Three dimensional band-filling control of complex oxides triggered by interfacial electron transfer

The d-band-filling of transition metals in complex oxides plays an essential role in determining their structural, electronic and magnetic properties. Traditionally, at the oxide heterointerface, band-filling control has been achieved via electrostatic modification in the structure of field-effect transistors or electron transfer, which is limited to the quasi-two-dimension at the interface. Here we report a three-dimensional (3D) band-filling control by changing the local lattice coordination in a designed oxide heterostructure. At the LaCoO3/LaTiO3 heterointerface, due to the Fermi level mismatch, electrons transfer from LaTiO3 to LaCoO3. This triggers destabilisation of the CoO6 octahedrons, i.e. the formation of lattice configurations with a reduced Co valence. The associated oxygen migration results in the 3D topotactic phase transition of LaCoO3. Tuned by the thickness of LaTiO3, different crystalline phases and band-fillings of Co occur, leading to the emergence of different magnetic ground states.


1) In
(f-g), the authors' schematic of the charge transfer from LaTiO<sub>3</sub> to LaCoO<sub>3</sub> is inconsistent with what was proposed in Ref. 12. Per the model put forward in that paper, the O 2p band in LaCoO<sub>3</sub> should also rise as the Fermi level shifts.
2) The authors state that they deposit a SrTiO<sub>3</sub> capping layer on top of the LTO layer to preserve the surface for measurements after atmospheric exposure. However, Fig. S3(a) shows no STO cap on C30/T15. Figure S2(b) shows the Ti oxidation state with position that includes an STO cap, but there is no explanation for how it was acquired. They should show a corresponding image that includes the STO cap.
3) Assuming the cap is there, I have several concerns about the capping layer and its affects on the spectroscopy measurements. For the L-edge XAS shown in Figure 2(d), which was acquired in total fluorescence yield, how did the authors deconvolute the expected Ti4+ signal from the STO cap from the Ti4+ in the LTO layer? The thicker the cap, the harder it would be to deconvolute. However, if the cap is very thin (a couple unit cells), it will not be a good barrier to protect the surface. 4) The apparent mechanism for the formation of the brownmillerite La<sub>2</sub>Co<sub>2</sub>O<sub>5</sub> phase is diffusion of the oxygen all the way through the LTO layer after electrons have been donated. The oxygen then desorbs from the film surface. I am skeptical that LTO would permit that level of oxygen out-diffusion and desorption without scavenging excess oxygen given the stability of the La2Ti2O7 phase. In the event that it does, however, the model that the authors have put forward does not conserve charge. The resulting La<sub>2</sub>Co<sub>2</sub>O<sub>5</sub>should be a stable phase with Co<sup>2+</sup> ions that does not depend on electrons donated from LaTiO<sub>3</sub>. Thus, why is the Ti in the 4+ charge state in LaTiO<sub>3</sub>? In theory the oxygen would move through the LaTiO<sub>3</sub> and then leave behind electrons when it desorbs to return the Ti to a 3+ state. 5) Fig. S3(a) shows that the LTO film is fairly rough on the surface, which is supported by the poor RHEED image at right in Fig. S1(b). Have the authors done spectroscopy on the LTO films to determine oxygen content under their growth conditions? The reference data in Figure 2(d) is taken from the literature, so I am curious what the valence of Ti is for a uniform LTO film from the group. The growth conditions (P ~1x10^-4 Pa = 7.5x10^-7 Torr) are not too different from the conditions in which STO is deposited in an MBE (low 10^-6 Torr of O2) or by high quality via PLD (see Lee et al. Scientific Reports 6, 19941 (2016), DOI: 10.1038/srep19941). 6) The efforts to create control samples with LAO and STO films are admirable and support their conclusions. However, I am still left wondering what is going on in the LaTiO<sub>3</sub> layer as I explained above. 7) The degree of electron diffusion into LaCoO<sub>3</sub> from LaTiO<sub>3</sub> that they postulate is far greater than any other groups have reported in similar systems. For example, Kleibeuker et al. PRL 113, 237402 (2014) DOI: 10.1103.237402 had a rather similar LaFeO<sub>3</sub>-LaTiO<sub>3</sub> interface and saw only 1-2 unit cells of Fe<sup>2+</sup> due to charge transfer. There are no reports that I'm aware of where charge transfer has occurred across films that are 15 unit cells thick because the screening length in complex oxides is so small. 8) Fig. S1 should include the time scale. It currently says Time(s) for the horizontal axis, but doesn't provide the actual times.
Reviewer #2 (Remarks to the Author): The work by Meng et al. reports the finding of restructurings of LaCoO3 (LCO) thin films when interfaced with LaTiO3 (LTO). These restructurings are shown to be driven by the (interface) electron transfer from LTO to LCO. Interestingly, the LCO restructuring results in the formation of CoO4 sublayers in the LCO matrix in a topotactic fashion, and even far from the actual interface. The LCO ferromagnetic (FM) ordering is apparently lost via this interface-triggered phase transition.
The findings of this work are highly interesting and may open a new pathway to the realization of intriguing oxide phases. The writing is sound, albeit somewhat technical at various places. It would be helpful for the reader to render the physics conclusions, impact and outlook of the given results more concise, if possible.
Concrete comments/questions are as follows: 1. It would be helfpful to display the differences in the different restructured LCO layers clearer. 2. FM order is reported lost from Fig. 4c, but how can the authors detect AFM order? Is the latter one a speculation or proven? The authors should discuss the (possible) magnetic order in some more detail.
3. Is there a possibility to increase the 30 layers substantially? Or in other words, how strongly '3D-like' to the authors expect the reconstruction to be? This is an interesting new work in the area of oxide heterostructures that deserves highlight recognition. After some refinements in the writing/presentation and addressing of the posed questions, the manuscript can be recommended for publication in Nature Communications.
Reviewer #3 (Remarks to the Author): Dear Editor, The manuscript of Meng et al. reports the epitaxial growth and charge reconstruction in a perovskite-based heterostructure of LaCoO3/LaTiO3. Besides interfacial charge transfer, the main finding is a reduction of the almost whole LaCoO3 perovskite to oxygen deficient phases. It is found that the content of the reduction can be controlled by the thickness of the LaTiO3 thin films, particularly the brownmillerite LaCoO2.5 structure is observed in a heterostructure of LaCoO3 (30 uc)/LaTiO3 (15 uc) as confirmed by both XRD and STEM measurements. However, both the conductivity and the magnetization of the heterostructures are worse than the bare LaCoO3 thin films. The results are of potential interest, but there are a number of key questions remain open.
1. The authors explain the finding is due to the charge transfer from LaTiO3 to LaCoO3. However, the EELS measurements in Fig.2 show hardly signature of Co2+. The XAS in Fig.s2a shows features of Co2+, but the analysis in Fig.s2b can not be consistent with the experimental data in Fig.s2a. I suggest the authors to move the XAS data to the main text and make a further careful analysis. Also, when electrons are transferred to the LCO, the fermi level is expected to be downshift instead of upshift as shown in Fig.1(g).
2. The pure electronic reconstruction at LCO/LTO interface should be of short range nature, see for example Ref.16. So it remains unclear how the electronic reconstruction induces the strong reduction of LCO films. In fact, good LCO growth prefers a high oxygen background pressure, while growing stoichiometric LTO needs a very low oxygen background pressure. To make a compromise, I noticed for LTO. It is no doubt that the LCO reduction could be due to the film growth of LTO process. The important questions remain are (1) whether the reduction of LCO is due to the annealing in high Redox reaction will depend on temperature, oxygen pressure, and time, and it has been found that Ti-perovskite could be even more reductive than LAO, see for example Chen et al. Nano letters, 11, 3774, (2011) and17, 7362 (2017). So control experiments have to be performed to check these two possibilities.

In this work, Meng et al employ interfacial charge transfer between LaTiO3 and LaCoO3 to induce a topotactic phase transition in LaCoO3 to
La2Co2O5. This work is very interesting and draws on recent DFT models that predict band alignment and charge transfer across interfaces using the alignment of the O 2p bands between layers. Charge transfer is expected from LaTiO3 to LaCoO3 to produce a Ti4+ and Co2+ formal charge near the interface after electronic reconstruction. References 15 and 16 demonstrate this phenomenon at the LaTiO3-LaCoO3 interface. While there is still much work to be done on these materials, the charge transfer is not sufficiently novel to merit publication in Nature Communications in its own right. Additionally, I have too many concerns regarding the origins of the topotactic phase transition to recommend the work for publication at this time.

et al
In Figure 1

(f-g), the authors' schematic of the charge transfer from LaTiO3 to LaCoO3 is inconsistent with what was proposed in Ref. 12. Per the model put forward in that paper, the O 2p band in LaCoO3 should also rise as the Fermi level shifts.
The authors state that they deposit a SrTiO3 capping layer on top of the LTO layer to preserve the surface for measurements after atmospheric exposure. However, Fig. S3(a) shows no STO cap on C30/T15. Figure S2(    are not too different from the conditions in which STO is deposited in an MBE (low 10^-6 Torr of O2) or by high quality via PLD (see Lee et al. Scientific Reports 6, 19941 (2016) Fig.2 show hardly signature of Co2+. The XAS in Fig.s2a shows features of Co2+, but the analysis in Fig.s2b can not be consistent with the experimental data in Fig.s2a. Fig.1(g (1) whether the reduction of LCO is due to the annealing in high vacuum (1×10 4 Pa) alone or due to the redox reaction or oxygen absorber of the LTO film. Redox reaction will depend on temperature, oxygen pressure, and time, and it has been found that Ti-perovskite could be even more reductive than LAO, see for example Chen et al. Nano letters, 11, 3774, (2011) and17, 7362 (2017 While I still have some questions regarding the physical cause of the LCO reduction, I believe that the authors have adequately addressed the concerns that I and the other reviewers raised in the first version. The control samples, modification of the discussion, and improvement of the XAS analysis are sufficient in my opinion to support their conclusions. If the other reviewers agree, I believe that the manuscript can be published in its current form.

I suggest the authors to move the XAS data to the main text and make a further careful analysis. Also, when electrons are transferred to the LCO, the fermi level is expected to be downshift instead of upshift as shown in
Reviewer #2 (Remarks to the Author): The authors have revised their manuscript according to the various questions/comments from the referees. There are many experimental details to be addressed and the assessment of the different issues appears nontrivial. Still seemingly, the authors provide a great deal of experimental data and insight to support their result and interpretation. Therefore, the main message and findings appear solid and interesting enough to now warrant support for publication in Nature Communications.

Reviewer #3 (Remarks to the Author):
Dear Editor, Most of my comments have been answered clearly. The results trun out to be interesting. However, I don't think the author has answered my key concern: it is the redox reaction rather than electronic charge transfer that controls the phenemena observed here. To make my comments clear: (1) Interfacial electronic charge transfer will mainly occur at the sample surface and/or the interface. So if the electronic effect plays the key role, even there is 3D reconstructon in the LCO layer, there should be some indication on the LTO layer, such as a difference in spatial profile for the Ti3+/Ti4+ should be present. Instead, the authors showed/mentioned that the LTO layer exhibits the Ti4+ oxidation state homogeneously, so the redox reaction most likely dominate the phenomena observed here and the capping layer of STO may also play an important role as indicated by other referees.
(2)More importantly, In the case of the pure elecronic charge transfer, the total amount of transfered charge is generally limited and should be not dependent on the thickness of LTO (or the transition should occur sharply). On page 8, line 185-186, the author mentioned that "the sample with a thicker LTO has a higher conent of Co2+"this seems consistent with the redox reaction picture, i.e. the chemical potential drives the reduction of LCO.
(3) The authors mentioned that they have grown amorphous LTO on high-quality LCO (30 u.c.) at ambient temperature and under h XRD. However, the reduction could be much enhanced at high temperatures. The authors are suggested to anneal the amorphous LTO/LCO at the condition similar to the high temperature deposition of LTO, and further checked the crystaline structure of the annealed sample.
There are also two minor issues: (1) Whether it is "topotactic phase transitions"or redox reduction of LCO? For me the topotactic phase transition often involves other strong reduction agent such as CaH2, so it might be more suitable to call it redox reaction.
(2) The first principle calculations in the Discussion part of the manuscript is not convincing.
Shortly, I think the above issues should be clarified before the publication of the manuscript.

Response to reviewers' comments Manuscript Number: NCOMMS-20-33853A
We thank all the reviewers for their careful reading and helpful comments of our manuscript. The fact that Reviewer #1 and #2 find our revision being well improved and acceptable for publication in Nature Communications is rewarding. Reviewer #3 further raises some issues about the mechanism of the transition of LCO. Our point-by-point responses to the comments are presented below in detail.

Reviewer #3
Comment 1: Interfacial electronic charge transfer will mainly occur at the sample surface and/or the interface. So if the electronic effect plays the key role, even there is 3D reconstruction in the LCO layer, there should be some indication on the LTO layer, such as a difference in spatial profile for the Ti3+/Ti4+ should be present. Instead, the authors showed/mentioned that the LTO layer exhibits the Ti4+ oxidation state homogeneously, so the redox reaction most likely dominate the phenomena observed here and the capping layer of STO may also play an important role as indicated by other referees. Reply: We totally agree with that the interfacial charge transfer between complex oxides should be quasi-two-dimensional (2D). This 2D charge transfer at the LCO/LTO interface triggers the structural transition in LCO, which is associated with the release of oxygen ions. The net charge on the LCO side is canceled and the electric field that confines the charge transfer within the 2D region is changed. Consequently, the LTO layers away from the interface would provide more electrons diffusing towards the interface. In brief, the initial 2D charge transfer at LCO/LTO triggers the structural phase transition, further changing the screening electric field, and resulting in the homogeneous valence states across the films at both sides. The key role of the initial 2D charge transfer at LCO/LTO (with the density as high as ~6.6×10 -14 cm -2 ) in triggering the structural phase transition has been evidenced by the reference samples of LCO/STO or LCO/LAO (Supplemental Material, Fig. S7), in which the mismatch of the oxygen concentration does introduce some oxygen vacancies (reduce) to the perovskite LCO, but the amount is far less than the extent to induce the structural phase transition.
It is worth noting that the Ti ions in LTO are in the valence of 4+, instead of 3+ as in the perovskite structure. We have presented a brief discussion based on the experimental observations: 1) The STO substrate and capping might be acting as electron reservoirs to accommodate excess electrons; 2) The charging of the LTO layer might be partially compensated by a certain amount of interstitial oxygen ions. In this sense, it might be important to cap the heterostructures with the amorphous layer, but not limited to STO specifically. For example, it was reported that LaNiO 3 could also be effective in compensating the excess charge of the underlying layer [Ref. 16].
To be clear to readers, we have made corresponding revisions to the first and second paragraphs of the Discussion part (Page 9&10) of the revised manuscript, and wish to motivate further investigations to clarify the in-depth mechanism.
Comment 2: More importantly, In the case of the pure electronic charge transfer, the total amount of transferred charge is generally limited and should be not dependent on the thickness of LTO (or the transition should occur sharply). On page 8, line 185-186, the author mentioned that "the sample with a thicker LTO has a higher content of Co2+"this seems consistent with the redox reaction picture, i.e. the chemical potential drives the reduction of LCO. Reply: Thanks for pointing out the important issue. The initial charge transfer at the LCO/LTO interface is indeed 2D. But since the structural phase transition of LCO is triggered, the valence state of Co varies, canceling the net charge on the LCO side and changing the electric field at the interface that confines the charge transfer within 2D. As detailed in the Reply-to-Comment#1, the LTO layers away from the interface would provide more electrons diffusing towards the interface, resulting in the scenario that the total amount of the transferred charge is proportional to the LTO thickness, consistent with the experimental observations. To clarify the issue, we have added the above descriptions to the first paragraph of the Discussion part (Page 9) of the revised manuscript.
In addition, we exclude the possibility of that "the chemical potential drives the reduction of LCO" as the only mechanism by three facts: (1) We have compared the formation and migration energy of oxygen vacancies in different oxides (Table R1). Since the structural phase transition of the LCO film is associated with the mobile oxygen ions diffusing out of the LCO through the capping layers (LTO in the current work), the upper layers with lower migration energy would lead to stronger reduction of LCO. As listed in Table R1, STO should have caused the strongest redox reaction among STO, LAO, and LTO. But experimentally, by capping LAO or STO layers, some oxygen vacancies are introduced to the perovskite LCO, but the amount is far less than the extent to induce the structural phase transition (Fig. S7 in Supplemental Material). (2) Similarly, by growing LCO films in the redox conditions, oxygen vacancies can be created but fail to obtain the La 3 Co 3 O 8 or La 2 Co 2 O 5 other than the perovskite phase. (3) The reference samples (Supplemental Material,Fig. S8) indicates that without the interfacial charge transfer, the structural transition of LCO will not occur. Pa) and only observed limited reduction by XRD. However, the reduction could be much enhanced at high temperatures. The authors are suggested to anneal the amorphous LTO/LCO at the condition similar to the high temperature deposition of LTO, and further checked the crystalline structure of the annealed sample. Reply: We have annealed the LCO/amorphous LTO sample at 670 °C for one hour. As shown in Fig. R1, LCO maintains the perovskite structure after annealing, evidencing that, without the triggering of the interfacial charge transfer, the treatment in a reductive environments alone cannot induce the structural phase transition of LCO. We have added above results to the revised Supplemental Material (Fig. S11). Figure R1. X-ray diffraction data of LCO(30 u.c.), as-grown and annealed LCO(30 u.c.)/amorphous LTO.

Comment 4:
Whether it is "topotactic phase transitions" or redox reduction of LCO? For me the topotactic phase transition often involves other strong reduction agent such as CaH2, so it might be more suitable to call it redox reaction. Reply: It is true that, in the current work, the phase transition of LCO is indeed a redox reaction since it is accompanied by the loss of oxygen and the change of Co oxidation state. And the crystalline orientation of the lattice is maintained in the phase transition, satisfying the definition of topotactic phase transition [Nat. Mater. 12, 1057(2013]. Therefore, to emphasize the ordering of LCO after phase transition, we are inclined to use "topotactic phase transitions". To be clear to the readers, we have added the descriptions about redox reaction to the Discussion part of the revised manuscript.
Comment 5: The first principle calculations in the Discussion part of the manuscript is not convincing.

Reply:
The results of the first principle calculations just provide the supplemental demonstration for the structural destabilization of LCO upon the electron transfer, which is mainly based on the experimental observations, i.e. the interfacial charge transfer does trigger the structural phase transition in LCO (see Reply-to-Comment 1 for details). STO (001)