ON THE MAGNETIC QUALITIES OF IRON. 1039 
experiment where the first application of the field was towards the side called 
negative. 
Again, using a model in which there were only 36 magnets, notes were made of 
the position assumed by each member of the group when the directing field was 
(1) applied, (2) reversed, (3) re-reversed, and so on. It was seen that the configura¬ 
tion assumed in the first application was not reproduced in the third, nor that of the 
second in the fourth. But the fifth application made all the memljers of the group 
take up the same position as they took in tlie third, and again the sixth agreed with 
the fourth. 
Further, tests of iron and steel show that if we have to deal, not with a previously 
unmagnetized piece but with a piece on which the process of “ demagnetizing by 
reversals ” has been performed, the phenomena which have just been described are not 
found, or, at least, are much less conspicuous than in virgin metal. 
They reappear after the primitive state is reproduced by annealing. In this respect ^ 
the molecular state of iron which has been demagnetized by reversals of a strong 
magnetic force (diminishing to zero) differs from that of a piece which has been de¬ 
magnetized by annealing—or even, it may be added, by mechanical vibration.* 
The same difference is found in the behaviour of the model after the two corre¬ 
sponding kinds of preliminary treatment. If a neutral initial state has been arrived 
at by a casual stirring up of the little magnets, the model behaves in the way shown 
in fig. 38, On the other hand, if the neutral state has been produced by repeated 
reversals of a strong deflecting field, very slowly performed, so as not to throw the 
group into general agitation, and gradually reduced in range to zero while the 
reversals go on, the displacements of the magnets under a subsequently applied 
weaker field are found to fall at once into a cyclic regime when the field is reversed. 
An example of this is given in figs. 40 and 41, which show experiments made with a 
group of 120 magnets. In fig. 40 the magnets had been shaken up to begin with, 
and, in consequence, when the magnetic field was applied the displacements were not 
at once cyclic. Fig. 41 shows the behaviour of the same group under precisely the 
same changes of deflecting field after the process of “ demagnetizing by reversals ” 
had been used to bring the group into a neutral state. Here we find a cyclic regime 
established at once, and as near an approach to symmetry in the curve as a model 
with a limited number of magnets can well be expected to show. Observations made 
on such a model are necessarily somewhat rough, and too much stress must not be 
laid on particular features of the curves. But in several repetitions of these experi¬ 
ments the same general distinction has always been apparent in the behaviour of the 
model after one and the other mode of preliminary treatment, a distinction which, as 
we have seen, forms a striking analogue to what is observed in actual iron or steel. 
* In illustration of this difference, reference may be made to figs. 26 and 27 in Plate 61, ‘ Phil. 
Trans.,’ 1885. 
