Engineered heterostructures designed for electric control of magnetic properties, the so-called magnetoelectric interfaces, present a novel route towards using the spin degree of freedom in electronic devices. Here, we review how a subset of such interfaces, namely ferromagnet–ferroelectric heterostructures, display electronically mediated control of magnetism and, in particular, emphasis is placed on how these effects manifest themselves as detectable spin-dependent transport phenomena. Examples of these effects are given for a variety of material systems on the basis of ferroelectric oxides, manganese and ruthenium magnetic complex oxides and elemental ferromagnetic metals. Results from both theory and experiment are discussed.
With the ever-approaching scaling and power consumption limit of current semiconductor device technology, the search is on for new material systems that could form the basis of the next-generation devices . Going beyond traditional semiconductors to more exotic materials and structures with extended functionality could lead to lower operating power and better scalability. One route based on magnetoelectric materials has recently attracted significant interest owing to the possibility of controlling magnetism by electric fields [2–4]. In a broad definition, magnetoelectric phenomena not only include the cross coupling between magnetic and electric order parameters , but also involve related effects such as electrically controlled magnetocrystalline anisotropy [6–9], exchange bias [10–12] and spin transport [13–16]. Tailoring these phenomena by electric fields opens exciting avenues for the design of new data storage and processing devices.
There are several mechanisms giving rise to magnetoelectric effects (for a recent review, see Velev et al. ). An intrinsic magnetoelectric coupling occurs in compounds with no time-reversal and no space-inversion symmetries . In such materials, an external electric field displaces the magnetic ions, eventually changing the exchange interactions between them and hence the magnetic properties of the compound. A well-known example of a magnetoelectric material is Cr2O3, where a rhombohedral unit cell and antiferromagnetic order break the space-inversion and time-reversal symmetries required for a linear magnetoelectric coupling to exist . Magnetoelectric coupling has the potential to be much larger in multi-ferroic materials—materials which simultaneously exhibit two (or more) ferroic orders, e.g. ferroelectricity and antiferromagnetism. A prominent example is BiFeO3, a room temperature multi-ferroic with antiferromagnetic Néel temperature TN=643 K and ferroelectric Curie temperature TC=1103 K.
Multi-ferroic and magnetoelectric materials open the doors to novel device functionalities, allowing the possibility of switching magnetization with electric fields, thereby offering a wealth of opportunities for data storage and processing applications. In particular, this property could remove the main hindrance in the miniaturization of magnetic random access memories where the write operation requires undesirable levels of power consumption to produce magnetic fields or large electric currents. Another advantage of magnetoelectric materials is the development of memory bits with multiple stable states [14,16] or functional switches for mixed memory logic functions .
The possibility of using bulk single-phase multi-ferroic–magnetoelectric materials for device applications, however, is limited by the strength of the effect and by their ability to maintain their magnetoelectric properties at room temperature. Composite multi-ferroics, such as layered heterostructures or nanostructured composites of ferroelectric and ferromagnetic compounds, broaden significantly the range of magnetoelectric systems. For example, one mechanism of magnetoelectric coupling may occur in composites of piezoelectric (ferroelectric) and magnetostrictive (ferro- or ferrimagnetic) compounds. In such structures, an applied electric field induces strain in the piezoelectric constituent, which is mechanically transferred to the magnetostrictive constituent, where it induces a magnetization [21–24]. The importance of composite multi-ferroics follows from the fact that none of the existing single-phase multi-ferroic materials combine large and robust electric and magnetic polarizations (ferromagnetism) at room temperature . Almost all practical structures, to date, use the elastic properties of the constituent materials to mediate the macroscopic magnetoelectric response.
A more interesting prospect, both from the point of view of the physics involved and for device applications, is the coupling occurring through primarily electronic mechanisms at magnetoelectric–multi-ferroic surfaces and interfaces. Three major types of such phenomena fit into this category: (i) effects on the interface or surface magnetization by electric fields, e.g. through proximity to an adjacent ferroelectric polarization; (ii) effects on surface magnetocrystalline anisotropy (MCA); and (iii) effects on electron and spin transport across ferromagnetic–ferroelectric interfaces in ferroelectric (multi-ferroic) tunnel junctions. Although all these effects are limited to interfaces, the electronic and magnetic properties of thin-film heterostructures are largely controlled by their interfaces, and therefore interfaces play an important role in determining their overall physical characteristics.
In the following sections, we explore these topics. In §2, we review the effects of electric fields on magnetic moment magnitude, orientation and order at surfaces and interfaces. In §3, we discuss tunnelling across thin-film ferroelectric materials where the interface magnetoelectric effects control spin transport. In §4, we summarize the discussions.
2. Electric field effects on magnetic moment magnitude, orientation and order at surfaces and interfaces
The surface (interface) magnetoelectric effect occurs in any surface (interface) of a ferromagnetic metal. When a metal surface is exposed to an electric field the induced surface charge, σ=ε0E, screens the electric field over a characteristic screening length of the metal. In a ferromagnetic metal, owing to the exchange splitting of the surface density of states (DOS), one spin DOS dominates over the other and therefore the screening charge is spin-dependent . As a result, the surface magnetization, Ms, changes with electric field, as shown from first-principles calculations of an Fe (001) surface in figure 1 . The induced surface magnetization is linear with respect to applied electric field E, i.e. μ0Ms=αsE, owing to broken space-inversion and time-reversal symmetries associated with surfaces (interfaces) of a ferromagnetic material. Calculations demonstrate, however, that the magnetoelectric response is small for elemental ferromagnetic metals, being about the same order of magnitude as for magnetoelectric Cr2O3 but limited to the surface. For example, the electric field as large as 1 V nm−1 leads to the magnetic moment change of the order of 10−3 μB per atom. The magnitude of the effect depends intimately on the Fermi-level spin polarization of the surface DOS, increasing with larger polarization to a maximum for 100 per cent spin-polarized materials, i.e. the so-called half-metals .
These magnetoelectric effects induced by interface screening can be substantially enhanced at the interface of a ferromagnetic metal with a dielectric material. This is due to the fact that, for a given electric field E in the dielectric, the induced surface charge σ=εE scales with the dielectric permittivity ε, which is larger than the vacuum permittivity ε0. Because the induced interface magnetization is proportional to the screening charge, the interface magnetoelectric effect is enhanced . For high-κ materials, the dielectric constant ε/ε0 can be of the order of 100 or even larger, increasing the magnetoelectric response by two or more orders of magnitude .
Electronically driven magnetoelectric effects can be further enhanced by using a ferroelectric material, rather than by using an external field alone, to induce a surface charge. In this case, the spin-dependent screening in the ferromagnetic material occurs in response to the polarization charge at the ferromagnet–ferroelectric interface, which can be significantly larger than those achievable with high-κ dielectrics with reasonable electric fields. In addition, the bi-stable nature of this charge opens the appealing possibility to change its sign by switching the ferroelectric polarization orientation with a transient electric field. Such a nonlinear magnetoelectric effect was predicated for a SrRuO3–BaTiO3 interface where a magnetic moment change of 0.32 μB per interface Ru atom results from ferroelectric polarization reversal . It was found that the effect arises basically from the build-up of screening charge, but goes beyond simply filling of rigid spin-dependent bands. Because SrRuO3 is an itinerant ferromagnet, there is a significant dependence of the exchange splitting on the band filling, i.e. the electron charge. Changing the electronic charge density, even only at the interface, affects the magnitude of the exchange splitting of the DOS itself (figure 2a), consistent with a simple Stoner model , enhancing significantly the magnetoelectric response.
This finding motivated the exploration of actually inducing magnetism in a non-magnetic material, elemental Pd metal that sits on the edge of itinerant magnetism. First-principle calculations of a Pd thin film revealed that it is possible to induce surface ferromagnetism by applying an external electric field large enough that the induced surface charge enhances the paramagnetic surface DOS at the Fermi level, ρF, beyond a critical point where it satisfies the well-known Stoner criterion, IρF>1, where I is a phenomenological parameter characterizing the strength of the intra-atomic Coulomb interaction between electrons . Beyond the critical electric field, Ec, the surface magnetization can be significantly modulated, with a simple dependence on electric field, E, of the surface magnetization being This prediction extends the previous studies of the interface magnetoelectric effects in materials at the proximity of a paramagnetic–ferromagnetic transition [34,35] to a realistic fully self-consistent calculation. Using a ferroelectric material to create a field effect may be advantageous to observe the predicted phenomenon due to a much stronger screening charge induced at the interface by ferroelectric polarization.
Another possibility for the build-up of screening charge is to induce a magnetic ordering transition, e.g. change the material interface from ferromagnetic to antiferromagnetic. It is well known that the doped La-manganites, La1−xAxMnO3 (A=Ca, Sr or Ba), possess a rich phase diagram as a function of hole concentration (x) and temperature that includes metal–insulator transitions as well as the colossal magnetoresistance effect . Of particular interest is the fact that the La-manganites go through a series of magnetic phases, from antiferromagnetic to ferromagnetic and back to antiferromagnetic as x is varied from 0 to 1. These transitions are generally attributed to the competition between the super-exchange interaction (which favours antiparallel alignment of neighbouring Mn magnetic moments) and the double-exchange interaction (which favours parallel alignment) . Changing the population of electrons that mediate the double exchange leads to a rich series of magnetic phase transitions. Carrier concentration can also be modulated electrostatically, opening the possibility to dramatically alter the properties of the manganite in a field effect device . By choosing x to reside near a magnetic phase transition, it should then be possible to drive the magnetic order back and forth between different phases through electrostatic screening.
This effect was demonstrated from first principles for a La1−xSrxMnO3 (LSMO)–BaTiO3 interface, where the screening charge density in the manganite near the interface depends on the orientation of the ferroelectric polarization in the adjacent BaTiO3 (BTO) layer . The calculations of total energy for different magnetic configurations revealed that switching ferroelectric polarization induces a change in magnetic order in the first two or three unit cells of the LSMO, as shown in figure 3.
Such a phenomenon constitutes a substantial and robust magnetoelectric effect resulting in a net change of interface magnetization approximately 7 μB per interface Mn atom. Recently, experimental measurements of electrical, magnetic and spectroscopic properties on the similar ferroelectric–manganite interface system of Pb1−xZrxTiO3/LSMO have shown that such a large magnetoelectric response does indeed occur . Furthermore, in addition to the change in magnetization, first-principle calculations recently established that this magnetoelectric effect can result in a profound change in resistance for ferroelectric tunnel junctions (FTJs) sporting this interface, as discussed in §3 .
In addition to the screening mechanisms described earlier, the changes in interface bonding during polarization reversal may play an important role in the magnetoelectric effect at ferromagnet–ferroelectric interfaces. The change in atomic displacement at the interface alters orbital hybridizations, affecting the interface magnetic moments. First-principles calculations for the Fe–BaTiO3 (001) interface show a large change in the interface magnetic moment, 0.25 μB per interface unit cell, when the ferroelectric polarization is switched by an electric field . This contribution largely comes from the induced magnetic moment on the interfacial Ti atom due to hybridization of the 3d orbitals of Ti and Fe across the interface. Very recently, this prediction was experimentally confirmed by the element selective X-ray resonant magnetic scattering spectra, which showed that the hysteresis loops obtained at the Ti L3,2-edges perfectly follow the hysteresis loops for Fe . A similar effect was also observed at the Co–BaTiO3 interface , further supporting the validity of the first-principles results . In addition to altering the interface magnetization, polarization reversal leads to a significant change in Fermi-level DOS (figure 4), which may play an important role in the ferroelectric control of spin polarization in tunnelling transport experiments, as described in §3. Similar effects were predicted for Co2MnSi–BaTiO3 , Fe3O4–BaTiO3  and Fe–ferroelectric interfaces [47,48].
Another route towards magnetoelectric control of magnetism is to control the MCA of a magnetic material, using an applied electric field. Because the MCA determines the stable orientation of magnetization, tailoring the anisotropy of a ferromagnetic film by electric fields allows for reorientation or even switching of magnetic moments. For metallic ferromagnets, the electrically driven magnetoelectric effect is confined to the interface, and, consequently, the electric field affects only the surface–interface MCA [8,27,29]. For 3d ferromagnets, the effect originates from the change in the relative population of the 3d orbitals, which contribute differently to the magnetic anisotropy energy. A strong effect of applied electric field on the interface MCA was demonstrated experimentally for Fe–MgO(001) interfaces . A change in the interface MCA energy can be used for switching magnetization by applied electric fields [49–54]. Very recently, electrically induced bi-stable magnetization switching between two easy magnetization directions was realized in magnetic tunnel junctions (MTJs) at room temperature [55,56]. These results demonstrate the potential of this method for magnetic data storage and spintronics applications . Alternatively, the MCA may be controlled at the ferroelectric–ferromagnet interface, as indicated by first-principles calculations for a Fe–BaTiO3 bilayer .
3. Magnetoelectric control of spin and electron transport in tunnel junctions with ferroelectric barriers
Multi-ferroic tunnel junctions (MFTJs) involve a new concept for a multi-functional device and have recently attracted significant interest . MFTJs exploit the capability to control electron and spin tunnelling via ferromagnetic and ferroelectric polarizations of the MFTJ constituents, as was first predicted by Zhuravlev et al. . MFTJs are a particular type of FTJ where a ferroelectric thin film serves as a barrier between two metal electrodes (figure 1b). The key property of an FTJ is the tunnelling electroresistance (TER) effect, meaning a change in resistance with reversal of ferroelectric polarization. The TER effect can arise from several origins, as was discussed in Tsymbal & Kohlstedt . These include changing (i) the electrostatic potential at the interfaces and throughout the barrier; (ii) bonding strength and electronic structure at the interfaces; and/or (iii) strain associated with the piezoelectric response.
The electrostatic effect, (i), results from the incomplete screening of polarization charges at the interface of FTJs . This creates finite-size charge depletion (accumulation) regions at the interfaces and hence an asymmetric potential profile in FTJs with dissimilar electrodes. The predicted TER effect becomes especially strong if an additional thin dielectric layer is placed at one or the other interface . The interface bonding effect, (ii), on TER becomes apparent in atomistic calculations  where it is shown that the presence of interfaces imposes restrictions on the atomic displacements responsible for the spontaneous polarization due to bonding with electrode atoms. The piezoelectric response of the ferroelectric barrier, (iii), may lead to a change in effective tunnelling barrier thickness with polarization reversal, giving rise to an exponential change in the tunnelling transmission of electrons. In particular, the change in atomic displacements throughout the barrier will influence the decay rate in the barrier and consequently the transmission through it .
Experimentally, the key problem is to reveal the correlation between ferroelectric polarization and tunnelling conductance. This was achieved in 2009 by three experimental groups reporting, independently, observations of the TER effect associated with the switching of ferroelectric polarization of BaTiO3 or Pb1−xZrxTiO3 ferroelectric films using scanning probe techniques [62–64]. As predicted [59,60], the observed effects were really giant, showing resistance changes by two or three orders of magnitude.
An MFTJ is an FTJ with ferromagnetic electrodes or, equivalently, an MTJ with a ferroelectric barrier (figure 5a–c). Electron tunnelling from a ferromagnetic metal electrode through any thin insulating layer barrier is spin polarized and, as a consequence, in an MTJ, the tunnelling current depends on the relative magnetization orientation of the two ferromagnetic electrodes, a phenomenon known as tunnelling magnetoresistance (TMR). For a review of TMR, see Tsymbal et al. . In an MFTJ, the TER and TMR effects coexist , making MFTJ a four-state resistance device where resistance can be switched by both electric and magnetic fields.
An important mechanism that influences transport properties of MFTJs is the interface magnetoelectric effect. This effect contributes to the change in the spin polarization of the tunnelling current and thus influences TMR. The interface magnetoelectric effect may also notably change the resistance of an MFTJ and thus be decisive for TER, in addition to the three mechanisms for TER discussed earlier. Later, we will give examples illustrating these phenomena.
First-principles transport calculations of SrRuO3–BaTiO3–SrRuO3 MFTJs show that the orientation of ferroelectric displacements in the BaTiO3 affects the transmission for parallel and antiparallel magnetization orientation of the electrodes, resulting in ferroelectric control of TMR . The asymmetric interface termination sequence (RuO2–BaO at one interface versus TiO2–SrO at the other) creates a different polarization profile when the ferroelectric polarization is switched (figure 6a), resulting in a substantial TER effect. For polarization to the right, the Ti–O displacements in the BaTiO3 responsible for the polarization are larger than those when polarization is to the left. This gives rise to a change in the magnitude of the band gap in BaTiO3 and therefore modifies the tunnelling decay rate, shown in figure 6b, for those states that are responsible for carrying most of the tunnelling current. In addition, the two states shown, the Δ1 and Δ5 bands, by symmetry carry the majority and minority states from the SrRuO3 electrodes. Because these states are affected differently, and the tunnelling transport depends exponentially on the decay rate, this leads to a significant change in TMR.
An important contribution affecting spin polarization of the tunnelling current (and thus TMR) in these junctions comes from the interface magnetoelectric effect. As is evident from figure 2, reversal of ferroelectric polarization of the BaTiO3 barrier changes the exchange splitting of the spin bands at the interface, the key property that controls tunnelling spin polarization and TMR . Thus, despite a relatively small change of the interface magnetic moment, the interface magnetoelectric effect is important for transport properties of MFTJs.
The interface magnetoelectric effect has been put forward to explain four resistance states observed experimentally in Fe–BaTiO3–La0.67Sr0.33MnO3 MFTJs . These junctions were fabricated in a nanoindentation geometry (figure 7a), and, at low temperatures, demonstrated both TMR and TER effects, reproducibly, as well as a significant change in TMR with ferroelectric polarization reversal (figure 7b). This effect on TMR was attributed to the interfacial magnetoelectric response at the Fe–BaTiO3 interface, revealing itself in a change of the minority-spin DOS with polarization reversal, as shown by the first-principles calculations in figure 4b. In addition, the large change in the minority-spin DOS at the Fe–BaTiO3 interface accounts for a significant contribution to the TER effect observed in this system, demonstrating the intimate coupling between order parameters responsible for the multi-ferroic tunnelling effects. Experimental evidence that the transport spin polarization can be controlled by the switchable ferroelectric polarization was also presented for MFTJs based on La0.7Sr0.3MnO3 electrodes with ferroelectric Ba0.95Sr0.05TiO3  and BiFeO3  tunnel barriers. The possibility of observing four resistance states due to the coexistence of TMR and TER effects was also predicted for MFTJs with organic ferroelectric barriers, such as poly(vinylidenefluoride) .
Another exciting prospect for the ferroelectric control of spin transport is to incorporate the interfacial magnetoelectric reconstruction effect that occurs at the LSMO–BTO interface, as described in §2 and illustrated in figure 3 [39,41]. As mentioned earlier, the doped manganite (La1−xSrxMnO3) exhibits a transition around x∼0.5, from a ferromagnetic phase to an antiferromagnetic phase . Important here is the fact that the ferromagnetic phase is half-metallic with electronic DOS only in the majority spin channel , whereas the antiferromagnetic phase has A-type magnetic order consisting of (001) planes of ferromagnetically ordered Mn moments that align antiparallel with neighbouring (001) planes. This magnetic structure leads to highly anisotropic ‘two-dimensional’ transport properties: metallic conductivity in the (001) plane and insulating behaviour along  .
Along with the change in magnetic order with polarization reversal at this interface comes the transition to the anisotropic metallic phase . This suggests an intriguing possibility to detect this behaviour: in an FTJ with such a magnetoelectrically active interface in the path of the tunnelling current, switching of the ferroelectric barrier is expected to change the ‘perpendicular metallicity’ on the interface of the La1−xSrxMnO3 electrode, effectively changing the tunnelling barrier thickness, and leading to a giant change in conductance. First-principle calculations of a junction consisting of approximately 1.9 nm of BaTiO3 as a tunnelling barrier with La0.7Sr0.3MnO3 and La0.6Sr0.4MnO3 as the left and right electrodes, respectively, predict an approximately 1800 per cent change of conductance with polarization reversal owing to the interface magnetic reconstruction.
The effect can be understood by comparing the spin-polarized local density of states (LDOS) on the interfacial layers of the right electrode for the two polarization states, plotted in figure 8a(i–iii). For polarization to the right, the LDOS at the Fermi level is non-zero only in the majority-spin channel owing to the half-metallicity. After polarization reversal, however, the antiferromagnetic transition sets in. This change in magnetic order reverses the moment of the second Mn layer (figure 8b), which interchanges the role of ‘up’ and ‘down’ spin states in the second unit cell (figure 8a(ii)(v)). Because of the large spin polarization of the Mn sites, the LDOS for majority-spin electrons at the Fermi level is very small in the second unit cell. Therefore, the first and second unit cells at the interface can be considered atomic-scale ‘electrodes’ separated by a monolayer-thin spacer layer, i.e. an atomic-scale spin valve. Because the spin polarization is nearly 100 per cent, this ‘spin valve’ exhibits a huge difference in conductance between parallel and antiparallel configurations, giving rise to the large TER effect.
Magnetoelectric phenomena at interfaces constitute an emerging field of research that has interesting fundamental physics and exciting prospects for device applications. Changing magnetic properties, such as surface magnetization, interface MCA, exchange bias and spin transport across interfaces by electric fields, reveals new functionalities of materials and thus stimulates new directions for research. Especially thrilling in our opinion are magnetoelectric phenomena that occur as a result of electronically driven mechanisms. Although typically these effects are limited to interfaces, in thin-film heterostructures, the interfaces largely control their properties and thus affecting electronic properties of interfaces alters dramatically physical characteristics of these heterostructures.
Using ferroelectric materials can significantly enhance electronically driven magnetoelectric effects owing to much larger intrinsic electric fields associated with the interface polarization charges. In addition, switchable electric polarization in ferroelectric materials allows for bi-stable operational properties needed for non-volatile memory and logic functionalities. In this article, we have reviewed a number of interface magnetoelectric phenomena and phenomena involving the interplay between the interface magnetoelectric effects and spin transport. A particular emphasis was made on MFTJs where the spin transport is largely affected by ferroelectric polarization through interface magnetoelectric effects. These include effects of ferroelectric polarization on exchange splitting of the interface spin bands and on the interface spin-dependent bonding states. We have also discussed the effect of magnetic reconstruction at the interfaces of complex oxides such as La1−xSrxMnO3, where ferroelectric polarization switching alters the magnetic order at the interface. This effect is especially interesting for controlling the resistance of MFTJs, where the electrically controlled atomic-scale spin valve changes the resistance by several orders in magnitude.
The authors acknowledge support from the National Science Foundation (NSF) through the Materials Research and Science Engineering Center at the University of Nebraska-Lincoln (NSF grant no. DMR-0820521) and the Experimental Programme to Stimulate Competitive Research (NSF grant no. EPS-1010674).
One contribution of 10 to a Discussion Meeting Issue ‘The new science of oxide interfaces’.
- This journal is © 2012 The Royal Society