Reversible electric control of exchange bias in a multiferroic field-effect device

Reversible electric control of exchange bias in a multiferroic field-effect device

(Parte 1 de 2)

Reversible electric control of exchange bias in a multiferroic field-effect device

Electric-field control of magnetization has many potential applications in magnetic memory storage, sensors and spintronics. One approach to obtain this control is through multiferroic materials. Instead of using direct coupling between ferroelectric and ferromagnetic order parameters in a single-phase multiferroic material, which only shows a weak magnetoelectric effect, a unique method using indirect coupling through an intermediate antiferromagnetic order parameter can be used. In this article, we demonstrate electrical control of exchange bias using a field-effect device employing multiferroic

(ferroelectric/antiferromagnetic) BiFeO3 as the dielectric and ferromagnetic La0.7Sr0.3MnO3 as the conducting channel; we can reversibly switch between two distinct exchange-bias states by switching the ferroelectric polarization of BiFeO3. This is an important step towards controlling magnetization with electric fields, which may enable a new class of electrically controllable spintronic devices and provide a new basis for producing electrically controllable spin-polarized currents.

New physical phenomena at artificially constructed heterointerfaces are at present an exciting area of condensed-matter science1. Oxides of 3d transition-metals, such as the cuprates, manganites and more recently the multiferroics, present a fascinating playground to explore the interactions of charge2, spin3, orbital4 and lattice degrees of freedom at such heterointerfaces, which eventually lead to new states of matter. One model system is the interface between LaAlO3 and

SrTiO3, nominally two band insulators, that exhibit unusual electronic reconstruction and transport phenomena when adjoined as a heterointerface2,5,6. The emergence of charge-transfer-driven orbital ordering and ferromagnetism in a (Y,Ca)Ba2Cu3O7 layer at the interface with the doped manganite La0.67Ca0.33MnO3 has been demonstrated, thus bringing to bear the role of the orbital degree of freedom4,7. Such an epitaxial heterointerface is illustrated schematically in Fig. 1, which describes the various degrees of freedom. Electric-field control of such an interfacial ferromagnetic state would be a significant step towards magnetoelectric devices8–12. The possible electronic reconstructions at the interface will undoubtedly be influenced by the ground-state electronic structure of the transition-metal species at the interface. For example, the d5 electronic state of Fe3+ is stable to perturbations and therefore is unlikely to undergo reconstruction13. With this as the background, we are exploring one such d5 model system, manifested in ferroelectric (FE) and antiferromagnetic (AFM)

BiFeO3 (BFO), which is epitaxially juxtaposed at the interface to a multivalent transition-metal ion such as Mn3+/Mn4+ in ferromagnetic La0.7Sr0.3MnO3 (LSMO). We have observed an unexpected ferromagnetic order that is induced in the Fe sublattice at the interface as a consequence of a complex interplay between the orbital degree of freedom and its coupling to the spin degree of freedom14.ThisferromagneticstateintheFesublatticegivesrisetoa significant exchange-bias interaction with the ferromagnetic LSMO (a shift in the magnetic hysteresis curve along the applied-field axis in a coupled ferromagnetic (FM)/AFM system, when magnetically field cooled through a blocking temperature (TB; refs 15,16)). The discovery of a correlation between the electronic orbital structure

1Materials Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA, 2Department of Physics, University of California, Berkeley, California 94720, USA, 3Department of Physics, University of California, San Diego, La Jolla, California 92093, USA. *

at the interface and exchange bias suggests the possibility of using an electric field to control the magnetization of the ferromagnet, which is precisely the focus of this paper.

Recently, there has been significant effort to electrically control the magnetization of a ferromagnetic thin film8,9. One technique has been to use a magnetoelectric multiferroic material with directly coupled FE and FM order parameters, such that there is direct control of magnetization by the electric field17. One problem with this type of direct coupling is that most singlephase FM–FE multiferroics show only a weak magnetoelectric effect10,1. In addition, there are only a small number of materials that can exhibit this form of direct coupling, because of a large number of material constraints, some of which are mutually exclusive18,19. There has been some progress in direct single-phase magnetoelectric coupling in multiferroic materials. For example, electrical control of magnetization in domain walls of GdFeO3 has been shown20. Although this is a significant achievement, the relative change in magnetization may not be sufficient for practical applications that require stronger magnetoelectric coupling or full reversalofmagnetization.Incontrasttothefewsingle-phaseFE/FM multiferroics, there are several known FE/AFM multiferroics. A different approach to achieve stronger magnetoelectric coupling, using FE/AFM multiferroics, employs a bilayer consisting of a thin FMmaterialthatiscoupledtothemultiferroic.Inthiscase,theAFM order parameter in the multiferroic acts as a medium that indirectly couples the FM ordering of an FM thin film and the FE ordering of the multiferroic. The mechanism for coupling between the AFM and the FM is exchange bias. Two ideal candidate materials for this type of indirect coupling are multiferroic (AFM/FE) BFO and ferromagnetic LSMO, which were previously mentioned to show significant exchange bias owing to interfacial orbital reconstruction. BFO is a suitable multiferroic for this application because it has strong coupling between its AFM and FE order parameters21. Controlling this exchange bias may allow for the manipulation of magnetization, by biasing at a magnetic field and electrically shifting the magnetic hysteresis curve beyond the coercive field in either direction.ThisbehaviourisschematicallyshowninFig. 2a.


Oxide A

Oxide B


LaAlO3 La0.67Ca0.33MnO3


Charge transferCharge transfer/orbital reconstruction Orbital reconstruction

High-mobility two-dimensionalelectron gasInterface ferromagnetism/ suppressed interfacial superconductivity

Interface ferromagnetism/ exchange bias

Orbital Lattice

Spin Charge

Emergent interfacial phenomenon

Figure 1 | The interplay between heterointerfacial degrees of freedom. a, A schematic showing the interplay between the different degrees of freedom at play (charge, orbital, spin and lattice) at heteroepitaxially grown interfaces between different oxide materials. b, Some canonical examples of interfacial reconstruction and the new emergent interfacial phenomenon that occur in these systems.

Electric control of magnetism has been attempted by using a field-effect device with the magnetoelectric material Cr2O3 as an AFM dielectric and an FM Co/Pt multilayer channel22. It was shown that through cooling with both electric and magnetic fields it was possible to affect the polarity of the exchange bias with respect to the direction of the magnetic cooling field. Later, it was demonstrated with multiferroic thin-film YMnO3 and permalloy bilayers23 that it was possible to set an exchange bias with field cooling and remove it with ferroelectric switching. After removal it was not possible to return to the field-cooled state. Using a different approach we have built a multiferroic field-effect device with a heteroepitaxiallygrown,ferromagneticmanganite(LSMO)channel layer. In our system we have been able to electrically switch between twoexchange-biasstatesreversibly,withoutmagnetic-fieldcooling.

To build our electric-field-effect device, a thin-film heterostructure of LSMO and BFO was heteroepitaxially grown by pulsed laser deposition on a SrTiO3 (STO) (100) substrate. The LSMO was chosen to be 3–5nm thick so that exchange-bias effects at the interface will dominate when probed with electrical-transport measurements. In addition, because of the FM-layer-thickness dependence in exchange-bias systems, minimizing the thickness of the LSMO layer also maximizes the magnitude of the exchange bias15. As a precautionary measure, the BFO layer was chosen to be 600nm thick to prevent pinhole shorts in the dielectric, which can cause gate leakage24. Structural characterization using X-ray diffraction revealed single-phase BFO and LSMO layers, and high-resolution transmission electron microscopy experiments confirmed a high-quality interface, as illustrated in Supplementary Fig. S1. Detailed electron energy-loss spectroscopy scans (Supplementary Fig. S1) revealed very little interdiffusion at the interface and hence the transition-metal site. The ferroelectric domain structure of the BFO layer exhibited a typical stripe-like structure consisting of a predominantly 71◦ domain-wall pattern, shown in Supplementary Fig. S2. After switching the FE polarization with an electric field, this domain pattern primarily switches by 180◦ from the original polarization state, retaining the stripe-like pattern (Supplementary Fig. S2).

Magnetization hysteresis curves for the as-grown films were measured with superconducting quantum interference device

(SQUID) magnetometry and exhibit exchange bias (Fig. 2b). After being cooled in a magnetic field of ±1T, applied parallel to the interface, there is a corresponding shift in the hysteresis loop. Cooling the sample in positive magnetic field results in a negative shift of the hysteresis loop and vice versa. These results were very reproducible; over 100 heterostructure samples with LSMO thicknesses in the range of 2–10nm were grown and exhibited exchange bias like the data presented in Fig. 2b. The magnitude of the exchange coupling systematically decreased with LSMO layer thickness, consistent with the conventional picture of exchange bias as originating from the interface15,16. To test that this shift is AFM–FM exchange bias, we measured a sample with a thin 2nm layer of STO inserted between the AFM and the FM. The magnetic hysteresis curve of this sample (Fig. 2b (inset)) shows no observable shift of the loop when field cooled. Furthermore, from these experiments the blocking temperature of this system was determined to be ∼100–120K. This is important because it sets an upper temperature limit for where exchange bias can be observed, and thus where devices can be operated.

To investigate electrical transport in these heterostructures three samples with 3-nm-thick LSMO and 600-nm-thick BFO were patterned with 45 gated Hall-bar structures schematically shown in Fig. 3. This LSMO thickness was chosen because it exhibitedthelargestexchangebiaswhilestillremainingconducting. The details of the device-fabrication process are described in the Methods section. After fabrication, sheet resistance as a function of temperature (RS–T) was measured to ensure that the material was not damaged during device processing and to verify that ohmic contacts were made to the buried LSMO layer. Figure 4a illustrates two typical RS–T plots for the LSMO layer, one for each FE polarization of the BFO. The inset shows a larger temperature range fordatatakenfortheas-grownBFOFEpolarizationstate.Theresult is consistent with measurements for similarly prepared bare 3nm LSMO films. It is interesting to note the similarity of the shape of the RS–T plots for the two gate polarities. These data show both a large vertical translation of resistance and a small multiplicative change in RS(T) between the two FE polarization states. We interpret the vertical translation as a change in residual resistivity

Rs0. From Matthiessen’s rule, this suggests a change in the density

ARTICLES NATUREMATERIALSDOI:10.1038/NMAT2803 Exchange-bias shiftM Ma

¬225 Oe

1 T field cool ¬1 T field cool

Magnetization (e.m.u./cm

Figure 2 | A schematic of magnetization inversion through exchange-bias modulation and magnetometry measurements. a, An example of how magnetization switching might occur in a system with reversible switching of exchange bias. Initially the magnetic layer is in the −MS state with positive exchange bias. When switched to negative exchange bias the magnetization switches to +MS. b, Magnetic hysteresis curves of the bulk film (BFO/LSMO/STO) obtained from SQUID measurements made at 7K.

Curves are shown for measurements after being field cooled from 350K in +1T (red) and in −1T (blue). The inset shows a magnetic hysteresis curve for a BFO/STO/LSMO/STO structure, with no exchange bias after field cooling.

or strength of scattering impurities where the electron gas resides. The multiplicative change probably arises from a direct change in carrier density due to doping, or a change in channel geometry. This multiplicative-change effect is smaller than results reported in heterostructures of Pb(ZrxTi1−x)O3/LSMO (ref. 25). More detailed analysisofthesedataispresentedinSupplementaryFig.S3.

For comparison of electrical-transport measurements with magnetometry measurements, the sample was mounted in a variable-temperature vacuum probe and cooled in a magnetic field parallel to the interface. On reaching the experimental temperature of 7K, the magnetoresistance (MR) was measured by sweeping the magnetic field between ±10kOe parallel to the direction of the current (Fig. 3). The data exhibit hysteresis when measured along the easy axis of the LSMO. From these data we define H+ as the location of the positive peak and H− to be the location of the negative peak (Fig. 4b). The coercive field (HC) is then defined to be halfthedistancebetweenthetwopeaks,HC=(H+−H−)/2,andthe exchange bias is defined to be the magnitude of the peak shift from zero, HEB =(H++H−)/2. For positive- and negative-field cooling we obtained essentially identical values to those measured with the

SQUIDatthesametemperatureasshowninFig. 2b. The resistance of the channel was hysteretic as a function of applied gate voltage (VG). An RS–voltage-pulse hysteresis curve, shown in Fig. 4c, illustrates this effect. Each data point in the hysteresis curve is the average of the two peak resistance values

Magnetic field


La0.7Sr0.3MnO3 Pd


Figure 3 | A schematic of the BFO/LSMO field-effect device. To change the BFO polarization, a voltage pulse VG is applied between the gate and the LSMO channel. The magnetic field for MR measurements is applied parallel to the direction of the current.

from a magnetoresistance sweep after pulsing the gate for 10ms with a voltage. The arrow on the curve shows the direction of the pulse sequence. For VG pulses greater than 17V the sheet resistance magnetic coercivity of the LSMO was also hysteretic with voltage applied to the BFO gate. A coercivity–VG hysteresis curve is shown in Fig. 4d. Like the RS–VG curve, it also has saturation values at 17 V and −17V of 1,550Oe and 1,000Oe respectively, suggesting that doping may play a role. Although the data in Fig. 4a suggest that doping is not the dominant effect in the modulation of RS, small changes in carrier density have been shown to modify the magnetic properties in LSMO (refs 26,27).

With the initial characterization of the device completed, we then studied the effect of BFO polarization on exchange bias. To do this the exchange bias in the LSMO layer was measured for the voltage-pulse sequence shown in Fig. 5a at 5.5K. This temperature was chosen to be well below the blocking temperature to maximize the magnitude of exchange bias. We use VG pulse voltages (with 10ms pulse widths), in zero applied magnetic field, that are large enough to fully saturate the BFO ferroelectric polarization, in this case ±60V as shown in Supplementary Fig. S4. The value of exchange bias is obtained after each of the pulses using the approachoutlinedinFig. 4b.Thepulsesequencealternatesbetween the two ferroelectric polarization states after every five pulses. We have observed in this system that the polarity of the exchange bias is opposite to the remanent magnetization MR in the LSMO before the gate pulse is applied. This remanent magnetization arises from the magnetoresistance sweep and can be set into either a negative or positive state, depending on the direction from which the sweep approaches zero applied magnetic field. The results of the electric-field-induced changes in the exchange bias (for both remanent magnetization polarities) are presented in Fig. 5b,c. HEB was normalized to the coercivity, HEB/HC, to eliminate complications that could arise from concurrent changes in the coercive field. The first interesting observation is that HEB can be modulated significantly by the application of the electric field. The magnitude of the exchange bias for both polarities was modulated by the electric field between a high and low value that corresponded to low or high resistance respectively. This change is reversible with electric field application, and requires no field cooling or applied magnetic field during voltage pulses.

The maximum exchange-bias modulation was ∼0.15 HC, which corresponds to an exchange-bias shift of ∼125Oe.


¬45 V

+45 V

Rs (k Ω/ )

Rs (k Ω/ )

(Parte 1 de 2)