Oxide dispersion-strengthened steel PM2000 after dynamic plastic deformation: nanostructure and annealing behaviour

The microstructure, texture and mechanical properties have been studied in PM2000 compressed via dynamic plastic deformation to a strain of 2.1. It is found that dynamic plastic deformation results in a duplex 〈111〉 + 〈100〉 fibre texture and refines the initial microstructure by nanoscale lamellae, which substantially increases the strength of the material, but decreases its thermal stability. In the as-deformed microstructure, the stored energy density is found to be higher in 〈111〉-oriented regions than in 〈100〉-oriented regions. Recovery during annealing at 715 °C reduces the energy stored in the deformed microstructure. This reduction is more pronounced in the 〈111〉-oriented regions. Orientation-dependent recrystallisation takes place in the recovered microstructure, leading to strengthening of the 〈111〉 fibre texture component at the expense of the 〈100〉 fibre texture component.


Introduction
Oxide dispersion-strengthened (ODS) steels are promising structural materials for the next-generation fission and fusion reactors because of their excellent resistance to both irradiation damage and high-temperature creep [1,2]. It has been reported that irradiation tolerance in steels and other materials may further be improved if their microstructures are refined to the submicrometre or nanometre scale [3,4]. One well-known way to refine the microstructure is via plastic deformation when original grains are subdivided by deformation-induced dislocation boundaries [5,6]. For example, structural refinement in the range of 0.3-0.6 lm was produced by equal channel angular extrusion (ECAE) in a ferritic/martensitic steel T91 and a 12Cr ODS steel [3,7]. The swelling rate during the irradiation of the submicrometre-grained T91 steel after warm ECAE to a strain of 2.3 was found to be three times lower than that of a coarse-grained sample [3]. Compared to the structural refinement achieved by ECAE in [3,7], a much finer boundary spacing can be obtained if the deformation is performed at high strain rates. For instance, a lamellar structure with a boundary spacing of only 0.1 lm developed in a modified 9Cr-1Mo steel (T91) [8] due to compression via dynamic plastic deformation (DPD) [9] to a strain of 2.3 at a strain rate of 10 2 -10 3 s -1 . A more effective structural refinement by DPD as compared to lowstrain-rate deformation has also been documented for other metals such as aluminium and nickel [10,11]. It is reasonable to expect that DPD can also be very effective in refining the microstructures of ODS steels. The primary aim of this work is therefore to investigate the effect of compression by DPD on the microstructure of an ODS material. For this purpose, we chose to study PM2000 [12], an iron-based ODS alloy developed for application in power plants. Since nanostructured materials produced by deformation are in general less thermally stable than their less-refined counterparts [13], the annealing behaviour of DPD-processed PM2000 is also investigated in the present work. Both transmission electron microscopy (TEM) and electron backscatter diffraction (EBSD) are used here to enable a detailed microstructural analysis of changes taking place during DPD and subsequent annealing. The microstructural analysis is complemented by hardness measurements and tensile tests.

Experimental
The nominal chemical composition of PM2000 [12] is shown in Table 1. The material was received in the form of a hot-extruded rod with a diameter of 13 mm. Two cylindrical specimens with a diameter of 6 mm and height of 9 mm were then machined with their cylinder axis along the extrusion direction (ED), see Fig. 1a. The specimens were compressed at room temperature by DPD in five steps to a final thickness of 1.1 mm, which corresponds to an equivalent strain, e vM = ln(h 0 /h) of 2.1, where h 0 is the initial height, and h is the final sample thickness, respectively. The diameter of the compressed samples was about 17.2 mm (Fig. 1b). Parts of the deformed samples were then annealed for 1 h at different temperatures between 500 and 800°C and for different time intervals at 715°C.
TEM foils from the longitudinal section in the asreceived sample and from the longitudinal section near the centre of the compressed disc (see Fig. 1c) were prepared by twin jet electropolishing in a solution of ethanol (70 vol%), water (12 vol%), 2-butoxy-ethanol (10 vol%), and perchloric acid (8 vol%). TEM images were obtained using a JEOL 2000FX transmission electron microscope operating at 200 kV.
EBSD analysis was performed using a Zeiss Supra 35 field emission gun scanning electron microscope equipped with a Channel 5 system. In the deformed sample, several regions with a total area of 750 lm 2 were investigated by the EBSD technique. Larger areas, at least 2500 lm 2 , were investigated in each annealed sample. Due to the limited angular resolution of this technique [14,15], misorientation angles less than 2°were not considered in the analysis. Boundaries with misorientation angles h between 2°a nd 15°were defined as low-angle boundaries (LABs), while boundaries with misorientation angles larger than 15°were classified as high-angle boundaries (HABs). Recrystallised grains in the orientation maps were defined as regions at least partly surrounded by HABs, having an equivalent circular diameter (ECD) above 3 lm and an internal point-topoint misorientation below 2° [16]. Area fractions of different fibre texture components were calculated allowing a 10°orientation deviation from the exact huvwi fibres. These area fractions are representative of volume fractions.
The boundary area density S V was also determined from the EBSD data [17,18], and the energy density stored in the form of boundaries in the deformed or recovered microstructure was estimated as u = c S V . The specific boundary energy of HABs was assumed to be c HAB = 617 mJ/m 2 [19]. The specific boundary energy of LABs with misorientation angles h \ h cr = 158 was calculated using the Read-Shockley equation [20,21]: Vickers hardness was determined using a load of 1 kg with a 10 s dwell time. Dog-bone shaped specimens with a gauge section of 5 9 1 9 0.5 mm 3 prepared from the as-received sample and the compressed sample were tensile tested at room temperature using an Instron 5848 MicroTester with an initial strain rate of 5 9 10 -3 s -1 . The tensile axis was aligned along the ED for the as-received condition and perpendicular to the compression axis (CA) for the compressed sample. A MTS LX300 laser extensometer was used for measuring strain during the test.

As-received PM2000
A bright field TEM image in Fig. 2a shows that the as-received material is characterised by a high dislocation density and contains well-dispersed oxide nanoparticles identified in our previous work [22] as orthorhombic yttrium-aluminium oxide YAlO 3 . The orientation map obtained by EBSD in Fig. 2b demonstrates both almost equiaxed and highly elongated grains. As is evident from the colour code in the inset, most grains have either h100i or h110i directions aligned along the extrusion direction. Approximately 50 % of all boundaries in this material are HABs. The average spacing d h [2°, defined as the mean distance between boundaries with misorientation angles greater than 2°, is 580 nm, and the HAB spacing is 940 nm as measured along the ED.
The hardness of this material is 332 HV1. The tensile test reveals a combination of moderate ultimate tensile strength (UTS) of 932 MPa and reasonably high ductility (see Fig. 3) with an elongation to failure of about 15 %.

PM2000 after compression by DPD
A characteristic deformation structure with lamellar boundaries almost perpendicular to the CA is observed after DPD (see Fig. 4). The mean thickness of the lamellae measured in TEM images (see an example in Fig. 4a) is 72 nm, and the d h[2°m easured along the CA in the orientation map in Fig. 4b is 76 nm, i.e. after compression by DPD the sample is  truly nanostructured. The fraction of HABs in this material is 46 %.
The orientation map in Fig. 4b indicates that lamellae are arranged in bands with either h100i directions (red in Fig. 4b) or h111i directions (blue in Fig. 4b) along the CA. The texture of the DPD-processed sample can therefore be described as a duplex h111i ? h100i fibre texture.
Compared to the as-received condition, the hardness and the UTS of the DPD-processed sample are increased to 443 HV1 and 1480 MPa, respectively. The elongation to failure of PM2000 after DPD is only 3 % (see Fig. 3).

Softening
The effect of annealing was first evaluated by conducting Vickers hardness measurements after annealing at different temperatures for 1 h. Results of these measurements (see Fig. 5) demonstrate that annealing for 1 h at temperatures up to 500°C does not appreciably affect the hardness reached by DPD. Within the temperature range, 500°C \ T B 700°C, the material softens gradually to 383 HV1. A sharp drop in hardness is observed when the annealing temperature exceeds 700°C (Fig. 5). For all the samples annealed in the temperature range, 715°C B T B 800°C, the hardness is below 300 HV1.
As the transition between slight and large reductions in hardness occurs at 715°C, the microstructural evolution during annealing was investigated at this temperature. Samples annealed for 10, 20 and 80 min at 715°C were chosen for a detailed microstructural analysis.
Evolution of microstructure and texture TEM images and orientation maps for the annealed samples selected for microstructural investigations are presented in Figs. 6 and 7, respectively. The TEM images provide evidence that the oxide nanoparticles have a certain pinning effect on dislocations (Fig. 6a) and migrating boundaries (Fig. 6b). The microstructure after 10 min of annealing at 715°C is considerably coarser (d h[2°= 176 nm) than that after DPD (see Fig. 8a). This coarsening, however, does not change the lamellar morphology induced by DPD (see Fig. 7a). The fraction of HABs after 10 min is similar to that in the as-deformed condition (see Fig. 8b). Recrystallisation nuclei are very rare after 10 min (one nucleus is marked by an arrow in Fig. 7a), and the area fraction of recrystallised material f RX is only 1 %.
After 20 min of annealing, f RX is much larger, 68 %, and the average boundary spacing d h[2°i s increased to 750 nm (Figs. 7b, 8). After annealing for 80 min,   (Figs. 7c, 8a, c). The fraction of LABs in this almost fully recrystallised microstructure is only 19 % (Fig. 8b). The orientation maps demonstrate that the majority of recrystallised grains have a h111i direction along the CA (blue in Fig. 7). In the sample annealed for 80 min (Fig. 7c), such h111ioriented grains occupy 67 % of the area. Figure 9 presents an example of the non-recrystallised regions still present in the microstructure after 80 min of annealing. A very fine lamellar structure with an average lamellae thickness of 165 nm and a high density of dislocations are preserved in such non-recrystallised regions. Analysis of selected area diffraction patterns from several lamellae in Fig. 9b indicates that their orientations belong to the h100i fibre texture.
The evolution of crystallographic texture measured by the EBSD technique in large sample areas is illustrated by inverse pole figures (IPFs) in Fig. 10. It is seen that the duplex h111i ? h100i fibre texture formed by DPD evolves during annealing into a welldefined h111i fibre texture, whereas the strength of the h100i fibre texture component decreases. Considering the area fractions of the individual fibre texture components in Fig. 11, it is apparent that during recrystallisation the h111i fibre strengthens at the expense of the h100i fibre, i.e. recrystallisation in PM2000 after compression by DPD is orientationdependent.

Nanostructure and strength
Compression of PM2000 by DPD to a strain of 2.1 results in nanoscale lamellar structures with a boundary spacing of 72 nm as measured along the CA using TEM. This spacing is considerably smaller than the value of 98 nm measured in a modified 9Cr-1Mo steel compressed by DPD to a similar strain [8]. The smaller spacing in PM2000 as compared to the modified 9Cr-1Mo steel may be attributed to a finer initial microstructure in PM2000 before DPD as well as to the presence of oxide nanodispersoids, which could make structural refinement of this ODS alloy more efficient. Such accelerated structural refinement has previously been reported for rolled aluminium containing small alumina particles [23]. Similar to other compressed body-centred cubic materials,  PM2000 develops a strong duplex h111i ? h100i fibre texture during compression by DPD [7,8,24,25].
The structural refinement by DPD makes PM2000 very hard and increases the UTS to 1480 MPa, which is significantly higher than the strength in the as-received condition and the UTS previously reported for PM2000 and other ODS steels such as mechanically alloyed MA956 and MA957 [26]. The strength of PM2000 after DPD is also higher than that of the modified 9Cr-1Mo steel processed by DPD to a similar strain, in which the UTS is only 1240 MPa [8]. The higher strength in the DPD-processed PM2000 is consistent with the finer microstructure in this material. The increase in strength of PM2000 during DPD is accompanied by a significant reduction in ductility-a phenomenon which has previously been observed in many cold-deformed materials due to their limited capacity of work hardening [8,27,28].

Recovery and recrystallisation
Our investigation demonstrates that the nanostructure formed in PM2000 after compression by DPD is fairly stable during annealing for 1 h at temperatures as high as 500°C. However, annealing at temperatures higher than 500°C makes the nanostructure unstable, as follows from the significant softening observed in Fig. 5 after annealing above 500°C. The shape of the hardness curve in Fig. 5 implies that the deformed microstructure only recovers within the temperature range from 500 to 700°C, whereas at temperatures higher than 700°C pronounced recrystallisation takes place. It is therefore apparent that compared to the as-received condition, in which no evidence of recrystallisation was observed even after annealing 1100°C for 1 h [29], the thermal stability of the nanostructured condition is considerably reduced.
The microstructural observations made in PM2000 after DPD and annealing at 715°C indicate that during recovery (within the first 10 min of annealing) the microstructure coarsens despite a certain pinning effect imposed by nanoparticles on boundary migration. The coarsened microstructure largely retains the lamellar morphology and maintains a similar proportion of LABs and HABs to that Figure 7 Orientation maps for PM2000 after compression by DPD and subsequent annealing at 715°C for a 10 min, b 20 min and c 80 min. The colours correspond to crystallographic directions along the CA according to the inverse pole figure in the lower right corner. One recrystallised grain found after 10 min of annealing is encircled in a. White and black lines represent LABs and HABs, respectively. In each map, the CA is horizontal.
observed in the as-deformed microstructure (see Fig. 8b). Although the density of interior dislocations was not determined quantitatively in the present experiment, it is apparent that the dislocation density in the TEM images taken after 10 min at 715°C is lower than that directly after DPD (cf. Figs. 4a, 6). This suggests that at 715°C dislocations annihilate despite the pining effect of nanoparticles.
Very few recrystallised grains appear within the first 10 min of annealing, but within the next 10 min (a total annealing time of 20 min) recrystallisation progresses very quickly and is almost complete after 80 min (see Figs. 7, 8c). As is evident from Fig. 7, recrystallisation proceeds in such a way that the frequency of h111i-oriented grains growing into the recovered environment is greater than the frequency of growing grains of any other orientation. This orientation-dependent recrystallisation leads to a clear predominance of the h111i fibre when the material is almost fully recrystallised (see Figs. 10, 11).
To understand the reason for the observed differences in the annealing behaviour of regions having different orientations, structural parameters of the deformed and recovered material are shown in Fig. 12 separately for the two different texture components. The data in Fig. 12a indicate that after DPD the average boundary spacing within the h100ioriented lamellae is larger (80 nm) than within the h111i-oriented lamellae (68 nm). After 10 min of annealing, however, the difference between the average boundary spacings d h[2°i n the recovered h111i and h100i-oriented regions is reduced, while after 20 min of annealing, d h[2°f or the h111i-oriented regions is even larger than that for the h100i-oriented regions. Figure 12b shows the stored energy density u estimated from the EBSD data for the h111iand h100ioriented regions in the same deformed and recovered samples. It is seen that in the deformed microstructure u h111i is considerably higher than u h100i . This difference can be attributed to the difference in the Taylor factor M between the h111i and h100i-oriented regions. According to the pencil-glide model proposed by Taylor, in body-centred cubic metals for grains compressed along h100i M is 2.1, while for compression along h111i the value of M is 3.2 [30]. Considering that the Taylor factor describes the ratio between the magnitude of accumulated plastic slip and the magnitude of applied strain, the higher Taylor factor for regions of the h111i fibre texture component suggests an increased slip activity in these regions. Consequently, a larger dislocation density and a higher stored energy are expected for the h111i-oriented regions compared to the h100ioriented regions, which is fully consistent with the stored energy values obtained for the as-deformed microstructure (see Fig. 12b). Such a higher stored energy in the h111i-oriented regions provides a higher driving force for nucleation of recrystallisation, and many more nuclei are formed within the h111i-oriented regions. These nuclei have h111i orientations, thus leading to the dominance of the h111ioriented recrystallized grains observed after annealing for 20 min (see Fig. 7b). The growth of abundant h111i-oriented nuclei results in the dominant h111i fibre texture and explains the orientation dependent recrystallisation behaviour. Such a correlation between the Taylor factor, stored energy and preferential nucleation is consistent with previous reports describing the annealing behaviour of body-centred cubic steels after rolling [31][32][33][34].  Nevertheless, it should be emphasised that the initially large difference in the stored energy density between h111iand h100i-oriented regions in the deformed condition becomes rather small after annealing for 10 min at 715°C. Moreover, after 20 min the stored energy density within the h111ioriented regions is even lower than of the h100i-oriented regions. Obviously, the stored energy density within the h111i-oriented regions is reduced during recovery much faster than within regions of the h100i fibre texture component (see Fig. 12b). For nucleation of recrystallisation beyond 10 min, when the stored energy density in both texture components is similar, there might be other reasons for preferred formation of h111i-oriented nuclei. One possible reason is that the morphological characteristics of recovered subgrains may be different for the h111i component and the h100i component [29]. Considering that rather straight lamellar boundaries are still observed within recovered h100i-oriented regions even after 80 min at 715°C (Fig. 9), it is reasonable to suggest that the recovered lamellar structure impedes the formation of nuclei in these regions.

Concluding remarks
Compression of PM2000 by DPD to a strain of 2.1 significantly refines the initial microstructure by nanoscale lamellae having a duplex h111i ? h100i fibre texture. The strength of material increases substantially, whereas the thermal stability of the nanostructured material is reduced. This reduced thermal stability is attributed to a high density of dislocations and dislocation boundaries induced by DPD. In the as-deformed microstructure, the stored energy density estimated from the EBSD data is found to be higher in the h111i-oriented regions than in the h100i-oriented regions. Recovery during annealing at 715°C reduces the energy stored in the deformed microstructure, and this reduction is more pronounced in the h111i-oriented regions. Orientation-dependent recrystallisation takes place in the recovered microstructure, leading to the strengthening of the h111i fibre texture component at the expense of the h100i fibre texture component. As a result, a strong h111i fibre texture is observed when recrystallisation is almost complete. Figure 11 Area fractions of the h111i and h100i fibre texture components in PM2000 after compression by DPD and subsequent annealing at 715°C.

Figure 12
Changes in deformed and recovered regions of either h111i or h100i fibre texture components in PM2000 after compression by DPD and subsequent annealing at 715°C: a average boundary spacing d h[2°a long the CA; b stored energy density u calculated from the EBSD data.