Introduction

Ferroelectric random-access memory (FeRAM) is emerging as a popular next-generation memory technology owing to its rapid read/write speeds, high-density storage, and low power consumption. With the advent of the information age, the demand for memory has dramatically increased and conventional memory technologies have struggled to meet the current requirements of high speed, large capacity, and low power consumption. With its unique working principle and superior performance, FeRAM has immense potential and broad market prospects in the information storage sector1,2,3,4. As electronic technology continues to advance, the need for flexibility, miniaturization, and portability in next-generation electronic devices continues to increase, further exciting interest in flexible and wearable electronics5,6,7,8. Conventional rigid substrates face challenges, such as those related to size reduction and increased power consumption, whereas flexible substrates offer a new solution. Particularly, flexible electronic devices are entering a period of rapid development, with numerous materials being integrated onto flexible substrates. Such substrates exhibit excellent mechanical stability, allowing electronic devices to function in various deformation environments, such as bending and stretching, thereby enabling integration into bendable and wearable electronic products and providing flexibility in design and application. However, flexible substrates still face challenges regarding their high-temperature stability and compatibility with epitaxial films. High-quality functional oxide films typically require specific lattice orientations and high deposition temperatures to achieve crystalline growth9,10,11,12,13. To address these challenges, two approaches have been developed: wet etching of the sacrificial layers and film transfer14,15,16,17. In 2020, Ruonan et al. used deionized water to etch away a Sr3Al2O6 sacrificial layer to obtain high-quality single-crystal BaTiO3 films and Pt/BaTiO3/PEDOT: PSS multilayer structures. Through their method, they achieved the heterogeneous integration of the film with a flexible substrate. The transferred BaTiO3 film retained its single-crystal state and exhibited significant ferroelectric hysteresis under both flat and bent conditions18. The second is direct deposition on flexible substrates, however, direct deposition of oxide films on flexible substrates such as polyimide (PI) is challenging because PI substrates tend to bend at high temperatures, which is not conducive to film crystallization and growth. Mica is a silicate mineral with a layered structure that offers several advantages as a flexible substrate. Mica has a high melting point (approximately 1200–1300 °C), good stability and mechanical strength at high temperatures, and excellent insulating properties19. Its layered structure provides good flexibility, giving it potential for applications in wearable electronics. Although conventional hafnium-based ferroelectric films possess excellent ferroelectric properties, their strong wake-up effect and fatigue effect limit their applicability20,21. Perovskite ferroelectric films, such as BaTiO3 films, are known for their outstanding ferroelectric performance and reliability. BaTiO3 is stable at high temperatures, exhibits excellent fatigue resistance, and has been successfully grown epitaxially on both rigid and flexible substrates22. However, its low spontaneous polarization value (~ 20 µC/cm2) limits its applicability23. Although Pb(Zr, Ti)O3 offers high polarization, a low coercive field, and fatigue resistance, it is not environmentally friendly owing to its lead content, and it poses contamination risks during fabrication24,25,26,27,28. BiFeO3 (BFO) has gained prominence among ferroelectric materials owing to its high spontaneous polarization (~ 100 µC/cm2) and lead-free composition29,30,31. Additionally, BFO films prepared on rigid substrates exhibit poor fatigue resistance. Addressing these challenges requires suitable bottom electrodes and buffer layer35. Flexible BFO ferroelectric memories have demonstrated significant application potential and research value. Their development addresses the key challenges in flexible electronic technology and offers a promising solution for future electronic device applications.

We have developed a flexible mica substrate-based BFO ferroelectric memory device by successfully fabricating a high-quality (111) BFO / (111) La0.65Sr0.35MnO3 (LSMO) / (111) SrTiO3 (STO)heterostructure. We conducted a comprehensive study on the ferroelectric properties and fatigue characteristics of the device under flat conditions and various bending radii. The results indicate that the ferroelectric properties and fatigue resistance of the device remained essentially unchanged when bent. After 108 bipolar switching cycles, the device showed no signs of fatigue and exhibited the best fatigue resistance among the reported BFO materials36,37. Moreover, the electric hysteresis loops of the device exhibited negligible changes under different bending radii, indicating that the fabricated device maintained its performance under bending or other complex conditions. Thus, this device has significant potential for application in flexible and wearable electronic devices.

Materials and methods

Device preparation process

The mica substrate was sequentially ultrasonically cleaned with acetone, absolute ethanol, and deionized water, 10 min for each cleaning process. After cleaning, the mica substrate was dried with a nitrogen gun and then placed into a pulsed laser deposition (PLD) growth chamber for the deposition of STO and LSMO. After the deposition, a portion of the LSMO was masked to serve as the bottom electrode. The substrate was then returned to the chamber to continue the growth of BFO. Lastly, radio-frequency magnetron sputtering was used to deposit a 70 μm × 70 μm Pt top electrode on the surface of the BFO film.

Film deposition parameters

The STO, LSMO, and BFO films were deposited on mica substrates by operating a PLD system (PLD-S40-L, Beijing Perfect Technology Co., Ltd.) with a KrF excimer laser at a wavelength of 248 nm. The buffer layer STO was first deposited on the substrate at 750 °C under an oxygen pressure of 10 Pa. Subsequently, LSMO was epitaxially grown on the STO film at 750 °C. The ferroelectric layer BFO was then epitaxially grown on the LSMO film at 650 °C. The heterojunction was annealed in situ at 600 °C under an oxygen pressure of 60 Pa for 30 min, followed by cooling to room temperature at a rate of 7 ℃/min.

Test equipment

The phase structure of the BFO/LSMO/STO/mica heterojunction was analyzed by applying X-ray diffraction (XRD, Bruker D8Advance, Germany) with Cu Kα radiation (λ = 1.5405 Å). The voltage and current were set at 40 kV and 40 mA, respectively, with a scanning step size of 0.01°. The optical transmittance of the device was measured at different wavelengths by applying UV-Vis spectrophotometry (Perkin Elmer Lambda 365). The surface morphology and roughness of the films were analyzed by using atomic force microscopy (AFM; Cypher S/Oxford Instruments Asylum Research); the ferroelectric domain switching and phase and amplitude measurements of the BFO films were performed under the SSPFM mode. Optical microscopy (Nikon Eclipse LV150N) was used to examine the integrity of the electrode patterns following ultraviolet lithography, and scanning electron microscopy (Zeiss Sigma 300/Nergy Oxford 30 Xplore) yielded high-resolution images of the device surface at an acceleration voltage of approximately 3 kV. Additionally, the elemental distribution and crystal structure within the device were determined by using transmission electron microscopy (TEM, FEI Tecnai G2 F20) and energy-dispersive X-ray spectroscopy (EDS, Bruker XFlash 5030), respectively. The current–voltage (I–V) characteristics of the device were measured by using a Keithley 2400 digital source meter to evaluate its electrical performance. The ferroelectric properties, which are key indicators of BFO films, were precisely quantified by using a HUACE FE-2000 ferroelectric analyzer.

Results and discussion

Fig. 1
Fig. 1
Full size image

(a) Schematic diagram of BFO/LSMO/STO/mica heterostructure. (b) XRD pattern of BFO/LSMO/STO/mica heterostructure. XRD ω-scan of (c) mica substrate, (d) STO thin film, (e) LSMO thin film, (f) BFO thin film.

Figure 1a shows the structural diagram of the BFO/LSMO/STO/mica heterojunction. The crystal orientation of the BFO/LSMO/STO/mica multilayer structure was determined by using XRD. Figure 1b presents a typical out-of-plane θ–2θ linear scan of the multilayer film. In addition to the (00 L) peaks from the mica substrate, only the diffraction peaks of (111) BFO/ (111) LSMO/ (111) STO were observed, with no other extraneous peaks, indicating that the prepared films had single-crystal structures. Moreover, based on the XRD results, we calculated the interplanar spacings of the various layers in the heterostructure; particularly, BFO (d (111) = 2.300 Å) had a mismatch rate of approximately 0.37% relative to the standard card, LSMO (d (111) = 2.250 Å) had a mismatch rate of approximately 0.45%, and STO (d (111) = 2.255 Å) had a mismatch rate of only 0.05%. These calculations indicate that our films achieved good epitaxial growth on the mica substrate, and that the slight differences in the actual interplanar spacing relative to the standard card values were attributable to in-plane tensile strain39,40,41. Figure S1 shows the XRD patterns of the BFO/LSMO/mica heterostructure, where only the diffraction peaks of the mica substrate are observed. This indicates that the BFO and LSMO films failed to crystallize, likely due to the significant lattice mismatch between LSMO and mica. By inserting an STO buffer layer, we successfully fabricated (111)-oriented BFO films. This method establishes a foundation for the good ferroelectric performance of flexible BFO devices. Supplementary Figure S2 complements the XRD plots of the BFO film at 600 ℃ versus 680 ℃ and the ω-scan test plot of the BFO film at 600 ℃. XRD φ scan patterns for STO (200) peak was showed in Supplementary Figure S3.

Supplementary Figure S4 complements the AFM plots of BFO films at 600 ℃, 650 ℃ and 680 ℃, showing that the BFO films prepared at 650 ℃ have the least roughness. The ω-scan was used to characterize the film’s full width at half maximum (FWHM). Figure 1c shows that the FWHM of the (005) mica substrate is only 0.08°, which provides a foundation for the quality of the subsequent epitaxial films. The FWHM results of epitaxial films was showed in Fig. 1d,e, indicating that the prepared films exhibited good in-plane and out-of-plane orientation consistency, demonstrating high crystallinity. Supplementary Figure S5 complements the SEM image of the film surface of the BFO film at 650 ℃.

Fig. 2
Fig. 2
Full size image

(a) Cross-sectional TEM image of BFO/LSMO/STO/mica heterostructure. (bd) EDS mappings of Bi, Mn, and Ti, respectively, in BFO/LSMO/STO/mica heterostructure. (e) HRTEM image and FFT pattern of mica/STO interface. (f) HRTEM image and FFT pattern of STO/LSMO interface. (g) HRTEM image and FFT pattern of LSMO/BFO interface. (hj) Atomic-scale HRTEM image of STO, LSMO, and BFO layers.

To further elucidate the epitaxial growth behavior of the BFO/LSMO/STO/mica heterojunction, we conducted high-resolution cross-sectional TEM (i.e., HRTEM) analysis. Figure 2a shows a cross-sectional image of the heterojunction in which the layers are tightly bonded with distinct interfaces. From bottom to top, the sequence of layers was as follows: mica substrate, STO, LSMO, and BFO. The clear interface between the STO buffer layer and mica substrate provides a robust foundation for the single-crystal epitaxy of BFO and LSMO on mica. To determine the elemental distribution within each film layer, we applied EDS to Bi, Mn, and Ti. Figure 2b–d present the EDS scan results, showing distinct elemental boundaries without significant diffusion, with the elements concentrated in their respective regions. Supplementary Figure. S6 shows the EDS scans for Fe, La, and Sr.

As shown in Fig. 2e–g, the HRTEM analysis of the BFO/LSMO/STO/mica heterostructure revealed clear and distinct interfaces between the different layers. A magnified view of the interface between the mica and STO (Fig. 2e) revealed a clear lattice structure, and the result of applying a fast Fourier transform (FFT) to the electron diffraction pattern confirmed the single-crystal nature of STO, revealing (001), (111), and (110) planes. Figure 2f presents an HRTEM image of the LSMO layer, where a clear lattice structure is visible in the magnified view and the corresponding electron diffraction pattern revealed (200), (220), and (020) planes. Figure 2g shows the BFO layer, highlighting a clear lattice structure, and with the corresponding electron diffraction pattern revealing (011), (111), and (001) planes. Figure 2h–j provide detailed HRTEM images of the STO, LSMO, and BFO layers, respectively, with marked interplanar spacings; particularly, in STO, the (110) and (001) planes had lattice spacings of 0.276 nm and 0.393 nm, respectively; in LSMO, the (020) and (200) planes had lattice spacings of 0.392 nm and 0.275 nm, respectively; in BFO, the (011) and (100) planes had lattice spacings of 0.285 nm and 0.395 nm, respectively. These detailed HRTEM images and the corresponding diffraction patterns are crucial for understanding the high-quality crystallinity of each layer in the BFO/LSMO/STO/mica heterostructure. HRTEM analysis of the BFO/LSMO/STO/mica heterostructure not only demonstrates the high-quality crystallinity and sharp interfaces of each layer, it also facilitated further understanding of the principles of growth and integration of complex oxide materials. This information is essential for advancing the development of next-generation multifunctional devices. Supplementary Figure S7 illustrates a test plot of the optical transmittance of the BFO/LSMO/STO/mica heterostructure.

Fig. 3
Fig. 3
Full size image

(a) I–V characteristics of BFO/LSMO/STO/mica heterojunction. (b) P–V hysteresis loops for flexible BFO thin-film memory. (c) PFM phase image of BFO thin film. (d) Phase–amplitude curve for BFO film.

BFO ferroelectric films, with their high spontaneous polarization at room temperature and good structural stability, have significant potential for application in electronic devices and non-volatile memories. However, a significant challenge at room temperature is the high leakage current of BFO films. This high leakage current can strongly suppress the ferroelectric polarization effect and degrade the related electrical performance. The key to overcoming this problem is the fabrication of high-quality single-crystal BFO films42,43,44,45,46. Figure 3a shows the I–V characteristics of the BFO/LSMO/STO/mica heterojunction. The fabricated devices exhibited low leakage currents, indicating that we successfully fabricated high-quality single-crystal BFO films by designing the appropriate buffer layers. The I–V curve for the device indicated pronounced hysteresis, which has been attributed to changes in resistance induced by the ferroelectric polarization characteristics. Under a strong electric field, the direction of the ferroelectric domains fully reversed, altering the polarization state and significantly increasing the current. The asymmetry in the I–V curve may be attributable to the different barrier heights at the electrode/BFO interface under varying polarization states. This asymmetry is also commonly observed in ferroelectric resistive switching behavior, which is consistent with previous reports47.

Figure 3(b) shows the polarization versus voltage (P–V) loops for the flexible BFO thin-film memory device at different voltages and a frequency of 1 kHz. The curves exhibited distinct rectangular hysteresis loops, indicating the significant ferroelectric properties of BFO. As the electric field changed from negative to positive, the polarization reversed, manifesting as symmetric loops in the P–V curves. Higher applied voltages contributed to circumvent the effects of the coercive field, leading to rapid polarization switching. By increasing the voltage, we observed the P–V curves to tend toward saturation polarization switching also became more pronounced. At 25 V, the flexible Pt/BFO/LSMO/STO/mica device exhibited a remnant polarization (2Pr) of 134 µC/cm2 and saturated polarization (2Pmax) of approximately 138 µC/cm2. Notably, these polarization values were higher than those for many previously reported (111) and (100) orientations BFO films grown on rigid substrates32,33,34,35,36,37. The polarization value of BFO films is closely related to their growth orientation, and the maximum theoretical polarization value of BFO films along the (111) orientation can reach 100 (µC/cm2)38.

Comprehensive characterization and analysis of heterostructures provide significant guidance for the subsequent fabrication of flexible smart wearable devices. Figure 3(c) shows the PFM image of the ferroelectric BFO film. The two different colors represent the different switched states of the ferroelectric domains with a 180° phase difference between them, confirming the switchable polarization and good nanoscale ferroelectricity of the BFO ferroelectric film48,49.

The amplitude–phase curve for the BFO film was showed in Fig. 3d. The amplitude curve exhibited a butterfly-shaped pattern, indicating bidirectional changes in polarization under the influence of the electric field. This result is consistent with those of previous studies, further confirming the ferroelectricity of BFO ferroelectric film50,51.

Fig. 4
Fig. 4
Full size image

(a) P–V hysteresis loop and transient current of BFO thin film. (b) Retention characteristic curves for flexible Pt/BFO/LSMO/STO/mica ferroelectric memories.

The P–V hysteresis loops and transient currents for BFO are shown in Fig. 4a. Near the coercive fields (Vc) of ± 15 V, distinct diametrically opposed transient current peaks can be observed. When the applied voltage reached the coercive field value, significant reorientation of the electric dipoles within the material occurred, leading to transient current spikes. This phenomenon occurred because the dipoles switched directions, causing a temporary increase in the current. The formation of these transient current peaks further confirms the ferroelectric properties of the flexible BFO memory devices42.

The retention characteristics directly affect the reliability of the ferroelectric memory data. Ferroelectric memory stores information in the polarization direction, and good retention characteristics indicate that the stored information can be preserved without loss or degradation over time. To comprehensively evaluate the reliability of the flexible Pt/BFO/LSMO/STO/mica device, we analyzed the retention characteristic curve under a bias voltage of 2 V and pulse width time of 100 ms (Fig. 4b). The results indicate that the flexible BFO ferroelectric memory demonstrates good stability and reliability over time. Across a long range of retention times, the polarization value remained consistent, highlighting the potential of this material for non-volatile memory applications for which data integrity and long-term storage are critical. These characteristics ensure that BFO memory devices satisfy the demands of various high-performance and high-reliability electronic applications.

Fig. 5
Fig. 5
Full size image

(a) P–V hysteresis loops for flexible BFO memory under different bending conditions. (b) Fatigue characteristics of flexible BFO memory device in unbent state at 100 kHz. (c) Fatigue characteristics of flexible BFO memory device with 3.5 mm bending radius at 100 kHz. (d) Fatigue characteristics of flexible BFO memory device with 6.0 mm bending radius at 100 kHz.

We conducted electrical tests on the device in both flat and bent states to evaluate its performance under bending conditions. Supplementary Figure S8 shows a physical diagram of a flexible device wearable. Supplementary Figure S9 shows the bending tests results for the flexible Pt/BFO/LSMO/STO/mica device on molds with bending radii of 6 mm and 3.5 mm, respectively. The P–V hysteresis loops for the flexible BFO device at a test frequency of 1 kHz and voltage of 20 V in the flat and various bending radii states are shown in Fig. 5a. The properties of ferroelectric materials are generally strongly dependent on their lattice structures. Stress can cause lattice distortion and alter the atomic positions and lattice constants, thereby affecting the arrangement and switching process of electric dipoles. Because ferroelectric polarization is induced by the directional alignment of electric dipoles, lattice distortion directly affects the polarization strength and shape of the hysteresis loop52,53. When bent to a radius of 3.5 mm, the hysteresis loop for the device broadened, indicating that the stress impacted the device, slightly weakening the polarization. At a bending radius of 6 mm, the hysteresis loop remained largely consistent with that in the flat state, with only minor changes. The results demonstrate that the fabricated flexible Pt/BFO/LSMO/STO/mica device exhibits good flexibility and mechanical robustness, maintaining its ferroelectric properties under various bending stresses, thus making it highly promising for flexible electronic applications.

In practical application, ferroelectric materials are often subject to repeated electric field applications, leading to the gradual degradation of their ferroelectric properties, known as the fatigue effect. Fatigue testing simulates this long-term usage, facilitating evaluation of the reliability and durability of the material under high-frequency switching operations. Understanding the fatigue behavior of materials ensures that the device will demonstrate stability throughout its lifespan. To assess the reliability of the fabricated flexible ferroelectric non-volatile memory device, we applied fatigue tests under 100 kHz square-wave conditions to devices in various bending states. Figure 5(b) illustrates the results for the unbent state; after 108 bipolar switching cycles, the device’s 2Pr value decreased from 135.33 to 135.09 µC/cm2, a reduction of only 0.28%. Additionally, the 2Pmax value decreased from 138.36 to 135.19 µC/cm2, a reduction of 2.30%. These results indicate good fatigue resistance.

Figure 5c,d show the fatigue test results for the device under different bending radii. These results were consistent with those for the unbent state (Fig. 5b). At a bending radius of 3.5 mm, after 108 cycles, the 2Pr value decreased from 135.99 to 135.07 µC/cm2, and the 2Pmax value decreased from 137.12 to 135.88 µC/cm2. When the bending radius was adjusted to 6 mm, after 108 bipolar switching cycles, the changes in polarization values were minimal, with the 2Pr value decreasing from 135.48 to 135.01 µC/cm2 and the 2Pmax value decreasing from 138.15 to 137.13 µC/cm2. The comparison of the P-V loops of flexible BFO devices after release from the bent state and in the unbent state was presented in Figure S10. Figure 6a shows that the polarization value and coercive field (Vc) of the device remain essentially unchanged for different bending radii. Thus, the device exhibits good fatigue resistance, stability, and reliability under various bending conditions. This is the first report of the fatigue characteristics of BFO films on flexible substrates. Compared with other reported ferroelectric memory devices, as shown in Fig. 6b, our flexible BFO ferroelectric memory device demonstrates superior fatigue resistance and larger polarization values. Even under bending stress, the polarization characteristics remained largely unaffected and retained their memory retention capabilities. The fabricated device has a wide range of application prospects in non-volatile memory and wearable flexible electronics technologies.

Fig. 6
Fig. 6
Full size image

(a) Comparison of the coercive field (Vc) and Pmax values for BFO/LSMO/STO/mica devices at different bending radii. (b) Comparison of ferroelectric memory devices.

Conclusions

We successfully obtained high-quality (111)-oriented BFO films on mica substrates by applying PLD with a rationally designed buffer layer. our fabricated Pt/BFO/LSMO/STO/mica device exhibited good performance, with 2Pr reaching 134 µC/cm2 and 2Pmax reaching approximately 138 µC/cm2. Previous studies have shown that the polarization capability of BFO ferroelectric films on rigid substrates begins to degrade after 106 cycles; however, our BFO films, which we directly fabricated on flexible substrates, maintained their polarization capability beyond 108 cycles, demonstrating exceptional fatigue resistance. During bending tests, the device performance did not significantly degrade; furthermore, the ferroelectric memory operated safely at a bending radius of 3.5 mm. In summary, owing to its high polarization, non-volatility, fatigue resistance, and flexibility, our flexible BFO ferroelectric memory device has significant prospects for application in flexible electronics and smart wearable devices.