Introduction

Flexible electronic/spintronic devices, with bendable and lightweight features, greatly expand the boundaries of spintronic/electronic applications and are gradually changing personal habits in our daily life1,2,3,4,5,6. As one of the core researching goals of the flexible spintronics/electronics, developing high-performance devices is highly coveted4,5,6. In this regard, integrating perpendicular magnetic anisotropy (PMA) to flexible devices shows tremendous potential. The PMA shows advantage in enhancing device density and lowering energy consumption while preserving superior thermal stability7,8,9. Moreover, controlling and tailoring the PMA on purpose plays a significant role in the formation of emergent spin textures such as chiral domain walls and magnetic skyrmions, which are considered as important ingredient for next-generation spintronic devices10,11. Currently, although a great number of the flexible PMA materials have been synthesized directly on flexible bases4,12,13,14, the limitations of the polymer substrates, such as low melting point and large roughness, hinder to obtain films with high-quality and robust PMA using the direct deposition technique15,16. Hence, it is desirable to develop facile method for the synthesis of high-quality flexible materials with a strong and controllable PMA.

SrRuO3 (SRO) is an attractive candidate for realizing flexible PMA-based spintronic materials and devices. It possesses itinerant ferromagnetism and strong spin-orbit coupling (SOC) simultaneously, resulting in an impressive magnetic crystalline anisotropy (MCA) with an anisotropy energy (MAE) above 106 erg cm−3 17,18,19,20. Moreover, the SRO is a 4d transition metal oxide (TMO), which has complex interactions between multiple degrees of freedom, and can trigger multiple types of phase transformations controlled by various external stimuli21,22,23,24. These features offer an opportunity to realize the strong and controllable PMA in the freestanding SRO membranes. Nowadays, for synthesizing the flexible films, one of the promising methods is to utilize sacrificial buffer layers through a combination of lift-off and thin-film transfer techniques25,26. Pioneered by Lu et al. in 201627, the water-soluble Sr3Al2O6 (SAO) sacrificial layer has been widely used to synthesize a variety of freestanding crystalline TMOs membranes with millimeter-scale lateral size, controllable thickness down to ultrathin limit, and extremely large tolerable strain28,29,30,31,32,33,34,35,36. Moreover, different crystalline-oriented oxide membranes can be readily fabricated by changing the substrate crystal orientation, which provides a feasible way to control the properties of oxide membranes24,35,37,38. In addition, by controlling the thickness and composition of the sacrificial layer, it can further modulate the properties of freestanding TMOs membranes39,40.

In this work, using the SAO and Ca1.5Sr1.5Al2O6 (CSAO) as sacrificial layers, we prepare freestanding SRO films with different growth orientations, and realize a strong and highly tunned PMA. Intriguingly, the crystal structures of flexible membranes are tetragonal and orthorhombic when deposited with the SAO and CSAO, respectively. The different structures have a great influence on the PMA and the effect becomes more prominent with the collaboration of the orientation. By altering the growth orientation of the films, first, the easy-axis of the tetragonal SRO is controlled. The [110]- and [111]-oriented films show in-plane MA (IMA) while the [001]-film owns a PMA. In contrary, for the orthorhombic SRO, the PMA is the protagonist in all of the three oriented films. Secondly, the PMA can be enlarged closely to 107 erg·cm−3 (7 × 106 erg·cm−3) in the [110]-oriented orthorhombic SRO. According the first-principle calculations, we find that the d-orbit band and MA of the Ru atom in the SRO are tightly affected by the orientation and crystal structure, leading the orientation- and structure-dependent MA. Finally, our films also exhibit reliable magnetism against the bending test. These results demonstrate that the sacrificial layer is a powerful approach to fabricate high-quality flexible oxide films, and the structure and orientation are of importance to realize the strong and controllable PMA in the prepare membranes.

Results and discussion

Sample synthesis

Figure 1a shows the schematic process for fabricating flexible SRO films with different orientations, i.e., [001], [110], and [111]. SRO/SAO bilayers are deposited on the [001]-, [110]-, and [111]-oriented SrTiO3 (STO) substrates using pulsed laser deposition (PLD) (seen details in the METHODS). During the deposition, the thickness of the film is monitored by the reflection high-energy electron diffraction (RHEED) and controlled with 30 unit cells (u.c.) and 40 u.c. for the SAO and SRO, respectively. Then, the bilayers are immersed in pure water to etch the sacrificial SAO layers and separate the SRO membranes. During the etching process, flexible PDMS is used as a support layer to mechanically stabilize the films27. After releasing, the SRO/PDMS structures are transferred on desired rigid or flexible substrates. Then the freestanding membranes remain on the fresh substrates after peeling off the PDMS via thermally releasing. Finally, we obtain millimeter-scale and complete flexible SRO films as shown in Fig. 1b. The surface and crystallization quality of the freestanding membranes are characterized by RHEED, atomic force microscopy (AFM), high-resolution transmission electron microscopy (HR-TEM) and selected-area electron diffraction (SAED) as shown in Fig. 1c–e and Supplementary Figure 1. These measurements show the SRO membranes are smooth with sub-nanometer roughness and high crystallization quality.

Fig. 1: Preparation of flexible SRO membranes with different crystal orientations.
figure 1

a Schematic illustrating the heterostructure growth, lift-off and transfer process for SRO membranes with different orientations. During the preparation, the water-soluble Sr3Al2O6 is used as sacrificial layers. b Flexible SRO membranes (gray area) transferred on soft PDMS. c An AFM image of the freestanding SRO membrane with a sub-nanometer roughness (RMS). Scale bar is 1 μm. d Low-magnification and e atomically resolved plane-viewed HR-TEM image of [111]-oriented freestanding SRO membrane transferred on a TEM grid with holey carbon film. Scale bars are 5 μm and 5 nm in (d) and (e), respectively. The inset in (e) shows the SAED pattern.

Full size image

Structural analyses

To explore the sacrificial layer influence on the magnetic properties of freestanding SRO membranes, we utilize different Ca3-xSrxAl2O6 as sacrificial layers. In Fig. 2a, we compare lattice parameters of various Ca3-xSrxAl2O6 sacrificial layers with bulk SRO and STO. We choose Ca1.5Sr1.5Al2O6 (x = 1.5, CSAO) with a lattice constant aCSAO = 4 × 3.87 Å, smaller than SRO and STO, while SAO has a larger lattice constant (aSAO = 4 × 3.96 Å) than SRO and STO27,37,40. These different sacrificial layers can impose dissimilar strain in the SRO films. Here, we denote the SRO grown with SAO and CSAO as SROSAO and SROCSAO, respectively. Supplementary Figure 2 and Supplementary Figure 3 show smooth surface and high crystalline quality of the SROCSAO, respectively, indicating the highquality of the freestanding SROCSAO (F-SROCSAO) as observed in the freestanding SROSAO (F-SROSAO). The X-ray diffraction (XRD) scans in Fig. 2b and Supplementary Figure 5 only show the peaks of the SRO, SAO/CSAO, and STO substrate, revealing epitaxial growth and single phase of the as-grown (strained) and released (freestanding) SRO films.

Fig. 2: Crystalline structure of epitaxial freestanding SRO films grown with different sacrificial layers.
figure 2

a Comparative lattice constants (Å) of the pseudocubic SRO, sacrificial layers (i.e., SAO and CSAO) and STO substrate. b XRD 2θ-ω scans of the strained and freestanding SRO films with the SAO (top) and CSAO (bottom). c Interplanar distance d of the strained and freestanding [001]-, [110]- and [111]-oriented SRO films grown on STO, SAO/STO (blue), and CSAO/STO (red). d–f Reciprocal space maps around the (103) and (013) peaks for the (d) [001]-oriented SRO films grown on STO substrate, [001]-oriented freestanding SRO films grown with (e) SAO and (f) CSAO. S-SRO, S-SROSAO, and S-SROCSAO represent the strained SRO without sacrificial layer, with SAO and CSAO buffer layers. F-SROSAO and F-SROCSAO represent the freestanding SRO grown with SAO and CSAO.

Full size image

More importantly, the systematical XRD measurements confirm the stress strain related to the sacrificial layers. According to the 2θ-ω scan curves, we calculate the interplanar distance (d) using the Bragg formula. As shown in Fig. 2c, for the [001]-case, the d of the strained SROSAO (S-SROSAO) is 3.918 Å which is smaller than that of bulk pseudocubic SRO18,41 (d = 3.93 Å, marked with a green line), indicating that the SRO undergoes a tensile strain induced by the large lattice constant of the SAO. In contrast, the strained SROCSAO (S-SROCSAO) grown with CSAO is compressed and its d is elongated out-of-plane to 3.95 Å. These observations are also observed in the [110]- and [111]-oriented SRO films as shown in Fig. 2c and Supplementary Figure 4, demonstrating that the SRO films are tensely and compressively strained when grown with the SAO and CSAO sacrificial layer, respectively. Next, we carry out the XRD scans of the freestanding SRO (F-SRO) films attached on the PDMS. Figure 2b demonstrates that the peaks of SRO membranes shift to left (right) after dissolving the SAO (CSAO) sacrificial layers, further confirming the tensile (compressive) strain in the as-grown SROSAO (SROCSAO) films, respectively. The interplanar distances d of the SROSAO and SROCSAO films increase and decrease to 3.926 Å and 3.929 Å after lift-off, respectively. Correspondingly, the similar changes of the d are also observed in [110]- and [111]-oriented membranes, as summarized in Fig. 2c. Obviously, there is a stress release in the freestanding SRO films. However, it should be noted that the F-SROSAO membranes have smaller d compared to the F-SROCSAO membranes regardless of the crystal orientation. Such a difference may suggest a structural transition in SRO films induced by different sacrificial layers.

To verify the structural transition, we perform X-ray reciprocal space mapping measurements. As shown in Fig. 2d, the mapping around (103)- and (013)-diffraction peaks of the SRO grown on STO without sacrificial layers reveal that (103)-peak of the SRO possess higher a Qz value than the (013)-peak (ΔQz ≠ 0). This indicates that the SRO is orthorhombic with a monoclinic distortion, in consistent with previous literatures18,42,43,44,45. Similar results are observed for the strained and freestanding SRO films grown with the CSAO, in which different Qz values for the (103)- and (013)-diffraction peaks are observed (Supplementary Fig. 6b and Fig. 2f). This implies S-SROCSAO and F-SROCSAO possess orthorhombic structure too. However, by comparing the peaks position for the films grown with the SAO (Supplementary Fig. 6a and Fig. 2e), it is found that the (103)- and (013)-peaks have same Qz values (ΔQz = 0), indicating that the S-SROSAO and F-SROSAO are tetragonal18,42,43,44,45. These results demonstrate that the structural transition can be engineered via altering the sacrificial layers, as shown in Fig. 2a. It is consistent with the literatures, in which the SRO films can form orthorhombic or tetragonal phases under the stress imposed by the substrate18,45, as we observed in the SAO/STO and CSAO/STO heterostructures. Both phases are almost degenerate in energy46,47. On the other hand, these two crystal structures are belonging to different space groups that prohibit direct transformation among each other47. Therefore, these different crystal structures can be stabilized by the strain imposed by the sacrificial layers, and the structural phases could be kept even being peeled off from the substrates, leading the structural transition observed above.

Magnetic anisotropy

The crystal orientations and observed structural transition have a remarkable influence toward the magnetic properties of flexible SRO films. To explore the impacts, magnetic hysteresis (M-μ0H) measurements were performed with the magnetic field in and out of the sample plane. As shown in Fig. 3, the magnetization is more easily reversed and saturated in [110] F-SROCSAO membranes (Fig. 3a–c) among the three orthorhombic samples, while it is easier in [001] tetragonal F-SROSAO membranes (Fig. 3e–f). Moreover, the SRO membranes exhibit large MAEs above 1 × 106 erg·cm−3 (Fig. 3d, h), confirming a robust MCA in SRO due to its strong SOC48. Intriguingly, the impacts of crystal orientations and structure on the MA are more pronounced. As shown in Fig. 3a–c, the orthorhombic F-SROCSAO membrane, obtained with the CSAO sacrificial layer, shows an orientation-dependent PMA. The [110]-oriented F-SROCSAO membrane owns the largest anisotropy energy (MAE, Ku) of 7 × 106 erg·cm−3, which is 4.46 and 3.15 times of the films grown along [001] (1.57 × 106 erg·cm−3) and [111] (2.22 × 106 erg·cm−3) directions, as summarized in Fig. 3d. In contrast, while in the tetragonal F-SROSAO membrane grown with the SAO, not only the MAE is orientation-dependent, but also the magnetic easy-axis is controlled by the orientation (Fig. 3e–h). At first, the | Ku | of the [110]-film is the largest of 6.05 × 106 erg·cm−3 which is 1.9 and 1.34 times of the [001]- and [111]-films. Moreover, by altering the orientation from [001] to [110] and [111], the easy-axis rotates from out-of-plane to in-plane direction. These results demonstrate that the MA of the freestanding SRO film is not only dependent on the crystal orientation, but also the crystal structure which can be controlled further by the sacrificial layers.

Fig. 3: MA of the freestanding SRO films modulated by the crystal orientation and structure.
figure 3

a–c and e–g MA of the (a) and (e) [001]-, (b) and (f) [110]-, (c) and (g) [111]-oriented freestanding SRO films grown with the CSAO and SAO, respectively. d, h MAE dependent on the orientation in the freestanding SRO films grown with the (d) CSAO and (h) SAO, respectively. PMA and IMA in (d) and (h) are the abbreviations of perpendicular and in-plane magnetic anisotropy, respectively.

Full size image

First-principle calculations

To understand the cooperative effects of crystal orientation and structure on the MA of freestanding SRO membranes, we perform first-principle density functional theory (DFT) calculations of the SRO MAE with the orthorhombic and tetragonal structure, as illustrated in Fig. 2a (see details in METHODS). The calculated MAE is defined as the energy difference between the magnetization along the out-of-plane and in-plane. The positive and negative values of MAE represent perpendicular and in-plane MA, respectively. According to Fig. 4a, d, firstly, the total energy calculation indicates that the SRO has a pronounced MCA with a high-energy magnitude around ~1 meV·u.c−1. (~107 erg·cm−3). Secondly, both of the [110] orthorhombic and tetragonal phases have the largest MAE among three different crystal orientations. More importantly, the magnetic easy-axis can been tuned by the crystal orientation and structure synergistically. For the orthorhombic SRO, it shows a MA with a vertical easy-axis, exhibiting PMA for all three orientations. While for the tetragonal structure, the easy-axis is flipped from the out-of-plane direction to the plane when changing the orientation from [001] to [110] and [111]. These results are well consistent with our experimental observations, demonstrating the cooperative effect of crystal orientation and structure on modulating not only the magnitude of magnetic anisotropy energy but also the direction of the magnetic easy-axis in SRO.

Fig. 4: Mechanism of the orientation- and structure-dependent MA.
figure 4

a, d The MAE of the (a) orthorhombic and (d) tetragonal SRO dependent on the orientation, respectively. b, e The d orbit-resolved MAE of the Ru atoms in [110]-oriented (b) orthorhombic and (e) tetragonal SRO. c, f The d orbit-resolved PDOSs of the Ru atoms in (c) orthorhombic and (f) tetragonal SRO. The insets in (c) and (f) shows the d-orbit energy band structures modulated by the structure.

Full size image

To gain further insight into the cooperative modulation effect, we have performed element- and orbital-projected MAE calculations for both orthorhombic and tetragonal SRO based on a second-order perturbation theory (see details in METHODS). The element-resolved calculations show large MAEs (~1 meV·u.c.−1) of the Ru atoms (Ru-MAEs) in both structural phases, as shown in Supplementary Figure 8. This indicates that SRO possesses a large single-ion anisotropy due to the strong spin-orbital interactions, in consistent with results of the SrIrO3/La0.67Sr0.33MnO3 heterostructures49. Fig. 4b, e show the corresponding orbital-projected Ru-MAEs for the [110]-oriented orthorhombic and tetragonal SRO, respectively. In the orthorhombic phase, the (dxz, dyz) matrix element makes an outstanding out-of-plane contribution to the Ru-MAE, leading a PMA (Fig. 4b) with a MAE of 0.937 meV·u.c.−1. However, for the tetragonal SRO, although the (dxz, dyz) keeps positive, most of the elements become negatively inducing an IMA with a MAE of −1.207 meV·u.c.−1 (Fig. 4e). It confirms that the tunable MA of SRO membranes are indeed controlled by Ru-MAEs through crystal orientation and structure. First-principle DFT calculations show the partial density of states (PDOS) around the Fermi level are solely derived from Ru atoms as shown in Fig. 4c, f. In SRO, there is four electrons occupied the t2g orbitals due to the crystal field splitting. The symmetrical reduction gives rise to the rotation and tilting of the RuO6 octahedron in orthorhombic SRO, alters the individual Ru-O-Ru bond length and angle, and then modifies the states of splitting and the electronic occupation, compared to the tetragonal phase. This structural modulation leads to a more uniform occupation of the t2g orbital in the orthorhombic phase rather than in the tetragonal one. As shown in the inset of Fig. 4c, in the orthorhombic structure, the t2g orbitals are almost three-fold degenerate. The degeneracy is partially lifted, however, in the tetragonal SRO. Two electrons are preferred occupied the dxy orbitals first, and the other electrons are located on the degenerate dxz/dyz orbitals as illuminated in the inset of the Fig. 4f. These structure-dependent orbital splitting directly influence the PDOS at the Fermi level, which in turn affect the Ru-MAEs.

Bending tests

For actual application, the flexibility of the freestanding films should satisfy the reliability demands of flexible devices50. To probe the reliability, a bending cycle test is taken on the [110]-oriented orthorhombic SRO sandwiched between two 50-μm PDMS films and adhered on a PI tape, forming a structure of PDMS/SRO/PDMS/PI. For the test, as shown in Fig. 5a, the testing structure is fixed by two molds which compresses the film with a radius of 2 mm, and then flattens without tensile stress as one bending cycle. Next, the magnetization-dependent field and temperature (M-μ0H and M-T) are characterized as a criterion for the endurance. Figure 5b shows the out-of-plane and in-plane M-μ0H curves after various bending cycles. During the testing, there is negligible change in the Ku, saturated and remanent magnetizations (MS and MR) as well as the Curie temperature (TC), indicating that the flexible membranes possess stable magnetic properties even after 10,000 bending cycles, showing great potential for flexible devices.

Fig. 5: The reliability of the flexible SRO films.
figure 5

a Sketch of the bending cycle test. b The M-μ0H loops of films in flat state measured after different bending cycles. OP and IP are abbreviations of out-of-plane and in plane. c The plots of Ku, saturated and remanent magnetizations (MS and MR) versus bending cycles. d The plots of out-of-plane coercive field and Curie temperature (HC and TC) versus bending cycles. During the test, the [110]-oriented flexible orthorhombic SRO film, sandwiched between two 50-μm PDMS films and adhered on a 55-μm PI tape, is used and bent to 2 mm radius.

Full size image

In summary, we have successfully fabricated high-quality flexible SRO membranes with a strong and tunable PMA using the method of sacrificial layers. It is observed that the PMA, including MAE and magnetic easy-axis of the films, can be manipulated cooperatively by changing the growth orientation and sacrificial buffers, i.e., CSAO and SAO. The reason is that the crystal structure of the SRO film can be controlled by the buffers, and with the collaboration of the orientation. As a result, by optimizing the orientation along [110], we achieve a flexible PMA film with a maximum MAE of 7 × 106 erg·cm−3 in orthorhombic SRO when grown with the CSAO. The first-principle calculations indicate that the d-orbit band and MA of the Ru atom in the SRO is tightly affected by the orientation and crystal structure, leading the orientation- and structure-dependent MA. In addition, by performing a bending cycle test, the flexible membranes also show a good magnetism reliability.

Our work exemplifies that the water-soluble sacrificial layer is a powerful approach to prepare high-quality flexible electric/magnetic films with reliable physics properties. Although, only millimeter-size samples have been fabricated currently29,30,31, this technique has shown a potential capability in large area and mass production in future, which is important to practical application. Additionally, the collaboration of the structure and orientation are efficiency to modulate the physics properties of functional materials, especially the PMA. The above strategies can be extended to a wide class of the oxide systems, and used to fabricate flexible oxide-based devices in future, especially to develop cryogenic electronics/spintronic, which is gaining great anticipation to traverse the power/memory wall of classical room temperature computing51,52.

Methods

Film preparation

The SAO/SRO and CSAO/SRO bilayers are deposited on different oriented STO substrates by pulsed laser deposition (PLD) using a KrF (λ = 248 nm) excimer laser. The SAO layer with a thickness of 30 u.c. is deposited on the [001]-STO substrate at a substrate temperature Tsub = 700 °C with an oxygen pressure Po2 = 2 × 10−3 mbar. The laser fluence is 2.0 J·cm−2, and the repetition rate is 4 Hz. Subsequently, 40 u.c. SRO film is grown with the same laser flux at Tsub = 700 °C and Po2 = 0.1 mbar. With same pulse counts and preparation condition of the SAO and SRO, the [110]- and [111]-bilayers are prepared. With same methods, the oriented CSAO/SRO bilayers are fabricated. Here, the CSAO layers are deposited with same condition as the SAO.

Film transfer

The films are adhered on clean polydimethylsiloxane (PDMS), then immersed in pure water at room temperature for 12 h to dissolve the SAO and CSAO buffers completely. Then, the SRO/PDMS structures are lifted out and washed clean with acetone. Finally, we obtain the different oriented flexible SRO films adhered on the PDMS. It should be noted that the SRO films can be transferred on other substrates or supports, such as the Si, STO, or TEM grids after being baked at 90 °C for 10 min.

Characterizations

The information on the surface morphology, epitaxy quality, and crystal structure are examined by atomic force microscopy (AFM, Dimension 3100, Bruker) and X-ray diffraction (XRD) scan (D8 Discover, Bruker). The high-resolution transmission electron microscopy and selected-area electron diffraction (HE-TEM and SAED, Talos F200CX, Thermal Fischer Scientific) are used to further characterize the crystalline quality. To perform the HE-TEM and SAED measurements for the flexible films, the freestanding SRO films are transferred on the TEM grids with holey carbon film. The magnetic properties are measured using superconducting quantum interference device (SQUID-VSM, Quantum Design) at 10 K. To measure the magnetism of the flexible SRO, the films are transferred on the clean STO substrates. During bending cycle testing, the [110]-oriented orthorhombic SRO is sandwiched between two 50-μm PDMS films then supported by a 55-μm PI tape, constructing a structure of PDMS/SRO/PDMS/PI. Then, the testing structure is fixed by two molds which compresses the film with a radius of 2 mm, and then flattens without tensile stress as one bending cycle. Next, the magnetization-dependent field and temperature (M-μ0H and M-T) are characterized as a criterion for the endurance.

First-principle calculations

First-principles calculations were performed within the Vienna ab initio simulation package (VASP) code53,54,55. In all the calculations, the exchange-correlation potential was treated with the generalized gradient approximation (GGA) with the Perdew-Burke-Ernzerhof (PBE) functional56. The strong correlation effects for the Ru 4d is taken into account by means of the GGA + U approach and the on-site effective Coulomb interaction parameter Ueff is taken as 2.0 eV57. The cutoff energy was 500 eV and 8 × 8 × 6 k-mesh was used in all simulations. To simulate the epitaxy, the horizontal lattice constants were fixed to those of the substrate, while the vertical one has been optimized until the force on each atom is smaller than 0.01 eV · Å−1. Furthermore, the magnetic anisotropy energy is calculated by using the force theorem7.