Effect of thermomechanical treatment and microstructure on pseudo-elastic behavior of Fe – Mn – Si – Cr – Ni-(V, C) shape memory alloy

This study systematically investigated the effects of heat and thermomechanical treatments on the pseudo-elastic behavior of an Fe-based shape-memory alloy (Fe – 17Mn – 5Si – 10Cr – 4Ni-1(V, C) %wt.). First, samples were solution-annealed at 1000 ◦ C for 2 h and aged at 760 ◦ C for 6 h. A thermomechanical treatment was then applied to the heat-treated samples. The microstructure and mechanical properties (i.e., 0.1% yield stress and pseudoelasticity) were characterized using X-ray diffraction, scanning and transmission electron microscopy, and uni- axial tensile tests. The results showed that decreasing the grain size and precipitation of VCs resulted in an improved pseudo-elasticity. The thermomechanical treatment decreased the number of thermal twins and developed a [111] texture in the austenite phase, which reduced the critical stress for inducing martensite. The presence of VC precipitates and texture formation in the thermomechanical-treated samples increased the pseudo-elastic strain more than twofold from 0.4 to 0.98%.


Introduction
Shape memory alloys (SMAs) are materials that can recover their initial shape after considerable strain by either removing stress (i.e., pseudoelasticity (PE)) or heating (i.e., shape memory effect (SME)) [1]. The PE behavior of SMAs has recently gained attention for applications in seismic-resistant structures. Recently, the PE characteristics of SMAs such as NiTi (also known as Nitinol) were used to reduce residual deformations in structures after seismic loading [2][3][4]. PE NiTi SMAs retain their original shape without permanent deformation and dissipate energy. Because of these outstanding characteristics, NiTi SMAs have been used as energy dissipation members [5], such as passive dampers [6], cross-bracing cables [7], and seismic isolators [6]. Nevertheless, the relatively high price of NiTi SMAs has hindered the widespread application of this material in the construction sector.
Iron-based SMAs (Fe-SMAs) have gained much attention over the last decade because of their good shape memory behavior, prominent mechanical properties (such as, stability of recovery stresses and acceptable elastic modulus), and low material manufacturing cost [8][9][10][11][12][13][14][15][16][17][18][19][20], justifying their large-scale application in civil engineering. Sato et al. [9] first reported the SME in Fe-Mn-Si-based SMAs in 1982, and these are regarded as a group of important shape memory materials. It has been shown that although Fe-SMAs have good energy dissipation characteristics and low-cycle fatigue performance, they exhibit a negligible PE.
It is commonly assumed that PE cannot occur in Fe-Mn-Si-based SMAs because of the non-thermo-elasticity of the martensitic transformation of γ to ε (face centered cubic (fcc) to hexagonal closed-packed (hcp)) in these SMAs [21]. However, the PE behavior in Fe-Mn-Si-based SMAs was reported to be influenced by the formation of NbC and subsequent thermomechanical treatment [22][23][24][25][26][27]. In recent years, a novel composition of Fe-SMA, namely Fe-17Mn-5Si-10Cr-4Ni-1(V,C), including nanosized VC, has been developed and optimized for civil engineering applications [28,29]. The alloy was primarily designed and optimized to provide a strong SME (and not PE), which can be used for the prestressed strengthening of civil structures. Investigations on the phase transformation [30], recovery stress and strain [31], creep and stress relaxation [32], corrosion behavior [33], elevated temperature [34,35], cyclic [35], and fatigue [8] behaviors of this Fe-SMA show that although it can be suitably used as a prestressing component for civil infrastructures [36][37][38][39][40][41], it offers only a negligible PE [35]. Hence, enhancing the PE behavior of Fe-SMAs can make this alloy a potentially interesting alternative to NiTi SMAs for seismic applications.
Although there are many studies on the SME of the Fe-17Mn-5Si-10Cr-4Ni-1(V, C) alloy (for example [40,41]), the PE of this alloy has only been scarcely investigated in the past [29,42,43]. Previous research demonstrated that a large elastic strain field forms near the VC precipitates, improving the SME and PE. Furthermore, VC precipitates can provide preferential nucleation sites for ε-phase and a high density of stacking faults in the austenite matrix [28,29,44,45]. Studies have also shown that special textures can improve the PE of this alloy [44,46].
Previous studies have reported that a thermomechanical treatment comprising a small deformation at ambient temperature and subsequent annealing above the A f temperature (finish austenite formation temperature) improved the SME of Fe-Mn-Si-based alloys [29,[47][48][49]. It was reported that the pre-rolling of solution-annealed austenite followed by an aging treatment of Fe-Mn-Si-based alloys containing C and Nb led to improved shape memory properties. The microstructure of the as-rolled specimens after aging had a good distribution of NbC particles and a high density of stacking faults in austenite. The interaction of NbC particles and stacking faults in austenite enhances the shape-memory properties [25]. In another study, a simple thermomechanical treatment of an Fe-Mn-Si-Cr SMA, which involved rolling at 600 • C followed by 10 min of aging at 800 • C, led to a relatively large PE behavior at approximately 100 • C. The creation of stacking faults and the reversible motion of correlated partial dislocations improved the PE [50].
Research on the enhancement of the PE of Fe-Mn-Si-based SMAs is limited. Therefore, there is a need for studies that provide a better understanding of the PE mechanisms in these alloys. This study aims, for the first time, to systematically investigate the influence of aging and thermomechanical training on the PE of the Fe-17Mn-5Si-10Cr-4Ni-1 (V, C) alloy.

Experimental procedure
The Fe-SMA investigated in this study was a 1.5 mm-thick hot-rolled (at 1100 • C) sheet with a chemical composition of Fe-17Mn-5Si-10Cr-4Ni-1(V, C) (wt. %) produced by re-fer AG. The specimens were cut into dog-bone shapes by electrical discharge machining. This hot-rolled material was designated as HR. The dimensions and geometry of the dog-bone-shaped samples was shown in supplementary I.
The hot-rolled samples were then subjected to heat treatment (aged at 750 • C for 6 h). The other samples were solution-annealed at 1000 • C for 2 h and then aged at 750 • C for 6 h, followed by air cooling and they were designated as HRA, S and SA, respectively (see Table 1). A quasistatic strain-controlled tensile test up to 2% and 4% strain with a constant displacement rate of 0.5 mm/min was performed to measure the PE. The hot-rolled and heat-treated samples underwent a thermomechanical treatment comprising quasi-static cyclic straincontrolled tensile loading-unloading up to 4% strain at room temperature and subsequent annealing at 500 • C for 15 min. The thermomechanical cycling was repeated four times to ensure complete forward and reverse transformation and rearrangement of martensite (training), then a tensile test was performed to measure the PE.
Quasi-static cyclic tensile tests were performed using a universal tensile testing machine (Z020; Zwick/Roell). The dog-bone-shaped samples were repeatedly loaded to a strain of 1% and unloaded with a constant displacement rate of 0.5 mm/min to a constant force of 10 N. Therefore, the samples were not tested under inelastic compressive strains. The strain increment was 1% for each cycle.
The recovered strain owing to PE (ε pr ) is characterized as the reversible strain during unloading because of the transformation of the ε-martensite (hcp) phase to austenite (fcc). ε pr was calculated by deducting the elastic strain (ε el ) from the total strain variation upon unloading (ε ul ) as follows: A typical stress-strain curve of the Fe-SMA with the associated definitions was shown in supplementary II. Microstructural characterization of the specimens was performed using scanning electron microscopy (SEM; FEI NanoSEM230), transmission electron microscopy (TEM; JEM-2200FS JEOL), and X-ray diffraction (XRD; Bruker D8) with Cu Kα radiation using a 0.012 mm-thick Ni filter. Image J software was used to analyze the microscopic images (for example, to measure the grain size and count the precipitates). The phase mapping and texture were evaluated on a SEM (Zeiss DSM96) using electron backscatter diffraction (EBSD) at 20 keV. Before the analysis, the surfaces of the specimens were ground and polished according to standard metallography preparation methods (ASTM E3) [51].

Influence of heat treatment on the pseudoelasticity
The effects of hot rolling and heat treatment on the microstructure were characterized using SEM. Fig. 1 shows SEM images of the samples (a) after hot rolling, (b) after hot rolling and aging at 750 • C for 6 h, (c) after solution treatment at 1000 • C for 2 h, and (d) after solution annealing at 1000 • C for 2 h and aging at 750 • C for 6 h. Fig. 1a suggests that the average grain size in the hot-rolled specimen was approximately 11 μm and that some grains had large bands, which were identified as thermal twins. These twins correspond to boundaries with low grain boundary energy [52]. Additionally, sporadic fine bands, assumed to be ε-martensite, were observed in some grains. The ε-martensite is considered to have formed during grinding. The average grain size increased to approximately 20 μm after the solution treatment ( Fig. 1b). After aging, elongated 100-300 nm Cr or V carbides precipitated inside the grains and along the grain boundaries ( Fig. 1b and d). Fig. 2 shows bright-field TEM image after aging at 750 • C for 6 h. Elastic distortion contrasts are visible around the precipitates, which illustrates that there is semi-coherency with the matrix, which causes an elastic strain field around the precipitates (Fig. 2). This elastic strain field near the precipitates is assumed to act as a spring back for the reverse transformation of ε (hcp) to γ (fcc) during unloading, leading to an improved PE. The high-resolution TEM image of the precipitate interface in (Supplementary III) confirms the semi-coherency of the precipitate interface with the matrix. Fig. 3 shows XRD patterns of the hot-rolled, solution-annealed, and aged samples. The XRD patterns of the hot-rolled and solution-annealed samples show that individual peaks related to the austenite phase were observed. As the ε-martensite bands were very thin, they could not be detected by XRD. Peaks related to the austenite and ε-martensite phases were detected in the XRD patterns of the undeformed aged samples. The presence of the ε-martensite phase in the undeformed sample can be Table 1 The yield stress (σ y0.1% and σ y0.2% ) and PE strain of the samples after tension up to 2% and 4% strain. identified by the presence of a residual stress field around the coherent precipitates.
When semi-coherent VC precipitates are formed at an aging temperature of 750 • C, the mismatch of lattice constants results in generation of stress fields around precipitates, as shown in Fig. 2. Furthermore, the austenite phase as a matrix material and carbide precipitates have different thermal coefficients during cooling, which result in thermal stresses and formation of the stress fields the vicinity of the precipitates.
These stress fields have an effect on the martensitic transformation temperature and according to the Clausius-Clapeyron equation, these stresses lead a shift in the phase transformation temperatures toward greater temperatures and also assist the formation of stress-induced martensite. Therefore, ε-martensite plates were created relatively easily during the aging process. Near these primary ε plates, the energy barrier for a greater phase transformation decreases. Meng et al. [53] reported that stress fields around the formed martensitic plate and crystal defects, such as dislocations, could reduce the martensite nucleation barrier. The number of carbide particles per area was estimated for the hotrolled and solution-annealed samples after aging by counting the   μm 2 ). The number density of the particles in the aged hot-rolled sample was 2-fold higher than that of the solution-annealed and aged sample. A number density of 1 (±0.05) × 10 12 particles/m 2 was measured for the aged hot-rolled samples.
The proper nucleation sites for heterogeneous precipitation are nonequilibrium defects, such as dislocations, stacking faults, grain boundaries, and excess vacancies. All of these factors cause an increase in the free energy of the material. The formation of a nucleus results in the elimination of defects; therefore, some free energies are released and the activation energy barrier either reduces or even vanishes [54]. According to the SEM images (Fig. 1), hot-rolled samples have a smaller grain size than the solution-annealed samples. Therefore, in the hot-rolled sample, the high fraction of grain boundaries, fine ε-martensite plates and some defects, such as stacking faults and dislocations, can act as a place for the heterogeneous nucleation of carbides during aging, leading to a greater density of precipitates, as compared to the solution-annealed sample.
Cyclic strain-controlled tensile tests (with displacement rate of 0.5 mm/min) were performed to characterize the mechanical properties of the different samples. Fig. 4 shows the stress-strain curves of the hotrolled, annealed, and aged samples under cyclic loading-unloading conditions. The figure shows that the unloading curves are non-linear owing to the PE behavior. The PE-recovered strain after unloading was calculated according to Eq. (1), and Table 1 presents the values for the 0.1 and 0.2% yield strengths of the material. It was previously found that the phase transformation from austenite to martensite initiated at stresses lower than σ y0.2% [55][56][57]. Therefore, σ y0.1%, which refers to 0.1% of the nonlinear strain, was considered as the limit of proportionality. Elastic deformation occurred up to a yield stress of σ y0.1% .
Beyond this point, the deformation could be due to the γ-austenite to ε-martensite transformation (because of the low stacking fault energy of this alloy) and dislocation glide. From the σ y0.2% point, the dislocation glide and slip had a more significant effect on the deformation.
The solution-treated samples showed a higher yield stress of 0.1% and lower yield strength of 0.2% than the hot-rolled samples, and the lowest PE strain of 0.3% is related to the solution-annealed sample. Stress-induced martensite formation occurred between the end of the linear elastic deformation and the beginning of the plastic deformation stages. Therefore, it is expected that a specimen with a higher σ y0.1% value and a lower yield strength (σ y0.2% ) will show a lower volume fraction of martensite and a more pronounced irreversible plastic deformation, which in turn decreases the PE. Therefore, it seems that this gap between the critical stress for transformation (σ y0.1% ) and the yield stress for slip (σ y0.2% ) has a direct influence on the PE behavior of the alloy. The stress-strain curves in Fig. 4 depict an improved strength for the hot-rolled samples, which could be attributed to the Hall-Petch effect as the grain size of hot-rolled samples is smaller than that of solution-annealing samples observed in Fig. 1.
The hot-rolled samples showed a more pronounced PE than the solution-annealed samples because of the smaller grain size. Reducing the grain size increases the austenite stability and PE owing to delayed slip deformation, and increases the interaction between the martensite bands and grain boundaries [46]. The aging treatment increased the strength and PE of the both hot-rolled and solution-annealed samples.
Residual stresses in the vicinity of the precipitates increase the ε transformation temperature and act as nucleation sites to simplify the growth of the ε phase. The precipitates in the austenite are in charge of reducing the martensite transformation stress (σ y0.1% ) and increasing the slip resistance of the matrix (σ y0.2% ) at room temperature, which has led to an increase in the PE in the aged samples. During the subsequent loading, the aged samples showed a more pronounced strain-hardening behavior, which is a characteristic of the fcc to hcp transformation. The interaction of different variants of martensite and the interplay of the martensite plate tips with grain boundaries and particles results in work hardening in the aged samples. The precipitates formed by increasing the yield strength of austenite (σ y0.2% ) hindered slip deformation in the aged specimens. Based on the results obtained from the tensile curves in Table 1, the hot-rolled aged sample (HRA) has the greatest stress gap between the transformation stress (σ y0.1% ) and the slip stress (σ y0.2% ), which has led to a significant increase in the PE. Fig. 5 shows the EBSD inverse pole figure (IPF) maps and the related phase maps of the microstructure of the solution-annealed and aged sample as well as the corresponding misorientation distributions. Fig. 5 shows the solution-annealed-aged (SA) sample was predominantly composed of γ-grains, but a small quantity of randomly distributed ε-phase was also detected in the microstructure. This small quantity of the ε-phase may be attributed to the formation of stress fields around the carbide precipitates because of the thermal expansion mismatch during the cooling procedure in the heat treatment. Furthermore, the misorientation distribution histograms for the solution-annealed and aged sample show that most of the boundaries are high angle due to the presence of grain boundaries and thermal twins. Because of the martensite formation in the aged sample, the number fraction of the lowangle boundaries (LABs) increased slightly. The EBSD analysis also shows that a very weak [001] texture along the rolling direction is observed in the austenite phase, which could be the remaining texture caused by rolling in the as-received samples. The EBSD inverse pole figure (IPF) maps and the related phase maps of the microstructure of the solution-annealed sample (S) as well as the corresponding misorientation distributions were presented in supplementary IV.

Effect of thermomechanical treatment on the pseudoelasticity
To improve the PE of the Fe-SMA, the samples were subjected to thermomechanical training. One cycle of thermomechanical treatment involved a deformation of up to 4% strain at room temperature and subsequent heating at 500 • C for 15 min. The stress-strain curves of each cycle were recorded for a detailed analysis and are shown in Fig. 6. Unlike NiTi and Cu-based SMAs with thermoelastic martensitic transformation, Fe-Mn-Si SMAs show pronounced work-hardening before training, as shown in Fig. 4.
The results showed that the yield stress increased after the thermomechanical training (Fig. 6). In the samples without precipitates (hotrolled and solution-annealed samples), the yield stress increased significantly with every training cycle (Fig. 6a and c). However, in the aged samples, a sharp increase in the yield stress occurred after the first cycle, and this increase was less pronounced in subsequent cycles. This increase was attributed to the work-hardening effect in each cycle. The results indicate that the minimum stress required to induce martensite (σ y0.1% ) was reduced in all the trained samples. The drop in the slope of the tensile curves indicates a work-hardening reduction and the easy formation of stress-induced martensite. Therefore, the thermomechanical treatment has resulted in widening of the stress gap between the transformation stress (σ y0.1% ) and slip stress (σ y0.2% ).   cycles, indicating that the martensite transformation was facilitated and the PE was enhanced. It is further presumed that thermomechanical training increases the dislocation density and the elastic strain field of the dislocations has resulted in a reduction of the nucleation barrier for the formation of coherent martensite nuclei.
Previous studies have shown that in Fe-Mn-Si SMAs with fcc to hcp transformations, work hardening cannot be fully ascribed to the interaction of stacking faults with martensite [58]. The cross-slip of dislocations is a significant factor that influences work hardening. However, in Fe-Mn-Si SMAs with low stacking fault energy, cross-slip does not occur. During the thermomechanical training of the Fe-Mn-Si SMA, the variation in the stress-strain behavior could be due to the interaction of the dislocation strain field and residual ε-martensite. Because of successive deformation and inadequate recovery, this results in pronounced work-hardening. Hence, a large amount of strain energy and a high internal strain accumulates in the microstructure.
The significant non-uniformly distributed internal strain in the Fe-Mn-Si SMA might support the hcp to fcc inverse transformation and could cause a lower A s (austenite start temperature) or a higher A f (austenite finish temperature). Past research showed that the residual inner stress in Fe-Mn-Si alloys can be released upon additional recovery at 420 • C [47,59]. It should be mentioned that two competing processes might occur during thermomechanical cycling, namely the formation of stress-induced martensite and multiplication of dislocations during deformation, and the reverse martensitic transformation and annihilation of dislocations during annealing at high temperatures. A past study on the training of Fe-Mn-Si alloys revealed that the dislocation accumulations during thermomechanical treatment became lower at higher recovery temperatures (500 and 600 • C) because of partial recrystallization [47].
Based on the results of the XRD pattern (Fig. 3), there are two types of martensite α ′ and ε in the HRA sample. The morphology of ε martensite is thin plate-type with planar interfaces. However, in the case of α ′ martensite, different types of morphologies have been reported, i.e., lath, butterfly and thin plate. Several previous studies [60,61] demonstrated that martensite has a hierarchical structure containing former austenite grains, packets, blocks, and laths or thin plates, as shown in Supplementary V. The martensite is a band restricted by LABs. Several martensite phases with similar crystallographic orientations form in a block restricted by boundaries with misorientations <15 • [62]. A packet is a collection of blocks with a similar plane. Wen et al. [63] recently has reported that if an austenite grain in an Fe-SMA is divided into finer domains by phases such as carbides or crystal defects, the martensite bands in the finer domains could be suppressed by these phases or defects [65]. Hence, stress-induced martensite bands were anticipated to form in a block, resulting in a reduction in collisions. Fig. 8a shows a SEM image of the aged sample after thermomechanical treatment. The figure shows that after thermomechanical treatment, an austenite grain is divided into packets and blocks, which contain untransformed martensite with the same orientation. Thus, the stress field of untransformed martensite reduces the barrier energy for new martensite transformation by providing an auto-catalyst mechanism, which in turn significantly improves PE.
The TEM images of the aged sample after the thermomechanical treatment in Fig. 8b shows a microstructure containing austenite grains consisting of ε-martensite and a high density of dislocations that form cell-like arrangements. Furthermore, Fig. 8b illustrates a prior austenite grain subdivided into finer domains, such that each new domains consists of blocks with single-oriented ε-martensite. Fig. 8b shows that aligned Cr or V carbides subdivide the grains into smaller domains. Thermomechanical training produced a uniform distribution of stacking faults and stress-induced martensite bands in the blocks.
Furthermore, the SEM and TEM images indicate that the thermomechanical treatment reduced the volume fraction of thermal twins, which was observed in the aged sample (Fig. 1c). Wen et al. [63] reported that the boundaries of the thermal twins act as obstacles to stress-induced martensite formation. Hence, stress-induced martensite forms more by decreasing the boundaries of the thermal twins after the thermomechanical treatments, leading to an increased PE. Fig. 9 shows the EBSD analysis results, PF maps, and the related phase maps of the microstructure of the thermomechanical-trained aged sample. The EBSD characterization revealed an almost random orientation before thermomechanical training (Fig. 5). Thermomechanical treatment resulted in the evolution of the texture structure of the alloy.
The volume fraction of ε-martensite significantly increased to 60% after thermomechanical treatment. A high fraction of LABs was observed in the thermomechanical-trained and aged sample after deformation up to 4% strain. The high fraction of LABs may be due to the compatibility and continuity of the plastic deformation between the austenite and martensite phases [64].
After recovery annealing at 500 • C for 15 min, the volume fraction of ε-martensite decreased to 5% (Supplementary VI), indicating that most of the stress-induced martensite reverted to austenite during recovery. Furthermore, a strong [111] texture in the austenite phase and [1011] in the martensite phase along the loading direction was observed after the thermomechanical treatment. The increase of the σ 0.2% yield strength for the HRA specimen strained along the rolling direction compared to the solution-annealed sample could be explained by the formation of this texture in the austenite phase. Moreover, it is known that the PE strain is dependent on the grain orientation. During fcc→hcp transformation, the fcc lattice can be transformed into an hcp lattice by forming partial dislocations separated by stacking faults on every second {111} layer. Previous studies have shown that there is a considerable influence of the grain orientation on the forward phase transformation fcc→hcp and the phase reversion hcp→fcc [44,46]. The hcp/fcc orientation relationship is given by Shoji-Nishiyama. There are four hcp variants which are orientated differently in the space, in which the fcc (111) planes are parallel to the hcp (0001) planes. It can be therefore concluded that the (111) texture apparently facilitates the growth of stacking faults and accordingly the phase transformation; a more detailed study on the underlying mechanisms using neutron diffraction experiments is ongoing at Empa, which will be published soon. If a grain containing many stacking faults has to choose between slip and phase transformation, it rather prefers the phase transformation as this means growth of existing stacking faults. Hence, the phase transformation is facilitated if there are already many stacking faults in a grain and the martensite thin plates could nucleate at these stacking faults. Furthermore, the reversible motion of partial dislocations could happen during unloading in these grains. After the thermomechanical treatment, the main (1011) texture was observed in martensite for the HRA sample (Fig. 9). The residual martensite phases after unloading in other planes show a (1011) texture, which can be transformed into austenite after recovery annealing at 500 • C as shown in Supplementary VI. Texture development coincides with the strain components of ε-martensite that improve PE, as reported for other ferrous alloys [65,66].
The results show that the combination of direct aging with thermomechanical treatment has led to an improvement of the PE by 125%, which is the greatest PE value reported for this alloy so far. This improvement is due to the control of the grain size and optimization of  the precipitation process through direct aging treatment, which has provided a sufficient increase in slip resistance with acceptable levels of transformation stresses. The thermomechanical treatment induced a gap between the transformation stress and the critical slip stress, which is helpful for the martensitic transformation and PE. Furthermore, the thermomechanical treatment resulted in formation of a texture in austenite and martensite. The results indicated that the slip resistance is dependent on the stress-state and grain orientation, and in order to achieve a great PE, the slip stress must be considerably greater than the transformation stress level.
In the case of Fe-17Mn-5Si-10Cr-4Ni-1(V, C) alloy investigated in this work, despite being able to achieve transformation stresses that are lower than slip levels, achieving a significant gap between the two stresses was possible through direct aging and thermomechanical treatment.

Conclusions
Although the FeMnSi SMA with a composition of Fe-17Mn-5Si-10Cr-4Ni-1(V, C) was suitably used as a prestressing component for civil infrastructures, it offers only a negligible pseudoelasticity (less than 0.4%). Hence, enhancing the pseudoelasticity can make this alloy a potentially interesting alternative to NiTi SMAs for seismic applications. Therefore, in this study, the effects of heat treatment (namely, solution annealing at 1000 • C for 2 h and aging at 750 • C for 6 h) and thermomechanical treatment (namely, cyclic tensile loading-unloading up to 4% strain at room temperature and subsequent annealing at 500 • C for 15 min) on the microstructure and mechanical properties (PE and yield stress) of this alloy were studied. The results indicated that both heat treatment and thermomechanical treatment led to a considerable improvement in pseudoelasticity of the FeMnSi SMA, which can be considered for the further development of this alloy. Based on the experimental results and microstructural evaluations, the following conclusions were obtained: 1. The formation of uniform distribution vanadium carbide precipitates after heat treatment led to the formation of stacking faults and ε-martensite in the microstructure and interaction of precipitates with the martensite plate tips provided spring back stress for revers transformation (martensite to austenite) resulting improved the pseudoelasticity and also increased work-hardening in the aged samples. 2. Microstructural engineering through heat treatment and thermomechanical training in the Fe-based SMA is employed to obtain a microstructure composed of austenite grains, a high number of stacking faults, precipitates, thin ε-martensite laths accompanied by partial dislocation arrays, which resulted to increase the pseudoelasticity by 125% (from 0.4 to 0.9%) after two cycles of training. 3. The critical stress for inducing martensite decreased with an increase in the number of training cycles. The reduction in the slope of the tensile curves indicates that the transformation stress was reduced, and stress-induced martensite formation became easier. On the other hand, thermomechanical treatment increased the stress fields of the untransformed martensite and dislocations and resulted to reduce the barrier energy for new martensite transformation by providing an auto-catalyst mechanism, which in turn significantly improves pseudoelasticity. 4. Thermomechanical treatment also developed a hierarchical structure containing former austenite grains, packets, blocks, and martensitic laths and resulted in the evolution of the texture structure of the alloy. Texture development (a [111] texture in the parent phase) coincided with the strain components of ε-martensite that improved pseudoelasticity 5. Hot-Rolled samples showed the higher pseudo-elasticity than solution annealed samples after heat treatment and training. The improved performance of hot-rolled specimens is related to their microstructure, which contain small grains that influence on the interaction between martensite -ε plates and the grain boundaries. 6. In fact, the grain refinements and precipitates improve the conditions for the back-and-forth movement of partial dislocations by increasing the back stresses. Therefore, methods that cause grain refinement, such as cold rolling or sever plastic deformation can lead to pseudoelasticity enhancement of this alloy and can be considered as future works.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
Data will be made available on request.

Acknowledgments
This study was funded by the EMPAPOSTDOCS-II program, which received funding from the European Union Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement number 754364. The contribution of re-fer AG in providing the test materials is also acknowledged. Any opinion and findings in this paper are those of the authors and do not necessarily reflect the view of the sponsors.