Shape memory alloys (SMAs), which display shape memory effect and superelasticity [[1], [2], [3]] and show great potential for applications in the aerospace [2], automotive [3], biomedical [4,5] and oil exploration industries [6] as compact actuators, have been widely investigated in recent decades. In most SMA actuators, a bias force is generally needed to reset actuation, which makes the design of actuators complicated [[7], [8], [9]]. Fortunately, two-way shape memory effect (TWSME) enables SMA actuators to remember both the high-temperature and low-temperature shapes without the need of a bias force, which simplifies the design and reduces maintenance costs [[10], [11], [12]]. The TWSME is usually ascribed to the oriented internal stress field [[13], [14], [15]] that is able to bias preferentially oriented martensite variants during martensitic transformation and thus bring about reversible macroscopic strain (two-way shape memory strain) upon cooling/heating under zero stress [13,16,17]. To obtain such an internal stress field, prior thermo-mechanical treatments, such as plastic deformation in martensite [18,19], superelastic training [20,21], stress-assisted ageing [22,23] and stress-induced martensite ageing [24], are usually required. Recently, metamagnetic shape memory alloys (MSMAs) have attracted great attention owing to their promising prospects for applications as magnetically driven actuators with the advantages of rapid response and large output stress (strain) [25,26]. The MSMAs exhibit magnetic-field-induced transformation in addition to temperature- and stress-induced martensitic transformation. In general, the strain associated with the magnetic-field-induced transformation (considering that the martensite is weak magnetic in typical MSMAs), namely the magnetic-field-induced strain, has a similar magnitude to that associated with the temperature-induced martensitic transformation [23,27]. Usually, the magnetic-field-induced strain in MSMAs is small due to the self-accommodated multi-variant state of martensite. It is expected that, if a MSMA shows TWSME with a large recoverable strain (owing to preferentially selected martensite variants under the internal stress field), a large magnetic-field-induced strain can be obtained provided that the magnetic-field-induced transformation can be triggered. Therefore, it is of great importance to seek MSMAs with large two-way shape memory strain in order to achieve a large magnetic-field-induced strain which is a significant and desirable property for applications in next-generation actuators.

Here in this work, we report such a metamagnetic shape memory microwire with large two-way shape memory strains of 1.0 % and 2.2 % in as-prepared and trained states, respectively. To the best of our knowledge, the two-way shape memory strain of 2.2 % is the highest value reported so far in metamagnetic shape memory alloys. Furthermore, a huge tensile recoverable strain of up to 11.2 % was achieved under a bias stress of 300 MPa in the trained microwire.

Polycrystalline button ingots with composition of Ni₅₀Mn_37.5Sn_12.5 (at.%) were prepared by arc melting using high-purity constituent elements. The glass-coated microwires with diameters of 80-200 μm were fabricated by Taylor-Ulitovsky method [28,29]. The samples were tested after removing the glass sheath but without post heat treatments. The in-situ synchrotron high-energy X-ray diffraction (HEXRD) technique [30,31] was used to determine the crystal structures of austenite and martensite of the microwire. All the (thermo-) mechanical tests were conducted under the tension mode in a dynamic mechanical analyzer (DMA, TA Instruments Q800) [32] equipped with a closed furnace. A microwire with diameter of 107 μm was used. The strain was recorded by a high resolution linear optical encoder with a displacement resolution of 1 nm. The detailed microstructure of the microwire was observed by transmission electron microscopy (TEM, Tecnai F30) [33] operated at 300 kV. The samples for TEM observations were prepared by argon ion-beam thinning using a Gatan Model 691 Precision Ion Polishing System. Further experimental details are provided in the supplementary information.

The one-dimensional (1D) HEXRD patterns collected at 298 K and 110 K for the as-prepared Ni₅₀Mn_37.5Sn_12.5 microwire are displayed in Fig. 1(a) and 1(b), respectively. At 298 K, the main diffraction peaks (Fig. 1(a)) can be indexed according to the cubic L2₁ Heusler austenitic structure with lattice parameter a_A =6.7696 Å. The appearance of superlattice reflections, such as {111}_A, {311}_A and {331}_A, indicates the second neighbor ordering of the L2₁ structure. However, besides the main peaks, there are several minor peaks in the diffraction pattern, as demonstrated in the enlarged view in the inset of Fig. 1(a). These minor peaks can be indexed according to the monoclinic five-layered modulated (5 M) martensitic structure as determined below from the HEXRD pattern shown in Fig. 1(b), clearly indicating that there is a trace of retained martensite in the austenite matrix at 298 K (higher than the reverse transformation finish temperature A_f of ~265 K as determined from Fig. 2(d)). The main diffraction peaks in the HEXRD pattern at 110 K (Fig. 1(b)) can be well indexed according to the 5 M martensitic structure with lattice parameters a_5M =4.8541 Å, b_5M =6.4480 Å, c_5M =24.4013 Å and β = 90.3272°. In addition, several other minor peaks can also observed in the HEXRD pattern at 110 K (Fig. 1(b)), which can be indexed according to the L2₁ structure, indicating the existence of retained austenite at this temperature (much lower than the martensitic transformation finish temperature M_f of ~177 K as determined from Fig. 2(d)). The geometric compatibility between austenite and martensite can be evaluated with the lattice parameters determined above. The middle eigenvalue λ₂ of transformation stretch matrix U, which is a quantitative measure of geometric compatibility, can be determined with the algorithms given in Refs. [34,35]. The λ₂ of this as-prepared Ni₅₀Mn_37.5Sn_12.5 microwire is determined to be 1.0128, which deviates considerably from unity. This suggests a poor geometric compatibility between the two phases. Such poor geometric compatibility renders the transformation between austenite and martensite difficult [36], which may account for the existence of retained austenite at 110 K (Fig. 1(b)) and retained martensite at 298 K (Fig. 1(a)). Furthermore, a stress transition layer and high internal stresses are anticipated owing to the incompatible interface between austenite and twined martensite [37].

For easy visualization, Fig. 2(a-c) shows the change of the macroscopic shape of the as-prepared Ni₅₀Mn_37.5Sn_12.5 microwire with temperature variation. The microwire was straight at 297 K (ambient temperature, above A_f) when the microwire was placed far above a cup of liquid nitrogen, as shown in Fig. 2(a). When the microwire was placed near the liquid nitrogen where it was cooled to 173 K (measured by a thermometer placed adjacent to the microwire, below M_f), the microwire bent to a semicircle shape (Fig. 2(b)). When the temperature recovered to 297 K, the microwire became straight again (Fig. 2(c)). During the subsequent cycles of cooling and heating, the microwire shape changed cyclically between the straight and semicircle shapes. This suggests an intrinsic TWSME in the as-prepared Ni₅₀Mn_37.5Sn_12.5 microwire.

To quantitatively study the TWSME, we measured the strain as a function of temperature during cooling and heating (at 3 K/min) under 1 MPa. By definition, the TWSME refers to the reversible spontaneous shape change by heating and cooling under zero stress [3] and in practice the tiny stress of 1 MPa was applied just to keep the sample straight during the measurement. The result for the as-prepared microwire is shown in Fig. 2(d). As can be seen, the microwire elongates from 227 K to 177 K (B→C, Fig. 2(d)) during cooling, as a result of martensitic transformation, and contracts from 226 K to 265 K (E→F, Fig. 2(d)), due to reverse martensitic transformation. The recoverable strain (ε_rec) amounts to 1.0 %. This, together with the reversible macroscopic shape change shown in Fig. 2(a-c), unambiguously confirms the TWSME in the as-prepared microwire. The reason for this intrinsic TWSME will be discussed later. We further performed superelastic training in the austenitic state at room temperature by loading/unloading the microwire for 30 cycles, with maximum applied stress of 330 MPa. The strain-temperature curve (under 1 MPa) for the trained microwire is also shown in Fig. 2(d). Clearly, the two-way shape memory recoverable strain ε_rec increases significantly to 2.2 % after training, indicating that superelastic training is a simple and effective way to improve TWSME. Since a large two-way shape memory strain of 2.2 % due to thermally induced martensitic transformation under zero stress is achieved, a large magnetic-field-induced strain as a result of magnetically driven martensitic transformation can be anticipated in the present Ni₅₀Mn_37.5Sn_12.5 metamagnetic shape memory microwire, which is beneficial for magnetic actuation in miniature devices.

In addition to the TWSME, we also investigated the load-biased shape memory effect of the as-prepared and trained microwire samples under high stress levels. The strain-temperature curves measured under constant stress levels of 100, 200 and 300 MPa are depicted in Fig. 3. The determination of the recoverable strain (ε_rec) and irrecoverable strain (ε_irr) is also illustrated in this figure. There is an apparent thermal hysteresis between the forward martensitic transformation and the reverse transformation, which leads to the difference in the cooling and heating curves. It is worth mentioning that the thermal hysteresis under 1 MPa (Fig. 2(d)) is higher than that under high stress levels (Fig. 3). This is because under a low stress the interactions between austenite and the self-accommodated and oriented martensitic variants lead to higher energy dissipation while under high stress levels the interactions between austenite and the oriented martensitic variants result in lower energy dissipation. It should also be noted that the reverse transformation start temperature A_s under 200 MPa is lower than that under 100 MPa (Fig. 3). This can be attributed to the higher elastic energy storage during thermal cycling under higher stress levels [38,39]. Since the stored elastic energy promotes the reverse transformation, lower superheating is enough to initiate the reverse transformation under 200 MPa, resulting in a lower A_s under 200 MPa. A similar phenomenon was also observed in previous studies [39,40]. As can be seen from Fig. 3, for both the as-prepared and trained samples, the strain associated with martensitic transformation (i.e. 2.4 % and 3.8 % for the as-prepared and trained samples, respectively) can be almost fully recovered after the cooling/heating cycle under 100 MPa. Irrecoverable strain (ε_irr) resulting from plastic deformation appears (the cooling and heating curves does not overlap at high temperatures) when the stress increases to 200 MPa and it increases with stress increasing. For both the as-prepared and trained samples, the recoverable strain (ε_rec) increases as the applied stress increases. It is found that the recoverable strain (ε_rec) in the trained sample is higher than that in the as-prepared one under the same stress level. Remarkably, a huge recoverable strain (ε_rec) of 11.2 % is achieved under 300 MPa in the trained Ni₅₀Mn_37.5Sn_12.5 metamagnetic shape memory microwire. Indeed, the huge recoverable strain is very important for realizing a large stroke in practical actuator applications [3,6,41]. The tensile stress-strain curves at different temperatures were also measured for the as-prepared microwire (see Fig. S1, Supplementary Information) in order to examine its thermo-mechanical response.

To gain a deep understanding of the underlying mechanism for the intrinsic TWSME in the as-prepared microwire and to investigate the microstructural change after training, TEM experiments were carried out at room temperature (above A_f). The bright-field TEM images for the as-prepared and trained microwire samples are shown in Fig. 4(a) and 4(b), respectively. Retained martensite (dark regions) in the austenite matrix is observed in the as-prepared microwire, as shown in Fig. 4(a). Close examination of the dark regions reveals thin martensite plates of nanometer scale (see the inset of Fig. 4(a)). The existence of retained martensite is in good agreement with the indication from the HEXRD experiment (Fig. 1(a)). After training, the amount of retained martensite increases, as seen from Fig. 4(b). The selected area electron diffraction (SAED) pattern taken from the austenite matrix in Fig. 4(b), along the [111] zone-axis of austenite, is displayed in Fig. 4(c), which can be indexed according to the L2₁ structure of austenite as determined from the HEXRD pattern (Fig. 1(a)). The SAED taken from the area containing both retained martensite and austenite matrix in Fig. 4(b) is shown in Fig. 4(d). In addition to the diffraction patterns of austenite, the diffraction patterns of martensite can be observed. Clearly, there exist four additional spots between the main reflections of martensite (see the inset of Fig. 4(d)), which unambiguously confirms the five-layered modulated (5 M) structure of martensite.

It is expected that the retained martensite in the austenite matrix may produce strong internal stress field at the martensite-austenite interface which gives rise to the TWSME. In order to get insights into the internal stress at the interface between the retained martensite and the austenite matrix, we further performed high resolution transmission electron microscopy (HRTEM) observations. Fig. 4(e) shows the HRTEM image of an area containing a phase interface in the trained sample. The magnified view of the interface area between the retained martensite and the austenite matrix is shown in Fig. 4(f). Local areas with severe lattice distortion, as denoted by the dashed ellipses in Fig. 4(f), can be observed, which clearly indicate the existence of internal stress fields at the interface between the retained martensite and the austenite matrix. The lattice distortion may be related to the poor geometric compatibility between the two phases as discussed before.

As is known, the prerequisite for TWSME in SMAs is the existence of oriented internal stress which favors the selection of specific martensite variant(s) during martensitic transformation under zero external stress [8], instead of formation of the self-accommodation microstructure [42]. In the present work, the retained martensite, which has a poor geometric compatibility with the austenite matrix, results in internal stress fields at the martensite-austenite interface. During cooling without external stress, the martensite variants that are favored by the internal stress fields tend to grow. This is different from the self-accommodated multi-variant martensite formed under zero stress. The transformation between austenite and the preferably selected martensite variants leads to reversible two-way shape memory strain. Since the retained martensite is inherent in the as-prepared microwire, it results in internal stress fields and gives rise to the intrinsic TWSME in the as-prepared microwire.

The fraction of retained martensite increases after training, which may lead to stronger internal stress fields. Thus, the fraction of preferably selected martensite variants becomes higher, giving rise to a larger two-way shape memory strain as compared with that in the as-prepared microwire. Furthermore, under the same external stress level, it is more prone to form single variant martensite in the trained sample owing to the stronger internal stress fields [43]. This is why the recoverable strain (ε_rec) in the trained sample is higher than that in the as-prepared one under the same stress level. The orientation change to form favorable orientations during training may also contribute to the higher shape memory strain in the trained sample.

In summary, intrinsic two-way shape memory effect (TWSME) with a recoverable strain of 1.0 % was achieved in the Ni₅₀Mn_37.5Sn_12.5 metamagnetic shape memory microwire. Unlike the TWSME in many other SMAs, the TWSME in the present as-prepared microwire does not need any post thermal-mechanical processing. The TWSME in the present microwire is mainly ascribed to the internal stress resulting from the retained martensite in the austenite matrix, as revealed by TEM and HEXRD experiments. After superelastic training for 30 loading/unloading cycles, the amount of retained martensite becomes higher and the recoverable strain of TWSME significantly increases to 2.2 %. To our knowledge, this is the highest two-way shape memory strain reported heretofore in metamagnetic shape memory alloys (MSMAs). Furthermore, a huge recoverable strain of 11.2 % was obtained under a bias stress of 300 MPa in the trained microwire. These shape memory properties confer the present microwire great potential for micro-actuation applications in intelligence systems.

Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.jmst.2019.10.042.