Abstract
After limit of strain hardening, fatigue crack was assumed to form as a result of large tensile stress by a piled-up group of dislocations plus externally applied tensile stress component, with the aid of other stress concentration factor associated with the obstacle or inclusion itself. Nucleation theory was applied to the initiation of the crack. The theoretical S–N curve is in quantitative agreement with the data. The scatter of the number of cycles for fracture, which is an inherent characteristic of fatigue fracture, is, mostly if not wholly, explained as a stochastic phenomenon associated with nucleation process of the fatigue crack. The theoretical temperature dependence of fatigue strength is in fairly agreement with the data in literature.