SHUDDEMAGEN. — DEMAGNETIZING FACTORS FOR IRON RODS. 237 



of the i\-curves is undoubtedly closely related to the change in the 

 pole-distance ratio //Z, which probably approaches the value unity for 

 complete saturation. The magnetization curves taken magnetometric- 

 ally tend to diverge, or spread apart, for high magnetizations, whereas 

 those taken ballistically all converge rapidly to the maximum ordinate 

 /^o- Figure 17 is reproduced from Mann's paper,^^ and shows the curves 

 from in = 5 to m = 50 obtained from his first cylinder. The method 

 by which Mann gets at the position of the " normal " magnetization 

 curve for an infinite rod is to assume that the magnetometric iV for a 

 cylindrical rod of in = 300 is the same as for an ellipsoid of the same 

 length and central cross-section, namely iV = 0.00075. 



In his investigation Benedicks obtained the value of JSf for only one 

 rod of hard steel (m = 25), but did this very thoroughly, using both 

 the ballistic step and magnetometric methods. His normal curve is 

 determined by transforming the steel cylinder into an ellipsoid of 

 in = 30, obtaining magnetometrically the magnetization curve for thi's 

 ellipsoid, and back-shearing this curve into the normal curve by means 

 of the known demagnetizing factor for this ellipsoid, which is 

 JSl = 0.0432. Theoretically the method is perfect, but we rather doubt 

 whether it can be depended upon to give uniformly agreeing results in 

 practice. The magnetization curves obtained by Benedicks are shown 

 in Figure 18, which has been reproduced from his article ^'^. The figure 

 shows the two types of iA"-curves, — the magnetometric and the balhs- 

 tic, — and their opposite behavior for high magnetizations. Benedicks 

 also publishes the A^-curves as he derives them from Ewing's original 

 six curves, all showing a behavior similar to that of his own curve 

 Xfiai- These iV^-curves are practically identical with those shown in 

 Figure 19 of this paper ; these were determined by our methods 

 directly from Ewing's curves shown in Figure 2, which were recon- 

 structed from the original figure ^^ in order to have both figures on 

 exactly the same scale as our own curves, for purposes of comparison. 

 See Figure 16, which shows the i\^-curves for our Bessemer steel rod of 

 diameter 1.905 cms. 



We might note that Benedicks gets no curvature in the iNT-curve 

 near the origin, because he takes his observations from hysteresis 

 cycles of magnetization, the maximum applied field being about 

 H' = 206 units. 



Benedicks criticizes Mann's assumption that N = 0.00075 for an 



" Phys. Rev., 3, 359-369 (1896). 



" Bihang Svenska, Vet.-Akad. Handlingar, 27 (1), Xo. 4, 14 pages (1902). 



18 Phil. Trans., 176 (1885), Plate 57, Figure 3. 



