A. LCF response

The cyclic stress response curves of MS XVI at 650 °C under three different strain amplitudes are shown in Figs. 1(a) and 1(b). In general, the total cyclic life could be subdivided into four distinct stages. In first stage, the cyclic stress response is almost stable for the first ten cycles [Fig. 1(a)]. Thereafter, in second stage, the material undergoes cyclic hardening where peak stresses increase gradually (30–55 MPa) until they reach a maximum [Figs. 1(a) and 1(b)]. This stage represents the most fraction of the cyclic life (up to ∼0.6N _f) depending on the loading amplitude. Here, the rate (to achieve maximum) and the amount of cyclic hardening increases with an increase in applied strain amplitude. Post-hardening, stress amplitudes are stable during the third stage and occupy a minor fraction of the cyclic life (up to ∼0.2N _f), as visible in Fig. 1(b). Finally, governed by the fatigue crack initiation and its propagation, peak stresses reduce gradually followed by a precipitous drop until failure.

FIG. 1.

Cyclic stress response curves [curves of tensile peak stress versus logarithmic number of cycles (a) and versus normalized cycles (b)] of MS XVI at 650 °C under three different strain amplitudes. (c) Comparison of the cyclic stress response curve of MS XVI with those of previously reported MS VI and MS VIII.

Figure 1(c) features the comparison of cyclic stress response of MS XVI with previously reported versions MS VI and MS VIII, see Ref. Reference Straßberger, Chauhan, Gräning, Czink and Aktaa10. It is evident that though MS XVI exhibits a similar cyclic stress response [Fig. 1(c)], it clearly offers highest cyclic strength and comparable cyclic life, which is quite remarkable. In addition to this, MS XVI undergoes a relatively higher second stage hardening than that observed for MS VI and MS VIII [Fig. 1(c)]. This behavior is also realized for other tested strain amplitudes (not shown here). Hence, from here, it can be concluded that MS XVI offers superior cyclic properties in comparison to previously reported MS VI and MS VIII which themselves showed better fatigue resistance in respect of their non-ODS counterpart.^{Reference Straßberger, Chauhan, Gräning, Czink and Aktaa10} The main LCF test results of MS XVI are summarized in Table III while those of MS VI and MS VIII can be found in Ref. Reference Straßberger, Chauhan, Gräning, Czink and Aktaa10. The values of the well-known LCF parameters obtained by fitting test data to the corresponding LCF relations^{Reference Ramberg and Osgood14–Reference Manson16} (see Fig. 2) are listed in Table IV.

TABLE III.

Summary of the LCF tests on MS XVI austenitic ODS steel.

^a Measured half-width of the hysteresis loop at zero stress at half-life.

^b Measured at half-life.

FIG. 2.

Bilogarithmic plots of the (a) stress amplitude $\left( {{{\Delta {\sigma _{\rm{t}}}} \over 2}} \right)$ versus inelastic strain amplitude $\left( {{{\Delta {\varepsilon _{{\rm{in}}}}} \over 2}} \right)$ fitted linearly with cyclic hardening relation^{Reference Ramberg and Osgood14} and (b) inelastic strain amplitude $\left( {{{\Delta {\varepsilon _{{\rm{in}}}}} \over 2}} \right)$ versus number of reversals to failure (2N _f) fitted linearly with the Coffin–Manson relation.^{Reference Coffin15,Reference Manson16}

TABLE IV.

Values of the LCF parameters obtained by fitting test data measured at half-life to the corresponding mentioned relations.

B. Microstructural observations

To understand deformation mechanisms, microstructures were characterized before and after cyclic straining.

The representative as-received/undeformed state microstructure, observed via TEM in the HAADF-STEM mode, is shown in Fig. 3. This mode shows contrast due to different chemical compositions (Z-contrast based on atomic number), crystallographic orientation, and dislocation density within the grains. As shown in Fig. 3, MS XVI in the as-received state manifests almost equiaxed grains which have a size distribution varying from 0.1 to 2 µm with dominance below 1 µm. In addition, some grains exhibit typical crystallographic defects such as dislocations, stacking faults, and twins, while most grains are almost defect-free. The misorientation analysis via electron backscattered diffraction (not shown here) revealed that the fractions of high-angle boundaries (>15°) and low-angle boundaries (2–15°) are 73% and 27%, respectively.^{Reference Gräning11} A small fraction of boundaries are the Σ3 twinning boundaries. Furthermore, various nanofeatures are also observed embedded in the matrix, which are planned to be reported in detail separately. Nevertheless, briefly, as shown in Fig. 3, some grain boundaries are decorated with a bright contrast Cr-enriched phase. Here, it is important to mention that this phase could not only form during MA due to the combination of chromium with carbon and oxygen to reduce the overall Gibbs formation energy but can also be an aftermath of segregation amid extrusion.^{Reference Gräning11} Besides, the majority is of the different dark-contrast nano-oxide particles, identified as Y₂O₃, Y₂Ti₂O₅, Y₂Ti₂O₇, and Y–Ti–Cr–O, which are distributed in both trans- as well as intergranular regions. These nanoparticles have a size distribution ranging from 1 to 70 nm with most exhibiting a diameter below 5 nm.

FIG. 3.

HAADF-STEM micrograph revealing MS XVI microstructure from the as-received state.

After cycling, the overall microstructure appears stable without any modifications in grain shape and size. This can be attributed to the presence of the intergranular obstacles which successfully pinned grain boundaries from migration (see the inset in Fig. 3). In addition, twinning and stacking fault fractions in general seem unchanged, as they are still perceived in only few grains. This is in accordance with LCF investigations on conventional austenitic 316L SS,^{Reference Srinivasan17} where also no deformation-induced twinning was reported. In general, deformation-induced twinning is considered as a stress-controlled process with a critical shear stress associated with it.^{Reference Byun18} Therefore, nucleation of twins can only take place if the externally applied stress is higher than the critical shear stress for twinning (which depends on the stacking fault energy, crystallographic orientation, grain size, strain rate, and temperature), otherwise slip mechanism will dominate.^{Reference Byun18–Reference Han, Cheng, Li, Wu and Zhang20} Indeed, after cyclic straining, the main difference was observed in respect of the dislocation density.

Figure 4 presents the comparison of the microstructure before and after cyclic straining at 650 °C under higher strain amplitude (±0.7%) in the inverted contrast HAADF-STEM mode. It is evident that in the as-received state [Fig. 4(a)], most grains appear clean with light contrast due to the low dislocation density. Whereas, after cycling [Fig. 4(b)], majority of grains manifest an overall homogeneous dark contrast due to high dislocation density (see the inset). Unfortunately, the quantitative dislocation densities could not be estimated due to uncertainty in the determined TEM sample thickness. However, though qualitative, the overall dislocation density is certainly increased, in comparison to that of the undeformed state. Hence, it can be concluded that the second stage hardening in MS XVI is due to the overall increment of the dislocation density. Moreover, as for hardening, the dislocation density is also a function of loading amplitude which increases with its magnitude. For example, apart from the usual deformed grains at lower strain amplitude (±0.4%), decent numbers of grains are still nearly free of dislocations [see numbered grains in Fig. 5(a)]. This indicates a comparatively inhomogeneous deformation. The dislocation-free regions/grains apparently lower the overall dislocation density and therefore, as a consequence, lower associated cyclic hardening at lower strain amplitudes. This is also related to the comparatively lower-induced inelastic strain at lower applied strain amplitude than at higher ones, see Table III.

FIG. 4.

Inverted contrast HAADF-STEM micrographs reveal microstructures (a) before (as-received state) and (b) after cyclic straining at 650 °C under ±0.7% strain amplitude. A relatively darker contrast grain in (b) is due to the presence of high dislocation density.

FIG. 5.

Inverted contrast HAADF-STEM micrographs revealing microstructures after cyclic straining at 650 °C under ±0.4% strain amplitude. (a) Numbered grains (1, 2, 3, 4, and 5) that are nearly free of dislocations and (b) grains at higher magnification with high density of short dislocations pinned at various nanoparticles.

During cyclic deformation, microstructural modifications transpire to achieve stable low energy configurations. Here, the modification rate and amount precisely depend upon the testing conditions such as deformation rate (strain rate), its extent (amplitude), and temperature. However, before the steady state is attained, the competition between dislocation generation/multiplication and their annihilation/storage (via dislocation–dislocation annihilation and dislocation–grain boundary interaction) occurs simultaneously. And the dominance of one mechanism over the other defines the evolution of dislocation density, and hence, the overall material behavior. For example, in case of MS XVI, a material with low initial dislocation density, the increment in dislocation density upon cycling suggests that the rate of dislocation generation/multiplication is obviously far greater than that of their annihilation/storage. By contrast, in ferritic–martensitic steels, which are known to undergo cyclic softening due to the overall reduction in dislocation density,^{Reference Chauhan, Hoffmann, Litvinov and Aktaa5} the rate of dislocation annihilation/storage is significantly larger than that of their generation/multiplication. Nevertheless, once the steady state is reached, the rate of dislocation generation/multiplication as well as their annihilation/storage is balanced. And hence, a stable dislocation density (which is apparently higher than in the undeformed state for MS XVI) leads to a stable cyclic response with constant peak stress (third stage).

In respect of the dislocation structures, mostly intragranular dislocation activity is observed along with the occasional dislocation pile up (Fig. 6). However, no noticeable three-dimensional (3D) dislocation arrangements such as the cells, ladder, vein, labyrinth, and persistent slip bands were identified. These typical 3D dislocation structures are commonly observed upon cycling conventional non-ODS austenitic steels, e.g., 316L SS^{Reference Srinivasan17,Reference Gerland and Violan21} which are associated with the formation of extrusions and intrusions where fatigue cracks nucleate.^{Reference Man, Obrtlík and Polák22} Their absence in MS XVI could be due to (i) embedded nano-oxide particles and/or (ii) its ultrafine grains (mostly below 1 µm).

FIG. 6.

Inverted contrast HAADF-STEM micrographs revealing dislocation structure after cyclic straining at 650 °C under ±0.7% (a) and ±0.4% (b) strain amplitudes. Clearly, high intragranular dislocation activity along with dislocation pile up is apparent from (a). The inset from (b) shows dislocation pinned at various nanoparticle–matrix interfaces.

Concerning oxide particles, in almost all grains, individual dislocations (short as well as long) appear pinned at the nanoparticle–matrix interfaces and under the effect of shear stresses, most of them are bowed out (Fig. 6). This indicates an attractive particle–dislocation interaction since pinning is realized on both arrival as well as departure side of the nanoparticles.^{Reference Chauhan, Hoffmann, Litvinov and Aktaa5} Therefore, there exists a threshold stress to detach dislocation out of this pinning, which must be overcome for the dislocation to glide; and therefore, plastic deformation to occur.^{Reference Nardone and Tien23–Reference Reppich25} This results in a retarded dislocation’s mobility, and hence, inhibiting strain localization by prohibiting 3D dislocation structure formation. Moreover, dislocation pinning provides a consequent strengthening which give ODS steels their superior monotonic as well as cyclic strength over non-ODS counterparts, e.g., see Refs. Reference Chauhan, Walter and Aktaa4, Reference Xu, Zhou, Li and He6, and Reference Straßberger, Chauhan, Gräning, Czink and Aktaa10. Furthermore, in comparison to the previous versions (MS VI and MS VIII), MS XVI higher cyclic strength and comparable cyclic life [Fig. 1(c)] can be attributed to its higher yttrium fraction (∼174%), which eventually resulted in a formation of higher density of impenetrable nano-oxides. This higher density of nano-oxide particles leads to a higher percentage of dislocation pinning which not only shows up in terms of their higher strength but also due to their higher restricted movement less number of dislocations reach the specimen’s surface for annihilation which retards the onset of damage.

However, Thiele et al. also pointed out that some minimum volume (∼1 µm in terms of grain size) is required for the creation of persistent slip bands or 3D dislocation structures.^{Reference Thiele, Holste and Klemm26} And since MS XVI is designed to have a submicron sized matrix, to achieve additional superimposed grain size strengthening without adversely affecting ductility and toughness,^{Reference Chauhan, Bergner, Etienne, Aktaa, de Carlan, Heintze, Litvinov, Hernandez-Mayoral, Oñorbe, Radiguet and Ulbricht27} the formation of above-mentioned 3D dislocation structures is also suppressed.^{Reference Chauhan, Litvinov and Aktaa28} On the contrary, ODS steels with particles deficient large grains and/or large volumes with similar orientations manifest 3D dislocation structures.^{Reference Chauhan, Litvinov and Aktaa28,Reference Kubena and Kruml29} Hence, due to both homogeneous particle distribution and ultrafine grains, the onset and progress of damage in MS XVI will also be constrained since no 3D dislocation structures evolve which are associated with the formation of extrusions and intrusions.^{Reference Man, Obrtlík and Polák22}

Besides, in comparison to the previously reported MS VI and MS VIII, MS XVI exhibits higher second stage hardening [Fig. 1(c)]. This can be attributed to its comparatively larger grain size. In terms of number fraction, the MS XVI average grain size (∼318 nm) is 172% and 189%, higher than that of the MS VI (∼117 nm) and MS VIII (∼110 nm), respectively.^{Reference Gräning, Rieth, Hoffmann and Möslang9} Large grains are known to have an excellent dislocation storage capacity and significant work hardening ability.^{Reference Wang, Chen, Zhou and Ma30} This significant work hardening ability of the large grains not only assists in minimizing the stored energy but also accumulates inelastic strain during cyclic straining.^{Reference Chauhan, Litvinov and Aktaa28} Hence, due to its comparatively higher average grain size, MS XVI exhibits higher cyclic hardening. At first glance, MS XVI higher average grain size appears counterintuitive with its comparatively higher density of impenetrable nano-oxides. Since higher density of dispersed nano-oxide particles are usually known to hinder grain growth more effectively.^{Reference Humphreys and Hatherly31} However, it is important to note that in comparison to previously reported MS VI (60 h) and MS VIII (70 h),^{Reference Gräning, Rieth, Hoffmann and Möslang9} MS XVI powder was milled for shorter duration, i.e., 35 h [as this milling time was sufficient to distribute oxide forming species (e.g. yttrium) into solid solution homogeneously^{Reference Gräning11,Reference Deschamps, de Geuser, Malaplate and Sornin13} ]. As a consequence, MS XVI powder underwent a comparatively lower degree of deformation which is manifested in terms of lower density of crystal defects such as dislocations and grain boundaries than MS VI and MS VIII powder. These crystal defects are known to be the nucleation sites for new grains during recrystallization.^{Reference Humphreys and Hatherly31} Therefore, due to its lower crystal defect density, MS XVI powder should have comparatively less newly nucleated grains that have grown during recrystallization amid hot extrusion, which is vice versa for MS VI and MS VIII. Furthermore, precipitation and coarsening of the nanometer-scale oxides also occur during hot extrusion.^{Reference Deschamps, de Geuser, Malaplate and Sornin13} Hence, this nano-oxide precipitation and the grain growth during extrusion are two competing processes (timewise). And if grain growth occurs before the formation of precipitates, these grains can grow freely until the precipitation has formed large enough precipitates to exert a pinning force, which is larger than the driving force for grain growth. It appears that the grain growth was relatively faster in the MS XVI than the precipitates formation, due to the shorter milling duration.