Introduction

Piezoelectric material can realize the conversion between electrical and mechanical energies, making them highly valuable in diverse electronic devices such as transducer, actuators, and sensors, among others1,2. In the past, lead-based PbZrxTi1-xO3 (PZT) ferroelectrics have dominated the piezoceramic market due to their outstanding piezoelectric and electromechanical performance, coupled with good thermal stability3,4,5. However, with growing concerns about the toxicity of Pb, there has been a resurgence in research focusing on lead-free ferroelectric materials6,7. Among the lead-free alternatives, potassium sodium niobate (K0.5Na0.5NbO3, KNN) based ferroelectric materials emerge as particularly promising for practical applications. They boast a high curie temperature (Tc > 300 °C), a reasonable piezoelectric coefficient (d33 ~ 80–120 pC/N), and can be synthesized via a straightforward, cost-effective high temperature solid-state process8,9,10,11,12,13,14,15,16,17,18.

The advancement of lead-free ceramics draws heavily from theories established for Pb-based ceramics, specifically the rhombohedral (R) to tetragonal (T) morphotropic phase boundary derived from PZT solid solution. According to the phenomenology theory, within the phase coexistence region, a flattened free energy profile with respect to polarization promotes both intrinsic lattice vibration and extrinsic domain wall motion, thereby enhancing the piezoelectric response19,20,21. Generally, the construction of room-temperature polymorphic phase transition (PPT) in KNN-based ceramics by chemical modification has been regarded as the most effective strategy to optimize the piezoelectric response22,23. Various dopants have been utilized to increase the R–orthorhombic (O) phase transition temperature (TR-O) or decrease the O–T phase transition temperature (TO-T), thus establishing the R-O, O-T, or R-T polymorphic phase boundaries in KNN system to acquire a high room-temperature piezoelectric coefficient exceeding 200 pC/N22,23. It’s noted that the structural changes caused by multi-element doping are often simplified as alterations in the phase constituent, obscuring the local structural changes associated with the doping elements themselves. For example, even within similar phase boundaries, the presence or absence of the element antimony (Sb) significantly influences the piezoelectric coefficient23,24.

For applications, a formidable challenge lies in the temperature sensitivity of the piezoelectric coefficient. In piezoelectric sensing applications, the key performance indicators for materials are the piezoelectric charge coefficient (represented by the small-signal d33) and the operational temperature range. It is crucial to achieve not just a high d33, but also stability of the piezoelectric properties over a wide temperature range16. This is essential for achieving improved performance with reliable and consistent sensing output across the operational temperature spectrum. However, in contrast to the morphotropic phase boundaries observed in PZT ceramics, the polymorphic phase boundaries within KNN are markedly temperature-dependent7,23,25, resulting in a highly temperature-unstable performance (Supplementary Note 1 and Fig. S1). Several alternative strategies have been proposed to enhance the temperature stability of KNN-based ceramics. These include the implementation of electrically-enhanced diffused PPT26 and the construction of diffused and successive PPT to mitigate the temperature sensitivity of their strain properties (i.e., the piezoelectric strain coefficient d33*)26,27,28,29. Nevertheless, the fluctuation of the small-signal d33 still ranges from ~30% to 70% over temperature range of room-temperature to 100 °C, which severely limits their practical applications30,31,32,33,34,35. In addition to that, the shift of O-T peak to higher temperature inevitably disrupts the room temperature phase boundary, resulting in a decrease in room temperature d33 (Supplementary Note 2 and Fig. S2). Therefore, it appears to be an insurmountable challenge to simultaneously improve d33 and temperature stability through compositional design if solely considering the influence of the average phase boundary.

In this study, we have tackled the aforementioned hurdles by finely adjusting atomic-scale local ferroelectric distortion through a straightforward composition design and solid-state synthesis method. We have theoretically unveiled distinct types of local ferroelectric distortions prompted by dopants and experimentally substantiated their impacts on overall properties. As a result, we have achieved an impressive temperature stability with minimal property variation across a broad temperature range up to 150 °C, while maintaining high piezoelectric performance.

Results

Atomic-scale physical images

We performed Density-functional theory (DFT) calculations using \(2\times 2\times 2\) supercells, incorporating prevalent doping elements in both C phase and O phase of KNN, namely Hf, Zr, Ta, Sb, Ba, Bi, and Li, to gain a preliminary understanding about the influence of dopants on the local ferroelectric distortion (Supplementary Note 3 and Fig. S3a). As illustrated in Fig. 1a, we describe the degree of Local Ferroelectric Distortion (LFD) at doped sites by assessing differences in bond lengths. The bond length difference is defined as the disparity between the longest and shortest bonds at doping site. In the paraelectric phase (C phase), this difference is null. In the ferroelectric phase, however, it is above zero, with the value corresponding to ferroelectric distortion, the larger value, the more pronounced ferroelectric distortion. Additionally, as illustrated in Fig. 1b, the barrier height (defined as the energy difference between the paraelectric and ferroelectric phase in \(2\times 2\times 2\) supercells) reflects the influence of LFD at doping site on the polarization response. A higher energy barrier indicates that polarization is less responsive to external fields (rotation and switching).

Fig. 1: Atomic-scale mechanisms.
Fig. 1: Atomic-scale mechanisms.
Full size image

The definition of (a) bond length difference and (b) energy barrier: The energy difference between the paraelectric phase and the ferroelectric phase (\(\pm {{\rm{P}}}\)) describes the polarization energy barrier, while bond length difference describes the magnitude of local ferroelectric distortion at doping site. c DFT calculations for pristine and Li, Ta, Zr, Hf, Sb, Bi, and Ba doped KNN, using a \(2\times 2\times 2\) supercell. Representative two distinct A site LFDs, where Ba (d1) induces less ferroelectric displacement along [011] direction than K/Na, while Li (d2) induces more displacement. The influence of two distinct A site LFDs on the surrounding Nb-O octahedral, where reduced LFD results in a decreased ferroelectric distortion of the nearest Nb-O octahedral (e1) and enhanced LFD results in an increased ferroelectric distortion of the nearest Nb-O octahedral (e2). Atom-scale physical insights of the impact of local (f1) reduced and (f2) enhanced LFD on the surrounding polarization response under external field.

The summarized findings of the calculations reveal two distinct variations (Fig. 1c):

  1. (I)

    Decrease in the LFD at doping site, manifested by a reduction in bond length difference at doping site and a decrease of the polarization energy barrier, as observed in Ba, Hf, Zr, Ta, and Sb.

  2. (II)

    Increase in the LFD at doping site, manifested by an augmentation in bond length difference at doping site, as observed in Bi and Li. Interestingly, the former results in a reduced energy barrier while the latter leads to an enhanced barrier.

To elucidate the mechanism how elements influence the LFD and energy barrier, we conducted calculations on charge density and density of states. Detailed analysis is provided in Supplementary Notes 3.2–3.4 and Figs. S4–S6. The increase in strain energy caused by ferroelectric distortion is balanced by enhancing the overlap of electronic clouds (thus reducing electrostatic energy). Factors including small ion radii (resulting in the same ferroelectric displacement but smaller strain), large electronegativity, hybridization of d orbitals of B-site ions, and the presence of lone pair electrons on the A-site cations (reducing electrostatic energy) facilitate the enhancement of LFD at doping site. Conversely, large ion radii (resulting in higher strain energy) and changes in electronic configurations (such as Sb at B-site and Ba at A-site, where Sb lacks d orbital hybridization and Ba lacks lone pair electrons, respectively) lead to a weakening of LFD. It is worth mentioning that, for the enhanced A-site LFD, Bi decreases the energy barrier, while Li increases the energy barrier. This can be attributed to their distinct influence on the surrounding Nb-O. Bi weakens the ferroelectric distortion of Nb (Fig. S5c3), while Li induces a stronger Nb ferroelectric displacement (Fig. S5a3).We constructed \(4\times 4\times 4\) supercells (Fig. S3b) for three representative structures, i.e., pristine, Ba doped (the lowest energy barrier), and Li doped (the largest energy barrier) KNN, aiming to analyze the influence of the two distinct LFDs at the doping site on the surrounding matrix. As depicted in Fig. 1d1, d2, akin to the results of \(2\times 2\times 2\) supercells, the DFT calculations for \(4\times 4\times 4\) supercells indicate that Li induces a larger ferroelectric distortion, whereas Ba has the opposite effect. Fig. 1e1, e2 are the \(4\times 4\times 4\) supercells after structural relaxation by DFT calculations, comparing the ferroelectric distortions between neighboring and distant Nb-O octahedra. For Ba-type LFD, the ferroelectric displacement of neighboring Nb is reduced (highlighted in red box and arrow). Conversely, for Li-type LFD, the neighboring Nb induces a larger ferroelectric displacement (highlighted in blue box and arrow) to hybridize with O, thereby mitigating the increased strain energy caused by enhanced LFD at doping site (Supplementary Note 4 and Fig. S7). For those distant Nb-O octahedra, on the other hand, the two types of LFDs exhibit no discernible impact, maintaining the same polarization as the pristine KNN (Fig. S8).

Based on the alterations in local ferroelectric structure and polarization energy barrier outlined above, we propose an atomic-scale physical picture to elucidate the effects of these two distinct LFDs on polarization response under external field. As illustrated in Fig. 1f1, the reduced LFD, accompanied by a reduced local energy barrier, intensifies the response to external field, consequently enhancing the rotation of the surrounding polarization. Conversely, as depicted in Fig. 1f2, the enhanced LFD, accompanied by an elevated energy barrier, results in a more subdued response to external field, thereby suppressing the response of the surrounding polarization.

Relationship between the two distinct LFDs and macroscopic performance

To experimentally validate our proposed physical images, we prepared pristine, individual Ba doped and Li doped KNN (1 mol%) (Abbreviated as KNN, KNNLi1, and KNNBa1) and measured their temperature-dependent properties. By in-situ x-ray diffraction (XRD) and refinements (Fig. 2a and Fig. S9), we found that the three systems exhibited similar phase components at both room temperature and PPT temperature of ~180 °C.

Fig. 2: Phase structure, ferroelectricity, dielectric, and piezoelectric properties of KNN, KNNLi1 and KNNBa1.
Fig. 2: Phase structure, ferroelectricity, dielectric, and piezoelectric properties of KNN, KNNLi1 and KNNBa1.
Full size image

a In-situ XRD, (b) Pr, temperature-related (c) dielectric constants and (d) d33 of KNN, KNNLi1, and KNNBa1, where the three samples show similar average phase constituent but different macroscopic performance. The enhanced LFD (Li doped KNN) leads to increased Pr and decreased dielectric constant, while reduced LFD leads to decreased Pr and increased dielectric constant. Due to the fact that the enhanced dielectric constant plays a major role in d33, the reduced LFD is beneficial for the enhancement of d33 no matter in single phase region or O-T phase boundary.

For perovskite ferroelectrics, the d33 can be expressed as:

$$\,{d}_{33}\propto {\varepsilon }_{33}{P}_{s}Q$$
(1)

Where \({{{\rm{\varepsilon }}}}_{33}\) represents the dielectric permittivity, Ps the spontaneous polarization and Q the electrostrictive coefficient. The parameter Q was reported to be closely related to the degree of order in cation arrangement36. Given the doping amount remains constant, the influence of Q can be considered negligible. As shown in Fig. 2b–d and Fig. S10, despite similar average phase components, KNNBa1 exhibits a reduced remnant polarization Pr, but an enhanced dielectric constant \(({{{\rm{\varepsilon }}}}_{{{\rm{r}}}})\) and higher d33, while KNNLi1 exhibits the opposite trend. The enhanced d33 in KNNBa1 can be attributed to the significantly enhanced \({{{\rm{\varepsilon }}}}_{{{\rm{r}}}}\) (easier polarization rotation due to the reduced LFD, as depicted in Fig. 1f1). Conversely, although Pr is augmented in KNNLi1 (due to enhanced LFD), the suppressed polarization rotation leads to reduced \({{{\rm{\varepsilon }}}}_{{{\rm{r}}}}\) thus lower d33. Furthermore, it’s noteworthy that at the PPT temperature of ~180 °C, the KNNBa1 demonstrates a significant enhancement in dielectric constant and piezoelectric coefficient, despite having a similar phase structure to KNNLi1 and pristine counterparts. Similar results can also be observed in Gao’s work17, where Bi or Zr dopant in KNN (both have reduced LFD as shown in Fig. S5c3 and Fig. S6a3) can lead a significantly enhancement in dielectric constants at O-T peak. Based on the experimental findings, we further propose the relationship between the two LFDs and macroscopic performance as follows,

  1. (I)

    In the single-phase region, a reduced LFD leads to decreased Pr but notably enhanced \({{{\rm{\varepsilon }}}}_{{{\rm{r}}}}\), which is advantageous for enhancing d33. Conversely, an increase in LFD yields the opposite effects.

  2. (II)

    At the phase boundary, the polarization potential of the matrix becomes flattened, and a decrease in LFD will further flatten the potential barrier landscape, resulting in a higher d33. Conversely, an increase in LFD will hinder the polarization response of the surrounding matrix, resulting in suppressed d33.

Strategy for synergistic improvement in d 33 and thermal stability

As previously mentioned, the temperature instability observed in lead-free ceramics arises from the room temperature PPT, which leads to a significant decline in performance as the temperature deviates from room temperature7,23,25. The construction of diffused phase transitions (O-T peak shifting towards higher temperatures) unavoidably results in a decrease in room temperature d33 value due to the disruption of the room temperature phase boundary26,27,28,29. We elaborate on how the above proposed physical insights have been effectively utilized to address this challenge. Specifically, the strategy involves reducing A-site LFD in a system exhibiting room temperature PPT (Fig. 3a1). On one hand, comparison between PbTiO3 and BaTiO3 has revealed that the reduced A-site ferroelectric distortion does not contribute to the stabilization of tetragonal phase37. Therefore, by reducing A-site LFD, we aim to shift the O-T peak above room temperature to establish a diffused PPT. On the other hand, the reduced A-site LFD can introduce a substantial number of regions with reduced energy barriers, facilitating polarization rotation (Fig. 3a2). These regions with reduced energy barriers can offset the decrease in \({{{\rm{\varepsilon }}}}_{{{\rm{r}}}}\) caused by the destruction of room temperature PPT (Fig. 3a3). Consequently, the possibility of a simultaneous improvement in d33 and temperature stability is expected.

To experimentally validate our proposed strategy, we selected (1-y)KNN-y(Bi0.5K0.5)ZrO3 as an example, which has an impressive d33 of ~360 pC/N38,39 and room temperature PPT because Bi doping can effectively decrease the O-T peak temperature while Zr doping can effectively increase the R-O peak temperature. According to the temperature dependent dielectric constants (Fig. S11a), by controlling the parameter y, we established a room temperature R-(O)-T phase boundary, achieving a high d33 ~ 370 pC/N when y was tuned to 0.55 (Fig. S11b). However, as shown in Fig. 3c, akin to other systems with room temperature R-T phase boundary30,31,32,33,34,35, a huge variation of ~50% was observed form room temperature to 100 °C, where a notably lower d33 of ~180 pC/N was recorded at 100 °C.

Fig. 3: The strategy schematic for synergistically improving the d33 and temperature stability and corresponding performances.
Fig. 3: The strategy schematic for synergistically improving the d33 and temperature stability and corresponding performances.
Full size image

Specifically, the strategy is achieved by reducing A-site LFD (a1). On the one hand, introduce a large number of areas with reduced LFD in the matrix to promote polarization rotation (a2). On the other hand, shift the O-T peak above room temperature to construct a diffused PPT phase boundary (a3). The red arrow indicate that the reduced local energy barrier can compensate the loss of \({{{\rm{\varepsilon }}}}_{{{\rm{r}}}}\) caused by destruction of room temperature PPT (i.e., shifting the O-T temperature above room temperature). Temperature-related (b) dielectric constants and (c) d33 of KNN-5.5(Bi0.5, K0.5)1-xBaxZrO3 (x = 0, 0.27, 0.36, 0.45).

Then we introduced Ba into this system to verify our strategy, i.e., 0.945KNN-0.055(Bi0.5K0.5)1-xBaxZrO3 (x = 0, 0.27, 0.36 and 0.45). As given in Fig. 3b, as x increased from 0 to 0.36, the room temperature O-T peak shifted from 45 °C to higher temperatures, reaching 82 °C, with a further enhancement of room temperature dielectric constant. When x = 0.27, we achieved a robust room temperature d33 of ~400 pC/N, with a decreased fluctuation of ~13% across the range of 25–100 °C. Upon further increase to x = 0.36, a higher room temperature d33 of ~430 pC/N was attained, with the fluctuation reduced to ~7%, resulting in a consistently high d33 exceeding 400 pC/N from room temperature to 100 °C. However, when x increases to 0.45, a decrease in dielectric constant and d33 was observed, possibly due to its higher O-T peak (>100 °C). In the following discussion, we focus on ceramics with x = 0 and x = 0.36, given their representative properties.

Mechanism responsible for the improved performance

As showcased in Fig. 4a and Fig. S12, the x = 0.36 sample exhibits a reduced Pr compared to x = 0.0. According to Eq. 1, the increased piezoelectric coefficient is ascribed to the enhanced dielectric constant, rather than the Pr. Consistent with the variation in the dielectric peaks depicted in Fig. 3b, XRD (Fig. 4b, c) and refinements (Figs. S13, S14) indicate the x = 0.0 composition exhibits a predominantly T-phase component (>90%). In contrast, the x = 0.36 sample exhibits a more diffuse phase coexistence of R, O, and T phases within the temperature range from 25 to 100 °C. This multiphase coexistence is associated with the shift of the O-T phase transition peak to higher temperatures, which is induced by the Ba addition (reduced A-site LFD). Previous reports have identified the average T-phase structure as a key factor in improving the performance of KNN ceramics17. However, in our findings, the increased dielectric constant in the x = 0.36 sample suggests the presence of other mechanisms promoting polarization rotation. TEM observations of the samples reveal that both x = 0 and x = 0.36 samples exhibit nano-domains (Fig. S15), thus ruling out the influence of domain walls. We conducted more direct characterizations of reduced local ferroelectric distortion on the A-sublattice with high-resolution high-angle annular dark-field (HAADF) STEM (Fig. 4d–e and Fig. S16). The local polarization length and direction can be determined based on the displacement of B-site atoms relative to the center of the four nearest A-site cations. Similar to the high performance KNN system17 and consistent with the XRD results (Fig. 4b), x = 0 ceramic exhibits an overall T-phase structure with local heterogeneity (Fig. 4d1). Conversely, as shown in Fig. 4d2, regions with significantly reduced polarization length were observed in the x = 0.36 sample (the yellow ellipses). Fig. 4e shows the STEM observation for x = 0.36 sample along [110] zone axis, further demonstrating the R-O-T multiphase coexistence and a large number of regions with significantly reduced polarization length. These findings indicate the successful implementation of our strategy in the KNN-5.5BKZ system. Additionally, the Ba doping results in larger and more uniform grains (Fig. S17), further contributing to the performance improvement. The underlying mechanism responsible for the performance changes above room temperature in x = 0.36 sample is analyzed as follows.

Fig. 4: Mechanisms analysis for enhanced piezoelectric coefficient and thermal stability.
Fig. 4: Mechanisms analysis for enhanced piezoelectric coefficient and thermal stability.
Full size image

aTemperature-related Pr of 0.945KNN-0.055(Bi0.5, K0.5)1-xBaxZrO3 (x = 0 and x = 0.36). Temperature-related phase constituent of (b) x = 0 and (c) x = 0.36. d, e Atom-scale HADDF STEM observations of the A-sublattice for samples with x = 0 and x = 0.36 along [001] zone axis: Polarization magnitude of (d1) x = 0 system and (d2) x = 0.36 system; (e) Atom-scale HADDF STEM observation for x = 0.36 system along [110] zone axis, exhibiting R-O-T multi-phase coexistence with a large amount of reduced LFDs (yellow cycle). f Schematic diagram of the potential energy curves at room temperature, at O-T peak, and beyond the O-T peak.

At room temperature, given the O-T peak of x = 0.36 sample is shifted to higher temperatures, the potential energy curve of the phase boundary region is steeper compared to the room temperature R-(O)-T boundary (x = 0 sample). However, the compensatory effect of the reduced LFD promotes an overall flatter potential energy curve for the system, resulting in a higher dielectric constant and piezoelectric coefficient for x = 0.36 sample. At the O-T peak, as shown in Fig. 4f2, the potential energy curve of the matrix becomes flatter, leading to an overall flatter potential energy curve of the system, thus enhancing d33. Since the O-T peak is only slightly higher than room temperature, the increase in d33 is not significant, maintaining a relatively stable state (i.e., ranging from ~430 pC/N to ~460 pC/N). Beyond the O-T peak but in close proximity to it, as shown in Fig. 4f3, the gradual disappearance of the phase boundary causes the potential energy curve of the matrix to steepen once again, but the reduced local energy barrier offsets this steepness, resulting in the system’s overall potential energy curve remaining relatively flat, thereby maintaining a high d33 > 400 pC/N.

Evaluation of the comprehensive performance

To elucidate the considerable potential of our strategy for lead-free piezoceramics, Fig. 5 compares the performance of the x = 0.36 sample with various lead-free systems40,41,42,43,44,45,46 and PZT-5A47. Fig. 5a presents a comparative analysis of the room-temperature d33 and Tc. We have attained an elevated d33 ~ 430 pC/N and Tc ~ 280 °C (Fig. S18), employing a simple solid-state synthesis approach. Regarding the room-temperature d33 and Tc, this achievement even surpasses some typical textured KNN8,44.

Fig. 5: Excellent comprehensive performance of x = 0.36 system.
Fig. 5: Excellent comprehensive performance of x = 0.36 system.
Full size image

a Comparison of d33 and Tc with lead-free systems40,41,42,43,44 and PZT-5A47; b Thermal stability Comparison with other typical lead-free system16,32,49,50 and PZT-5A48.

Fig. 5b provides a comparative analysis of the temperature stability of d33. Compared to the classic thermal stable PZT-5A ceramic, which has a high d33 of ~375–385 pC/N with a negligible variation (<10%) across room temperature to 150 °C47,48, lead-free piezoceramics exhibit significantly inferior performances. The barium titanate system, due to its low Tc, experiences a drastic reduction in d33 and approaches zero at 100 °C, with △d33 exceeding 100%49. The bismuth ferrite (BF) system demonstrates a high Tc but low d33 of ~160 pC/N, along with a compromised temperature stability (△d33 ~ 44%)50. The actively studied temperature-stable KNN-CZ5 system, though presenting a moderate d33 of ~340 pC/N at room temperature, also experiences huge decline to ~180 pC/N from room temperature to 150 °C, with △d33 of 47%32. Through a dedicated composition stacking process by tape casting16, excellent temperature stability can be achieved in KNN, with d33 of ~330 pC/N and Δd33 of ~6% from room temperature to 100 °C. In contrast, our work consistently maintains a d33 above 400 pC/N from room temperature to 100 °C, showcasing superior temperature stability (△d33 ~ 7%), by simple composition design. Empirically, annealing treatment after poling will stabilize the piezoelectric properties. Here, we poled the x = 0.36 ceramic at room-temperature and subjected it to a 5 h annealing at 150 °C. Leveraging annealing approach after poling, this temperature range can be extended to 150 °C with piezoelectric variation △d33 being on the order of 8%, while maintaining a high d33 ~ 380 pC/N, comparable to the classical lead-based PZT-5A ceramics47,48. In addition to the temperature stability of dielectric and piezoelectric properties, the stability of electromechanical coupling factors (kp, kt and k33) are also critical performance metric, especially for piezoelectric transducer applications. These couplings for x = 0.36 sample before and after annealing were also assessed (Fig. S18). After annealing, the x = 0.36 ceramic demonstrated high kp,kt and k33 values of ~ 50%, 48%, and 65%, respectively, with variations below ~10% from room temperature up to 150 °C, (Fig. S19). Additionally, the thermal cycling stability of the piezoelectric d33is equally crucial for applications. We conducted both in-situ and ex-situ tests on the cycling performance (ten cycles) from 50 to 100 °C and from 50 to 150 °C, respectively. As showcased in Figs. S20 and S21, the samples exhibited excellent temperature stability throughout the in-situ cycling process, while Fig. S22 gives the normalized values to room temperature d33 for the samples after annealing at elevated temperatures (ex-situ measurement), also revealing excellent thermal stability during repeated heating and cooling processes. The comprehensive performance demonstrates that the x = 0.36 ceramics hold great potential for piezoelectric applications, particularly in scenarios where ultrahigh property stability across a broad temperature range is demanded.

Through in-depth analysis of the alterations in local structure induced by dopants, we theoretically demonstrate two distinct atomic-scale ferroelectric distortions concealed within the average phase structure. Additionally, we elucidate how these distortions interplay with the phase boundaries. Building upon this physical understanding, we propose strategies to address practical performance challenges in lead-free piezoceramics. Remarkably, employing a simple solid-state synthesis approach has enabled us to achieve comprehensive performance that has been unprecedented during the past two decades, highlighting significant potential for piezoelectric sensing applications. We anticipate that the atomic-scale physical insights presented in this work will not only inspire advancements in the piezoelectric field but also serve as a catalyst for the design of other polar functional materials.

Methods

Sample preparation

A series of ceramics of (1-y)KNN-y(BK)1-xBxZ, (K0.45Na0.55)NbO3, (K0.45Na0.55)0.99Li0.01NbO3, (K0.45Na0.55)0.99Ba0.01NbO3 were synthesized via a solid-state reaction method using Na2CO3 [99.7%], K2CO3 [99%], Nb2O5 [99.99%], ZrO2 [99.99%], Bi2O3[99.9%] and Ba2CO3[99.7%]. Firstly, Na2CO3 and K2CO3 powders were dried at 300 °C for 2 h. Then, the raw materials were weighed based on the stoichiometric ratios of the above nominal compositions and milled for 24 h in ethanol. The powders were synthesized after calcination at 850–980 °C for 6–8 h. Subsequently, the powders were milled for another 24 h in ethanol. 1.5 wt% polyvinyl alcohol (PVA) as a binder was added into the dried powders and then the green ceramic disks were pressed with the thickness of 1.0 mm and diameter of 10.0 mm. After burning out the PVA, the disk samples were sintered at 1100–1195 °C for 3–4 h in air. As-sintered ceramics were polished, coated with silver paste, and fired at 600 °C for 30 min to form electrodes. To ensure a sufficient polarization reversal, the poled samples were acquired by poling the ceramics at room temperature and 100 °C for doped and pristine KNN, respectively, for 30 min in silicon oil bath under a DC field of 2–4 kV/mm. To further extend the temperature range to 150 °C, the sample (x = 0.36, y = 0.55) were annealed in silicon oil for 5 h at 150 °C.

Characterizations

The room-temperature and in-situ temperature-dependent piezoelectric coefficient (d33) measurements were conducted by a quasi-static piezoelectric d33 meter (ZJ-3A, CAS, Shanghai, China and GDPT-900A, CAS, Shanghai, China). The in-situ temperature-dependent ferroelectric hysteresis loops (P-E loops) for fresh samples were measured using a ferroelectric test system (TF Analyzer 2000, AixACCT systems, Germany). The temperature dependence of relative permittivity and electromechanical coupling coefficients were measured by LCR meters, equipped with a computer-controlled temperature chamber (DMS-500, Balab, Wuhan, China).

The crystalline structure of the ceramics was examined by the in-situ X-ray diffraction method (XRD, SmartLab, Rigaku, Tokyo, Japan). The surface morphology of the specimens was analyzed through scanning electron microscopy (SEM, JP/JSM5200). A spherical aberration-corrected transmission electron microscope (Thermo Fisher Spectra 300, USA) was used to acquire the high-angle annular dark-field scanning-transmission electron microscopic (HAADF-STEM) images and bright-field images. The TEM sample was prepared by the focused ion beam (FIB) milling technology. The HAADF-STEM image analysis was based on the customized MATLAB scripts51. For each atomic column (A or B site), the coordinate, intensity, and polar vector were recorded for further analyses, including drawing the intensity, polar vector, and length maps.

DFT calculations

DFT calculations were performed by using the Vienna ab-initio simulation package and the specific computation procedure can be found in Supplementary Note 3.