190 
MR. A. A. GRIFFITH ON 
Remembering that the surface tension of a substance is the work done in forming unit 
area of new surface, it will be seen that the tension of any surface of a crystal must 
depend on the angle it makes with the crystal axis. Thus the surface tension parallel 
to the planes of least strength must be less than that in any other direction. 
Consequently, in a body composed of a number of crystals there must exist a mutual 
surface tension at each intercrystalline boundary. Now, the theory of surface tension 
shows that the magnitude of such a mutual tension is greatly diminished by making the 
transition between the two bodies more gradual. Hence the formation of the amorphous 
boundary layer involves a reduction in the surface energy of the crystals, and this is 
shown in the experiments by a drop in the stress. If this account of the phenomenon is 
complete, the drop in stress must be determined by the condition that the loss of strain 
energy equals the reduction in surface energy. The mechanism of the process appears 
to be that the breaking up of the boundary, which must accompany yield, is resisted 
by the surface tension, and yielding therefore requires a higher stress for its initiation 
than for its maintenance. 
According to this view, the loss of strain energy should be inversely proportional to 
the linear dimensions of the crystals. Hence the results of different experiments should 
show considerable variation in the magnitude of the drop in stress. This is actually the 
case ; a single series of experiments on mild steel, by Robertson and Cook, gave drops 
varying from 17 per cent, to 36 per cent., while in other experiments as little as 7 per 
cent, has been observed. 
In the above series of experiments the average loss of strain energy was about 
12 inch-lbs. per cubic inch. Assuming, for simplicity, that the crystals were cubes, 
of, say, 0-001-inch side (which is a fair value for well-treated mild steel), the area of the 
intercrystal surface was 3000 sq. inches per cubic inch. These figures give the average 
intercrystal surface tension as 0-004 lbs. per inch. This is certainly of the right order 
of magnitude. 
Many of the phenomena discussed above will be more complicated, in practice, if 
the coefficient of expansion of the crystals is not the same in all directions. In such 
an event, internal stresses will be set up in cooling, on account of the random arrangement 
of the crystals, and these stresses must be taken into consideration in applying the theory. 
There remains for consideration the problem of the fracture of metals under alternating 
stress. It is known that fatigue failure occurs as the result of cracking after repeated 
slipping on gliding planes, and the theory has been advanced* that this cracking is due 
to repeated to and fro sliding and consequent attrition and removal of material from 
the gliding planes. This theory presents some difficulties, in that it does not explain 
how the attrition can occur, or the method of disposing of the debris. 
A theory which is free from these objections may be constructed if it is supposed 
that a change in volume occurs on the passage of the metal from the crystalline to the 
amorphous state. This assumption is, of course, known to be valid for many substances. 
* Ewing and Humfrey, ‘ Phil. Trans.,’ A, 1902, p. 200. 
