122 Proceedings of the Eoyal Society of Edinburgh. [Sess. 
ing poles of the turning octet lie in planes inclined to one another at 60° 
round the axis about which it turns, and the successive positions of stability 
are consequently 120° apart. Hence we have only to think of the crystal 
as made up of successive layers parallel to the magnetic plane and in twin 
relation to one another (each turned through 180° with respect to the 
adjoining layer), to obtain abrupt magnetic changes at intervals of 60° in 
the rotation of the crystal, though in each iron atom of the compound the 
electron grouping remains substantially cubic. 
29. The forms of the new model that have so far been described may be 
said to be based on an atom of the Lewis-Langmuir type, in which the 
electrons are conceived to have small orbits or to be small rings. But the 
same ideas may be embodied in another form, based on an atom of the 
Rutherford-Bohr type, with large 
electron-orbits encircling the nucleus. 
Thus in 6g. 19 the Weber element 
is represented by a central electron- 
orbit W, while A and B are a pair 
of coplanar elliptic orbits with a 
common focus at the nucleus of the 
atom, which is also the centre of W. 
The plane of A and B is to be thought 
of as fixed ; the plane of W may turn. 
Under the influence of a deflecting field it turns stably through a small 
angle and then becomes unstable. 
Fig, 19. 
The model photographed in fig. 20 (PI. I) shows a corresponding arrange- 
ment in which the electron -orbits are represented by coils carrying currents 
which have the directions indicated by the arrows in fig. 19. In the 
model the controlling elliptical coils are set as nearly as may be in one 
plane, and the central coil is pivoted to turn about the axis ah, with only 
a small amount of clearance on each side. A complete model would of 
course have several pairs of controlling coils forming a symmetrical system 
in three dimensions, and its Weber element (which might consist of one or 
more coils) would have freedom to turn in any direction. It would 
obviously pass from one to another position of stability in such a way as to 
exhibit hysteresis and the other characteristics of the magnetising process. 
In any one position the stability would depend on an inequality of controls^ 
just as in the model with magnets, and there would be a correspondingly 
narrow range of stable deflection. Generally, what has been said about 
the model witu magnets will apply to this one. 
If the atomic shell or system of fixed elliptic orbits is to have cubic 
