Robust isothermal electric control of exchange bias at room temperature

Robust isothermal electric control of exchange bias at room temperature

(Parte 1 de 2)

Robust isothermal electric control of exchange bias at room temperature

Voltage-controlled spin electronics is crucial for continued progress in information technology. It aims at reduced power consumption, increased integration density and enhanced functionality where non-volatile memory is combined with highspeed logical processing. Promising spintronic device concepts use the electric control of interface and surface magnetization. Fromthecombinationofmagnetometry,spin-polarizedphotoemissionspectroscopy,symmetryargumentsandfirst-principles calculations, we show that the (0001) surface of magnetoelectric Cr2O3 has a roughness-insensitive, electrically switchable magnetization. Using a ferromagnetic Pd/Co multilayer deposited on the (0001) surface of a Cr2O3 single crystal, we achieve reversible, room-temperature isothermal switching of the exchange-bias field between positive and negative values by reversing the electric field while maintaining a permanent magnetic field. This effect reflects the switching of the bulk antiferromagnetic domain state and the interface magnetization coupled to it. The switchable exchange bias sets in exactly at the bulk Néel temperature.

Spintronics strives to exploit the spin degree of freedom of electrons for an advanced generation of electronic devices1,2. In particular, voltage-controlled spin electronics is of vital importance to continue progress in information technology. The main objective of such an advanced technology is to reduce power consumption while enhancing processing speed, integration density and functionality in comparison with presentday complementary metal–oxide–semiconductor electronics3–6. Almost all existing and prototypical solid-state spintronic devices rely on tailored interface magnetism, enabling spin-selective transmission or scattering of electrons. Controlling magnetism at thin-film interfaces, preferably by purely electrical means, is a key challenge to better spintronics7–10.

The absence of direct coupling between magnetization and electric field makes the electric control of collective magnetism in general, and surface and interface magnetism in particular, a scientific challenge. The significance of controlled interface magnetism started with the exchange-bias effect. Exchange bias is a coupling phenomenon at magnetic interfaces that manifests itself most prominently in the shift of the ferromagnetic hysteresis loop along the magnetic-field axis and is quantified by the magnitude µ0HEB of the shift11. The exchange-bias pinning of ferromagnetic thin films is employed in giant magnetoresistance and tunnelling magnetoresistance structures of magnetic-field sensors and modern magnetic read heads12.

Electric control of exchange bias has been proposed for various spintronic applications that go beyond giant magnetoresistance and tunnelling magnetoresistance technology5. One approach to such voltage control requires a reversible, laterally uniform, isothermal electric tuning of the exchange-bias field at room temperature, which remains a significant challenge.

Early attempts in electrically controlled exchange bias tried to exploit the linear magnetoelectric susceptibility of the antiferromagnetic material Cr2O3 as an active exchange-bias

1Department of Physics & Astronomy and the Nebraska Center for Materials and Nanoscience, University of Nebraska, Lincoln, Nebraska 68588-0111, USA, 2Department of Physics, 257 Flarsheim Hall, University of Missouri, 5110 Rockhill Road, Kansas City, Kansas 64110, USA, 3Brookhaven National Laboratory, National Synchrotron Light Source, Upton, New York 11973, USA. *

pinning system13. In a magnetoelectric material an applied electric field induces a net magnetic moment, which can be used to electrically manipulate the magnetic states of an adjacent exchange-coupled ferromagnetic film14. The small value of the maximum parallel magnetoelectric susceptibility better suited for this purpose. Such materials have two or more ferroic order parameters, such as ferroelectric polarization and (anti)ferromagnetic order16.

The potential for an increased magnetoelectric response, for the multiferroic materials, was dictated by the maximum possible value of αmeij. It is determined by the geometric mean of the ferroic susceptibilities, both of which can individually be very high in multiferroics17–20. Coupling between these order parameters has been demonstrated21. However, it is typically weak, and the theoretical upper limit of αmeij is rarely reached16. Artificial two-phase multiferroics have been studied extensively.

Such piezoelectric/ferromagnetic heterosystems allow for electric control of anisotropy22,23. However, strain-induced non-hysteretic magnetoelastic effects are often not stable (persistent) in the absence of an applied field (that is, volatile). Removing the electric field from a linear piezoelectric element releases the strain in the ferromagnetic component and hence restores the anisotropy of the piezoelectrically unstrained film. When using a ferroelectric material, to induce piezoelectric strain control, one may take advantage of the ferroelectric hysteresis to impose some residual strain that will persist after removing the electric field. In contrast to this electric control of magnetic anisotropy in two-phase multiferroics, we report on a non-volatile electric control of unidirectional magnetic anisotropy. The most promising multiferroic single-phase materials used for electrically controlled exchange bias are YMnO3 and BiFeO3 (refs 24,25). Complete suppression of the exchange bias has been


ARTICLES NATUREMATERIALSDOI:10.1038/NMAT2785 achieved at 2K in an YMnO3/NiFe (permalloy) heterostructure. This effect, however, is irreversible. Moreover, the limitation of low temperatures makes YMnO3 unsuitable for applications. The situation is better with BiFeO3. In BiFeO3/CoFe heterostructures, local magnetization reversal on a lateral length scale of up to

2µm has been demonstrated25,26. However, global magnetization reversal, which could be revealed in macroscopic magnetic hysteresis, has not been achieved. Global, but not isothermal magnetoelectric switching has been achieved in the pioneering

Cr2O3/CoPt heterostructure13. However, each sign reversal of the exchange-bias field required a new magnetoelectric annealing procedure, in which the pinning layer is cooled from T > TN to T < TN in the presence of both electric and magnetic fields. Isothermal electric control of exchange bias has been attempted by various groups, but with only marginal success27,28. The result was that reversible and global electrically controlled exchange bias carriedoutisothermallyatroomtemperatureremainedelusive. Here we reveal an unconventional ferromagnetism at the

(0001) surface of the magnetoelectric antiferromagnet Cr2O3 and demonstrate its suitability for electrically controlled exchange bias and magnetization. New insights were achieved by combining first-principles calculations, general symmetry arguments, spinresolved photoemission spectroscopy and magnetometry (see

Supplementary Information) for the Cr2O3 (0001) surface and its interface in an exchange-bias heterostructure. We used a molecular beam epitaxy (MBE)-grown chromia thin film (see the Methods section) for the spin- and energy-resolved ultraviolet photoemission spectroscopy (UPS), whereas the isothermal electric control of exchange bias was done on a heterostructure involving an oriented chromia single crystal with (0001) surface. The choice of a high-quality single crystal for the exchange-bias system completely rules out sample heating induced by leakage currents because of the virtually perfect insulating properties of singlecrystalline chromia. The UPS measurements have been carried out in zero electric field after magnetoelectric initialization of the antiferromagnetic domain state. The non-zero conductivity of thin films is a well-known experimental advantage used for the photoemission investigation of samples that otherwise are virtually perfectly insulating in the bulk. The finite conductivity of the thin film prevents charge accumulation, which could lead to misrepresented photoelectron energies. On the basis of the understanding of the surface ferromagnetism of Cr2O3 (0001), a new concept of Cr2O3 (0001)-based exchange bias is implemented. As a result, a reversible, isothermal and global electric control of exchange bias is demonstrated at room temperature by reproducible electrically induced discrete shifts of the global magnetic hysteresis loop along the magnetic-field axis (see Supplementary Movie).

Magnetically uncompensated surfaces of antiferromagnetically ordered single crystals have been a subject of intense investigations, in particular in the framework of exchange bias11. The surface magnetization of an uncompensated antiferromagnetic surface with roughness usually averages out, so that only a small nonequilibrium statistical fluctuation remains for exchange coupling with the adjacent ferromagnet29.

The surfaces of single-domain antiferromagnetic magnetoelectrics, such as the (0001) surface of the antiferromagnetically long-range ordered Cr2O3, are remarkable exceptions. The free energy of this system, with a boundary, depends on the polar vector n (external normal) as a macroscopic parameter. The existence of the magnetization at the boundary can be deduced from the reduction of the bulk magnetic point group by the presence of an invariant vector n. As both n and E are polar vectors, the boundary reduces the symmetry in a similar way to the electric field, E, in the bulk. An equilibrium magnetization must therefore exist at the surface of a magnetoelectric antiferromagnet, or at its interface with another material. This argument automatically includes equilibrium surface roughness; a more detailed analysis will be published elsewhere.

Both the bulk single crystal and the thin-film sample are confirmed to be (0001) oriented by X-ray diffraction. The surface topography of the bulk and the thin-film sample are mapped using atomic force microscopy (AFM). Figure 1 is organized in such a way that structural data of the bulk sample are shown in the upper panels of a and b. The corresponding data of the thin-film sample are shown in the lower panels of Fig. 1a,b. The (0001) orientation of the bulk surface is independently corroborated by the hexagonal reflection pattern obtained in low-energy electron diffraction. The prominent (0006) and (00012) X-ray peaks of the bulk sample are virtually identically reproduced in the thin film (compare peak positions in the upper and lower panels of Fig. 1a). The surface topography of the samples reveals a plateau with a root-mean-squared (r.m.s.) roughness of 0.88nm for the surface of the bulk crystal (upper panel of Fig. 1b) and an even lower r.m.s. roughness of 0.19nm (lower panel of Fig. 1b) for the thin-film sample measured along selected lines.

Figure 1c illustrates a configuration of the Cr2O3 (0001) surface. It is seen that the particular antiferromagnetic domain has an uncompensatedsurfacemagneticmonolayerwithalignedmoments on all surface Cr3+ ions, even if the surface is not atomically flat. Twofeaturesconspiretoproducethisproperty.First,thecorundum lattice of Cr2O3 can be imagined as a layered arrangement of buckled Cr3+ ions sandwiched between the triangular layers of O2− ions30. The electrostatically stable charge-neutral surface of this crystal is terminated by a semi-layer of Cr; this termination can be viewed as the cleavage of the crystal in the middle of the buckled Cr3+ layer31. Second, Cr ions, which are structurally similar with respect to the underlying O layer, have parallel spins. As a result, a single-domain antiferromagnetic state has all surface Cr spins pointing in the same direction. Note that although we have shown thesurfaceCrionsinbulk-likepositionsinFig. 1,thisassumptionis immaterial for the existence of the surface magnetization, as follows from the general symmetry argument.

In single-crystalline Cr2O3, the antiferromagnetic order allows two degenerate 180◦ antiferromagnetic domains14 (see Fig. 1 and Supplementary Fig. S1). These two domains have surface magnetizations of opposite sign. If the degeneracy of the two domain types is not lifted, the system develops a random multidomain state with zero net surface magnetization when it is cooled below TN. However, magnetoelectric annealing allows for preferential selection of one of these 180◦ domains by exploiting the free-energy gain 1F = αEH (ref. 14). As a result, even a rough Cr2O3 (0001) surface becomes spin-polarized when an antiferromagnetic single-domain state is established. Evidence of this roughness-insensitive surface magnetism is revealed by magnetometry (Supplementary Fig. S2 and Discussion) as well as spin-resolved UPS. Interpretation of the latter is supported by calculations of the site-resolved density of states (DOS) revealing a spin-polarized surface band above the valence-band maximum, in agreement with experimental findings. The UPS carried out on our MBE-grown Cr2O3 (0001) sample is sensitive to occupied surface electronic states.

Figure 2a shows the spin-polarized photoelectron intensity versus binding energy measured at 100K. First, the MBE-grown

Cr2O3 (0001) thin film has been cooled from T > TN in a small magnetic field of 30mT alone, into a multidomain antiferromagnetic state. Spin-up and spin-down photoelectron intensities /↑,↓ (red circles and blue squares) are virtually identical, indicatingnegligiblenetsurfacemagnetizationandpolarization.

Furthermore, multiple measurements were undertaken for the single-antiferromagnetic-domain states, each with a fresh sample preparation. Subsequent sample preparations involve alternating



Al O(0006) Al O

200 nmr.m.s. within (0.15 μm, 0.81 μm) 0.8 nm r.m.s. within (0.04 μm, 0.50 μm) 0.19 nm

Height (nm) 200 nm

Height (nm)

Cr O

Figure 1 | Structural characterization. a, θ–2θ X-ray diffraction pattern of chromia bulk single crystal (upper panel) and thin film (lower panel) showing the chromia (0006) and (00012) peaks, respectively. The film is deposited on a sapphire (0001) substrate, giving rise to (0006), (00012), Kα and Kβ(∗) peaks and a weak structure-factor-forbidden (0009) peak. The inset shows a room-temperature low-energy electron diffraction pattern of the hexagonal chromia (0001) surface measured at an electron energy of 140eV. b, Real-space topography of the chromia (0001) surface of bulk single crystal (upper panel) and thin film (lower panel) measured by AFM. The respective main frames show cross-sectional analysis along indicated lines. A r.m.s. roughness of 0.88nm is calculated in the region between scanning position 0.15 and 0.81µm for the bulk single crystal. The r.m.s. roughness of 0.19nm of the thin film is measured between 0.04 and 0.50µm. c, The spin structure of a Cr2O3 single crystal with a stepped (0001) surface is shown for one of its two antiferromagnetic single-domain states. Up (red) and down (dark blue) spins of the Cr3+ ions (green spheres) point along the c axis.



Polarization Binding energy E¬E(eV)

Binding energy E¬E(eV)

Intensity (arb. units)

Intensity (arb. units)DOS (states/atom, eV)

Figure 2 | Spin-polarized UPS measurements and layer-resolved DOS. a, The intensity of photoelectrons (occupied states) versus binding energy from a Cr2O3 (0001) surface measured at T =100K after cooling in µ0H=30mT and E=0 from T>TN. Spin-up and spin-down intensities are shown by red circles and blues squares, respectively. Inset: The result of a first-principles calculation of the layer-resolved DOS. Colour code follows the experiment. The green line indicates a surplus surface state with spin-up polarization. b, Spin-up (red circles) and spin-down (blue squares) intensities after magnetoelectric annealing in E=3.85×10−4 kVmm−1 and µ0H=30mT. The lines are best fits of multiple-peak Gaussian functions. The diamonds show Cr 3d spin-up (red and green) and

spin-down (blue) contributions extracted from the fits. The Gaussian fit shown by the green diamonds reflects specific surface states. Colour code matches the theoretical DOS data. The triangles show the contrast, P, in the spin-dependent intensities versus binding energy. The green triangles highlight the contribution from the surface state. Maximum absolute errors in P are indicated by bars.

magnetoelectric field cooling from above the Néel temperature to the target temperature of T =100K. The magnetoelectric annealing in alternating fields gives rise to alternating surface magnetization. This leads to a reproducible reversal of the spin-polarization measured at T = 100K. Such data are then summed to provide an averageoverallnetpolarization,independentofinstrumentalasymmetry, as is the standard practice in spin-polarized photoemission and spin-polarized inverse photoemission. The signal is clearly spin-split after magnetoelectric annealing in E = 3.85 × 10−4 kVmm−1 and µ0H = 30mT (compare red circles and blue squares in Fig. 2b), demonstrating high net spin- polarization at the surface. The spin contrast P =(/↑−/↓)/(/↑+/↓), exhibited by triangles in Fig. 2b, is seen to increase significantly close

to the valence-band maximum, EF, (green triangles). We identify this feature with the contribution from the spin-polarized surface states of Cr 3d character. To corroborate this interpretation, we decompose the spin-dependent photoemission spectra /↑,↓ into contributions from Cr 3d bulk and surface states. The contribution above the valence-band maximum (green) is interpreted as an extra spin-polarized surface state. This interpretation is in accordance with our first-principles calculation of the layer-resolved DOS of the Cr2O3 (0001) surface shown in the inset of Fig. 2a. The DOS of a representative central layer with spin-up sublattice (majority/minority in red/blue) magnetization is shown by the lower two curves of the inset. The DOS of the surface layer is shown by the two upper curves in the inset of Fig. 2a. Note that in addition to the bulk states, a surplus spin-up density of states (green) appears above the valenceband maximum. This is consistent with our experimental findings in photoemission (Fig. 2b) and magnetometry (Supplementary Fig. S2 and Discussion).

Experimental and theoretical evidence together point very strongly to the existence of a roughness-insensitive ferromagnetic state at the Cr2O3 (0001) surface when the underlying Cr2O3 single crystal is in an antiferromagnetic single-domain state. Our findings indicate that the surface has a magnetization and is spin-polarized, despite the roughness that is evidently present according to our AFM investigations. Although the roughness may have some effect on the magnitude of the surface magnetization, its mere presence is unusual, and further supported by the experiments on electrically switched exchange bias.

The ferromagnetic surface moment can be isothermally switched by electrical means, giving rise to reversible switching of large exchange-bias fields in our perpendicular exchange-bias

The Cr2O3 substrate in the exchange-bias heterostructure is a (0001) bulk single crystal. The temperature dependence of the exchange bias and its relation to the temperature dependence of the chromia surface magnetization are discussed in the Supplementary Discussion.

Figure 3 demonstrates large isothermal electric switching of the exchange-bias field. It is achieved by leaving the realm of the linear magnetoelectric effect, which gives rise to only a minuscule electric control effect27,28. In contrast to this small linear effect, significant electrically controlled switching requires that a critical threshold given by the product |EH|c, where E and H are isothermally applied axial electric and magnetic fields, is overcome. Initially the het- erostructure has been magnetoelectrically annealed in EH >0 with red squares in Fig. 3a,b show the same virgin loop with positive ex- changebiasofµ0HEB=+6mT.Next,withoutchangingthetemperature,afieldproductEH <0ofindividualfieldsE=−2.6kVmm−1 and µ0H =+154mT is applied for less than a second. During the time when an electric field is applied, the electric current is moni- tored to stay below 0.01µA, resulting in virtually zero sample heating. After applying the electric- and magnetic-field product a magnetic hysteresis loop is measured in E =0. Green triangles (Fig. 3a) show the resulting loop with a pronounced negative exchange bias of µ0HEB ≈ −13mT. The same field product is achieved with

E=+2.6kVmm−1 andµ0H =−154mT,havingthesameeffecton the exchange bias as shown in Fig. 3b by blue circles. The isothermal switching of the exchange-bias field implies a field-induced switch- ing of the antiferromagnetic single-domain state of Cr2O3 into the opposite antiferromagnetic registration. This switching is accom- panied by a reversal of the interface magnetization. Figure 3c shows a sequence of switched exchange-bias fields obtained by switching the electric field back and forth between E = +2.6kVmm−1 and no signs of ageing. The asymmetry between positive and negative




(Parte 1 de 2)