366 REPORTS ON THE STATE OF SCIENCE, ETC. 
The electro-magnetic fatigue-testing machine designed by the author and used in 
these tests has been described elsewhere (Journal of the Institute of Metals, 1917; also 
in The Engineer, 1920), and it will suffice here to mention that this type of machine 
offers great advantages for measurements of hysteresis, because a considerable body 
of metal is subjected to the full range of stress—not merely a thin film as in 
rotating cantilever tests—and because the range can be measured with accuracy, and 
the frequency of reversal is rapid (2000 cycles per minute). 
The test-piece used was fitted with a special form of sensitive thermocouple 
arranged in such a manner that errors due to the conduction of heat from extraneous 
sources were almost wholly eliminated, so that the gradient of temperature in the 
test-piece was measured with accuracy. The galvanometer used, therefore, gave a 
direct measure of the rate at which heat was being given out by the metal at any 
moment, under any desired range of stress. The thermocouple was attached to the 
test-piece before the commencement of the test, and the readings of the galvanometer 
were observed at intervals during the continuance of the fatigue run, 
The changes of hysteresis during such fatigue tests differ widely in different metals 
and with the range of stress. The more important conclusions indicated by a study 
of many records, for test-pieces that did or did not break under the applied loads, are 
summarised below : 
(1) In many metals, particularly those which have been annealed, the great 
hysteresis commonly observed in static tests dies away rapidly and gradually during 
the first few minutes or hours of a continued fatigue test: Such hysteresis may be 
termed ‘transient’ or ‘ primary,’ and is ascribed to gliding motion in the crystals 
resulting in hardening action akin to that of cold-work of any other kind. This 
primary hysteresis disappears long before fracture in a test under moderate stress, and 
is not regarded as the cause of fracture, but rather as a protective action that allows 
of beneficial redistributions of stress within the metal, particularly in parts of complex 
shape, such as the plates of riveted joints studied by Wilson and Haigh (British 
Association : 1922). 
(2) In test-pieces that are not going to break under the applied range of stress, 
hysteresis decreases continuously during a very long period, but does not completely 
disappear. Moreover, the definite decrease may not show itself, in some cases, until 
several millions of repetitions of stress have been imposed. This ‘ continuous’ or 
“secondary ’ hysteresis is regarded as evidence of the irreversible thermodynamic 
action that has been described. 
(3) In test-pieces that are going to break under the range of stress applied in the 
test, hysteresis continues to increase gradually, after the disappearance of the initial 
transient effect, sometimes for many millions of cycles before fracture. This pheno- 
menon is regarded as evidence of the gradual unstable extension of the zones at which 
‘secondary ° hysteresis is occurring. 
(4) Immediately before fracture hysteresis increases rapidly, and often to very 
high values, such that the heating effect is obvious and occasionally painful to the sense 
of touch. This final stage may be described.as ‘ tertiary > hysteresis, and is ascribed 
to the rapid extension of cavities, probably with gliding motion, associated with the 
formation of open cracks in the metal. 
(5) In many metals, particularly cold-worked or otherwise hardened, hysteresis — 
that is initially slight or almost imperceptible increases slowly but greatly (e.g. in ratio — 
20 : 1) during the first few million cycles of stress, and thereafter remains more or less 
constant, or fluctuates, until the rapid rise sets in after many more millions have been 
imposed. 
(6) If, at any stage in a long fatigue test under constant range of stress, the 
hysteresis is measured for a series of different ranges—and plotted against these 
different ranges—the result is a smooth graph with no distinct indication of the 
fatigue limit. Cursory inspection may give the impression that fatigue increases 
rapidly near the known fatigue limit ; but the impression is illusory, being dependent 
on the scales employed for plotting. 
(7) The hysteresis observed under a given range of stress, and the form of the 
‘ hysteresis-range of stress’ graph, vary widely in any given metal according to the 
treatment given to the metal immediately before the tests. Thus static elongation, — 
or a short fatigue test under a high stress, greatly increases the hysteresis ; and con- 
versely, boiling a steel test-piece, or applying a moderate range of stress fora fairly long 
period, tends to reduce the hysteresis observed under this and other ranges. 
