The mechanism for the enhanced piezoelectricity in multi-elements doped (K,Na)NbO3 ceramics

(K,Na)NbO3 based ceramics are considered to be one of the most promising lead-free ferroelectrics replacing Pb(Zr,Ti)O3. Despite extensive studies over the last two decades, the mechanism for the enhanced piezoelectricity in multi-elements doped (K,Na)NbO3 ceramics has not been fully understood. Here, we combine temperature-dependent synchrotron x-ray diffraction and property measurements, atomic-scale scanning transmission electron microscopy, and first-principle and phase-field calculations to establish the dopant–structure–property relationship for multi-elements doped (K,Na)NbO3 ceramics. Our results indicate that the dopants induced tetragonal phase and the accompanying high-density nanoscale heterostructures with low-angle polar vectors are responsible for the high dielectric and piezoelectric properties. This work explains the mechanism of the high piezoelectricity recently achieved in (K,Na)NbO3 ceramics and provides guidance for the design of high-performance ferroelectric ceramics, which is expected to benefit numerous functional materials.

doped samples sintered by the same procedure as the multi-elements doped sample.
The KNN precursor powders were prepared at 850 C and it can be questioned if all the dopants were dissolved in the KNN lattice prior to sintering. This would influence on the distribution of the dopants after sintering, particularly with respect to the two sintering procedures used.
The doping of KNN require also knowledge about the defect chemistry of the material, which also may influence on the electrical properties. The authors should consider the following points related to the defect chemistry caused by the doping: • Zr has a lower formal charge than Nb and this difference need to be charge compensated for by an additional point defect(s), which one(s)? Correspondingly and more challenging, the Bi doping was obtained by simply substituting a K+/Na+ with Bi3+. This is not possible without creating secondary phases or other charge compensation point defects. These considerations need to be commented in the manuscript and also discussed in relation to the heterogenous distribution of the dopants and the properties.
• How was the dopants charge compensated for in the DFT calculations and how did the authors distribute K and Na on the A-lattice in the supercells used? Would the distribution of K and Na affect the findings from DFT? • How was loss of alkali-oxide during sintering compensated for or was this not considered (only nominal compositions are given)?
The co-existence of two phases in the multi-element doped sample point to chemical heterogeneity also evidenced by STEM. The authors should also provide X-ray diffraction data above Tc to demonstrate if this heterogeneity remains above Tc showing apparently two cubic KNN phase present. It would also be useful to report the unit cell parameters and volume as pseudo cubic values to make an easier comparison of the two KNN phases.
No ferroelectric properties were reported. It would be very useful for the quality of the paper to also report ferroelectric hysteresis loops.
In the evaluation of the STEM data the effect of the distribution of Bi and Zr were disregarded. Since the atomic number of Zr is so close to Nb this can be argued for, but the heavy element Bi with atomic number 83 cannot be disregarded. It is also confusing that no A-site and B-site intensity map is not reported for KNN-Bi,Sb,Zr. To make this analysis more convincing the same set of analysis are needed for both materials (KNN-Sb and KNN-Bi,Sb,Zr). Doping has also recently been shown to affect the piezoelectric properties of PMN-PT single crystals, see Le et al. Science 364 (2019), 264. The authors should consider to make comparison with this work and similar works in other systems where minor doping have been shown to enhance piezoelectricity.
Reviewer #3 (Remarks to the Author): The current contribution discusses the impact of a multiple of elements on the atomic level structural aspects and the piezoelectric properties of KNN ceramics. It is clearly noticed that the basic underlying idea is an extension from their earlier claims that unit cell level chemical heterogeneity that disturbs a long range polar order and consequently brings about an angular distribution in the direction of polar vector of each unit cell. I see that the collection of experimental data is of high quality, quite complete, and appropriate in supporting their claims, though I think I have to raise a coupling of questions to be clarified.
First, the presence of the claimed nano-heterogeneity is obvious from their extensive TEM studies, but a demonstration on how these nano-heterogeneity is contributing to the piezoelectricity seems missing. One could assume that the low-angle polar vectors contribute to the piezoelectricity when an electric field is applied through polarization rotation, but this is just a possibility. It could be that or could be a shear contribution from misoriented polar vectors or maybe something else. I do not see a decisive experimental verification on the claimed scenario here. In fact, their first principles calculation suggested that the coexistence of those multi-elements tends to reduce the anisotropy that is directly proportional to the magnitude of polar vectors. If this is so, the induced angular variation in the direction of polar vectors has a high chance not to make any useful contribution to the piezoelectricity. Please note that the piezoelectricity, especially the electromechanical strain, is a function of both the easy rotation and the magnitude of polar vectors. As noted, one trades off the other. In this sense, I think it would be better for the authors to tone down their claim a little by dropping the word "origin" in the title. Second, the authors are consistent in emphasizing the effect of multi-elements to get the currently observed enhanced piezoelectricity by comparing the singly doped ones with the triply doped one. I agree that the latter leads to better enhanced piezoelectricity than the former. Then, how about introducing only two different elements? In fact, it has been well known that doubly introduced elements, Li and Sb, are also highly effective in enhancing the piezoelectricity. Is the underlying mechanism for the doubly doped KNN different from what is claimed in this work? How about more than three elements? Do the authors expect any room for further enhancement in piezoelectricity? In relation to this issue, I am also curious why the authors chose a mixture of A-site substituting Bi and B-site substituting Sb and Zr. Does the proposed mechanism only work with the choice of this type of mixture or also work with any combination of elements? manuscript and supplementary materials We would sincerely thank the referees for their time and efforts in careful reading of the manuscript and in preparation of the review reports. We truly appreciate their positive comments, valuable questions and suggestions. We have revised our manuscript accordingly, we believed its quality had been greatly improved. The point-by-point responses to comments were enclosed in the following. We hope we have satisfactorily addressed all referees' concerns and questions. Also the revisions in the manuscript were highlighted.
Response to Reviewer #1 Comment 1: The manuscript by Gao et al. reports on the mechanism responsible for the enhanced piezoelectricity in multi-elements doped KNN-based ceramics. Authors have employed atomic-scale imaging, first-principle, and phase-field calculations, among other tools to identify the low-angle polar nano-domains as the critical factor in improving the dielectric and piezoelectric properties. There are some reports on the similar lines on lead-based perovskites (Adv. Funct. Mater. 2020, 2006823;Nature Materials 17 (2018), 349;Nature 546 (2017), 391, etc.) to explain the large piezoelectric response in relaxor ferroelectrics but such studies on KNN-based piezoelectric are lacking. The results, explanations there of presented in the manuscript are interesting and are worth publishing in Nature Communications. However, the following points must be addressed in the revised manuscript: Reply: We thank the reviewer for his/her positive and valuable comments. We also thank the reviewer for providing references with similar topic on lead-based perovskite, we added the reference in our revised paper and discussed the mechanism explaining the large piezoelectric response in different relaxor ferroelectric systems. In addition, we will respond to the comments one by one in the following.
Comment 2: Reg. ceramic compositions: Authors have synthesized Sb, Bi, and Zr doped and co-doped samples. There is no mention of the formation of extrinsic defects in these samples. For instance, the substitution of 4% of K + with Bi 3+ in KNN would introduce additional oxygen in the structure (i.e., suppresses the oxygen vacancies or oxygen at interstitials type-defects if structure permits). Accordingly, the nominal composition should be K 0.48 Bi 0.02 Na 0.5 NbO 3.02 . Zr 4+ replacing Nb 5+ would create oxygen vacancies (assuming A-site or B-site vacancy formation is not favorable for 1% Zr doped KNN). Similarly, for the multi-elements doped composition K 0.48 Bi 0.02 Na 0.5 Nb 0.92 Sb 0.04 Zr 0.04 O 3 has 0.02 / .. and 0.04 defects.

Reply:
We truly appreciate the valuable comments. We totally agreed with the reviewer that the dopants can induce extrinsic defects, which play important roles in impacting the macroscopic properties. As the reviewer pointed out, A-sites and oxygen vacancies will be introduced by donor dopant on A-site and acceptor dopant on B-site of KNN, respectively. We performed the first-principle calculations based on density functional theory to optimize the structure of A-site vacancies and oxygen vacancies to demonstrate the contributions of the defects on microstructure, as show in the following Fig. R1 (Supplementary Fig. 7 in the revised manuscript). We added the following paragraphs in the main text to discuss the contributions of the defects. "It should be noted that the additions of Bi 3+ and Zr 4+ to A-and B-sites of KNN ceramics will inevitably introduce vacancy defects such as A-site vacancies and oxygen vacancies, respectively. The contributions of the A-site and oxygen vacancies to the structure were also studied based on first-principle calculations and given in Supplementary Fig. 7. Interestingly, the difference in the six B-O bond lengths of the octahedron, after adding different dopants (Fig. 1e) and oxygen vacancies, is obviously reduced compared with the undoped KNN. On the contrary, the A-site vacancies including Na-vacancy and K-vacancy have much weaker effect on the length of the B-O bonds." "It should be noted here that dopants generally induce vacancies, for example, the Bi 3+ occupying A-site K + /Na + position will cause A-site vacancies (including or ), while the Zr 4+ occupying B-site Nb 5+ position will lead to oxygen vacancies ( .. ), which are expected to impact the dielectric and piezoelectric properties. In our studied multi-elements doped KNN, we attempted to keep the charge balanced by tuning the composition with the appropriate doped levels, thus the effect of defects induced by dopants is minimal and is not considered in our mechanism." Comment 3: What is the rationale behind choosing the Sb, Bi, and Zr as the dopants? Reply: Thanks for the good question. The piezoelectricity of KNN was reported to be effectively enhanced by multi-elements dopant and ascribed to the construction of room temperature rhombohedral-tetragonal morphotropic phase boundary [Annu. Rev. Mater. Res. 48, 191-217 (2018)]. Unfortunately we cannot detect the rhombohedral phase in any of the systems, therefore the motivation of our present work is to explore the real underlying mechanism responsible for the high piezoelectricity in multi-elements doped KNN, we chose the Sb, Bi and Zr doped KNN as an example system. We performed synchrotron X-ray diffractions, TEM experiments and first-principle calculations on several multi-elements doped KNN systems, try to understand the relationship between the macroscopic properties and microstructure.

Comment 4:
The relative density(ies) of the pellets used in this study should be explicitly stated. Figure S1 suggests significant porosity in some of these samples.

Reply:
We appreciate the valuable comment. We have added the following statements in the method part to explain the way we measured the relative density: "The densities of the ceramics were measured by the Archimedes method, where the relative density was calculated based on the theoretical density of KNN. 39 " We are sorry about the low quality SEM images given in Supplementary Fig. 1      kV/cm, the d * 33 value goes down above 10kV/cm because higher electric field will clamp the domain wall motion and lead to the decreased piezoelectricity, showing a high non-linearity.
Comment 8: Fig. S6: What was the peak shape function used for the Rietveld refinement of the XRD data? How was the FWHM of XRD peaks plotted in Fig. S4 and S6 calculated? (110)pc is a doublet for Tetragonal crystal structure and a triplet for Orthorhombic crystal structure, whereas (111)pc is a singlet in both Tetragonal and Orthorhombic crystal structures. So, is the value of FWHM (let us say at 300 K for peak (100) ~ 0.115 degrees for the entire broad peak (tetragonal + orthorhombic multiples) or individual peak due to reflection from one set of planes [e.g., Tetragonal (100)].
Reply: Thanks to the reviewer for the valuable comments. We used the pseudo-Voigt convoluted with axial divergence asymmetry function for the Rietveld refinement of the XRD data. The FWHM of XRD peaks plotted in Supplementary Figs. 4, and 6 were derived from the peak analysis using Origin software built-in package. The values of FWHM shown in Supplementary Figs. 4, and 6 stand for the entire broad peak (tetragonal+orthorhombic). This gives us an indication of how the crystal structure changes with temperature and where the transition temperature may locate. To make this clear, we added the following sentence in the method of Structure characterization by synchrotron radiation X-ray diffraction (synchrotron XRD) section: "The values of full width at half maximum stand for the entire broad peak, derived from peak analysis using the Origin software built-in package." Comments 9: Page 14 and Figure S6: What does the phrase "structural distortion" in the context of line 279 mean? A decrease in the degree of tetragonality (c/a ratio) is often labeled as a decrease in structural distortion. Further, in a system with tetragonal and orthorhombic phase coexistence, FWHM of (200)pc peak could decrease if a) one of the phases disappears, or b) phase fractions remain constant, but the degree of tetragonality or orthorhombicity decreases. Please elaborate.
Reply: Thanks for the valuable comments. The structural distortion here represents how far the unit cell deviates from Cubic unit cell. We agree with the reviewer that the c/a ratio closer to 1 (e.g. the decrease in the degree of tetragonality (c/a ratio)) means a decrease in structural distortion. We also agree that the decrease in FWHM of (200)pc peak could have other possible sources in a system with the coexistence of tetragonal and orthorhombic phases. Here, based on our refinement of XRD for KNN-Bi,Sb,Zr as a function of temperature, we can conclude that the decrease of the FWHM for (200)pc peak in Supplementary Fig. 6 is due to the disappearance of orthorhombic phase (please see Fig. 1) with increasing temperature. By comparing different compositions, the decreased FWHM of (200)pc means the degree of tetragonality or orthorhombicity decreases. We deleted "structural distortion" from the results discussion for a clear expression, we also changed the caption of Supplementary Fig. 6, as shown in the following: "It is interesting to note that the FWHM values of KNN-Bi,Sb,Zr sample are lower than those of KNN-Sb counterpart ( Supplementary Fig. 4), indicating the tetragonality or orthorhombicity of KNN-Bi,Sb,Zr sample is decreased comparing to the KNN-Sb sample. " Comments 10: Is there any noticeable dispersion in Tm (temperature of dielectric anomaly corresponding to orthorhombic to tetragonal transition) with frequency for KNN-Bi,Sb,Zr? Reply: Thanks to the good question. Yes, we observed noticeable dispersion around the temperature of dielectric anomaly corresponding to orthorhombic to tetragonal transition. To make this clear, we added the Fig. R7 (Supplementary Fig. 11 in the revised manuscript) in the supporting information. Reply: We appreciate the valuable question. In our opinion, the magnitude of the polar vector angle mainly corresponds to the dopant element distribution and concentration, where the nature of the dopants plays important roles. This can be confirmed by the first principle calculation, where the nature of the dopants, including the size and charge valence, are already considered during the calculations, and we can see that the dopants will cause the B-O bonds of the octahedrons approaching the same length, as shown in Figs. 1e and R1. Based on the current techniques, however, it is hard to distinguish the impact of size/charge on the polar vector which might require three dimensional structure evolution on atomic scale. In our STEM observation instead, we used the 2D mapping of the polar vector angle evolution considering the dopant distribution and concentration.
Comments 12: Can we expect similar results if KNN-LS is co-doped with Sc, Ta, and Sn?
Reply: We appreciate the valuable question. Based on the references and our recent works, we thought the KNN-LS with the addition of Sc/Bi (we need adding the Bi 3+ on A-site to keep charge neutrality if we dope Sc 3+ on B-site) had minor contribution to the piezoelectricity while the addition of Ta 5+ would greatly benefit the piezoelectricity enhancement. Meanwhile, very minor Sn 4+ addition will increase the piezoelectricity because of enhanced local structure heterogeneity, however large amount of Sn 4+ dopant will decrease the piezoelectricity since Sn 4+ with full d 10 electronic configuration will not favor the ferroelectricity based on the pseudo Jahn-Teller effect. The detailed discussions are shown in below: (1) Effect of Ta 5+ . There were many publications reporting the similar multi-elements doped KNN ceramics with desired piezoelectricity.  Fig. R8 (a).
(3) Effect of Sn 4+ . The Li, Sb, Sc, Bi, Sn doped KNN was also prepared, with d 33 value of 340 pC/N. The temperature dependent dielectric property is shown in Fig.  R8 (b). (4) Of interest is that the d 33 value of Li, Sb, Sc, Bi, Ta doped KNN is above 400 pC/N, with temperature dependent dielectric property given in Fig. R8 (c). Response to Reviewer #2 Comment 1: Gao and co-workers have submitted a paper entitled "Origin of high piezoelectricity in multi-element doped KNN ceramics". In the recent couple of decades intense research efforts have been focused on the development of lead-free piezoelectric materials as alternatives to state-of-the art PZT materials. One of the most promising lead-free materials are KNN boosted by the seminal paper by Saito et al. published in Nature in 2004. In this work unusually high piezoelectricity in multi-element doped KNN materials is reported for the first time, which only have been shown previously in highly textured KNN ceramics. In this work KNN ceramics with different levels of Bi, Zr, Sb or Zr/Sb/bi doping were synthesized by conventional ceramic processing. The KNN ceramics were characterized by combination of temperature-dependent synchrotron x-ray diffraction, atomic-scale transmission electron microscopy, dielectric spectroscopy and measurement of piezoelectric coefficient, combined with first principles and phase field calculations. The unusually high piezoelectric coefficient was rationalised by a mechanism involving heterogeneity with respect to crystal symmetry and chemical composition. Nano-scale heterogeneity with different polar vectors were responsible for the high dielectric and piezoelectric responses. The unusual high piezoelectric properties of these KNN ceramics is particularly interesting and the paper has therefore the potential to be published in Nature communication. Before publication the authors should address the following remarks to the manuscript.

Reply:
We thank the reviewer for the positive comments and valuable suggestions.

Comment 2:
The last paragraph before the Results chapter, contains results on both the microstructure, dielectric (including a Rayleigh analysis) and piezoelectric properties. These important data should be moved to the results chapter or is simply the heading (Results) displaced in the manuscript?
Reply: Thanks for the valuable suggestion. We have moved this paragraph to the result part according to the suggestion and highlight the change in the main text.
Comment 3: Before the authors start reporting on the properties of the KNN ceramics, the microstructure (supplementary Figure 1) and the density (Supplementary Table 1) of the materials needs to be reported in the first paragraph in the Results chapter. How was the density measured? The images for the single-doped samples gives an impression of a significantly higher porosity than inferred from the relative density. The authors should also provide SEM images of polished samples in order to give evidence of the low porosity/high density.

Reply:
We appreciate the valuable comments. We added the following sentences in the first paragraph in the results chapter, reporting the microstructure and density, we also added the density measurement details in the section of the method. "The multi-elements doped KNN is found to possess much larger grain size, being on the order of 25μm, one order larger than the single element doped counterparts, though all of the samples have a similar density of 94-96%." "The densities of the ceramics were measured by the Archimedes method, where the relative density was calculated based on the theoretical density of KNN. 39 " In order to clearly show the microstructure and grain size, the SEM images of the polished and thermally etched samples were given in the revised manuscript, as shown in Fig. R9 (Supplementary Fig. 1 in the revised manuscript). It should be noted here that the fissures between different grains in Figs. R9 (b) and (d), were associated with the high etching temperature, other than the voids or pores in the samples. Here we also gave the SEM images before the thermal etching treatments, as in Fig. R10, showing a very low porosity/high density. Many small pores were observed in KNN-Bi,Sb,Zr (Fig. R10a), this was associated with the abnormal grain growth [J. Mater. Chem. C, 8, 7606-7649 (2020)].

Comment 4:
The multi-elements and single element doped samples were not prepared by the same sintering procedure. The mechanism presented is based on the relative comparison of the single and multi-elements doped sample, but the different processing procedure makes this questionable. How important is the sintering procedure for the properties for the multi-elements doped sample? The SEM image also give evidence for a very different microstructure of the multi-elements doped sample. (supplementary Figure 1). The work would be strengthened by also report on the single-element doped samples sintered by the same procedure as the multi-elements doped sample. The KNN precursor powders were prepared at 850 C and it can be questioned if all the dopants were dissolved in the KNN lattice prior to sintering. This would influence on the distribution of the dopants after sintering, particularly with respect to the two sintering procedures used.

Reply:
We appreciate the valuable comments and questions. The multi-elements and single element doped samples were not prepared by the same sintering procedure, since different dopants doped KNN ceramics were sintered according to their optimized sintering conditions. This is because the sintering temperature window of KNN-based ceramic is very narrow [J. Am. Ceram. Soc. 94, 3659-3665, (2011)]. Following reviewer's suggestion, we tried to use the same procedure to prepare different elements doped KNN ceramics, as shown in Fig. R11. The single element doped KNNs were not sintered or over-sintered, the samples had very high dielectric loss and could not be poled completely, with piezoelectric coefficients below 50 pC/N. We checked the precursor powders of our studied compositions by XRD. According to the results shown in Fig. R12, the precursor powders did not show any impurity phase or secondary phase, demonstrating that all the dopants were dissolved in the KNN lattice. We totally agree with the reviewer that different microstructures exist in the multi-elements and single element doped KNNs. In order to study the role of different microstructures impacting the piezoelectric properties, we performed TEM to study the domain structures of KNN-Bi,Sb,Zr and comparing to KNN-Sb, as shown in the Fig. R13 (Supplementary Fig. 14 in the revised manuscript). As shown in Fig. R13, the grain size of KNN-Bi,Sb,Zr is much bigger than that of KNN-Sb. The domain size is proportional to the square root of grain size, thus the domain size of KNN-Bi,Sb,Zr was found to be about 500 nm, five time that of KNN-Sb (~100 nm). Of particular importance is that the hierarchical domain structure was observed in the large laminar domain of KNN-Bi,Sb,Zr ceramics, accounting for the greatly enhanced piezoelectric properties [Natl. Sci. Rev. 7, 355-365 (2020)]. Based on the above discussion, we also added the domain structure observation in the revised manuscript, as shown in the following. "In addition, the high mobility of the 90 o domain walls and hierarchical domain structure ( Supplementary Fig. 14) in the large grain of KNN-Bi,Sb,Zr ceramics may also be an important contributor to the enhanced extrinsic piezoelectricity. 22,33,34 " Comment 5: The doping of KNN require also knowledge about the defect chemistry of the material, which also may influence on the electrical properties. The authors should consider the following points related to the defect chemistry caused by the doping: Zr has a lower formal charge than Nb and this difference need to be charge compensated for by an additional point defect(s), which one(s)? Correspondingly and more challenging, the Bi doping was obtained by simply substituting a K+/Na+ with Bi3+. This is not possible without creating secondary phases or other charge compensation point defects. These considerations need to be commented in the manuscript and also discussed in relation to the heterogenous distribution of the dopants and the properties.

Reply:
We truly appreciate the valuable comments. We agree that the defects induced by the dopants are very important impacting the microstructure and properties. In this system, the Zr 4+ has lower valence thus generating oxygen vacancy when replacing Nb 5+ . While the Bi 3+ was used to provide the charge compensation on A-site by A-site vacancies because the Bi 3+ has higher valence comparing to Na + /K + . We performed the first-principle calculations based on density functional theory to optimize the structure of A-site vacancies and oxygen vacancies, tried to understand the impact of oxygen vacancy and Na/K vacancies on the phase structure, as shown in Fig. R14 ( Supplementary Fig. 7 in the revised manuscript). Based on reviewer's suggestion, we added the following paragraphs in the main text to describe the contribution of the defects. "It should be noted that the additions of Bi 3+ and Zr 4+ to A-and B-sites of KNN ceramics will inevitably introduce vacancy defects such as A-site vacancies and oxygen vacancies, respectively. The contributions of the A-site and oxygen vacancies to the structure were also studied based on first-principle calculations and given in Supplementary Fig. 7. Interestingly, the difference in the six B-O bond lengths of the octahedron, after adding different dopants (Fig. 1e) and oxygen vacancies, is obviously reduced compared with the undoped KNN. On the contrary, the A-site vacancies including Na-vacancy and K-vacancy have much weaker effect on the length of the B-O bonds." "It should be noted here that dopants generally induce vacancies, for example, the Bi 3+ occupying A-site K + /Na + position will cause A-site vacancies (including or ), while the Zr 4+ occupying B-site Nb 5+ position will lead to oxygen vacancies ( .. ), which are expected to impact the dielectric and piezoelectric properties. In our studied multi-elements doped KNN, we attempted to keep the charge balanced by tuning the composition with the appropriate doped levels, thus the effect of defects induced by dopants is minimal and is not considered in our mechanism." Comment 6: How was the dopants charge compensated for in the DFT calculations and how did the authors distribute K and Na on the A-lattice in the supercells used? Would the distribution of K and Na affect the findings from DFT? How was loss of alkali-oxide during sintering compensated for or was this not considered (only nominal compositions are given)?
Reply: We appreciate the valuable questions and comments. For doped KNNs, the A-site or O vacancies may exist due to charge compensation and the volatile nature of K/Na atoms. However, considering the finite size of the supercell in the DFT calculations (a 2×2×2 40 atoms supercell that has been used to improve the computational efficiency while still maintaining the local structural characteristics), it is improper to incorporate one dopant atom and one A-site or O vacancy together in such a cell, as this would introduce an unrealistic high vacancy defect concentration. Besides, the effect of the vacancy defects induced by doping in experiments is regarded as minimal by carefully tuning the composition with appropriate doped levels. In view of these, the dopant charge compensation was not considered explicitly with the charge balanced automatically in the calculations, and more attention has been paid to understand the local structural properties of the regions around the dopants, i.e., the lengths of the six B-O bonds. We also added the following sentence and Fig. R16 (Supplementary Fig. 15 in the revised manuscript) in the method and supporting information to describe the distribution of K and Na on the A-lattice in the supercells. "The supercell with the uniform distribution of A-site cations along the a, b, and c directions was used in the calculations, as shown in the supplementary Fig. 15." Fig. R16 The supercell of pure KNN that used to perform first-principles calculations. The purple, yellow, green, and red atoms are K, Na, Nb, and O atoms, respectively. The supercell with the uniform distribution of A-site cations along the a, b, and c directions was used in the calculations ( Supplementary Fig. 15 in the revised manuscript).
We added the following sentence to describe the method used to perform the DFT calculations for doped KNN: "For the doped KNNs, the models were built by substituting K, Na, and Nb atoms with in-principle equivalent dopant atoms, i.e., Bi on the A site, Sb and Zr on the B-site." Based on the comments, we performed the DFT calculations with another distribution of A-site cations supercell (The supercell of nonuniform distribution as shown in Fig.  R17a), as shown in the Fig. R17b. The results shown that the nonuniform distribution supercell had slightly influence on the length of B-O bonds comparing to the supercell of uniform distributed K and Na elements. But the evolution of B-O bonds length of nonuniform distribution supercell, after adding different dopants, was similar with the uniform distribution supercell results. In fact, after adding different dopants, the difference of B-O bonds length of nonuniform and uniform distribution supercell was all obviously reduced compared with the undoped KNN. In view of these, the distribution of K and Na elements in DFT calculations may have small effect for the trend, that the difference of six B-O bond lengths of the octahedron, after adding different dopants, is obviously reduced compared with the undoped KNN. We didn't consider the loss of alkali-oxide during sintering, and only used the nominal composition in the article. In fact, the loss of alkali-oxide during sintering will lead to A-site vacancies, including K-vacancy and Na-vacancy, have very weak impact on the B-O bond lengths based on first-principles calculations, as discussed above.

Comment 7:
The co-existence of two phases in the multi-element doped sample point to chemical heterogeneity also evidenced by STEM. The authors should also provide X-ray diffraction data above Tc to demonstrate if this heterogeneity remains above Tc showing apparently two cubic KNN phase present. It would also be useful to report the unit cell parameters and volume as pseudo cubic values to make an easier comparison of the two KNN phases.

Reply:
We thank the reviewer for the valuable suggestions. Based on the comments, we performed in-situ XRD as a function of temperature, and did the refinement to exhibit the phase structure evolution at high temperature (the X-ray diffraction data near and above Tc). We used two models (P4mm and Pm-3m) for the refinement. The database CIF files were ICSD 173741 and ICSD 186364 for P4mm and Pm-3m, respectively. The FWHM was found to slowly decrease at temperature above Tc. The refinement results shown that the Tetragonal and Cubic phases coexisted over temperature range from 500 K to 700 K, above which, the pure Cubic phase could be observed. The XRD data were given in Fig. R18 (Supplementary Fig. 10 in the revised manuscript). We revised the following paragraph in the main text to describe the evolution of phase structure at temperature above Tc. " Fig. 2a gives dielectric properties of KNN-Bi,Sb,Zr ceramics over a temperature range of 130-700 K, with dielectric anomalies at 550 K and 290 K, which correspond to Curie temperature and O-T phase transition temperature (Fig. 1d), respectively. It should be noted that the tetragonal phase still exists above the Curie temperature over a broad temperature range even the volume fraction is well below 10% ( Supplementary Fig. 10), which can also be confirmed by the broad dielectric peak at 550 K as shown in Fig. 2a." Comment 8: No ferroelectric properties were reported. It would be very useful for the quality of the paper to also report ferroelectric hysteresis loops.

Reply:
We thank the reviewer for the good suggestion. The ferroelectric hysteresis loop was added in Fig. R19 (Supplementary Fig. 2 in the revised manuscript). In addition, the coercive fields of KNN-Sb, KNN-Zr, KNN-Bi, were found to be on the order of 11, 13, and 18 kV/cm, respectively, and added in the modified Supplemental Table 1.

Comment 9
In the evaluation of the STEM data the effect of the distribution of Bi and Zr were disregarded. Since the atomic number of Zr is so close to Nb this can be argued for, but the heavy element Bi with atomic number 83 cannot be disregarded. It is also confusing that no A-site and B-site intensity map is not reported for KNN-Bi,Sb,Zr. To make this analysis more convincing the same set of analysis are needed for both materials (KNN-Sb and KNN-Bi,Sb,Zr). Reply: We thank the reviewer for the valuable comments. As the reviewer pointed out, it is hard to identify Zr on the B-site of KNN because the atomic number of Zr is so close to Nb. We also did the A-sublattice intensity analysis, as given in Fig. R20, unfortunately, it is difficult for us to separate the Bi dopant rich areas, even the atomic number of Bi is much larger than both K and Na elements, due to the following reasons: In the Sb doped KNN sample, K and Na are not randomly distributed. Local K and Na segregations were observed in Fig. R20 (a), where the high intensity (bright) areas belong to K rich regions because K has atomic number 19, being larger than that of Na (atom number 11). As the Bi was added, high contrast areas were also observed, as shown in Fig. R20 (b). However, it is difficult to identify the actual element that contributes to the observed STEM-HAADF contrast. With the coexistence of Bi, K and Na elements, the brighter contrast areas can be either K-rich or Bi-rich areas that cannot be well defined. In addition, the amount of Bi-dopants is very small, which might not contribute to the contrast of the STEM-HADDF image too much considering the large amount of K element. Therefore, we could not give a solid proof of the Bi rich area from STEM-HAADF images.  (2019), 264. The authors should consider to make comparison with this work and similar works in other systems where minor doping have been shown to enhance piezoelectricity. Reply: Thanks for the valuable suggestion. We added recently research on lead-based relaxor ferroelectrics in our paper and make comparisons accordingly [Science 364, 264-268, (2019); Kumar, A., et al. Nat. Mater. (2020);and Adv. Funct. Mater. 2006823, (2020)]. Please check the following paragraph. "Interestingly, the rare-earth doped PMN-PT ceramics with enhanced piezoelectric coefficient also exhibit increased volume fraction of tetragonal phase with reduced Curie temperature and broad relaxation peak. 31,36 Tetragonal phase with low-angle domain wall was also observed in the structure evolution of PMN-PT solid solution with increasing Ti concentration based on the molecular dynamic calculation and STEM observation, being considered responsible for its high piezoelectric properties. 26,37,38 " Response to Reviewer #3 Comment 1: The current contribution discusses the impact of a multiple of elements on the atomic level structural aspects and the piezoelectric properties of KNN ceramics. It is clearly noticed that the basic underlying idea is an extension from their earlier claims that unit cell level chemical heterogeneity that disturbs a long range polar order and consequently brings about an angular distribution in the direction of polar vector of each unit cell. I see that the collection of experimental data is of high quality, quite complete, and appropriate in supporting their claims, though I think I have to raise a coupling of questions to be clarified.

Reply:
We thank the reviewer for his/her positive comments and valuable questions. We will respond to the questions in the following.
Comment 2: First, the presence of the claimed nano-heterogeneity is obvious from their extensive TEM studies, but a demonstration on how these nano-heterogeneity is contributing to the piezoelectricity seems missing. One could assume that the low-angle polar vectors contribute to the piezoelectricity when an electric field is applied through polarization rotation, but this is just a possibility. It could be that or could be a shear contribution from misoriented polar vectors or maybe something else. I do not see a decisive experimental verification on the claimed scenario here.

Reply:
We thank the reviewer for the valuable comments. We agree with the reviewer that many potential contributors exist for the enhanced piezoelectricity in multi-elements doped KNN ceramics, including a shear contribution from the misoriented polar vectors. Based on our microstructure observations, theoretical calculations and property measurements, we are confident that the high piezoelectricity is from the rotation of low-angle polar vectors, because of the following reasons, hope the reviewer is happy with our explanations.
1. The observation of relaxor feature from dielectric measurement indicates the existence of local structure heterogeneity on nanoscale [Nature, 546, 391-395 (2017)]. 2. There is large amount of local structure heterogeneities with low-angle polar vectors confirmed by STEM. 3. The phase field modeling reveals the contribution from low-angle polar vectors to dielectric/piezoelectric properties is evident, where the interface energy of the local structure heterogeneity with low-angle polar vector is competitive with the Landau energy thus facilitates the polarization rotation. The rotation of polar vectors is actually corresponding to the shear contribution. Based on the above discussions, we proposed the mechanism, i.e., the dopants induced average tetragonal phase coupled with large amount of local structure heterogeneities with low-angle polar vectors, is responsible for the enhanced properties in multi-elements doped KNN ceramics.

Comment 3:
In fact, their first principles calculation suggested that the coexistence of those multi-elements tends to reduce the anisotropy that is directly proportional to the magnitude of polar vectors. If this is so, the induced angular variation in the direction of polar vectors has a high chance not to make any useful contribution to the piezoelectricity. Please note that the piezoelectricity, especially the electromechanical strain, is a function of both the easy rotation and the magnitude of polar vectors. As noted, one trades off the other.
Reply: Thanks for the valuable comments. We are sorry about the confusion. The first principles calculation emphasized that the existence of the dopant tended to reduce the bond-length difference between the six B-O bonds in the octahedrons. All the six B-O bonds in the octahedron will decide the magnitude of the polar vectors. We actually observed the overall spontaneous polarization was decreased in KNN-Bi,Sb,Zr comparing to pure KNN. However, the reduced anisotropy of the octahedron will impact the direction of polar vectors, leading to the low-angle polar vectors deviating from the spontaneous polarization direction of the tetragonal phase (confirmed by Synchrotron XRD and STEM). The polar vectors with lower deviation angles are prone to rotate under applied electric field, contributing to the enhanced dielectric permittivity. We agree with the reviewer that the piezoelectricity and electromechanical strain are a function of both the easy rotation and the magnitude of polar vectors and one trade off the other. According to the equation: d 33 ∝ ε Q P s , where ε is dielectric constant (3000 for KNN-Bi,Sb,Zr and 400 for pure KNN) and P s is spontaneous polarization (20 µC for KNN-Bi,Sb,Zr and 30 µC for pure KNN), while the Q value is electrostrictive coefficient which is insensitive to dopant [Appl. Phys. Review, 1, 011103, (2014)], the piezoelectricity and electromechanical strain of KNN-Bi,Sb,Zr are expected to increase because the increase in ε is much greater than the decrease in P s .

Comment 4:
In this sense, I think it would be better for the authors to tone down their claim a little by dropping the word "origin" in the title.
Reply: Thanks for the valuable suggestion. We agree with the reviewer that a decisive experimental verification is important on our claimed mechanism, where there are other mechanisms might also contribute to the properties. Following the suggestion, we revised the title as "The mechanism for the enhanced piezoelectricity in multi-elements doped (Ka,Na)NbO 3 ceramics".
Comment 5: Second, the authors are consistent in emphasizing the effect of multi-elements to get the currently observed enhanced piezoelectricity by comparing the singly doped ones with the triply doped one. I agree that the latter leads to better enhanced piezoelectricity than the former. Then, how about introducing only two different elements? In fact, it has been well known that doubly introduced elements, Li and Sb, are also highly effective in enhancing the piezoelectricity.

Reply:
We thank the reviewer for the valuable comments. We agree with the reviewer that the doubly introduced element, Li and Sb, can effectively enhance the piezoelectricity [J. Appl. Phys. 100, 104108 (2006)]. The dielectric properties and phase structure of Li and Sb doped KNN were shown below in Fig. R21.
We can see that the KNN-LS5.2 ceramics possess mixed phases at the region of polymorphic phases at room temperature. The maximum d 33 value of this system is about 260 pC/N. In this work, we tried to increase the amount of Li and Sb dopants in the KNN system, attempting to introduce the average tetragonal phase in the doped KNN. Unfortunately, secondary phase and impurity were generated due to the solubility limit of these dopants in KNN system, actually each individual dopant in KNN has solubility limitation and cannot be miscible with KNN. On the other hand, if we add the third dopant element, for instance Ta, we can further increase the overall dopant concentration which is enough to induce the average tetragonal phase, in this way, we can increase the piezoelectricity above 400 pC/N.