Permanent ferroelectric retention of BiFeO3 mesocrystal

Non-volatile electronic devices based on magnetoelectric multiferroics have triggered new possibilities of outperforming conventional devices for applications. However, ferroelectric reliability issues, such as imprint, retention and fatigue, must be solved before the realization of practical devices. In this study, everlasting ferroelectric retention in the heteroepitaxially constrained multiferroic mesocrystal is reported, suggesting a new approach to overcome the failure of ferroelectric retention. Studied by scanning probe microscopy and transmission electron microscopy, and supported via the phase-field simulations, the key to the success of ferroelectric retention is to prevent the crystal from ferroelastic deformation during the relaxation of the spontaneous polarization in a ferroelectric nanocrystal.

4. It is difficult to understand "The energy barrier of the 71{degree sign} switching is much lower than that of the 180{degree sign} switching" As authors argued, 70 domain switching involves ferroelastic deformation, and thus the penalty should be higher. Why opposite? Any physical explanation? 5. The fact that 40nm thick system requires larger coercive field to switching seem to contradict the claim that it relaxes faster. Larger coercive field usually mean more stable system. Any clarification?
Reviewer #3 (Remarks to the Author): This paper presents a novel and interesting study of the effect of crystal size and geometry and retention in self assembled BFO-CFO mesocrystalline systems.
Although the results are interesting, the authors dramatically oversell their significance. All ferroelectrics under appropriate boundary conditions should have permanent ferroelectric polarization. It is thus somewhat of a "straw man" argument to create a system where you have poor retention and then tune it until you have good retention. Further, while measuring a years retention is very good, I do not think the authors should describe it as "everlasting". This seems quite over-dramatic, "extremely long" would be sufficient Another serious deficiency of the paper is what seems to be a willful neglect of the previous literature on BFO-CFO nanocomposite systems. There have been literally hundreds of papers on this since the 2005 Nanoletters paper of Zavaliche et al (Nanoletters, 5 1793(Nanoletters, 5 (2005) from Ramesh's group (and of course the 2004 pioneering paper in Science from Zheng et al in 2004 on BTO-CFO, also from Ramesh's group). One especially relevant to this paper is Zheng et al, Advanced Materials 18,2747Materials 18, (2006, where the fabrication of the 111 oriented structures studied here was described. It is inconceivable that the authors are not aware of this because the corresponding author of this work is on that paper! There are numerous investigations of these kinds of systems using PFM (unsurprising given that the chief interest is coupling between polarization and magnetism through strain in this system and the lead driver has been the Ramesh group in Berkeley who are experts in PFM), none of which are cited. One in particular stands out, Zheng et al, Nanoletters, 6, 1405 (2006), where the PFM response of BFO-CFO heterostructures on 111 substrates was measured. Curiously, when this was published 10 years ago, nothing was mentioned about the retention, either good or bad.
As the corresponding author was a postdoc in the Ramesh group and indeed co-authored many of the relevant works in the area, I can only conclude the large body of work on BFO-CFO nanostructures has been intentionally neglected to create an exaggerated sense of the novelty of this work.
My recommendation is that this work not be accepted for publication in Nature Communications. I would like to see the authors rewrite it without all the unnecessary spin, and include instead a sufficient discussion and citations to previous work. they should then submit it to a more specialized journal. If they do this, I do think the community will find the results to be of interest. I should note that in the process the authors would do well to seek a proofread from a native speaker of English as this paper is literally full of grammatical mistakes. As some of the authors have US and UK addresses, I am sure that this would not be too difficult to do and really should be done any time that a group of non-native speakers submits a paper for publication.
After the revisions by the authors I can recommend this paper to be published.
Reviewer #2 (Remarks to the Author): I have reviewed the revised manuscript, as well as the authors' response, and I am happy with the authors' revision. I would suggest that the authors include more references on PFM, phase field simulations, and BFO, to make it more representative of the field.

Reviewer Comment:
The authors suggest to rely on elastic constrains by a matrix surrounding BFO mesocrystals in order to address the challenge with poor polarization retention in multiferroic BFO. The working hypothesis is that the matrix prevents ferroelastic deformation, something that takes place during switching. Although an important problem to solve, and an original and relevant approach combining phase field simulations with thin film analysis is employed, if the paper is to be published in Nature Communications I suggest that the following points are addressed.

Response:
We thank the referee for the support on this manuscript. The questions are addressed in the following response.

Response:
We do agree with the reviewer that the pyramid shaped protrusions on top of the mesocrystal have no lateral constrains. Thus, there is a possibility of mediating backswitching via the nucleation of opposite domains by these protrusions. However, the lateral constraints in the bottom of BFO mesocrystal can still prevent the growth of opposite nucleated domains since an elastic deformation still needs to be involved to reverse the ferroelectric polarization in the constrained BFO. In addition, the relative portion of the protrusion is small, therefore, even they can be back-switched, the domains are not stable. We have included the corresponding discussion in the revised manuscript.

Response:
We thank the reviewer for this relevant question. Actually, the out-of-plane strain is highly coupled with the in-plane one, since BFO nanocrystal is treated as an elastic object with the strain imposed by the CFO matrix. Thus, it is more important to compare the in-plane strain state in the as-grown state. In order to realize this, we have carried out additional TEM and Geometric Phase Analysis (GPA) 1 analysis on the BFO nanocrystals with different sizes. Fig. R3 (a) and R3 (f) are the plane-view HRTEM images of a 120 nm-thick BFO mesocrystal. The GPA 1 method, which is an image processing technology used for mapping lattice displacement and has been successfully applied to characterize interface or defect structures such as misfit dislocations and their associated strain fields, was carried out based on the HRTEM images at the heterogeneous interfaces. The strain maps of the BFO-CFO heterointerface shown in Fig. R3 (b), (c), (g), and (h) were performed at the interface between CFO matrix and BFO nanocrystals of diameter ~15 nm and ~170 nm, respectively. For the e xx , e yy distortion maps (<110>, <1-12> directions) of small BFO nanocrystal (~15 nm) as shown in Fig. R3 (b) and R3 (c), the uniform color contrast was revealed with a lattice misfit of ~2.0% in e xx (Fig. R3 (d)) and ~2.5% in e yy ( Fig.   R3 (d)) between the CFO and BFO phases. This suggests that the BFO nanocrystal is constrained by a strong strain from the matrix. While for the distortion maps (e xx , e yy ) of large BFO nanocrystal (~170 nm) as shown in Figs. R3 (g) and R3 (h), the obvious color contrast was revealed with a lattice misfit of ~6.0% in e xx and e yy maps (Fig. R3 (i)) between the CFO and BFO phases. The value of lattice misfit is close to the calculated misfit using their bulk values, revealing that the lattices of BFO nanocrystal are almost relaxed. In addition, dislocations can be found at the interface between CFO matrix and large BFO nanocrystals (Fig. R3 (j)), suggesting the strain relaxation is mediated by the formation of dislocations, which cannot be found in the vicinity of small BFO nanocrystals (Fig. R3 (e)). This explains why large BFO nanocrystals relax much faster than small ones. We have included these results in the support materials. The corresponding discussion has also been incorporated in the revised manuscript.

Reviewer Comment:
This is a very interesting work on the retention of ferroelectric state of BFO mesocrystal through compressive stress, and will pave the way for practical memory devices. I recommend its publication, though there are number of points that need clarification.

Response:
We sincerely appreciate the support on this manuscript. All the points have been carefully revisited and addressed in the following response.
Reviewer Comment: 1. On page 5, "Such a trajectory results a reversal of ferroelectricity and a deformation of the crystal in the field direction as illustrated in Fig. 1a." This is not entirely accurate, as the final state after switching, compared to initial one, has no macroscopic deformation.

Response:
Thank you very much for pointing out this. There is no macroscopic deformation after the ferroelectric switching. We are very sorry for the mistake and we have modified it to "Such a relative displacement of the Zr/Ti from the centrosymmetric positions results a reversal of ferroelectricity as illustrated in Fig. 1a." Reviewer Comment: 2. The intermediate state in Fig. 1c is not entirely convincing. Is there other evidence supporting this claim? One possibility would be measuring strain versus field during the process.

Response:
We thank the reviewer for this suggestion. We have thought to measure the strain versus field during the switching process. However, the ferroelectric switching typically happens in a time scale less than 1 micro-second, therefore, it is very difficult to pick up this signal. In addition, due to the leakage problem, it is very hard to do a large area poling. More importantly, the retention behavior strongly depends on the size of BFO nanocrystal. Therefore, a technique with spatial resolution is required. In order to verify this, we have conducted PFM to switch the ferroelectricity of BFO nanocrystals as a function of voltage pulse duration. The OOP and IP phase images before and after the application of voltage pulse are shown in Fig. R4 (a). For the analysis of polarization direction and switching modes, the OOP and IP phase images in Fig. R4 (a) are combined and the results are presented as the schematics shown in Fig. R4 (b). Here, we compared the change of polarization and found the

Response:
We are sorry for this confusion. According to the early study by Kubel [1] .
Reviewer Comment: 5. The fact that 40 nm thick system requires larger coercive field to switching seem to contradict the claim that it relaxes faster. Larger coercive field usually mean more stable system. Any clarification?

Response:
Although the 40 nm thick system does show a larger coercive field, the deviation of the center of the hysteresis loop from the origin is also larger compared to the other two systems as shown in Fig. R5. The deviation is considered as a driving force of the ferroelectric relaxation since it suggests an existence of the depolarization field.

Reviewer Comment:
This paper presents a novel and interesting study of the effect of crystal size and geometry and retention in self assembled BFO-CFO mesocrystalline systems.

Response:
We thank the referee for the positive response.

Reviewer Comment:
Although the results are interesting, the authors dramatically oversell their significance. All ferroelectrics under appropriate boundary conditions should have permanent ferroelectric polarization. It is thus somewhat of a "straw man" argument to create a system where you have poor retention and then tune it until you have good retention. Further, while measuring a year retention is very good, I do not think the authors should describe it as "everlasting". This seems quite over-dramatic, "extremely long" would be sufficient.

Response:
In principle, there is no ferroelectric retention problem if appropriate electrical boundary conditions can be implemented. In a typical metal-ferroelectric-metal capacitor, the problem can be solved when a structure of symmetric electrodes is used 1 . However, there are several other configurations for practical applications, such as metal-ferroelectric-semiconductor and AFM tip/ferroelectric/metal, in which the symmetric electrodes cannot be made. Therefore, a severe retention problem occurs.
We have tracked this problem on BFO for a while 2-4 and a serious retention problem can be observed on BFO films with various orientations. Fig. R6 shows the comparison on the ferroelectric retention of BFO films, our BFO mesocrystal, and other ferroelectric materials with asymmetric electrical boundaries, suggesting that the ferroelectric retention is a generic problem. A new mechanism should be incorporated to provide a solution to this problem. This sets the novelty of this work.
We provide a possible solution to this long-term issue, which hinders the applications of ferroelectrics in the past and multiferroics in the future. The relaxation of ferroelectric polarization follows certain physical principles. Typically, the characteristic time of relaxation can be extracted based on the initial relaxation trend via various theoretical model. However, in our system, it doesn't show any degradation. Thus, we can say the retention time is much longer than our measuring period, one order longer at least. We call it "permanent", because it can last longer than the regular device life. However, if the reviewer insists, we will modify it. We have also included Fig. R6 in the supporting materials.  Mixed phase region, Diameter~95nm [4] T-matrix, Diameter~95nm [4] Unstrained BFO, Diameter~80nm [4] PZT/LSCO, Diameter~200nm [5] PZT/LSCO, Diameter~120nm [5] PZT/Pt, Diameter~160nm [6] LNO, Diameter~750nm [7] PZT/LNO, Area~20um 2 [8] PZT/Pt, Area~20um 2 [8] PZT/LSCO, Area~9um 2 [9] Polycrystalline PZNT [10] BMF, Diameter~1um [11] SBT/Pt, Diameter~180nm [12] PZT/LSCO, Area~0.06um 2 [13] PMN/LSCO [14]  One especially relevant to this paper is Zheng et al, Advanced Materials 18,2747Materials 18, (2006, where the fabrication of the 111 oriented structures studied here was described. It is inconceivable that the authors are not aware of this because the corresponding author of this work is on that paper! There are numerous investigations of these kinds of systems using PFM (unsurprising given that the chief interest is coupling between polarization and magnetism through strain in this system and the lead driver has been the Ramesh group in Berkeley who are experts in PFM), none of which are cited. One in particular stands out, Zheng et al, Nanoletters, 6, 1405 (2006), where the PFM response of BFO-CFO heterostructures on 111 substrates was measured. Curiously, when this was published 10 years ago, nothing was mentioned about the retention, either good or bad.

Response:
As the reviewer said, I am in this field for a while. I am aware of all the publications on the similar systems. In the early studies, the focus has been paid on the growth control and the magnetoelectric coupling on this system. Most of the studies were conducted on the BFO-CFO system on STO(001), in which BFO is the matrix, CFO forms nanopillars opposite to our case. Although PFM has been shown on this system, there is no study addressing the ferroelectric retention issue, which remains as a critical issue to practical applications. I will be happy to withdraw this paper if the reviewer can point out a study on the same issue with the similar system.

Reviewer Comment:
As the corresponding author was a postdoc in the Ramesh group and indeed coauthored many of the relevant works in the area, I can only conclude the large body of work on BFO-CFO nanostructures has been intentionally neglected to create an exaggerated sense of the novelty of this work.

Response:
I am very sorry and surprised that the reviewer has such an impression on my study.
The major reason that I did not cite these papers is that they are not very relevant to this study (not related to the issue of ferroelectric retention). Moreover, based on my training, we always avoid the self-citation to increase our own citation numbers. So, definitely there is no such an intention to ignore previous work to create an exaggerated sense of the novelty of this work. In order to fix this issue, some relevant references related to the growth part have been included in the revised manuscript. I believe the novelty of this study has been addressed very carefully in the introduction part. We suggested a possible solution of ferroelectric retention problem, which is a very critical issue to solve if one wants to use BFO in practical applications.

Reviewer Comment:
My recommendation is that this work not be accepted for publication in Nature Communications. I would like to see the authors rewrite it without all the unnecessary spin, and include instead a sufficient discussion and citations to previous work. they should then submit it to a more specialized journal. If they do this, I do think the community will find the results to be of interest. I should note that in the process the authors would do well to seek a proofread from a native speaker of English as this paper is literally full of grammatical mistakes. As some of the authors have US and UK addresses, I am sure that this would not be too difficult to do and really should be done any time that a group of non-native speakers submits a paper for publication.

Response:
We have revised the manuscript and included some previous studies in the references.
The proof reading and clarity in language have been improved. We hope with the improvement the reviewer can recommend the publication of this manuscript in