114 Proceedings of the Royal Society of Edinburgh. [Sess. 
diagonal of the cube, namely 
2 ’ 
and this is the limiting (greatest) length 
that can be assigned to the atomic magnets. If we treat each atomic 
magnet as a rod with point poles, its half length r must be less than 
but will approach that value very nearly, since i is small. Hence, 
taking r = 
hjli 
4 ’ 
M 64 M 32 I, 
Now, the most favourable condition for rupture will occur when a row 
with unsupported ends lies across the direction of the deflecting held, and 
in that case, as we saw in § 18, the value of H which initiates rupture is 
nearly the same as for an isolated pair, namely (§ 15), 
H = ^fF 
On substituting the above value of M/r^ this becomes 
In iron I^ is about 1700 c.g.s. units, and accordingly we should have 
H = 654[Fd 
as the held at which instability begins. Now [F^] — as was shown in 
§ 8 — is of the order of 2500, when the magnetic poles are near enough 
together to satisfy the condition that the elastic dehection which precedes 
rupture shall not exceed about half a degree (see § 2). Thus in soft 
iron the calculated held required to break up the rows is more than a 
million times as great as the held which actually suffices to make the 
Weber elements become unstable, when that metal is magnetised. It is 
obviously necessary to look for an arrangement in which the Weber 
elements will have a far less stable control. 
21. These considerations have led me to devise a new model which, 
while retaining the idea of magnetic control (a feature that accords well 
with all the complex phenomena of ferromagnetism), escapes the excessive 
stability of the old model, and at the same time satishes the essential con- 
dition that the Weber element in each atom may turn only through a very 
small angle before becoming unstable. 
The new model is based on the idea that the system of electrons in 
each atom comprises two groups : an inner group which constitutes the 
