary but a very few per cent from the values of the N's, 

 for the corresponding elHpsoids. Mann does not believe that ballistic 

 and magnetometric determinations of N will give comparable results. 



The most recent work on the demagnetizing factor which I have 

 seen, is embodied in a short but extremely suggestive paper published 



2 Magnetische Kreise, Berlin, 1894, pp. 36-45; Wied. Ann., 46, 485-499 (1892). 



3 Phil. Trans., 176, II, 535 (1885). ''^ 



4 Phil. Mag., 26, 450(1888). 



5 Dissert., Berlin, 1895; Phys. Rev., 3, .359-369 (1896). 



