364 REPORTS ON THE STATE OF SCIENCE, ETC. 
fracture might be taken, at first sight, to be that of a more brittle quality of steel. 
The pulsating stress fracture shows the cup and cone, but less markedly than the 
tensile fracture. 
Through all the variations of form of the fatigue fractures produced by different 
combinations of steady and alternating stresses, the flat face of the fatigue crack 
may be identified as corresponding, in some measure, to the flat inner zone of the 
tensile fracture ; while the final distorted tear of the fatigue fracture corresponds to 
the regular cup and cone of the tensile. It is inferred that the determining cause 
of fracture is the same in the two cases: that fatigue is associated with the formation 
of small cavities within the metal, without the protective action of gliding. Whereas, 
in the tensile test, these cavities form only when the metal becomes brittle—through 
the agency of triple-tensile stress in metals that are normally ductile—their formation 
occurs under alternating stress even while the metal retains its full ductility. General 
observation shows that fatigue is associated with little if any of that hardening effect 
which, otherwise, is invariably associated with such gliding movements as produce 
ductile strain. It is inferred that fatigue is due to the formation of cavities under 
conditions such that the openings cannot be healed by gliding movements ; that the 
action originates from numerous zones of exceedingly small dimensions, zones at 
which undue stress concentrations are present, and are so small that the numbers of 
molecules involvedinsimultaneous motion do not suffice to allow of gliding—probably 
because their movements are too rapidly quelled by loss of thermal energy by con- 
duction. Itis, of course, to be understood that gliding motion may occur under alter- 
nating stresses as well as in simple tension—for example, when the fatigue limit lies 
near to or above the tensile yield-point, or when a certain amount of slipping occurs 
in the initial stages of a fatigue test; but this slipping or gliding movement is not 
regarded as the cause of that fatigue which, after the repetition of many millions of 
cycles of stress, breaks the metal in such a characteristic manner. 
The development of a fatigue crack is attributed, therefore, to the formation of 
cavities in the metal, these cavities being produced initially by the contraction of 
volume that accompanies change of state, and extended gradually at each application 
of stress. The detailed action may be pictured as follows: In a test-piece composed 
of crystalline grains and intermingled vitreous matter (or other constituents) present 
before the start of the test or produced during the test by gliding movements that may 
have converted crystalline to vitreous, certain molecules are nearly unstable. Some 
are nearly ready to spring out of the crystalline mass into the vitreous, while others 
are nearly ready to fall into the lattice from the vitreous assemblage. Both actions 
may be precipitated by a change of energy, due to change of stress or temperature ; 
but both are very local, involving only small numbers of molecules. When the 
temperature of the metal is raised slightly, as by boiling, in Dr. Muir’s well-known 
experiments, these molecules fall into positions of greater stability ; and, in subse- 
quent mechanical tests, only still smaller numbers are influenced by any change of 
stress that may be imposed upon the metal. Thus the metal is rendered more nearly 
elastic, in respect that hysteresis is reduced. 
It is conceived that this dual action may occur indefinitely without necessarily 
causing fatigue, provided that the effects of crystallisation completely neutralise 
those of decrystallisation: that is, if the action is mechanically reversible—the 
molecules that are displaced at one extreme of the stress cycle falling back into their 
original places at the opposite extreme, or at the mid-stroke. But it is clear that 
such a dual action, occurring with sufficient violence, would speedily open a crack if 
the two parts of the action did not exactly compensate. The nature of the action 
is such that it must either die away gradually, leaving the metal unbroken, or increase 
gradually, leading to fracture; but its increase or decrease may be so gradual that, 
in tests continued only for moderate periods, it may appear constant in magnitude. - 
We may next ask what evidence, if any, may be expected of such an action as has 
been described above—a dual change in which the two parts nearly neutralise one 
another. If the change involves no gliding motion, we can expect no changes of 
shape or volume such as would afford a basis of measurement. But since the two 
parts of the change involve the giving out and absorption of quantities of heat, and 
occur at different stages of the stress cycle, we may expect that the action will be 
irreversible in the thermodynamic sense ; and accompanied, therefore, by a conver- 
sion of mechanical work to heat, on an overall balance. The irreversible action 
must be accompanied by an increase. of entropy. Evidence of such an action is 
at: hand in:the phenomena of hysteresis, described in’ the following section: 4 
