4.2.1. At Nominally Ambient Temperature

To characterize the cyclic deformation and transformation behavior during cyclic loading in the HCF regime, specimens taken from batch B were cyclically loaded in stress-controlled single step tests with stress amplitudes in the range 225 MPa ≤ σ<sup>a</sup> ≤ 280 MPa. In Figure 8a the resulting developments of plastic strain amplitude εa,p are plotted against the number of cycles N. The εa,p,N curves illustrate pronounced cyclic hardening after a short period of initial cyclic softening. Cyclic hardening is caused by an increase in dislocation and stacking fault density, as well as by formation of deformation-induced α´-martensite (Figure 8b). A positive influence of α´-martensite formation on fatigue life was observed, because austenite-α´-martensite transformation causes cyclic hardening, which significantly reduces plastic strain amplitude. Similarly, total-strain controlled fatigue tests in the LCF regime (Figure 6), after an incubation period dependent on stress amplitude, resulted in a continuously increasing volume fraction of α´-martensite over the number of cycles. With increasing stress amplitude, the onset of α´-martensite formation was shifted to lower N and higher α´-martensite fractions were measured at

specimen failure. This development of α´-martensite in stress-controlled tests was comparable to the results given in [16] as well as for total-strain controlled fatigue tests shown in Figure 6c. However, with 1.7 FE% < ξ < 8.8 FE% the maximum values of α´-martensite at specimen failure/ultimate number of cycles were significantly lower compared to total strain-controlled LCF tests. The change in the specimen temperature during the fatigue tests is shown in Figure 8c. As in the LCF tests, changes in temperature were used to detect cyclic deformation behavior. However, in the stress-controlled fatigue tests, an increase in temperature correlated with an increase of the σ-ε hysteresis loop area, indicating cyclic softening, whereas a decrease in temperature correlated with a decrease of the σ-ε hysteresis loop area, and characterized cyclic hardening. Besides using ΔT to characterize cyclic deformation behavior, information about the general amount of self-heating is provided (Figure 8c). In consideration of Equation (4), a maximum specimen temperature of 102 ◦C was achieved while the value of ΔT signal shows 53 K. As is known, temperature has a significant influence on deformation-induced α´-martensite formation and consequently fatigue life. In literature [5,72–79] the fatigue life of metastable austenitic stainless steels is given as S-N curves obtained by fatigue testing with load frequencies up to 150 Hz. Obviously, specimen temperatures in HCF tests with frequency higher as 2 Hz were higher than ambient temperature; however, the information about the specimen temperatures is not at all times given. To obtain HCF results at true AT or at temperatures around 25 ◦C–30 ◦C, fatigue tests with significantly lower frequency, e.g. f = 0.2 Hz, have to be performed. This requires 115 days of test duration to achieve NL = 2 <sup>×</sup> 106, i.e. a duration that is impractical for systematic investigation of HCF behavior due to high resource expenses. In conclusion, to obtain representative results in the HCF regime in a reasonable time, it is necessary to perform fatigue tests of metastable austenite with variable test frequencies during single step fatigue tests, which effectively suppress self-heating of e.g. ΔTmax < 10 K. Therefore, it has to be noted that the presented results (and many others given in literature [5,72–79], especially regarding the development of α´-martensite and its influence on fatigue life) are influenced by specimen temperature, which needs to be taken into consideration.

**Figure 8.** Development of (**a**) plastic strain amplitude εa,p, (**b**) α'-martensite and (**c**) change in the specimen's temperature ΔT versus load cycles N during HCF tests at AT with f = 5 Hz of batch B.
