1904 - 5 .] Magnetic Quality in Molecular Assemblages. 1051 
bounding plane. Thus, in the data given in § 22, we have to 
omit the reckoning for negative values of v. We thus find, for 
the parallel component on a magnet at the boundary, the quantity 
V . ^f ![18-6(a 4 + 13* + y *) - 2] 
approximately. This is always positive. In the same way the 
transverse component is found to be 
^~“ 2 1 8 -6(a« + + / - Q 2 ) 1/2 . 
The longitudinal effect has therefore the same sign as it had in the 
interior, while the sign of the transverse effect is reversed. Thus 
the turning moment is, in the neighbourhood of a quaternary 
axis, always directed towards that axis, which is therefore a 
direction of stable equilibrium. In the neighbourhood of a ternary 
axis, the moment of the transverse component overcomes that of 
the parallel component, so that this axis is one of unstable 
magnetisation. In fact, the conditions of stability are the reverse 
of those which hold in the interior. 
How the values of H(2p) _5/2 and HA. 4 (2p) -9/2 , given in § 22, 
show that molecular control in the interior is effectively determined 
by very near neighbours. The effect of those at a greater distance 
than seven times the least distance between molecules in the 
postulated homogeneous arrangement does not seem to be capable 
of making serious change in the values of the constants given 
above. If it does not, it follows that flaws, of thickness far 
smaller than the range of molecular forces, effectively introduce 
boundaries. Consequently, so far as the terms now under dis- 
cussion go, a considerable amount of magnetisation parallel to a 
quaternary axis may be retained. And this effect of the flaws 
tends also to diminish the residual internal magnetisation parallel 
to the binary and the ternary axes. 
In the case of a cubic crystal, on removal of the magnetising 
force, the tendency is to retain on the whole magnetisation along 
the ternary axis most nearly coincident with the original direction 
of magnetisation. At a surface of the cube, on the other hand, 
the magnetisation tends to be normal ; and the change from the 
one condition to the other takes place in an extremely thin film. 
