Introduction

Dielectric ceramic capacitors are widely utilized in numerous electronics and electrical devices because of their rapid charging/discharging speed, high power density, and excellent stability, this includes, but is not limited to, high-frequency switching power supplies, defibrillators, power inverters and conversion systems, and power management systems1,2,3,4. Particularly, dielectric ceramic capacitors have immense application potential in electrical equipment operating under high electric fields (E, up to ~ 1000 kV·cm−1) and wide temperature ranges (−55 ~ 150 °C)5,6,7. However, their low energy storage densities (Wrec) fall short of meeting the ever-increasing needs for miniaturization and integration of power electronic devices8,9,10. Therefore, developing lead-free dielectric ceramic materials with high Wrec and efficiency (η), as well as good temperature stability, is an urgent prerequisite for the development of high-performance electronic devices11.

Na1/2Bi1/2TiO3 (NBT) ceramic is a perovskite ferroelectric material with a large room temperature spontaneous polarization close to 40 μC·cm−2, which is attributed to the hybridization between Bi 6s and O 2p orbitals12,13,14. However, the large remnant polarization (Pr) and coercive field (Ec) exhibited by pure NBT ceramic present challenges for direct application in the field of dielectric energy storage. Designing ceramics with a high energy storage density generally involves three key approaches: increasing the maximum polarization (Pmax), reducing the Pr, and increasing the breakdown electric field (EB)15,16,17,18,19.

The Pr can be significantly reduced through the design of a solid solution with a relaxor or paraelectric end member; however, this often leads to a reduction in Pmax. Previous research has confirmed that within the (1-x)Na0.5Bi0.5TiO3-xSr0.7Bi0.2TiO3 (NBT-SBT) binary solid solution ceramics, an increase in SBT relaxor content from x = 0.30 to 0.35 leads to a decrease in hysteresis (Hysteresis = ΔP1/2E / Pmax, at 100 kV·cm−1) from 19.3% to 13.5%, i.e., a decrease of 30%20. In comparison to other binary solid solutions such as NBT-SrTiO35,21, NBT-BaTiO322,23,24, NBT-CaTiO325, etc., NBT-SBT demonstrates effectiveness in reducing Pr while maintaining Pmax. However, it is noteworthy that during the high-temperature sintering process, the volatilization of Na+ and Bi3+ leads to the formation of A-site vacancies and oxygen vacancies, thereby promoting the reduction of Ti4+ ions to Ti3+. This valence reduction can disrupt ferroelectric ordering and thus reduce the Pmax. Therefore, approaches that can inhibit the reduction of Ti4+ as well as further enhance the polarization for NBT-SBT ceramics are highly required.

It has been actively documented that doping with the multivalent element Mn can effectively suppress the reduction of B-site ions in many perovskite materials26,27,28,29. Inspired by these findings, we propose a strategy to enhance the polarization of NBT-SBT ceramics by doping with Mn. This dopant, with its variable valence states, is anticipated to prevent the reduction of Ti4+ ions to Ti3+, thus mitigating the negative impact of Ti3+ on polarization.

Results and discussion

To validate the effect of Mn ions, we first performed first-principles density functional theory (DFT)30,31 and density functional perturbation theory (DFPT)32,33 calculations using the ABINIT software package34,35,36. As shown in Fig. 1, a 2 \(\times\) 2 \(\times\) 2 supercell of the R3 phase of Na1/2Bi1/2TiO3 structure was constructed based on the R3c phase of NaTiO3 by alternately replacing Na’s nearest neighbors with Bi (see Fig. 1a, lattice constants and angles: a = b = c = 7.76388 Å,α = β = γ = 89.5191°). The DFT results are shown in Fig. 1f and Supplementary Table1. The polarization of Na1/2Bi1/2TiO3 is P = (31.88, 31.52, 31.20) µC·cm−2, and the Born effective charge of Ti4+ is 6.26 e. If the valence state of a Ti4+ ion is reduced to Ti3+, the polarization is marginally reduced to (31.50, 31.06, 30.69) µC·cm−2, and the overall spontaneous polarization decreases from 54.62 to 53.84 µC·cm−2.

Fig. 1: First-principles density functional theory and density functional perturbation theory calculations of Na1/2Bi1/2TiO3-based ceramics.
figure 1

Crystal structure diagrams of a Na1/2Bi1/2TiO3 in R3 phase, b Na1/2Bi1/2TiO3 with Ti3+ replacing Ti4+, c Mn4+-doped Na1/2Bi1/2TiO3, d Mn3+-doped Na1/2Bi1/2TiO3, and e Mn2+-doped Na1/2Bi1/2TiO3; f Polarization and g Displacements from density functional perturbation theory calculations.

Full size image

For Mn substitution on Ti, we consider three different valence states of Mn, viz., Mn4+, Mn3+ and Mn2+, as shown in Fig. 1c–g and Supplementary Tab. 1. Mn ions exhibit different valences at different temperatures, and under high-temperature sintering conditions, Mn4+ ions will be reduced to Mn3+ or Mn2+ ions37,38. Substitutions with Mn4+and Mn3+ yield a smaller polarization compared to pure counterpart, with values of (30.15, 29.89, 29.70) µC·cm−2 and (28.93, 28.68, 28.47) µC·cm−2, respectively, as illustrated in Fig. 1f. In contrast, Mn2+ substitution leads to a substantial increase in polarization to (34.38, 33.71, 33.19) µC·cm−2, with spontaneous polarization increases to 58.48 µC·cm−2.

To understand such behavior, we decomposed the polarization into three contributions, from the A-site (Na/Bi), the B-site (Ti and substituted), and the O-site, as listed in Supplementary Table 2. For the Ti3+ case, its Born effective charge 5.96 e is slightly smaller than that of Ti4+ (6.26 e), and it causes decreased off-centering displacements of all ions along the X, Y, and Z directions (see Fig. 1g and Supplementary Table 2), ultimately resulting in a small reduction in polarization. Mn4+ and Mn3+ substitutions also result in decreased off-centering displacements of all ions, together with the smaller Born effective charges of Mn4+ (4.98 e) and Mn3+ (2.95 e) than Ti4+, leading to decreased polarization. In contrast, albeit the smaller Born effective charge of 2.66 e for Mn2+, it yields substantial enhancement of the off-centering displacements of all ions, in particular for the A-site and O-site, which over-compensate for the reduced contribution from the B-site, leading to an overall increase in polarization. The displacements of the A-site ions are believed to be closely related to the ionic radii of the dopants. Specifically, Mn2+ significantly increases off-centering displacements compared to Mn4+ and Mn3+, as Mn2+ has the largest ionic radius. The correlation analysis can be found in Supplementary Fig. 1, along with the corresponding explanation. Therefore, we aim to mitigate the adverse effects of Ti3+ on polarization in NBT-SBT ceramics by doping with variable valence element Mn, which suppresses the reduction of Ti4+ and further enhances polarization by doping Mn2+ into the B-site to increase off-centering displacements.

To experimentally validate this approach, we selected 0.65Na0.5Bi0.5TiO3−0.35Sr0.7Bi0.2TiO3 composition as an example for our present work, since it meets the requirements of low hysteresis and high polarization (13.5%, 30 μC·cm−2, at 100 kV·cm−1)20. Figure 2a presents the X-ray photoelectron spectroscopy (XPS) results of Ti 2p for a 0.5 mol% Mn doped 0.65NBT-0.35SBT ceramic. Comprehensive XPS results for all components are available in Supplementary Fig. 2, where the Ti4+ and Ti3+ ratios are summarized in Fig. 2b. In the Ti 2p XPS spectrum, a peak appears at 457.8 eV alongside a broad feature near 465.0 eV. Fitting analysis reveals that the broad peak arises from the partial overlap between Ti4+ 2p at 463.7 eV and Ti3+ 2p at 466.0 eV39,40. Notably, the component with a Mn content of 0.5 mol% exhibits the highest proportion of Ti4+ content. Further increases in Mn content gradually reduce the Ti4+ content but remain higher than that of the undoped component. This variance underscores the impact of the varying Mn concentrations on inhibiting the reduction reaction from Ti4+ to Ti3+, where Mn impedes the reduction of Ti4+ to Ti3+ by reducing its own valence state28. At lower Mn doping levels, it effectively suppresses the majority of Ti4+ reduction to Ti3+, consequently resulting in a significant increase in polarization value. At higher Mn contents, the slight decrease in Ti4+ concentration may be associated with a very small amount of Mn ions exceeding the doping limit and beginning to induce the formation of a secondary phase. The extremely low content of this secondary phase may escape detection by SEM and XRD techniques. The onset of the second phase may weaken the impact of Mn doping on the reduction of Ti4+. Upon excessive Mn doping, besides Mn ions replacing Ti4+, there is an Mn surplus, leading to the formation of a substantial secondary phase that can be detected by SEM and XRD (The characterization and analysis of the secondary phase will be introduced in detail later). The presence of this secondary phase inevitably contributes to the increased polarization loss.

Fig. 2: X-ray photoelectron spectroscopy (XPS) and Electron paramagnetic resonance (EPR) characterization results of Mn-doped 0.65NBT-0.35SBT-based ceramics.
figure 2

XPS characterization results of Ti in a x = 0.5 mol% ceramic and b 0.65NBT-0.35SBT-xMn ceramics; EPR characterization results of Mn in c 0.65NBT-0.35SBT-xMn ceramics and d x = 0.5 mol% ceramic.

Full size image

The electron paramagnetic resonance (EPR) results are depicted in Fig. 2c, d. The undoped component did not exhibit any discernible EPR signal for Mn (illustrated by the orange curve in Fig. 2c), whereas the three doped components exhibited six Mn EPR signals (attributable to the hyperfine splitting of the Mn isotope with nuclear spin of 5/2), with signal intensity escalating alongside higher Mn content. Utilizing the component with 0.5 mol% Mn content as an example, it is obvious that there is only one type of signal, i.e., the presence of Mn2+. In contrast, the 1 mol% Mn doped component shows two distinct types of signals: Mn2+ and Mn4+. This is evidenced by the observed splitting of the EPR signal peaks in the magnified view of Supplementary Fig. 3. The Mn3+, often referred to as the EPR silencing ion41, manifests notable zero-field splitting, rendering its EPR spectrum undetectable at low frequencies in the conventional X-band (~ 9.5 GHz)42,43,44.

Based on first principles analysis of the Mn valence effect on polarization, it can be concluded that doping with 0.5 mol% Mn2+ is beneficial for improving the polarization of NBT-SBT ceramics. However, excess Mn doping leads to the presence of excessive Mn3+ or Mn4+, which is not conducive to increasing polarization.

Figure 3a illustrates the unipolar hysteresis loops of NBT-SBT-xMn ceramics. The hysteresis loop of undoped NBT-SBT ceramic exhibits the characteristic shape of a relaxor ferroelectric, featuring nonlinear hysteresis and small values of Pr. Upon doping with 0.5% Mn, the Pmax at 120 kV·cm−1 shows a significant increase compared to the undoped sample, rising from 36.8 μC·cm−2 to 43.7 μC·cm−2, marking a nearly 20% increment. While Pr decreases from 1.4 μC·cm−2 to 1.1 μC·cm−2, exhibiting a 21% decrease, thus a significant reduction in energy loss is expected. When the doping amount increases to 1%, the Pmax is 39.4 μC·cm−2, which falls between the values of the undoped and 0.5 mol% Mn-doped samples. At doping levels of 0.5 mol% and 1 mol%, the shape of the hysteresis loops does not change significantly. The Pmax does not increase monotonically with the Mn content. Excessive Mn, whether in the + 3 or + 4 valence state, tends to reduce polarization. Upon reaching a doping level of 1.5%, the hysteresis becomes larger, and the Pr suddenly increases from ~ 1.6 μC·cm−2 to 5.9 μC·cm−2. The current density-electric field curve presented in Supplementary Fig. 4 demonstrates that excessive Mn doping in NBT-SBT-based ceramics leads to a significant broadening of current density peaks due to the formation of a secondary phase, resulting in a pronounced alteration in ferroelectric behavior, that is, the hysteresis loss significantly increases.

Fig. 3: Ferroelectric, dielectric, and structural properties of 0.65NBT-0.35SBT-xMn ceramics.
figure 3

a Room temperature unipolar hysteresis loops of 0.65NBT-0.35SBT-xMn ceramics. b, c Dielectric temperature spectra as a function of frequency for x = 0.5 mol% and 1.5 mol% ceramics. d The dielectric constant and loss as a function of temperature at 1 kHz for samples with various Mn doped levels. e, f X-ray diffraction for x = 0.5 mol% and 1.5 mol% ceramics as a function of temperature. g Room temperature X-ray diffraction. h Rietveld refinement of X-ray diffraction for x = 0.5 mol% ceramic. i Raman spectra.

Full size image

The dielectric temperature spectrum and X-ray diffraction (XRD) analyses also reveal distinct influences of varying Mn contents on the perovskite structure. As depicted in Fig. 3b–d and Supplementary Fig. 5, at Mn contents of 0.5 mol% and 1 mol%, the dielectric constants at identical frequencies are enhanced compared to the undoped counterpart, with the enhancement being more pronounced in the 0.5 mol% Mn-doped sample, as illustrated in Fig. 3d. Upon increasing the Mn content to 1.5 mol%, the T1 peak associated with rhombohedral-tetragonal phase transition broadens in shape and frequency dispersion, while the T2 peak corresponding to the tetragonal-cubic phase transition shifts toward lower temperature, suggesting a contraction of the temperature range associated with the tetragonal phase. This observation is consistent with the changes observed in the XRD (200) diffraction peaks depicted in Fig. 3e, f. Specifically, the rectangular box indicative of the tetragonal phase shifts to lower temperatures, and its height diminishes, underscoring a compression of the temperature range. The dielectric relaxation coefficients (γ) of all components at 1 MHz are between 1.85–1.98, indicating that the samples are characteristic of typical relaxor ferroelectrics. The P-E loops of undoped, 0.5 mol% Mn-doped, and 1 mol% Mn-doped components are very similar in shape and exhibit low hysteresis. In contrast, the markedly increased hysteresis observed in the hysteresis loop of 1.5 mol% Mn-doped components is primarily attributed to the abnormal ferroelectric phase transition induced by the presence of a secondary phase. Furthermore, within the temperature range from room temperature to T1, the dielectric loss decreases with increasing Mn content, whereas beyond T2, the trend reverses, revealing the ionic conduction at elevated temperatures. Although the disparity in losses between T1 and T2 is insignificant, it indicates that Mn doping contributes to the reduced polarization losses.

Figure 3g illustrates the room temperature XRD test results with corresponding Rietveld refinement presented in Fig. 3h, Supplementary Fig. 6, and Supplementary Table 3. The components with Mn at 0, 1 mol%, and 0.5 mol% exhibit standard diffraction peaks indicative of the pure perovskite structure, whereas the component with Mn at 1.5 mol% manifests a distinct second phase diffraction peak. Based on comparison with the standard PDF card and Supplementary Fig. 7, it is conjectured that the secondary phase component comprises an oxide containing Mn and Ti elements. The dielectric constants of the compositions doped with 5 mol% and 10 mol% Mn are enhanced, due to the doping of Mn preventing the reduction of Ti4+ to Ti3+, thereby reducing the concentration of Ti3+. In contrast, the dielectric constant of the 15 mol% Mn doped composition shows a marked increase near the temperature T2, attributed to structural anomalies, including the formation of secondary phases induced by the excess Mn. In addition, as depicted in Fig. 3i and Supplementary Tab. 4, the Raman spectra of components with Mn contents of 0, 0.5 mol%, and 1 mol% exhibit striking resemblance, reaffirming the standard perovskite structure45,46,47,48. However, in the Raman spectra of the component featuring Mn content of 1.5 mol%, the Raman peaks corresponding to B-O and [BO6] experience significant broadening, indicating pronounced alterations in the arrangement of B-site atoms under excessive Mn content. The emergence of a secondary phase elucidates the substantial increase in hysteresis observed.

In order to further optimize the energy storage performance of Mn-doped 0.65NBT-0.35SBT, we fabricated multilayer ceramic capacitors (MLCCs) with three distinct grain orientations, namely < 111 > orientation, < 001 > orientation, and random orientation, utilizing the component exhibiting the highest polarization value with lowest hysteresis loss, containing 0.5 mol% Mn. This investigation aimed to explore the impact of different grain orientations on the breakdown electric field. Achieving micrometer-scale and large aspect ratio plate-like templates is essential for preparing textured perovskite ceramics49. Traditional BaTiO3 templates tend to react with bismuth-based matrix and dissolve at elevated temperatures, posing challenges in completing the process of aligned grain growth. Hence, the utilization of SrTiO3 templates with higher melting temperatures is desired for guiding grain orientation growth7. To acquire the < 111 >-oriented SrTiO3 flake microcrystalline template, we first synthesized a flaky Ba6Ti17O40 precursor with a layered structure parallel to the surface of the oxygen octahedron via the molten salt method (refer to Fig. 4a and Supplementary Fig. 8a). Then, utilizing a two-step topological chemical microcrystalline transformation, the precursor Ba6Ti17O40 underwent initial conversion into a flaky < 111 >-oriented BaTiO3 microcrystalline template (Fig. 4b), followed by transformation into a flaky < 111 >-oriented SrTiO3 microcrystalline template (Fig. 4c). Regarding the < 001 >-oriented SrTiO3 flake microcrystalline template, a Bi4Ti3O12 precursor with an Aurivillius structure was initially synthesized using the molten salt method (refer to Fig. 4d and Supplementary Fig. 8d). This precursor was directly transformed into the < 001 >-oriented SrTiO3 flake microcrystalline template via a topological chemical transformation method (refer to Fig. 4e and Supplementary Fig. 8e). The length-to-thickness ratios of the < 111 > and < 001 >-oriented SrTiO3 flake microcrystalline templates are predominantly greater than 10. As confirmed in Fig. 4m, the prepared templates achieve their respective specific orientations, rendering them suitable for use as texture templates.

Fig. 4: Morphology and orientation of templates and MLCCs.
figure 4

Surface scanning electron microscopy images of a Ba6Ti17O40, b < 111 >-BaTiO3, c < 111 >-SrTiO3, d Bi4Ti3O12 and e <001 >-SrTiO3. f Schematic diagram of MLCC; Scanning electron microscopy images of the cross-section. g Nontextured MLCC, h < 001 >-textured MLCC, and i < 111 >-textured MLCC; Electron backscatter diffraction of the cross-section. j Nontextured MLCC, k < 001 >-textured MLCC, and l < 111 >-textured MLCC. m XRD of templates, corresponding precursors, and ceramic layers.

Full size image

The nontextured, < 001 >-textured, and < 111 >-textured MLCCs were prepared using tape casting technology, with 5% SrTiO3 flake microcrystalline templates to induce the textured MLCCs. Their structural diagram is shown in Fig. 4f, and the cross-sectional morphology and electron backscatter diffraction (EBSD) patterns of the ceramic layer are illustrated in Fig. 4g–l. The cross-section of the nontextured MLCC is polished, while the cross-sections of the < 111 > and < 001 >-oriented MLCCs remain untreated. Each MLCC consists of 5 effective dielectric layers, with a single ceramic layer thickness ranging from 18 to 20 μm and a total effective electrode area of 48 mm2 (4 mm × 2.4 mm × 5 layer). From Supplementary Fig. 9, it is evident that the SrTiO3 templates are discernible, indicating that the majority of templates remain insoluble in the Mn-doped NBT-SBT matrix at high-temperature treatment. Compared with the nontextured sample, the textured samples exhibit a significant increase in the number of specifically oriented grains. XRD characterization results presented in Fig. 4m confirm the presence of pure perovskite phase in the ceramic layers of all three MLCCs. As expected, the intensity of the (111) diffraction peak in the < 111 >-textured sample is notably higher, while the intensity of the (002) diffraction peak in the < 001 >-textured sample exhibits significant enhancement compared to the nontextured counterpart. This accounts for a satisfactory Lotgering factor F111 of 87%, and excellent F001 reaching 99%, signifying strong < 111 > and < 001 > preferred orientations of the grains, respectively.

Figure 5 presents the characterization results for polarization, strain, and energy storage performance of the three types of MLCCs. Remarkably, all hysteresis loops exhibit similar shapes, where the field-induced polarization shows the highest value in the nontextured sample and lowest value in < 111 >-textured one at the same electric field (Fig. 5a). This is associated with the presence of SrTiO3 templates, which possess a smaller but more linear dielectric property as a function of electric field when compared to the 0.5 mol% Mn-doped 0.65NBT-0.35SBT matrix (Supplementary Fig. 10). Under 500 kV·cm−1, the strain is notably high for < 001 >-textured MLCCs, reaching 0.79%, followed by nontextured MLCCs at 0.60%, and the lowest for < 111 >-textured MLCCs, at only 0.38%. This disparity can be attributed to the lower electrostriction coefficient (Q33) of ABO3 perovskite crystals in the < 111 > direction compared to other directions. The variations in the octahedral oxygen structure within perovskite materials as a function of polarization along different crystallographic directions are shown in Fig. 1150 of the Supplementary Information. When Ti4+ ions at the B site move along the [001] direction, the oxygen ion O1 is pushed upwards, releasing energy due to the compression of the Ti-O1 distance. Consequently, the lattice parameter along the [001] direction is extended by a distance d. The strain along the [001] direction is expressed as d/a. When Ti4+ ions move along the [111] direction, the oxygen ions can move in a direction perpendicular to the [111] direction, releasing the Ti-O pairs potential and thereby reducing Q33 along the [111] direction. Compared to the [001] direction, it is easier for oxygen ions to move perpendicularly along the [111] direction to release compressive energy. Thus, the maximum Q33 is observed along the < 001 > direction, while the minimum Q33 is along the < 111 > direction. A substantial reduction in tensile strain mitigates the likelihood of microcrack initiation and crack-driven dielectric breakdown, as well as slowing the growth rate of partial discharge trees, which can be exacerbated by tensile strain and stress. The Weibull distribution of EB is given in Fig. 5c. As expected, the < 111 >-textured sample attains the highest value, reaching 855 kV·cm−1, followed by the nontextured sample at 604 kV·cm−1. The < 001 >-textured sample exhibits the lowest EB, registering only 555 kV·cm−1. The test results of the hysteresis loop show an approximate critical electric field (Fig. 5d). The calculated Wrec are 15.6 J·cm−3, 12.2 J·cm−3, and 9.70 J·cm−3, for < 111 >-textured, < 001 >-textured, and nontextured samples, respectively. Although the polarization of the < 001 >-textured samples is slightly higher than that of the < 111 >-textured samples under the same electric field, the < 111 >-textured samples exhibit a higher Wrec. This is attributed to the delayed saturation polarization of the hysteresis loops in the < 111 >-textured samples. Under an applied electric field, < 001 >-oriented rhombohedral ceramics exhibit a 4R engineered domain configuration, where the polarization reversal process involves 71° and 109° polarization rotations. In contrast, < 111 >-oriented rhombohedral ceramics display a single-domain structure, with an ideal polarization reversal involving a 180° rotation. Therefore, compared to < 100 >-oriented samples, < 111 >-oriented samples are more challenging to polarize, resulting in a delayed saturation polarization. < 111 >-textured samples achieve the highest Wrec, benefiting from both enhanced EB due to reduced tensile strain and a significant increase in Pmax induced by a high electric field. Notably, the η of all three MLCCs exceeds 90%, with a value on the order of 94-95% for < 111 >-textured MLCCs. Remarkably, Mn-doped < 111 >-textured MLCCs achieve both high Wrec and η, surpassing those of the extensively studied NBT-SBT samples7,13,20.

Fig. 5: Ferroelectric properties, electrostriction characteristics, energy storage performance, and charge-discharge behavior of MLCCs.
figure 5

a Room temperature unipolar hysteresis loops of 500 kV·cm−1. b Electrostrain. c Weibull distribution. d Room temperature unipolar hysteresis loops of breakdown electric field. e Room temperature energy storage density. f Room temperature energy storage efficiency. g Room temperature pulse discharge curves. h Room temperature discharge energy density of nontextured, < 001 >-textured and < 111 >-textured MLCCs. i Comparison of energy storage performance.

Full size image

The charging and discharging performance is a critical factor in assessing the practical energy storage capabilities of dielectric capacitors. The results of pulse charging and discharging tests (Fig. 5g, h, and Supplementary Fig. 12) reveal that the critical EB of the studied three types of samples closely mirrors the EB achieved by Weibull distribution. Under the critical pulse test electric field, the discharge energy density (Wdis) is comparable to the Wrec obtained from hysteresis loops. In resistance-capacitance (RC) circuits, the Wdis of < 111 >-textured samples reaches up to 15.7 J·cm−3. Compared with dielectric ceramic materials, which have been extensively studied for energy storage applications in the past many years (listed in Supplementary Table 5)24,51,52,53,54,55,56,57,58,59,60,61,62,63, the samples we prepared not only achieve ultra-high Wrec under moderate electric fields but also maintain exceptionally high η. Therefore, they have a significant competitive advantage in the field of environmentally friendly energy storage materials.

From an application perspective, temperature stability and fatigue performance tests were conducted on < 111 >-textured samples. Figure 6a, b, and Supplementary Fig. 13 illustrate the stability test results across an ultra-wide temperature range of −100 ~ 200 °C and at 600 kV·cm−1. The Wrec of < 111 >-textured samples remains around 10 J·cm−3 across the temperature range of −70 ~ 200 °C, with a variation of less than 15%, together with its satisfactory η on the order of > 80%, demonstrating excellent temperature stability of the energy storage performance. Figure 6c, d presents the fatigue performance test results at 600 kV·cm−1 after 106 electrical cycles. The Wrec of the < 111 >-textured samples consistently exceeds 10 J·cm−3, while the η remains above 95%, highlighting its superior fatigue performance.

Fig. 6: The stability of < 111 >-textured MLCCs.
figure 6

a Hysteresis loops and b Energy storage performance in the range of −100 ~ 200 °C. c Room temperature hysteresis loops. d Room temperature energy storage performance as a function of cycling numbers up to 106 cycles.

Full size image

In summary, we enhanced the Pmax of 0.65NBT-0.35SBT relaxor ferroelectric ceramics by Mn doping, focusing on polarization engineering. In addition, we reduced tensile strain through grain alignment to increase the EB, together with its enhanced polarization, significantly improving energy storage performance. We fabricated < 111 >-oriented MLCCs utilizing this component, achieving an ultra-high energy storage density of 15.7 J·cm−3 and an excellent η exceeding 95% at 850 kV·cm−1. The variation of Wrec across a wide temperature range of −70 ~ 200 °C is less than 15%, demonstrating a superior temperature stability characteristic. This work confirms that the synergistic optimization strategy of polarization engineering and grain alignment engineering can simultaneously achieve high Wrec and η in dielectric materials. The newly developed MLCCs in this study have emerged as promising candidates for high-power energy storage applications.

Methods

Fabrication of bulk ceramics

Bulk ceramics with the chemical composition of (1-x)(0.65Na0.5Bi0.5TiO3-0.35Sr0.7Bi0.2TiO3)-xMnO2 (x = 0, 0.5 mol%, 1 mol%, and 1.5 mol%) were synthesized using the solid-state reaction method. High-purity chemicals, including Na2CO3 (Aladdin, 99.99%), Bi2O3 (Aladdin, 99.99%), SrCO3 (Rhawn, 99.95%), TiO2 (Aladdin, 99.99%), and MnO2 (Rhawn, 99.95%), sourced from Aladdin, were precisely weighed according to stoichiometric ratios. The mixture was then dispersed in absolute ethanol and milled in a planetary ball mill for 24 h at a rotation speed of 250 rad·min−1 using zirconium dioxide balls. The resulting slurry was dried and calcined at 800 °C for 4 h in an air atmosphere. After calcination, the powder was ball milled again for 24 h, dried, and sieved through a 120-mesh screen to obtain unsintered powder with a pure perovskite phase. Small discs with a diameter of 10 mm were formed by pressing a measured amount of unsintered powder at a pressure of 2 MPa using a uniaxial tablet press. These discs were vacuum packaged and subjected to cold isostatic pressing at a pressure of 240 MPa for 10 min to enhance their mechanical strength. Finally, the discs were sintered at 1190 °C for 2 h to produce bulk ceramics in disc form.

Synthesis of < 111 > and < 001 >-oriented SrTiO3 templates

The synthesis of < 111 >-SrTiO3 flake templates involves three sequential steps. Firstly, high-purity BaTiO3 (Aladdin, 99.99%), TiO2 (Aladdin, 99.99%), NaCl (Sinopharm, 99.50%), and KCl (Sinopharm, 99.50%) were homogeneously mixed and subjected to a reaction at 1150 °C for 5 h. The molar ratio of BaTiO3 to TiO2 was 6:9, and the mass ratio of NaCl to KCl was 1:2. The combined mass ratio of BaTiO3 and TiO2 to NaCl and KCl was maintained at 1:2. The resultant Ba6Ti17O40 sheet precursor was purified by successive washing with hot deionized water to remove NaCl and KCl, which acted as a molten salt medium. The metastable Ba6Ti17O40 with monoclinic structure can spontaneously grow in 2D morphology in a molten salt environment and can be used to synthesize microcrystalline templates with pure perovskite phase through a reasonable topological microcrystal conversion method. In the second step, Ba6Ti17O40 was mixed with high-purity BaCO3 (Aladdin, 99.80%) and NaCl molten salt, ensuring a molar ratio of 1:12 between Ba6Ti17O40 and BaCO3, and a mass ratio of 1:2 between the combined mass of Ba6Ti17O40 and BaCO3 to NaCl. The mixture was reacted at 1150 °C for 4 h, followed by the removal of NaCl using hot deionized water to yield the < 111 >-BaTiO3 sheet templates. Lastly, < 111 >-BaTiO3 was mixed with high-purity SrCl2·6H2O (Sinopharm, 99.80%) and NaCl molten salt in a homogeneous distribution. The molar ratio of BaTiO3 to SrCl2·6H2O was 1:10, with their total mass equating to the mass of NaCl. The mixture was reacted at 1100 °C for 5 h, and subsequent removal of NaCl with hot deionized water produced the final < 111 >-SrTiO3 templates. For the synthesis of < 001 >-SrTiO3 templates, two steps were employed. Initially, high-purity Bi2O3 (Aladdin, 99.99%), TiO2 (Aladdin, 99.99%), NaCl (Sinopharm, 99.50%), and KCl (Sinopharm, 99.50%) were uniformly mixed and reacted at 1100 °C for 4 h. The molar ratio of Bi2O3 to TiO2 was 2:3, while the mass ratio of NaCl to KCl was 1:1. The combined mass ratio of Bi2O3 and TiO2 to NaCl and KCl was maintained at 1:1. The resulting Bi4Ti3O12 sheet precursors were purified by washing with hot deionized water to remove NaCl and KCl. In the subsequent step, Bi4Ti3O12 was combined with SrCl2·6H2O (Sinopharm, 99.80%) and KCl molten salt, ensuring a molar ratio of 1:10 between Bi4Ti3O12 and SrCl2·6H2O, and a mass ratio of the total mass of Bi4Ti3O12 and SrCl2 to KCl at 1:1. The mixture was reacted at 1000 °C for 2 h. Bi compounds were then removed using pure nitric acid, followed by the removal of KCl using hot deionized water to obtain the < 001 >-SrTiO3 sheet templates.

Fabrication of nontextured, < 111 >-textured, and < 001 >-textured MLCCs

The preparation of multilayer ceramic capacitors (MLCCs) utilized ceramic powder with the composition 0.995(0.65Na0.5Bi0.5TiO3-0.35Sr0.7Bi0.2TiO3)-0.5 mol% MnO2. The ceramic powder, together with polyester/polyamide copolymer dispersant (KD-1), absolute ethanol, xylene, polyvinyl butyral binder (PVB), butyl phenyl phthalate plasticizer (BBP) plasticizer, and polyalkylene glycol plasticizer (PAG), was mixed at a mass ratio of 40:1:10:10:2:1:1 and ball milled for 24 h. The resulting slurry was degassed in a vacuum for 0.5 h to eliminate air bubbles. For MLCCs requiring texturing, < 111 >-SrTiO3 or < 001 >-SrTiO3 templates, constituting 5% of the ceramic powder mass, were added before vacuum degassing. This was followed by an additional low-speed (200 rad·min−1) ball milling for 0.5 h. The defoamed slurry was cast into green film tape using a casting machine, and the platinum slurry was screen-printed onto it. The resulting layers were shaped according to the platinum electrodes and stacked layer by layer, with each MLCC comprising five effective layers. The facing area of two adjacent platinum electrodes measured 5 mm × 3 mm. To prevent delamination and bulging, warm water isostatic pressing (75 °C, 30 MPa, pressure held for 25 min) was employed. Subsequently, the samples underwent sintering under different conditions: Nontextured MLCCs were sintered at 1150 °C for 2 h, while < 111 >-textured and < 001 >-textured MLCCs were sintered at 1250 °C for 20 h. To enhance density, the sintered MLCCs were further processed in a hot isostatic pressing furnace at 1000 °C for 1 h in a mixed atmosphere of argon and oxygen. The SEM images of the cross-section of the < 001 >-textured MLCCs before and after hot isostatic pressing in a mixed atmosphere are shown in Supplementary Fig. 14. The overlapped area of two adjacent platinum electrodes in the sintered MLCCs measured 4 mm × 2.4 mm.

Structure, morphology, and texture characterizations

The microstructure of the material was analyzed using X-ray diffraction (Smartlab, Rigaku, Japan) and Raman spectroscopy (Renishaw invia Qontor, UK). Rietveld structure refinement was performed with GSAS-II software to accurately determine the crystal structure. The degree of texture was assessed from the XRD data. Scanning Electron Microscopy (TM3030Plus3843, HITACHI, Tokyo, Japan), Energy Dispersive Spectroscopy (UltimMax100, Oxford EDS System, UK), and Electron Backscattered Diffraction (Oxford EBSD System, UK) were employed to characterize the micromorphology, component distribution, and grain orientation of the material. In addition, the valence states of elements were examined using X-ray photoelectron spectroscopy (Thermo Fisher ESCALAB Xi + ) and Electron Paramagnetic Resonance (ER200-SRC-10/12). It is noteworthy that the EPR measurements for Mn were conducted at liquid nitrogen temperature (77 K).

First-principles density functional theory (DFT) and density functional perturbation theory (DFPT) calculations

First-principles DFT and DFPT calculations were performed using the ABINIT package. The projector augmented wave (PAW) method was used for describing electron-ion interactions, and the generalized gradient approximation (GGA) parametrized by Perdew-Burke-Ernzerhof solid (PBEsol) was used for treating electron-electron exchange-correlation functionals64, taken from the Pseudo Dojo65 website with the JTH (v1.1) type66. The orbitals of 2s22p63s1, 6s26p3, 3s23p64s23d2, 2s22p4, and 3s23p64s23d5 were explicitly treated as valence electrons for Na, Bi, Ti, O, and Mn, respectively. To better account for the localized Mn and Ti 3d electrons, the GGA plus Hubbard U (GGA + U) method was employed67. The U values were 4 eV and 1 eV for Mn (Mn4+, Mn3+, and Mn2+)68 and Ti3+,69, respectively. A cutoff energy of 30 Hartree for the plane-wave basis set and Γ-centered 6 × 6 × 6 k-point mesh were used. For structural optimization, the crystal structures are relaxed until all the forces are smaller than 2 × 10−7 Ha/Bohr. The convergence threshold of symmetries of the self-consistent field iteration was set to 10−8. The electrical polarizations were calculated based on the DFPT-computed Born effective charges and atomic displacements with respect to the high-symmetry phase. Various defects were considered, including Mn (Mn4+, Mn3+, and Mn2+) and Ti3+ replacement of Ti4+. For charged defects, a compensating jellium background of opposite charge was used, except for Ti3+, where Constrained Density Functional Theory (cDFT) was used to impose the desired valence state70,71.

Electrical and energy storage performance characterizations

For electrical performance evaluation, circular gold electrodes with a diameter of 2 mm were sputtered onto both sides of the bulk ceramic samples, while silver paste was applied to the ends of the MLCCs to form terminal electrodes. The dependence on the dielectric constant, dielectric loss, and temperature was measured using an LCR meter (E4980AL, Keysight). Hysteresis loops (P-E), field-induced current density curves (J-E), and field-induced strain curves (S-E) were recorded using a ferroelectric analyzer (TF analyzer 2000E, aixACCT, Aachen, Germany). Pulse discharge curves were obtained using a charge and discharge tester (PKCPR1701, PolyK, USA). The total energy density (W), recoverable energy density (Wrec), and energy storage efficiency (η) were calculated utilizing formulas 1-372,73,74, while pulse discharge energy density (Wdis) was determined using formula 418,75,76:

$$W=\int _{0}^{{P}_{\max }}{EdP},$$
(1)
$${W}_{{{{\rm{rec}}}}}=- \int _{{P}_{\max }}^{{{{{\rm{P}}}}}_{{{{\rm{r}}}}}}{EdP},$$
(2)
$$\eta=\frac{{W}_{{{{\rm{rec}}}}}}{W}\times 100 \%,$$
(3)
$$ {W}_{{{{\rm{dis}}}}}=\frac{{{{\rm{\it{R}}}}}\int {i}^{2}(t){dt}}{V}.$$
(4)

where P signifies polarization, Pmax denotes maximum polarization, Pr represents residual polarization, E stands for electric field strength, R indicates the external resistance value (The applied resistance value for this test is 1 kΩ), i represents the current value, t signifies time, and V denotes the effective volume of the sample.