1893 - 94 .] Prof. C. G. Knott and A. Shand on JSfidkd Tubes. 293 
lated for the ellipsoid, and is 
7T in the case of the sphere. 
O 
It 
vanishes for very elongated bodies, like wires. In such cases, when 
P is negligible, the ratio l/H measures the susceptibility. 
When, however, the magnetised body is not very elongated, the 
value of P becomes appreciable ] and Lord Rayleigh has shown 
how, by a simple graphic construction, we can pass from the 
magnetisation curve of a long thin wire to that of an oblate spheroid 
of any eccentricity. Kow, the susceptibility of iron in vanishingly 
low fields has a value of about 30 or 40. Hence, in bars whose length 
is not more than 12 times the breadth, the ratio I/H is a measure 
rather of l/P than of k. Por this condition of things we have 
approximately 
PI = H. 
If a is the area of section of the metal of the tube or bar, the 
magnetic moment is 
Ia = H^* 
But this is practically the same for all the tubes ; consequently 
the demagnetising factor is proportional to the area of section of the 
metal. 
If we try to apply the same reasoning to the case of the nickel 
tubes, we are met by two difficulties. In the first place, there is a 
considerable residual magnetism even in the lowest fields — as much 
as 52 per cent, in field 4. [This residual magnetism has a well- 
marked maximum — 66 per cent, in the tube of widest bore — in a 
field of 25 or 30, just a little lower than the field for which I/H 
has a maximum value (Wiedemann’s Wendepunkt).'] Then, in the 
second place, the susceptibility of nickel in vanishingly low fields is 
probably not greater than 5 or 6. Thus we cannot hope to find the 
relation 
I- 
k 
1+kT 
H 
even approximately applicable; and still less can we regard the 
ratio I/H as measuring the reciprocal of the demagnetising factor. 
Experimentally there is a further source of discrepancy in the 
results ; for it is very difficult to get nickel (which has once been 
