Judith L.
MacManus-Driscoll
*a,
Rui
Wu
*ab and
Weiwei
Li
*ac
aDepartment of Materials Science and Metallurgy, University of Cambridge, Cambridge, CB3 0FS, UK. E-mail: jld35@cam.ac.uk; ruiwu001@scut.edu.cn; wl337@nuaa.edu.cn
bSpin-X Institute, School of Physics and Optoelectronics, State Key Laboratory of Luminescent Materials and Devices, Guangdong-Hong Kong-Macao Joint Laboratory of Optoelectronic and Magnetic Functional Materials, South China University of Technology, Guangzhou 511442, China
cMIIT Key Laboratory of Aerospace Information Materials and Physics, State Key Laboratory of Mechanics and Control of Mechanical Structures, College of Physics, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China
First published on 21st February 2023
Interfaces in complex oxides give rise to fascinating new physical phenomena arising from the interconnected spin, lattice, charge and orbital degrees of freedom. Most commonly, interfaces are engineered in epitaxial superlattice films. Of growing interest also are epitaxial vertically aligned nanocomposite films where interfaces form by self-assembly. These two thin film forms offer different capabilities for materials tuning and have been explored largely separately from one another. Ferroics (ferroelectric, ferromagnetic, multiferroic) are among the most fascinating phenomena to be manipulated using interface effects. Hence, in this review we compare and contrast the ferroic properties that arise in these two different film forms, highlighting exemplary materials combinations which demonstrate novel, enhanced and/or emergent ferroic functionalities. We discuss the origins of the observed functionalities and propose where knowledge can be translated from one materials form to another, to potentially produce new functionalities. Finally, for the two different film forms we present a perspective on underexplored/emerging research directions.
While there have been many important reviews on emergent functional properties of SL1 and VAN,2 a broader article comparing the phenomena in these different structures has not been presented. The two fields have typically followed quite separate paths by quite separate groups. At first glance, this seems somewhat surprising since the tools for creating the films, the functional materials being studied, and the desired multifunctionalities being aimed for are either the same or are similar. On the other hand, there are subtle differences in approach and knowledge-sets for creating each film form which has precluded overlap between them.
In this review we focus on ferroic properties in the two materials as these are particularly exciting and because they exemplify the salient effects offered by each platform. We aim to present the key differences in the functionalities achieved by exploiting interface effects in the two thin film forms, comparing the merits and drawbacks of each and the reasons for those differences. We also aim to give insight into what can be learned from one form for the benefit of the other. We highlight novel phenomena and promising areas for future studies. We first compare the basic growth and structural features of each form and then we compare their exemplar ferroic properties. After that, we then explore in more detail the individual ferroelectric, ferromagnetic and multiferroic properties of SLs and follow this by exploring the same three phenomena for VANs. After each SL ferroics section, we discuss what can be learned for VAN from the SL field. Where appropriate, we do this also for each VAN section also. Finally, we propose future perspectives for each, including underexplored/emerging research directions.
Ferroics are fascinating exemplar materials to explore for our different film forms as they exhibit very wide ranging and tunable properties with huge future applications prospects. Ferroics refer to materials presenting at least one intrinsic long-range ordering such as magnetization, electric polarization, etc. The first application of a ferroic was in lodestone, which was used as a simple magnet and later as a compass, more than 2000 years ago. An understanding of the basic origin of the ferroics was not achieved until the beginning of the last century with the help of quantum mechanics. Over the past century, and in particular in the last decades, many new materials, of metallic, semiconducting, insulating and even organic materials classes, have been added to the ferroics collection. This endows them with very important applications in a range of current electronics devices.
Ferroelectrics have a spontaneous electric polarization that can be switched by an external electric field,3 or by mechanical or optical stimuli.4 They are used in a wide range of radio-frequency and microwave devices, as well in thermal sensors, actuators, and transducers, etc.,5 and there is also strong interest for non-volatile memory.6 Magnetic thin films are used in magnetic recording media, the leading technology for mass data storage.7 Also, magnetic random-access memory (MRAM) could be the winning technology for next-generation ‘universal’ memory.8 Another important ferroic form is multiferroics where two ferroic properties coexist, giving promise of magnetoelectricity for multi-state memory in which there is control of magnetism with an electric field. Magnetoelectric random access memory (MeRAM) possesses all advantages of MRAM, but crucially at much lower energy consumption.9
For the field of ferroic electronics, thin films of transition metal oxides (TMOs) are of particular interest as they show a very wide range of ferroic properties which are linked to the d orbitals in TM ions. These orbitals, in the presence of a strong ligand field from oxygen ions bonded to the TM ions, produce a wide range of excited states, polarisation effects, and electronic properties. These effects are highly tunable owing to the strong correlation between charge, orbital and structural degrees of freedom of the TM–O bond. Of the TMOs, oxide perovskites have been most widely studied owing to their wide ranging functionalities, ease of growth by pulsed laser deposition (PLD) and compatibility with widely available perovskite single crystals.
If one adds an interface between different TMOs, then symmetry breaking effects, strain and strain-relieving defects6 at the interface add yet another handle of tunability, coupling and complexity to TMO functionality. Interfaces and the consequent coupling between two different TMOs can be accomplished effectively in either SL or VAN forms. Conventional SLs have interfaces parallel to the substrate and are grown one layer at a time using vacuum deposition methods. There are a couple of notable exceptions where SLs form by self-assembly from a combined target material, i.e. TiO2/VO210 and YBa2Cu3O7−x/BaZrO311 the first formed by spinodal decomposition and the second by phase separation.
The schematics of SL and VAN epitaxial thin film forms, grown mainly by PLD, but less often also by molecular beam epitaxy (MBE) for SL films, are shown schematically in Fig. 1, with structural characteristics and interface-related phenomena shown. For far-from-equilibrium processing methods (including PLD but also sputtering), while vertical columnar grains form in standard planar films, these are different to the VAN columnar grains. For VAN columnar grains, their growth is controlled by vertical epitaxy with the matrix phase in the film and there is vertical epitaxial strain associated with this, as well as potential for interface phenomena at these interfaces, whereas for standard planar films such effects do not occur. The remaining figures in the paper go into more depth than Fig. 1, explaining in detail the origins of the interface phenomena in both VAN and SL films.
Fig. 1 Schematic diagrams showing structural characteristics (inner ring) and observed interface-related phenomena (outer ring) in epitaxial ferroic thin films formed from (a) SL structures and (b) VAN structures. Tricolor multiferroism is a kind of multiferroism induced in a three-component SL. See Fig. 2–8 for more details about how both structural and physical effects link to the observed interface-related phenomena. |
We note that multilayer (ML) films are similar to SL films, except they contain fewer layers (<∼5) than SL (>∼5), and the individual film layers are typically thicker than for SL (e.g. >15–30 unit cells for ML, and <15 unit cells for SL). In this paper, MLs are discussed where SLs have not yet been presented.
Both SLs and VANs have large area interfaces which enable physical phenomena to be both precisely tuned, e.g. using external stimuli such as magnetic or electric fields, and probed, e.g. using spectroscopic tools. Table 1 compares the growth and structural features of the two different materials forms.
SL | VAN | |
---|---|---|
a The substrate influences strain only for ≤20 nm film thickness. Above this thickness, the vertical strain dominates.12 Hence, strain can be 3-dimensional with auxetic-like effects possible.13 b >1000 nm thicknesses possible. c Insufficient time in SL to nucleate and grow dislocations between layers so other defects form to alleviate strain. d Growth of vertical interfaces slow, enabling misfit dislocations to form.14 e Interfaces become less 2D-like as the number of superlattice layers increases above ∼10–20 layers and then the layers roughen. f Interface directions can deviate from the vertical direction, and hence can be tortuous. g In VAN, there can be chemical segregation of species at an atomic level right at the interface.15 This is not the same as chemical intermixing which occurs by reaction between two materials at either side of the interface. This is the case for SL where you have two separate materials grown from separate targets and which can then react at the interface during growth. In VAN, if intermixing is thermodynamically favoured it will have already occurred when the materials are all mixed together in the ceramic target. It means there is not additional chemical reaction/chemical intermixing at the interfaces between the two materials. | ||
Complexity of growth process? | Higher | Lower |
Very wide variety of materials combinations possible? | Y | N |
Intermixing at interfaces? | Few unit cell level | Unit cell levelg |
Substrate control of strain | Y | Na |
Easy and uniform control of in-plane strain | Y (in films <20 nm) | N |
Easy and uniform control of out-of-plane strain | N | Y (in films >20 nm)b |
Misfit dislocations at interfaces | Nc | Yd |
Dimensionality of interfaces | 2D | Non-ideal 2D |
Deviation from dimensionality | Very lowe | Widef |
The overall drive in research on VANs has been more applied owing to simplicity of growth, and the ability to grow relatively thick, highly strained films resulting from a high density of vertical interfaces, whereas for SL films the focus has been on basic science, enabled by the growth of well-defined 2D interfaces. The growth of SLs is more complex than VAN, requiring reflection high energy electron diffraction (RHEED) and precise substrate heating to carefully control the thickness and qualities of the individual layers. On the other hand, SL growth produces much more well-defined 2D interfaces than VAN, making them ideally suited to exploring emergent interface effects.
VANs grow in a one-shot process from a single ceramic target, and the VAN structure forms epitaxially by self-assembly in the film. The growth of VANs does not need RHEED or very precise control of growth conditions as for epitaxial SLs where each film layer must be very smooth. However, to create new interface-coupled effects in well-aligned VANs, one needs to predict which phases will self-assemble in a given VAN film, starting from the target of all the mixed cations. This requires a greater knowledge of the stabilising forces of epitaxy with the substrate and thermodynamics of the possible phases that can arise from the global cation mix. A simple example is as follows: you may want to grow a VAN of NiO and Fe3O4 on a standard single crystal substrate. Depending on the Ni:Fe ratio in the target, and assuming standard growth conditions, you will achieve either NiO + NiFe2O4 or NiFe2O4 + Fe3O4, and hence not the desired NiO and Fe3O4 mix. This example does not mean you will always achieve a ternary phase and a binary phase when the aim is for two binaries. VANs of binaries are indeed possible where intermediate ternaries do not form, e.g. TiO2 + VO2.16
The basic ground rules for growing VANs have been set out in comprehensive reviews, the first in 201017 with others building on the understanding subsequently.18 Suffice for this article is to note that in VAN films a matrix phase forms in the film with an inclusion phases. The inclusion phase should be structurally dissimilar to the matrix phase to prevent ready mixing of the two materials during growth. This is a different situation to SLs where structurally similar phases are required so that the artificial interfaces that form between them are more perfect. If there is a sufficiently high-volume fraction of the inclusion, it typically forms as nanopillars which grow embedded in the matrix phase. The nanopillars are fine, typically <20 nm in diameter and ∼5–50 nm pitch, controllable by selecting the appropriate growth parameters.2b
In VAN films the vertical interfaces formed with the matrix can be faceted semi-coherent interfaces or cylindrical surface incoherent interfaces. We term both types of interfaces non-ideal 2D. Owing to the fine pillar dimensions and pitch, it is possible for large vertical and uniform strains to be induced in the matrix from the pillar scaffolds.17,19,20 More details about the large vertical strains for VAN films are discussed in Section 6. The most perfect VANs are achieved when the matrix material is structurally the same or very similar to the substrate.15b Here, a single interface type (lowest lattice mismatch one) will form, enabling the high density, highly vertical interfaces to be created. For this situation, strain control to >1 micron film thickness has been shown.14,21 On the other hand, when the matrix material is lattice mismatched to the substrate and also to the second phase in the VAN film, the films grow less perfectly.22 Here, the interfaces have higher energy than when films are isostructural. To minimise the vertical interface energies, a range of different interfaces with similar interfacial energy can form, which leads to meandering of the vertical interfaces. Examples are shown in later figures in this paper. More work needs to be done to give a more quantitative description of VAN microstructures in relation to the properties of the two materials in the film and the growth conditions, and the consequences of the nature of the meandered microstructure on the strain levels in the films and thus the functional properties.
The kinetics of interface formation are very different for SL and VAN films which influences the perfection of the interfaces (and regions in the vicinity of the interfaces). For VAN films the vertical interfaces grow slowly, allowing misfit dislocations to form to partially relieve the strain, whereas for SLs ‘forced’ artificial interfaces form very rapidly, which can lead to the formation of point-like defects to relieve any interfacial strain effects.14 Also, since the VAN interfaces are formed at equilibrium and are self-assembled, they are not chemically intermixed as SL film interfaces often are.1a This, however, comes at the expense of a lower possible number of materials permutations that can be combined together.
The most enhanced or novel ferroic properties from SL and VAN films, i.e. the maximum possible values or novel ferroic properties achieved for each materials form, are summarised in Table 2. Different materials pairs between SL and VAN are compared because different pairs are compatible for each thin film form. Hence for VAN, one typically requires structurally dissimilar materials to prevent a single mixed phase from forming, while for SL structurally similar materials are required to enable high quality interfaces to form. From Table 2, it is clear that the two materials forms offer quite different maximum or novel effects. An important question is what can be learned from one materials platform and translated to another? For example, could a topological Hall effect be achieved in VAN similar to SL, or could a self-biased magnetoelectric effect be achieved in SL similar to VAN?
SL | VAN | |
---|---|---|
FM – ferromagnetism. FE – ferroelectricity, ME – magnetoelectricity. DME = direct magnetoelectric coefficient, CME = converse magnetoelectric coefficient. | ||
FE | In 10–200 nm thick films | In ∼micron thick films |
T C of BaTiO3/SrTiO3 increased by several hundred K23 | T C of BaTiO3–Sm2O3 VAN increased by several hundred K21 | |
P r ∼ 16.5 μC cm−2, the value is doubled in SrTiO3/BaTiO3/CaTiO3cf. plain BaTiO3 film24 | P r ∼ 12.1 μC cm−2, the value is trebled in (Ba,Sr)TiO3–Sm2O3 VAN film ∼ cf. plain (Ba,Sr)TiO3 film25 | |
Tunability values ∼40% at 200 kV cm−1 and 300 K in SrTiO3/BaTiO3/SrO26 which is >3 higher than for BaTiO3 plain films | Tunability values are ∼75% at 200 kV cm−1 and 300 K in (Ba,Sr)TiO3–Sm2O3 VAN film which is ∼40% higher than for (Ba,Sr)TiO3 plain film.27 Also, loss scales with tunability, which is opposite to plain films. | |
Negative capacitance observed in Ba0.8Sr0.2TiO3/LaAlO3 and PbTiO3/SrTiO3 at 300 K28 | Negative capacitance not yet observed | |
Polar vortices observed in PbTiO3/SrTiO328b,29 | Polar vortices not yet observed | |
FM | In-plane exchange bias as high as 3 kOe induced in CaRuO3/CaMnO3 at 10 K30 | Perpendicular exchange bias of ∼0.91 kOe at room temperature obtained in NiO–NiFe2O4 VAN22cf. ∼0.1 kOe in NiO–NiFe2O4 bilayer at room temperature.31 |
High interfacial magnetization of up to 2.5μB (396 emu cm−3) achieved in LaFeO3/LaCrO332 | Saturation magnetisation in Fe3O4–BiFeO3 VAN film ∼900 emu cm−3cf. ∼300 emu cm−3 in pure Fe3O4 film with similiar thickness33 | |
High TC ferromagnetism of up to 375 K created at interfaces from non-magnetic parent materials in LaFeO3/LaCrO332 | Ferro/ferrimagnets with several-hundred-K TC created from AFM parent material ZnFe2O434 | |
Perpendicular magnetic anisotropy (Keff) of 4 × 106 erg cm−3 (∼13 kOe) realized in La1−xSrxMnO3/SrIrO3 at 10 K35 | Perpendicular magnetic anisotropy of 85 kOe induced in BaTiO3–CoFe2O4 at 300 K via strain36 | |
Topological Hall effect up to 75 μΩ cm observed in LaMnO3/SrIrO3 at 10 K37 | No topological Hall effect observed | |
DME coefficient αME up to 55 V cm−1 Oe−1 at 300 K observed in BaTiO3/BiFeO338 | DME coefficient αME up to 20 V cm−1 Oe−1 at 300 K measured in BiFeO3–CoFe2O439 | |
ME | No sizable CME yet observed | Self-biased CME effect at room temperature with ME coefficient of ∼1.4 × 10−9 s m−1 at 300 K achieved in Na0.5Bi0.5TiO3–NiO–NiFe2O440 |
The rest of the paper expands on the information in Table 2, discussing the different ferroic phenomena of ferroelectricity, ferromagnetism and multiferroicity observed in SL films (part A) and VAN films (part B), with potential synergies between the two film forms noted. Finally, an overall summary, perspective and conclusions are presented (part C).
Fig. 2 Ferroelectric phenomena in SL films. (a) P–E hysteresis loops measured from (NdMnO3)n/(SrMnO3)n/(LaMnO3)n superlattices. Reproduced with permission.41 Copyright 2012 Springer Nature. (b) FE TC in (BaTiO3)n/(SrTiO3)m superlattices as a function of n and m layer numbers. Reproduced with permission.23 Copyright 2006 AAAS. (c) Left panel: STEM image and diagram of (SrTiO3)x/(BaTiO3)y/(CaTiO3)z superlattices. Right panel: P–E hysteresis loops measured from (SrTiO3)2/(BaTiO3)4/(CaTiO3)2 and (SrTiO3)10/(BaTiO3)10/(CaTiO3)10 superlattices. Reproduced with permission.24 2005 Springer Nature. (d) Left: Schematic of crystal structure of (SrTiO3)n−1/(BaTiO3)1/SrO superlattices and STEM images the Ti, Ba, Sr and (Ba + Sr) elements. Middle: Real and imaginary parts of dielectric constant (K, the ratio of the material's permittivity ε to the permittivity of vacuum ε0) as a function of frequency for the n = 6 superlattice film. The inset shows the film loss tangent. Right: The ratio of K(Ebias, applied electrical field)/K(0, without applied electrical field) (left-hand axis) and relative tunability (right-hand axis) of the n = 6 superlattice film. Reproduced with permission.26 2020 Springer Nature. (e) Temperature dependent capacitance for superlattices grown on SrTiO3, DyScO3, and GdScO3 substrates. Reproduced with permission.28a Copyright 2014 American Chemical Society. (f) Left panel: STEM image of vortex structures formed in a (SrTiO3)10/(PbTiO3)10 superlattice. Reproduced with permission.29a 2016 Springer Nature. Right panel: Optical and thermal field control of the coexistence phase and superlattice phase. Reproduced with permission.42 2019 Springer Nature. |
No such emergent interface FE has been reported in VAN films to date. While it is certainly harder to grow such perfect vertical interfaces of desired compositions in VAN films, it is possible in some systems where the matrix film is highly similar in both structure and lattice parameter to the substrate.44 In fact, the effect has likely never been searched for, and is a fruitful avenue of future research, particularly in manganite-based VAN systems where mixed Mn valence is induced by the presence of oxygen vacancies. Indeed, interface structures in VAN systems predicted using DFT calculations show strongly distorted oxygen sublattices and higher oxygen vacancy concentrations.15b Hence, in a SrMnO3/CeO2 VAN film, interface-induced oxygen vacancies are anticipated. These would reduce some of the Mn4+ to Mn3+. Hence, it is possible that emergent FE could arise from this mixed valency in VAN films, just as in SL films. The signal would, in theory, be much higher than in SL films, owing to the higher density of interfaces. We note that a reduced overall cation valence at interface structures in VAN films would not be limited to manganites. Other transition metal oxides would also likely give mixed valent cation interfaces and hence wide-ranging novel interfacial properties. This is a rich area for future exploration.
Another proven route to inducing ferroelectricity in SLs is to induce octahedral rotations into a SL layer from another SL layer. This leads to hybrid improper ferroelectricity.45 It has been shown in perovskite SLs with the same B ions but different A (A and A′) ions, hence A/A′ SLs, e.g. LaGaO3/YGaO3. The octahedral rotations induce a sizeable polarization of 0.65 μC cm−2. This same effect would likely not be possible in VAN films, owing to the larger lateral pitch of the vertical structures, i.e. in the range 5–50 nm, rather than <1–3 nm as required here. We note also that ferroelectricity across the interfaces in the SLs is predicted to originate from charge transfer,46 charge order,47 and element intermixing.48 Similar effects are likely also taking place in VAN structures but, as far as known, have not been probed yet. This would be an interesting area of future research.
As we show later, the TC in VAN films is highly tunable, and, just as for SL films, several hundred K TC enhancements are readily achievable. The biggest differences with SLs is the ease of growth of VAN films combined with the ability to maintain vertical strain to > micron thicknesses,21,25 whereas in SLs the total film thicknesses are only several 10's unit cells maximum. VANs are therefore more suited to actuator/energy harvesting/dielectric applications where larger volumes of material are needed, whereas SL films are more suited to Si-based electronics where charges at planar interfaces play an important role in switching for processor or memory applications.
(SrTiO3)n−1(BaTiO3)1SrO SLs with n = 6 (single unit cell thick BaTiO3) (Fig. 2d) have lower dielectric loss than standard films. This is because the single unit cell thick BaTiO3 in the SL is under chemical pressure from the SrTiO3 layers. This likely reduces oxygen loss at the same time as increasing the tetragonality and hence also the ferroelectric polarization. Overall, at a temperature T = 300 K and an electric field E = 400 kV cm−1, a dielectric constant (K) tunability, i.e. the ratio of K(E)/K(0) (right panel in Fig. 2d), of around 46%, and a loss tangent (tanδ) of ∼0.06 were achieved at a 110 GHz frequency.26 Higher tunability values were achieved in much thicker (∼μm) VAN films (Section 6), i.e. dielectric constant tunability of up to 75% and tanδ of 0.01,27,54 although the frequency studied so far was lower than for SL films, at ∼1 MHz.
Another form of ferroelectricity in SLs has been demonstrated via improper ferroelectricity. In (PbTiO3)n(SrTiO3)m SLs with very short periods of n = 2 and m = 3, improper ferroelectricity55 is achieved resulting from strong competition between ferroelectric PbTiO339 and antiferrodistortive SrTiO356 distortions across the interface, causing antiferrodistortive rotations of the oxygen atoms in PbTiO3. In turn, a relatively high dielectric constant of 600 is achieved at 300 K and, from 300 K to 550 K, the dielectric constant is independent of temperature. No improper ferroelectricity has been observed in VAN films to date but there is the possibility to observe it in the most perfect VAN structures. The PbTiO3/SrTiO3 VAN system would be a good starting point. While phase intermixing between PbTiO3 and SrTiO3 could occur in the VAN films, there is precedent for lack of mixing between similarly lattice matched perovskite VAN materials.57 Thus for PbTiO3/SrTiO3 VAN, a similar situation could come about. Less (or no) mixing compared to the bulk thermodynamic case related of the role of epitaxy at the substrate film interface. Hence pure SrTiO3 on SrTiO3 is the lowest interfacial energy situation and may prevent intermixing with Pb. More experiments and calculations are highly required to confirm this possibility.
Another novel phenomenon observed recently in PbTiO3/SrTiO3 SLs is the formation of polar vortices in PbTiO3. In fact, ferroelectric vortex states were theoretically predicted in nanostructures,59 but not observed in thin films until 2016. Long-range and chiral ferroelectric vortex structures have been observed in PbTiO3 in (PbTiO3)n/(SrTiO3)n SLs with n = 10 (left in Fig. 2f).29a A combination of substrate induced strain and confinement from the SrTiO3 layers produces the effect leading to a clockwise vortex-like pattern, paired with every other vortex in an anti-clockwise manner. The coexistence phase of ferroelectric a1/a2 domains and vortex states can be transferred to a single 3D SL phase by optical and thermal manipulation (right in Fig. 2f).42 Room-temperature polar-skyrmion bubbles have further been observed in these SL systems,28b giving the possibility of highly efficient skyrmion memory.
There have been no such reports thus far of polar vortices in VAN films. However, the ∼5 nm length scales, strain and confinement effects involved in the above SL example, can also be readily achieved in VAN films, even in the PbTiO3/SrTiO3 system. Hence, PbTiO3 pillars should form in a SrTiO3 matrix when the film is grown on SrTiO3. The Pb and Sr would be unlikely to mix based on the differing ion sizes and charges, and also because it would not be preferred because pure SrTiO3 forms a perfect epitaxial match on SrTiO3. Thus, it may be possible to achieve few nm sized nanopillars of PbTiO3 as can be achieved in other VAN systems60 and thus create vertical polar skyrmions throughout the whole film thickness.
At SL interfaces, the following effects have been shown: the enhancement of TC, modified electronic conduction, induced ferromagnetism from non-ferromagnetic layers, exchange bias, magnetic anisotropy, and topological Hall effect, etc. (Fig. 3). In Section 7, we discuss the comparative magnetic effects in VAN films.
Fig. 3 Magnetic phenomena in superlattice films. (a) Magnetic hysteresis loops of (CaRuO3)3/(CaMnO3)N superlattices. Reproduced with permission.30 Copyright 2012 American Physical Society. The inset shows the temperature dependent HC and HEB for (CaRuO3)3/(CaMnO3)8. (b) Temperature dependent HE for (LaNiO3)7/(LaMnO3)7 superlattices after field-cooling in a +0.4 T field (green circles) and in a −0.4 T field (orange diamonds). Reproduced with permission.62 Copyright 2012 Springer Nature. (c) Schematic diagram of (LaFeO3)1/(LaCrO3)1 superlattices grown on (111)-oriented substrates and the resultant spin structures. Reproduced with permission.63 Copyright 2014 Annual Reviews. (d) Left panel: STEM image (top) and octahedral tilt angles (bottom) of (La0.9Ba0.1MnO3)8/(SrTiO3)8 superlattice. Right panel: Ferromagnetic TC as a function of layer thickness in (La0.9Ba0.1MnO3)n/(SrTiO3)n superlattices. Reproduced with permission.64 Copyright 2020 Wiley-VCH. (e) Magnetic anisotropy of (La0.7Sr0.3MnO3)3/(SrIrO3)n superlattices as a function of SrIrO3 layer thickness. Reproduced with permission.65 Copyright 2016 PNAS. (f) Left panel: Schematic diagram of the topological Hall effect originating from the Berry phase and a chiral spin texture. Right panel: The topological Hall resistivity of the (LaMnO3)5/(SrIrO3)5 superlattice. Reproduced with permission.37 Copyright 2020 AAAS. |
Exchange bias is also achieved in (111)-oriented LaMnO3/LaNiO3 (ferromagnetic/paramagnetic) SLs (Fig. 3b). Here, interfacial interactions, e.g. quantum confinement effects,68 induce a complex magnetic structure in a non-magnetic material. Hence, magnetism resembling a spin density wave in the paramagnetic LaNiO3 layers in contact with the ferromagnetic LaMnO3 is created.62 An exchange bias as high as 90 Oe is induced at 5 K. Large vertical exchange bias effects have been demonstrated at much higher temperatures in VAN films, namely at room temperature, and the reasons for this are discussed in Section 7.
Similar to the (CaRuO3)3/(CaMnO3)n SLs, in (LaMnO3)2n/(SrMnO3)n SL, a metallic ferromagnetic state emerges at the interface from the parent materials which are not ferromagnetic, namely an antiferromagnetic Mott insulator (LaMnO3),69 and a band insulator with G-type antiferromagnetism (SrMnO3).70 Again, interfacial charge transfer between Mn3+ and Mn4+ cations is responsible for emergent FM, i.e. electrons ‘leak’ from the Mn3+ in LaMnO3 into the Mn4+ in SrMnO3.70,71 The metallic state is apparent for n ≤ 2 (when the occupied Mn eg electrons spread over the entire lattice), but for higher n, the bulk of the film (insulating antiferromagnetic) dominates the properties.
Emergent ferromagnetism and metallicity should also be easily achievable in VAN films, as charge transfer across interfaces is not restricted to planar interfaces. To date, in several VAN systems interesting additional magnetic and conducting effects have been observed (separate to the expected contributions from the bulk properties to the film).15a,72 However, the origins of these effects, whether they be from charge transfer effects, orbital reconstructions, etc., have not been widely explored. Further studies of such effects in perfectly aligned VAN structures, including understanding of the phenomena using high resolution microscopic and scanning probe tools to probe the structures and interface charge states is a fruitful area of future research. Also, since the vertical interfaces in VAN films intersect the film surface, the conducting properties of VAN interfaces are readily measurable giving an advantage over SL where interfaces are buried.73
Another mechanism to induce ferromagnetism at interfaces is by tuning the relative fractions of B–O–B, B–O–B′, and B′–O–B′ bonds, In (ABO3)/(AB′O3) SLs, there are different magnetic exchange interactions couplings across the interfaces for different orientation films. Hence, different spin structures have been observed for (LaFeO3)1/(LaCrO3)1 SLs grown on (001)-, (110)-, and (111)-oriented SrTiO3 substrates.32a,74 As shown in Fig. 3c, based on Anderson, Goodenough, and Kanamori rules, ferromagnetic exchange interactions across Fe–O–Cr bonds can be realized in the (111)-oriented SL films. A high TC of 375 K and high interfacial magnetization of 2.5μB were achieved in (111)-oriented (LaFeO3)1/(LaCrO3)1 SL.
A further interfacial phenomenon which induces modified interfacial properties in magnetic perovskite SLs is orbital reconstruction. This occurs when TMO thin films are strained leading to different BO6 octahedral rotations.61,64 For example, the ferromagnetic TC in La2/3Sr1/3MnO3 is significantly enhanced from 370 K in bulk La2/3Sr1/3MnO3 to 650 K in (La2/3Sr1/3MnO3)3/(BaTiO3)3 SLs owing to straining of the La2/3Sr1/3MnO3 by BaTiO3.75 The change of the Mn orbital occupation from dx2−y2 within the film to d3z2−r2 at the interface is responsible.
A further mechanism for tuning the magnetic coupling within SL layers containing magnetic moments is via non-trivial oxygen octahedral coupling (OOC) across the heterointerfaces which produces different BO6 octahedral rotations in the layer of interest.61c,76 For instance, in non-magnetic (LaMnO3)n/(SrTiO3)n SLs, magnetisation was induced in the interfacial layers in LaMnO3 with a maximal saturation magnetization of 2.97μB per Mn for the n = 6 SL film.77 It should be noted that while interactions between layers in the films lead to the modulation of BO6 octahedral rotations, epitaxial strain also play an important role. To largely eliminate the strain effect, (La0.9Ba0.1MnO3)n/(SrTiO3)n SLs were studied,64 where there is only 0.2% strain between the La0.9Ba0.1MnO3 and SrTiO3. It was possible to engineer the BO6 octahedral rotations up to 12 unit cells, larger than previous reports by 6 unit cells. Consequently, the out-of-plane Mn d3z2−r2 orbitals became occupied, the Mn eg bandwidth reduced, a ferromagnetic insulating state in La0.9Ba0.1MnO3 created, and TC enhanced to 235 K from 185 K in bulk (Fig. 3d).
As already noted, in SL films the substrate plays a role in controlling the perovskite octahedral rotations and oxygen octahedral coupling. The substrate exerts strain on the film which induces octahedral rotations in the film, but there is also coupling of the film octahedra to substrate octahedra. The question is whether similar interface coupling effects could be created in VAN films but without the interference from the substrate as is the case of SL films. In fact, in VAN the substrate influence is weak above around 20 nm film thickness and since micron thick films are relatively easy to grow and indeed preferred for VAN, any coupling effects across the vertical interfaces in VAN are largely independent of the substrate. On the other hand, as shown in Table 1, the challenge for VAN is creating highly faceted interfaces which are also perpendicular to the substrate without meandering. As already noted, this is highly possible when the matrix material is the same as the same or a very closely lattice and structurally matched material as the substrate material, e.g. SrTiO3-based VAN film on SrTiO3 substrate,54 and hence there is wide scope for exploring new interface coupling effects in VAN films with the most ideal interfaces.
Broken inversion symmetry and strong spin–orbit coupling in 3d–5d interfaces can also lead to large chiral Dzyaloshinskii–Moriya interactions (DMIs), inducing non-trivial spin textures, i.e. topological phenomena. In LaMnO3/SrIrO3 SLs (Fig. 3f), a highly robust chiral magnetic phase with large topological Hall effect (THE) of 75 μΩ cm was observed.37 The present research on the interfaces with iridate oxide systems is still in infancy, and many open and intriguing questions still remain to be explored.
Finally, we note that the skyrmion structures formed in SrIrO3 superlattices incorporate insulating or metallic layers are formed in films of only 2–8 unit cells (or 1–4 nm) thickness.37,80 Thus, as mentioned for the polar skyrmions in Section 2.4, it is highly possible to grow VAN structures of this dimension and potentially grow, by self-assembly, skyrmion SrIrO3 columns in a film matrix, the route for growing perovskite nanocolumns in VAN films having already been demonstrated in several systems, e.g. BiFeO3, BaTiO3, and Ba1−xSrxTiO3 systems.25 Indeed, it is possible that SrIrO3 would grow more readily in VAN films compared to plain epitaxial films. This is because the phase could be stabilised using vertical epitaxy with a matrix phase in the VAN film.19b Also, to maintain stoichiometry of the SrIrO3, a matrix material containing either or both Sr and Ir species could be used so as to produce a source material to prevent loss of these volatile cations during growth.
Two key examples of multiferroism induced in SLs are in La2/3Sr1/3MnO3/BaTiO3 (non-polar-tilt ferromagnet/polar ferroelectric) and La2/3Sr1/3MnO3/BiFeO3 (non-polar-tilt ferromagnet/ferroelectric antiferromagnet). In the first case (Fig. 4a), due to proximity effects a structural phase transition from non-polar to unconventional polar (hence giving both ferroelectricity and ferromagnetism, thus multiferroicity) was induced in ultrathin 4-unit-cell La2/3Sr1/3MnO3.83 In the second case, magnetism (up to 200 K) was induced in ultrathin 5-unit-cell ferroelectric BiFeO3,87 to a level of 1.83 ± 0.16μB per Fe, increased from 0.03μB per Fe in bulk BiFeO3, and attributed to strong orbital reconstruction between Fe and Mn across the interfaces.
Fig. 4 Multiferroic phenomena in SL films. (a) Top panel: STEM images of (La2/3Sr1/3MnO3)N/(BaTiO3)21 superlattices. Bottom panel: M vs. H loops (left) and M vs. T curves (right) for (La2/3Sr1/3MnO3)N/(BaTiO3)21 superlattices. Reproduced with permission.83 Copyright 2017 PNAS. (b) Top panel: A tricolor superlattice for designing the multiferroicity. Reproduced with permission.84 Copyright 2007 Elsevier. Bottom panel: Temperature dependent polarization measured under electrical and magnetic field for NdMnO3/SrMnO3/LaMnO3 superlattices. Reproduced with permission.41 Copyright 2012 Springer Nature. (c) Top panel: STEM images of end-members LuFe2O4 (left) and LuFeO3 (right); middle panel: (LuFeO3)9/(LuFe2O4)1 superlattice. Bottom panel: XMCD PEEM images of (LuFeO3)9/(LuFe2O4)1 superlattice measured at 200 K and 320 K. Reproduced with permission.85 Copyright 2016 Springer Nature. (d) Top panel: Schematic of growth and transfer of a freestanding thin films. Middle panel: STEM images of 3-unit-cell BiFeO3 film before and after etching Sr3Al2O6 layer. Bottom panel: Piezoelectric properties of 4-unit-cell BiFeO3 film after etching Sr3Al2O6 layer. Reproduced with permission.86 Copyright 2019 Springer Nature. |
The creation of ferroelectricity, multiferroicity, and a DME effect has been achieved through simultaneous breaking of time-reversal and spatial-inversion symmetries in “tricolour” ABCABC SLs composed of three sequential kinds of perovskite layers88 (top plot in Fig. 4b). The polarization at the AB (or BC) interfaces has the same sign leading a non-vanishing toroidal moment equivalent to a built-in vector potential under the presence of spin–orbit coupling, thereby producing ME coupling. For example, a ferroelectric TC of 40 K was achieved in NdMnO3/SrMnO3/LaMnO3 SLs, even though individual layers have no polar moment. Also, ME coupling was achieved by magnetic modulation of the polarization (bottom plot in Fig. 4b).41 Density functional calculations show that the combination of broken space inversion symmetry and the presence of Mn3+/Mn4+ mixed valence, arising from cationic asymmetry and interfacial polar discontinuity, produce emergent ferroelectric and ME coupling. However, the interface-induced ME coupling is very weak and occurs below 55 K.
Other works on creating magnetoelectricity in SLs have centred around a wide range of SL combinations, e.g. in BaTiO3/BiFeO3 (ferroelectric/ferroelectric and antiferromagnet) SLs a direct ME coefficient αME up to 55 V cm−1 Oe−1 at 300 K was measured,38b higher than in single-phase BiFeO3 (αME = 4.2 V cm−1 Oe−1), which was explained by the diffusion of Ba and Ti into the BiFeO3 layers. Hence, the influence of unwanted chemical intermixing across the interface and impurities buried within superlattices is uncertain.
Another different mechanism to achieve multiferroicity and ME coupling was demonstrated using symmetry breaking across a SL interface of (LuFeO3)m/(LuFe2O4)1. Rumpling (periodic buckling of lattice planes) in LuFeO389 imposed a local distortion in one unit cell LuFe2O4, thus removing the mirror symmetry in the LuFe2O4 and enabling ferroelectricity to emerge in this ferrimagnetic phase85 (Fig. 4c). The ferrimagnetic TC of the LuFe2O4 was increased from 240 K to 281 K because of a local distortion applied by LuFeO3. Since the ferroelectric order couples to the ferrimagnetism, electrical field control of magnetism, as measured using X-ray magnetic circular dichroic photoemission electron microscopy (XCMD PEEM), was achieved at 200 K (Fig. 4c). The clamping of the films to the substrate or leakage effects possibly limit macroscopic measurements using electrodes applied across the whole film.
Finally, we note that there has been more success in achieving direct ME (DME) in all-metallic systems90 than in oxides. An example is (Ta/Cu/IrMn/FeCoSiB)n/AlN/Si,90c where a relatively large direct ME effect of 96 V cm−1 Oe−1 was observed at 300 K. However, the requirement to use precious metal Ir, the need to use a relatively complex stack of layers, and no observation of CME, are limiting factors for this system. Hence, arguably, there is more promise for oxide systems, although controlled interface composition and structure, and the elimination of substrate clamping effects are important areas of future study. The latter effect prevents intrinsic displacement of ferroelectric and magnetostrictive materials.
Very promisingly, in recent years much progress has been made with lift-off and transfer of films to practical substrates.86,91 As far as known, the transfer process has not been applied to multiferroic SL films but it has been used for single layer films and it is expected that multiferroic SL films would behave similarly. The advantage of lifting off a multiferroic SL film is that it would be free from substrate clamping. This could enable strain coupled magnetoelectricity to be realised. Fig. 4d shows an example of using water-soluble Sr3Al2O6 as the sacrificial buffer layer to produce freestanding BiFeO3 multiferroic thin films with a high crystalline quality that can be fabricated down to one unit cell. Also, robust and switchable ferroelectric polarization was demonstrated for a four unit-cell, transferred BiFeO3 film on Si, as confirmed by STEM and piezoelectric measurements, respectively. We note that VAN films have also been successfully lifted-off and transferred to Si. Owing the nanoscale nature of VAN films, they appear to be more resistant to crack propagation yielding a higher success rate of film transfer compared to plain films.92 More work is needed to explore the performance of lifted-off ferroic and multiferroic VAN films.
Contrasting with much of the SL work describe above, artificial multiferroic effects in VAN films have not focused on purely interfacial phenomena, but on more macroscopic effects caused by strain coupling of phenomena across vertical interfaces which take effect in the whole body of film. An advantage of VAN films over SL is the lack of a substrate clamping effect. On the other hand, leakage currents can be problematic if the right combinations of materials are not used. Strategies for overcoming the leakage while also boosting interface coupling are discussed for VAN multiferroics in Section 8.
We recall that in standard planar SL ferroelectric films, epitaxial strain from the underlying substrate is often used to tune properties.1e,24,93 However, this substrate-induced interfacial strain can only extend to tens of nanometers thickness and hence the TC enhancement is largely confined to the interfacial region. It has been shown in epitaxial perovskites that from above the first tens of nanometers thickness up to ∼100 nm, the strain is relieved by a range of crystalline defects to compensate the strain.94 Such defects can degrade the functionality of epitaxial perovskite devices which exploit the functionality of the whole film rather than just the interface, e.g. a strain-enhanced-TC ferroelectric actuator. In VAN films, on the other hand, by incorporating stiff nanopillars into a ferroelectric matrix large vertical strains (of >1%) can be induced throughout the whole film and hence the thickness-limiting substrate strain effect is not relied upon.21 This enables wide-ranging ferroelectric property tuning in micron thick films. Since the lateral distance between the nanopillars is typically only ∼20 nm, the strain is not relaxed to a large extent as for epitaxial films, and so the film perfection and crystallinity remains high for 100s nm thick VAN films.13,95 Hence, enhanced TC, remnant polarization, retention, dielectric tunability, and reduced loss tangent and leakage current are achieved.25,27,54,96
In section A, we presented several examples of VAN compositions that could be explored to try and mimic and also new opportunities the exciting emergent effects achieved in SL films. In the following sections we focus on VAN films. Before, we go into details of each of the enhanced properties, explaining how the strain in VAN films contributes to these properties, we first consider how research on SLs could benefit from VAN research. A key processing advantage of the VAN approach over SL is it is relatively fast and simple for exploring new combinations of materials and their interfaces. Hence, VAN can serve as fast screening method to predict which phases and which ratios of phases could also give rise to novel interface related functionalities in SL structures. For example, in Section 8, we show how a triple VAN system enables a self-biased magnetoelectric to be achieved. For SL films, it would take longer to explore the right thicknesses of the layers to use, to be sure there are no interface mixing effects, and also to know if the magnetoelectric effect were present (as it could obscured by substrate clamping effects). Hence, with the knowledge gained from VAN, SL films of the demonstrated compositions could now be tested, and films lifted off to remove any clamping effects.
Fig. 5 Enhanced ferroelectric properties achieved using VAN structures. (a) Cross-sectional TEM image of SrTiO3–Sm2O3 VAN; (b) remnant polarization of (Ba,Sr)TiO3–Sm2O3 VAN film as a function of the tetragonality of (Ba,Sr)TiO3 phase. Reproduced with permission.25 Copyright 2017 Wiley-VCH. (c) Giant piezoelectric Na-deficient NaNbO3–NaNbO3 (NPR-NNO) VAN film. Reproduced with permission.97 Copyright 2020 AAAS. (d) Comparison of normalized retained polarization versus temperature for a pure (Ba,Sr)TiO3 film and for (Ba,Sr)TiO3–Sm2O3 VAN films with different Sm2O3 contents (x). Reproduced with permission.25 Copyright 2017 Wiley-VCH. (e) The evolution of polarisation retention in a FE domain of a BiFeO3 (111) mesocrystal with time. Reproduced with permission.96b Copyright 2016 Springer Nature. (f) High tunable relative permittivity, εr, as shown by the large change in capacitance value with electric field in a (Ba,Sr)TiO3–Sm2O3 VAN film, measured at 1 MHz. An enhanced commutation quality factor (CQF) is also achieved. Reproduced with permission.27 Copyright 2012 American Chemical Society. (g) Reduced leakage current of a BiFeO3–Sm2O3 VAN film compared to plain BiFeO3 or plain Sm2O3 films. Reproduced with permission.96a Copyright 2008 AIP Publishing. |
As shown in Table 2, similar ferroelectric property enhancements are only achievable in planar or SL films if they are ∼10's nm thick or less.93a Such high TCs in micron thick films are necessary for advanced device applications, such as the electrocaloric devices,101 actuators,102 and microwave tuneable filters,53a and cannot be achieved in planar films.
The remnant polarization (Pr) has also been systematically increased using (Ba,Sr)TiO3-based VAN structures (Fig. 5b), both by increasing the fraction, x, of nanopillars and increasing film thickness (to ∼μm's),25 both of which increase the film tetragonality. The Pr values are about ∼3× larger compared to (Ba,Sr)TiO3 planar films (Table 1). A near doubling of Pr has also been shown in BaTiO3–NiO VAN films.103 A giant enhancement of piezoelectric response is also possible in VAN films (Fig. 5c). Liu et al.97 showed in nanopillar-like sodium-deficient NaNbO3 VAN a piezoelectric coefficient d33 of ∼1098 pm V−1, which compares to ∼22 pm V−1 for stoichiometric NaNbO3 films. The TC of the VAN film is ∼723 K, which is similar to that of the stoichiometric NaNbO3 film. Local heterogeneity of sodium-deficient NaNbO3 and stoichiometric NaNbO3 phases was deemed to be responsible.
Temperature-dependent and time-dependent changes of the polarization state of ferroelectric films, i.e. thermal and temporal retention, are of great importance for non-volatile nanoelectronic devices based on ferroelectrics,25,96b,104e.g. for above-room temperature-operation, high-stability memory devices. The retention parameters are sensitive to film thickness and microstructure, electrodes, etc.105 Much enhanced thermal and temporal polarization retention has been demonstrated in (Ba,Sr)TiO3–Sm2O3 and CoFe2O4–BiFeO3 VAN films (Fig. 5d and e) compared to the pure film counterparts of (Ba,Sr)TiO3 and BiFeO3, respectively. Ultra-dense vertical interfaces lead to a strong tetragonal distortion and vertical ferroelectric domains25,96b which are ‘pinned’, thus retarding relaxation processes.
Leakage is a long-standing problem in ferroelectric oxide thin films, commonly caused by oxygen deficiency and the presence of point or line defects. As already mentioned above, leakage can be reduced using VAN films. Taking the example of BiFeO3, a promising candidate for multifunctional non-volatile memory devices because of the coexistence of both the antiferromagnetic order with Néel temperature (TN) of 643 K106 and ferroelectric order with TC of 1123 K,107 leakage from a range of defects is common and can give either p-type or n-type behaviour.108 In BiFeO3-based VAN films, the leakage was reduced by several orders (Fig. 5g).72a,96a This was ascribed to the compressive out-of-plane (OOP) strain of −1.46% that was induced in the BiFeO3, induced by the embedded nanopillars. It was postulated that such a compression would reduce the oxygen vacancy concentration and hence carriers associated with these vacancies.109 However, some chemical intermixing, leading to doping of Sm into the BiFeO3 could also contribute to the lowering of the leakage. Of course, not all VAN films can have judicious compression engineered into them as for this BiFeO3 example.
One example of the benefits of mesoporous VAN is the doped/undoped SrTiO3–MgO VAN film system. Here, the MgO is chemically etched out of the film, leaving a mesoporous SrTiO3 matrix (Fig. 6a) with the out-of-plane tensile strain in the SrTiO3 matrix largely intact. More understanding of the mechanisms and universality of matrix strain retention after etching pillars out of the VAN films is required. This mesoporous matrix leads to photoelectrochemical (PEC) water splitting compared to literature (Fig. 6b and c),110 arising from (a) an order of magnitude larger surface area for light absorption, (b) optimised carrier concentration from Ag and Nb of the doped SrTiO3,111 and (c) above-room-temperature ferroelectricity from the vertical strain effect induced by the VAN structure, which produced a reduced flat-band potential and enhanced charge extraction (Fig. 6d–f)110,111
Fig. 6 Doped/undoped SrTiO3-based VAN porous ferroelectric films for water splitting applications. (a) SEM image of the surface of a undoped mesoporous SrTiO3 film. Reproduced with permission.110 Copyright 2016 American Chemical Society. (b) Schematic of the water-splitting by the undoped mesoporous SrTiO3 film. Reproduced with permission.110 Copyright 2016 American Chemical Society. (c) Enhanced incident photon to current efficiency (IPCE) in the undoped mesoporous SrTiO3 film compared to both a continuous SrTiO3 film and the Nb-doped SrTiO3 substrate. Reproduced with permission.110 Copyright 2016 American Chemical Society. (d) The photoelectrochemical (PEC) response of a mesoporous (Ag,Nb)-co-doped SrTiO3 film at different polarization states. Reproduced with permission.111 Copyright 2019 American Chemical Society. (e and f) Schematic of the modified band alignment at the interfaces in the mesoporous (Ag, Nb)-co-doped SrTiO3 film and electrolyte solution with different polarization states and the impact on the PEC. Reproduced with permission.111 Copyright 2019 American Chemical Society. |
Another area which has been unexplored so far is the filling of epitaxial mesoporous films with different phases, e.g. using atomic layer deposition or solution processing. This can produce new composites with metallic or polymeric phases which can couple to the mesoporous matrix either to achieve multifunctional properties, or to simply exploit large surface areas and reduced dimensionality (important for strong light absorption and high charge extraction in optoelectronic systems, for example). These areas remain largely unexplored but there is strong potential for new multifunctional systems to emerge.
A key benefit of VAN over SL resides in the potential for high density magnetic memory and storage applications. Radically new approaches are required to increase magnetic recording densities beyond 1 Tbit in−2 and to lower energy consumption.113 While self-assembled magnetic nanoparticles, such as the self-assembled FePt nanoparticle arrays shows a high area density up to 50 Tbit in−2 for <4 nm particle sizes,114 difficulties exist for anchoring and ordering these nanoparticles. Self-assembled VAN films, with ∼Tbit in−2 density (∼10 nm sized features with ∼10 nm spacing) represent a promising approach. The vertical alignment of magnetic nanocolumns anchored in a non-magnetic matrix gives exchange decoupled, well-separated perpendicular magnetic recording bits. A range of other enhanced magnetic properties in VAN films also have potential benefit for magnetic storage/memory applications, as outlined below.
Similar magnetisation enhancement was found in Pr0.5Ba0.5MnO3–CeO2 VAN films115 due to the enhanced tetragonality in Pr0.5Ba0.5MnO3 from the stiff CeO2 nanopillars, although other structural effects could be at play. However, in La0.7Sr0.3MnO3–MgO VAN films, the magnetization was suppressed, possibly because of octahedral tilting/rotations.116
Antiferromagnetic-to-ferromagnetic property switching has also been demonstrated in different VAN systems, e.g. in spinel ZnFe2O4 based VAN films,34 a TC of ∼500 K was achieved with 4.8μB per Fe moment, and moderate coercivity (Fig. 7a (bottom panel)), overwhelmingly larger than the TN (∼10 K) of bulk ZnFe2O4 (Fig. 7a) (top panel). The VAN films induced stronger intersite A–B exchange interactions of the Fe3+ cations than the intrasite interactions at the A- or B-sites. Mössbauer spectra (Fig. 7b) showed that half of the Fe3+ ions go into the A sites in the ZnFe2O4 phase. This is similar to the case for nanocrystalline powder form of ZnFe2O4.117
Fig. 7 Enhanced magnetic properties achieved using VAN structures. (a) The magnetic hysteresis loops of a ZnFe2O4 plain film and a ZnFe2O4–SrTiO3 VAN film grown on a SrTiO3 (001) substrate, measured at different temperatures. The magnetic field was applied in the OOP direction. Reproduced with permission.34 Copyright 2018 Wiley-VCH. The insets show schematics of a plain ZnFe2O4 film and a ZnFe2O4–SrTiO3 VAN film, respectively. (b) Mössbauer spectra of the Fe element in the ZnFe2O4 plain film and the ZnFe2O4–SrTiO3 VAN film. The insets show schematics of the magnetic structures of the ZnFe2O4 phases. Reproduced with permission.34 Copyright 2018 Wiley-VCH. (c) Tunable magnetic anisotropy and coercivity in BiFeO3–CoFe2O4 VAN films grown on SrTiO3 substrates with magnetic field applied either in-plane (IP) or OOP. Reproduced with permission.120 Copyright 2013 American Chemical Society. (d) Enhanced perpendicular exchange bias effect in a NiO–NiFe2O4 VAN film with magnetic field applied OOP. Reproduced with permission.22 Copyright 2018 American Chemical Society. (e) Temperature dependence of the resistivity measured at different magnetic fields applied OOP. Reproduced with permission.121 Copyright 2019 AIP Publishing. (f) Plan-view (top panel) and cross-sectional (bottom panel) high resolution TEM images of a CeO2–Ni VAN film. Reproduced with permission.122 Copyright 2016 Royal Society of Chemistry. |
Similar antiferromagnetic-to-ferromagnetic property switching has also been also observed in other VAN films, e.g. SmMnO3–Bi2O3,96d (Sr,Sm)MnO3–Sm2O3,118 and EuTiO3–Eu2O3.119 In all cases, large vertical strains were induced, e.g. in the latter case the value was ∼3.15% in the EuTiO3 phase. The strain decreases the Eu–Ti–Eu bond angle along 〈111〉 up to 1°, leading to a weakening of the antiferromagnetic interactions and switching from antiferromagnetic to ferromagnetic behaviour.
In VAN films the total anisotropy is controlled by three kinds of magnetic anisotropy, i.e. magnetocrystalline, shape and magnetostrictive. While magnetocrystalline anisotropy is an inherent property of the material, shape and magnetostrictive anisotropy are highly tunable in VAN structures. Because of the large aspect ratio of the magnetic nanopillars which grow in VAN films, a large shape anisotropy is produced with an easy axis along the OOP direction, producing the PMA. A shape anisotropy field Hshape = 2.0 kOe was estimated in the BaTiO3–CoFe2O4 VAN films, where the aspect ratio of magnetic pillars was around 10.36 The shape anisotropy is controlled by substrate orientation (as well as other growth parameters). The substrate orientation effect on the magnetic anisotropy and coercivity has been demonstrated for CoFe2O4 pillars in BiFeO3–CoFe2O4 VAN films (Fig. 7c) as well as also for other VAN systems.120,128 Since BiFeO3 is an antiferromagnet, it has been proposed that spin-flop exchange coupling between the compensated BiFeO3 and the CoFe2O4 spins at the interfaces might contribute to the observed PMA in BiFeO3–CoFe2O4 VAN films.129
Strain-induced stress anisotropy in VAN films also plays a significant role via the magnetoelastic effect. Here, the magnetoelastic energy depends on both the magnetostriction coefficient of the magnetic material as well the strain in it. CoFe2O4 has a large magnetostrictive coefficient with a large strain-induced magnetoelastic effect. In BaTiO3–CoFe2O4 VAN films, a compressive strain in the CoFe2O4 columns of ∼0.8% is achieved, with corresponding anisotropy field of Hstress = 34 kOe, matching the coercivity of the measured magnetic hysteresis.130 The magnetostrictive and magnetocrystalline anisotropies have been systematically tuned by varying the chemical composition of the CoxNi1−xFe2O4 column phases in the CoxNi1−xFe2O4–BiFeO3 VAN system.131 Large tunability of magnetic anisotropy and coercivity has also been achieved in La0.7Sr0.3MnO3–MgO VAN, where a switch from the easy-plane magnetic anisotropy to PMA occurs as the volume ratio of MgO increases, thus leading to different strain values in the La0.7Sr0.3MnO3 matrix.60a
Potential systems for high-density perpendicular recording media have also been shown in ferromagnetic metal pillar-oxide VAN systems.132 Favourable properties such as ultra-high bit densities of ∼Tb in−2 (Fig. 7f shows an example for a CeO2–Ni film),133 uniaxial anisotropy, strong OOP magnetization, high coercivities, anisotropic electric conductivity,132b and higher TCs and lower intrinsic resistances than ferromagnetic oxides133,134 have all been demonstrated.
The first trial PEB system was a (BiFeO3)0.9–(Fe3O4)0.1 composite film (Fe3O4 nanoparticles in a BiFeO3 film). A PEB field of HE ∼ 0.04 kOe was obtained at room temperature.112b Later, larger PEB values were obtained in LaFeO3–La0.7Sr0.3MnO3 and BiFeO3–La0.7Sr0.3MnO3 VAN systems with EB fields up to ∼1 kOe,112c,d although this was achieved at a relatively low temperatures (<10 K). Most recently, a large PEB of ∼0.91 kOe at room temperature was obtained in the NiO–NiFe2O4 VAN system (Fig. 7d).22 With the benefits of both high TC and high TN of the FM and AF materials as well as TC > TN, the blocking temperature (TB) in this system reached above 400 K.
A strong advantage of VAN artificial multiferroics over SL are minimal clamping from the underlying substrate. Clamping prevents effective vertical strain and strain coupling of a ferroelectric with a magnetostrictive material to produce a change in magnetisation with applied voltage to give the converse magnetoelectric effect (CME), or change in polarisation with change in magnetisation to give the direct magnetoelectric effect (DME).
BaTiO3–CoFe2O4 was the first VAN film explored for achieving the ME effect.130 However, owing to poor electrical insulation of most ferromagnetic materials in multiferroic nanocomposites, only thermally-driven ME coupling was demonstrated. On the other hand, it has been shown in many different VAN systems e.g. BaTiO3–CoFe2O4,130 Ni0.5Zn0.5Fe2O4–BaTiO3136 and Y3Fe5O12–BaTiO3137 (Fig. 8a). However, thermal ME provides only indirect evidence of ME coupling, as the strain is not induced by an electric field.
Fig. 8 Enhanced magnetoelectric properties achieved using VAN structures. (a) Thermally-driven ME effect in Y3Fe5O12–BaTiO3 VAN films detected by the FMR. Reproduced with permission.137 Copyright 2017 American Chemical Society. (b) XLD asymmetry study of the lattice distortion induced by magnetic field in a BaTiO3–CoFe2O4 VAN film, the upper plot showing magnetic field applied OOP and the lower plot showing the magnetic field applied IP. Reproduced with permission.138a Copyright 2013 Springer Nature. (c) The effect of the application of IP magnetic field on the amplitude–voltage butterfly loop (DME effect) in BaTiO3–CoFe2O4 VAN films. Reproduced with permission.138b Copyright 2019 Wiley-VCH. (d and e) Charge mediated ME to give the CME effect in La2CoMnO6–ZnO VAN film. Reproduced with permission.142 Copyright 2013 American Chemical Society. (d) Cross-sectional (false-colour) TEM image of the La2CoMnO6–ZnO VAN film. (e) Electric field dependence of the magnetic hysteresis loops of La2CoMnO6–ZnO VAN film where the magnetic field is applied IP. (f) In situ electric field control of magnetization (CME effect) in Na0.5Bi0.5TiO3–CoFe2O4 VAN film where the magnetic field is applied OOP. Reproduced with permission.143 Copyright 2018 American Chemical Society. The inset shows a schematic of the application of an in situ electric field. (g) The magnetoelectric coefficient of the Na0.5Bi0.5TiO3–CoFe2O4 VAN film calculated from the magnetic hysteresis loops given in (f). (h) Cross-sectional TEM and EDX image of a 50 nm-thick Na0.5Bi0.5TiO3–NiO–NiFe2O4 film. Reproduced with permission.40 Copyright 2021 Springer Nature. (i) In situ electric field control of magnetization (CME effect) in a 200 nm-thick Na0.5Bi0.5TiO3–NiO–NiFe2O4 VAN film where the magnetic field is applied OOP. The inset shows the magnetoelectric coefficient, calculated from the magnetic hysteresis loops. Reproduced with permission.40 Copyright 2021 Springer Nature. |
In terms of DME, the BaTiO3–CoFe2O4 system is a key exemplar system.138 Here, magnetic field induced strain has been observed using different methods, X-ray linear dichroism (XLD) (Fig. 8b),138a second-harmonic generation,138b and by measuring the modified PFM amplitude upon applying a magnetic field along both IP and OOP directions (Fig. 8c).138b,139 However, possibly owing to leakage problems, a direct electric readout of DME has not been reported.
The first direct measurement of the CME effect was shown in BiFeO3–CoFe2O4 VAN.140 Here, magnetic hysteresis loops were measured to detect the change of magnetization upon applying ex situ electric field pulses. However, this system suffered leakage as revealed by conductive atomic force microscopy (c-AFM) and cross-sectional scanning tunnelling microscopy (XSTM).15a,141
In La2CoMnO6–ZnO VAN films, in situ measurements showed the existence of the CME effect as well as tunability of magnetization and coercivity (Fig. 8d and e).142 It was determined that voltage-controlled charge trapping-detrapping in the ZnO at the high-density vertical interfaces led to a modified transition metal ion valence of Mn or Co in the La2CoMnO6 and hence to a change in the magnetic state of the ferrimagnetic La2CoMnO6. However, the low TC of La2CoMnO6 (<210 K) in the thin films meant that the CME was observed only at a relatively low temperature (120 K).
The first direct measurement of a large CME effect at room temperature was observed in a VAN system of Na0.5Bi0.5TiO3–CoFe2O4. By using the wide bandgap (3.3 eV), high TC (603 K) Na0.5Bi0.5TiO3 ferroelectric instead of the commonly studied, lower TC and lower bandgap BaTiO3 and BiFeO3 ferroelectrics, a large-scale in situ electric field control of the magnetic anisotropy was obtained at room temperature (Fig. 8f).143 The relatively low leakage was linked to (a) the lower leakage ferroelectric, (b) reduction of VAN vertical interface leakage which can arise when mixed valence transition metal cations are present in the system, and most importantly (c) the formation of a rectifying interface between the film and the substrate, enabling current blockage. An CME coefficient of around 1.25 × 10−9 S m−1, at a bias field of ∼0.3 T, was achieved (Fig. 8g).
Taking the above result one step further, a function-oriented composition design approach was used to reduce leakage even more. Hence, an antiferromagnetic (AFM) phase was incorporated into a similar VAN composition as discussed above,40 making a triple composite system. Here, a triple VAN 3-1 structure of FE-AFM-FM Na0.5Bi0.5TiO3–NiO–NiFe2O4 was created (Fig. 8h). We note that AFM materials have so far played an important role in traditional spintronic devices as pinning layers,144 and as emerging AFM spintronic devices, offering ultrafast (∼THz) dynamics, high stability to external fields, and absence of stray fields.145 However, they have largely been overlooked in multiferroic VANs. In the triple VAN system, the high TC (603 K) of Na0.5Bi0.5TiO3 was again exploited, as above. But in addition, the soft magnetism of NiFe2O4, the high Néel temperature (525 K) of NiO, and the large exchange coupling between the FM and AFM material was also capitalised upon.
A large, self-biased (zero applied magnetic field) CME effect at room temperature with a ME coefficient of up to 1.38 × 10−9 s m−1 was achieved (Fig. 8i). The self-biased CME effect is ascribed to 3 critical features introduced by the addition of NiO: the leakage is strongly suppressed by the introduction of p-type NiO; the antiferromagnetic spins are rotated by the strain due to the large magnetostrictive effect of NiO, as indicated by the electric controllable EB effect; and the exchange coupling between the AFM NiO spins and the FM NiFe2O4 spins enhance the overall ME effect. This one-shot growth film, which does not rely on precise control of composition since if one phase has a slightly different ratio than another it will not be of detriment to the coupling effect, holds much promise for wide ranging applications relating to non-volatile memory and magnetic sensors.146 The next steps for this work are to demonstrate growth on TiN/Si.
Also, to minimise strain and defect formation, the choice of more closely lattice matched layers in the SL films should be explored, where possible. To minimise cation non-stoichiometry in films, growth under lower oxidation conditions should be studied.148 To minimise defects, films could be grown higher temperatures (i.e. at above 1000 °C rather than the conventional 700–800 °C).149 However, considering a key driver is to make oxide SLs CMOS compatible, other strategies to minimise defects at ∼500 °C and below must also be considered, e.g. atomic layer growth methods where epitaxial perovskite films have been grown at <300 °C,150 as well as novel ways to enhance growth kinetics and also carefully control composition.14
2. Most ferroic SLs have been grown mainly on perovskite substrates with (001)-orientation. Epitaxial integration of ferroic superlattices with different crystal orientation other than (001) should be explored. This field has grown rapidly in the past decade but there are many more materials combinations to be explored. Non-(001) orientations give different surface/interface terminations, neutral charge states and 2D-like structures (e.g. (111) orientations giving buckled honeycombs). Also, growth on a range of buffered semiconductor substrates beyond the well-studied Si, e.g. on Ge, Si–Ge, GaAs, GaN, and for flexible electronics, flexible substrates, e.g. mica151 or analogous crystals needs to be explored.
3. Room-temperature ferroelectricity and ferromagnetism have been widely achieved in superlattice films. However, room-temperature multiferroic superlattices with strong and robust magnetoelectric coupling have still not been realized. Since lift-off procedures for oxide films are now quite well established,86,91,152 this allows clamping effects to be overcome and so more work could be directed here to achieve robust, crack-free, non-wrinkled and well adhered films. Also, the vast majority of work on superlattices has focused on A/B superlattices. However, there is a largely unexplored area of multicomponent superlattices with complex periodicities. These could offer much further potential for achieving interesting ferroic behaviour. An example was shown for tricolor superlattices in Fig. 4, where three constituent materials forming an ABC periodicity. With the ABC periodicity, inversion symmetry is broken, enabling interesting novel magnetoelectric phenomena to be achieved.
4. There remains much to be understood about emergent phenomena, such as polar skyrmions, vortices, negative capacitance, and so on. Real-space topological textures such as polar skyrmions and vortices have been observed, but we are still in the early stages of the exploration of these phenomena. For example, the dynamics of such skyrmions structures and vortices must be explored in detail. In addition, these topologies demonstrate chirality and if this can be manipulated by an external field (such as an electric field), then this would provide a powerful way to use the chirality as an externally controllable order parameter.
5. The interfaces between 3d TMOs and 5d TMOs provide a new fruitful research area. Again, current research is still in the early stages and limited to a few systems, so more systematic investigations are highly desired to fully unravel the unique role of 5d TMOs which in turn could provide more candidates to search for novel magnetic textures and topological phenomena. Moreover, owing to the inherent mixture of spin and orbital degrees of freedom in 5d TMOs, these superlattices may open a new pathway to achieve robust electric field control of magnetism at room temperature and above.
2. As highlighted in Table 1, the wide ranging emergent interface phenomena, topological, skyrmion and polar vortices observed in SL films have not been observed in VAN films. Considering the very high epitaxial quality of VAN films, the ultraclean self-assembled interfaces, and the ability to control dimensionalities down to a few nm via controlling growth conditions, there are unique opportunities to discover and control these novel effects in VAN.
3. Combining VAN and SLs to form 3D film architectures has not yet been widely studied.154 However, these structures could offer the chance to combine the complementary advantages of both methods and should be explored further.
4. Pulsed laser deposition is the main method to make VAN films with good quality. However, the method is area-limited. However, promising growth of a limited number of VAN systems by sputtering155 and chemical methods156 show the possibility for growth by methods of industrial interest. More work should be focused in these areas.
5. As already noted, compared to SL films, VAN films can exhibit more bulk-like effects, i.e. interface-driven effects throughout the whole film, not just right at the interfaces. Their simplicity of fabrication is also an important aspect. However, typically they have a random column distribution. For some applications, such as data storage, a highly ordered arrangement is desired. Templating by a variety of ways can produce the desired order,2b,155a,157 recently spontaneous order has been shown.2b More work is needed to determine which methods are most practical for applications.
6. Just as for SLs, mentioned above, most VAN film work has been undertaken on single crystal substrates although progress has been made with STO/TiN-buffered Si158 and with mica,159 and lifted-off films onto Si.92 More work is required on these’other substrates’ as well as new ways to grow directly on Si, e.g. using remote epitaxy.
7. As shown in Section 6, high crystalline quality mesoporous films derived from VAN films give enhanced ferroelectric effects and they also exploit high area surface effects for photocatalysis. There are many new opportunities to exploit the large surface area effects or to create new composites (not just oxide–oxide but also oxide–metal and oxide–organic systems) by chemical coating of mesoporous films made by selective chemical etching of VAN films. With this broader range of compositions, new coupling effects in wide ranging functional systems can be realised.
This journal is © The Royal Society of Chemistry 2023 |