B.—CHEMISTRY. 4] 
vacuum and extracting the dissolved gases as completely as possible 
The hysteresis loss is reduced to a quarter or less of its original value by 
this treatment. Pure iron so freed from gases is almost as soft as copper. 
The magnetic properties seem to be more profoundly altered than any 
others, but there is evidently a wide field here for investigation. 
Impurities other than gases may exert an influence out of all proportion 
to their quantity if concentrated in the boundaries between the crystal 
grains. When the added element is insoluble, or practically so, in the 
metal, the effect is obvious, as in the famous instance of gold to which 
0.01 per cent. of bismuth has been added, the soft and ductile gold becoming 
excessively brittle, as shown by Roberts-Austen as a lecture experiment. 
Even when the two metals are miscible in the solid state it is quite possible 
that there may be a concentration of the impurity at the boundaries, if 
the addition be one which lowers the surface tension of the metal, it 
having been shown that surface tension plays an important part in the 
determination of those boundaries. Traces of oxide or sulphide are 
naturally rejected in freezing, and Tammann has found that when cadmium 
is dissolved in a solution of ammonium nitrate without the evolution of 
gas, a fine network of insoluble matter is left, representing the outlines 
of the crystal grains. Even if we imagine a metal so carefully purified 
that all these possibilities have been eliminated, it still does not follow 
that the mass is chemically homogeneous. There must be some change 
in the condition of the space lattice as the boundary is approached, and 
whether we suppose that this disturbance is limited to a layer a few atoms 
thick, or assume, as Brillouin and Rosenhain have done, that there exists 
an amorphous intercrystalline layer of appreciable thickness, one must 
conclude that there will be some chemical difference at the boundaries, 
and this is confirmed by the effect of etching reagents, which commonly 
indicate a difference in the rate of etching between the mass of a erysta’ 
grain and its boundary. Chemical reagents differ widely in this respect. 
Some brasses are readily brought into a state of brittleness, in which the 
crystal grains break away from one another under shock or alternating 
stresses, and it is usually possible to bring about the separation by contact 
with a suitable chemical reagent. It appears likely that failure in practice 
most often begins under the influence of chemical corrosion. The remark- 
able feature of this kind of failure is that it is only caused by a few 
chemical reagents, and that others will attack the metal generally without 
_ any selective action on the boundaries of the grains. Two reagents have 
in a striking degree this property of attacking the boundaries first— 
ammonia and the salts of mercury. The latter act with extraordinary 
rapidity, so that specimens of brass may be found which will disintegrate 
completely into a mass of loose grains, like sand, within a few seconds 
_ after immersion in a solution of mercurous nitrate. On the other hand, 
_ nitric acid or ferric chloride will attack the same brass uniformly, as if the 
condition of inter-crystalline brittleness were totally absent. 
When a face of a crystal is brought into contact with an etching reagent, 
such as water for rock salt, hydrofluoric acid for quartz, or cupric 
ammonium chloride for iron, the surface is not dissolved away evenly, 
leaving it smooth, but characteristic etching pits are produced, the sides 
of the pits being evidently crystal faces. This shows that chemical 
action proceeds more readily along certain planes of a crystal than along 
