Epitaxial, high-quality Sr3Al2O6/BTO/LSMO heterostructures were grown by pulsed laser deposition (PLD) on SrTiO3 (STO) substrates in sequence, as shown in Fig. 1a. The water-soluble Sr3Al2O6 (SAO) layer acts as a sacrificial buffer layer during the lift-off process. Synchrotron-based reciprocal space mapping (RSM) shows asymmetric RSM (-103) reflections with the four-fold splits, which verifies the fully epitaxial growth of LSMO, BTO, SAO, and STO (Fig. 1b), comparable with the results obtained from symmetric RSM (002) reflections and θ-2θ scans (Supplementary Fig.S1 and S2). The a- and c-axis lattice parameters of the LSMO/BTO/SAO heterostructures are calculated (Supplementary Table S1). The SAO buffer layer was etched with equal velocity in all directions (Fig. 1e) in deionized water, and the LSMO/BTO heterostructure was released, maintaining good heteroepitaxy, as shown in Fig. 1c. While the LSMO/BTO heterostructures were transferred onto the polydimethylsiloxane (PDMS) support, they broke and self-assembled into regular nanostripes along the [110] crystallographic direction of BTO (Fig. 1d and f).
The LSMO/BTO heterostructured interface was characterized by the aberration-corrected scanning transmission electron microscopy (STEM) (Supplementary Fig. S3, S4). The dislocation spacing is about 40 nm at the interface, and the width of LSMO/BTO nanostripes is 2 ~ 4 µm (Supplementary Fig. S5, S6). These results indicate that the accumulation of mismatch stress is released by breaking the ionic bond at the mismatch dislocation during the peel-off processing. With desorption from the PDMS support, the LSMO/BTO nanostripes coil themselves up into nano-springs because of the built-in strain from the lattice mismatch (Fig. 1g and h). The inside and outside of the nano-springs are LSMO and BTO layers, respectively. Figure 1i presents a relaxed, straight, and left-handed nano-spring lying on a Si substrate, with ∼6 µm in diameter, ∼2.5 µm in width, and ∼12.5 µm in pitch distance (Supplementary Fig. S7). Naturally, the construction technology of bulking structures is universal, especially in epitaxial oxide heterostructures, such as Mn0.5Zn0.5Fe2O4/BTO and Mn0.5Zn0.5Fe2O4/PMN-PT (Supplementary Fig. S8). Moreover, the spring geometry parameters can be adjusted by the lattice mismatch, thickness, and patterning process25,26.
In the LSMO/BTO nano-springs, the strain distribution, crystal structure, and polarization are highly coupled. In Fig. 2a, a single LSMO/BTO spring is selected and prepared for cross-sectional TEM imaging (Supplementary Fig. S9). The LSMO (34 nm thick) and BTO (36 nm thick) layers are distinguished in the curved LSMO/BTO heterostructure by low-resolution cross-sectional STEM-HAADF (high-angle annular dark-field) imaging (Fig. 2b and c). Figure 2d shows the high-resolution STEM-HAADF image of the LSMO/BTO interface along the [100] direction, confirming the epitaxial interface. To assess local variations of polarization, we map the off-center displacement of Ti atom relative to the corner Ba atoms. Figure 2e and f show the atomic displacement and polarization arrow overlaid on the STEM-HAADF images focused on the interface of BTO/LSMO and the surface of the BTO layer, respectively. In the BTO layer, the tensile strain near the surface zone gives rise to polarization along the in-plane direction, which then rotates toward the out-of-plane direction due to the gradually increased compressive strain induced by lattice mismatch when moving towards the BTO/LSMO interface (Fig. 2e). Near the interface, the compressive strain dominates, leading to most polarizations pointing toward the out-of-plane direction (Fig. 2f). Here, the nearly continuous polarization rotation is originated from atomic displacement and lattice tilting along the z-direction. These results suggest that the local polarization lamination and rotation in the BTO layer of the nano-springs are very different from the bent BTO film in which large bending can promote the local polarizations to merge and transform into in plane direction under normal stress16.
We investigate the mechanical properties of the LSMO/BTO nano-springs in situ using a nano-manipulator introduced inside SEM. The elongation and compression tests of nano-springs are carried out, as shown in Fig. 3. In the compression test, one end of the LSMO/BTO nano-spring is bonded onto a mobile tungsten tip. Under compression, the nano-spring is shortened and twisted while keeping its integrity (Fig. 3a). The length of the nano-spring is decreased from 23.8 µm to 6.6 µm, i.e. by 72.3%. After removing the applied stress, the nano-spring returns to its initial state immediately, demonstrating a total elastic deformation during the compression. With the other end of the spring fixed onto lift-out grid, we also pull the nano-spring extensively to reach its limit (from spring structure to concentric winding structure), as shown in Fig. 3b. During the elongation process, the initial dimensions of the nano-spring are ∼ 90 µm in length, ∼7.1 µm in diameter, while the length of the nano-spring is increased to ∼137.2 µm and the diameter is decreased to ∼3.6 µm (Supplementary Video 2). Such a dramatic deformation results in a 500 % length-diameter aspect ratio change, as shown in Fig. 3e (see Supplementary Fig. S10 and Supplementary Video 3). Additionally, the last image in Fig. 3b shows that once it breaks down, the nano-spring returns to its initial form immediately, indicating a full recoiling capability. It is well-known that oxides are brittle with poor flexibility and elasticity due to their strong ionic or extended covalent bonds. However, these elongation and compression tests reveal that that the ferroelectric oxide nano-springs made of LSMO/BTO exhibit strong super-elasticity and super-scalability. A custom-built in situ TEM holder is used to measure the Hooker coefficient. Figure 3c shows the TEM bright-field images of the initial state and the mechanically loaded state, and a continued compressing manipulation can be found in Supplementary Video 4. In situ mechanical compression tests provide the variation of compressive force vs. displacement shown in Fig. 3d. Upon linear approximation, the Hooker coefficient of the spring is calculated as 0.45 N/m, which is higher than that of SiGe/Si/Cr nano-spring27.
Figure 4a,b and e,f show the strain component distribution εz and the ferroelectric/ferroelastic domain structures of nano-springs from the phase-field simulations under compression and elongation along the z-axis, respectively. The ferroelectric strip domains around the surface of the nano-springs can be obtained since the complex strain field of mixed bending and twisting is induced by the uniaxial compression and tension (Supplementary Fig. S12-13). When the nano-spring is compressed (stretched), the εz on the outer surface is mainly tensile (compressive) strain, and the εz on the inner surface is mainly compressive (tensile) strain, which results in the polarization of the outer surface deflected parallel (perpendicular) to the z-axis, while the polarization of the inner surface deflected perpendicular (parallel) to the z-axis. Figure 4c and g show the domain structure and polarization states on the outer surface of the nano-spring under compress and elongation tests, respectively. And Fig. 4d and h show the corresponding polarization component Pθ and strain distribution εθ along the middle line on the outer surface of the nano-spring, respectively (The θ-axis comes from the natural coordinate system introduced to the middle line of the outer surface, which is detailed in the Supplementary Materials). It is observed that when the super-elastic BTO thin films are twisted into the self-assembled nano-springs, the electromechanical coupling behavior is different from the direct correlation between polarization and strain in the previous ferroelectric thin film state (Supplementary Fig. S14g-i, S15g-i). The same relationship between polarization components Pr, Pφ and strain components εr, εφ can be obtained in Supplementary Fig. S14 and S15. The ferroelectric nano-spring system shows a very different domain structure and polarization orientation under compression and elongation compared with the corresponding thin film28. The simulated results corroborate the experimental observation that these nano-springs present excellent elasticity and recovering capability under both elongation and compression (Supplementary Fig. 11). The corresponding domain evolution shows that the polarization is strongly coupled with the bending or twisting strain29–31, arising from the ferroelectric nature of the nano-springs.