Report

# Electric Field-Induced Modification of Magnetism in Thin-Film Ferromagnets

See allHide authors and affiliations

Science  19 Jan 2007:
Vol. 315, Issue 5810, pp. 349-351
DOI: 10.1126/science.1136629

## Abstract

A large electric field at the surface of a ferromagnetic metal is expected to appreciably change its electron density. In particular, the metal's intrinsic magnetic properties, which are commonly regarded as fixed material constants, will be affected. This requires, however, that the surface has a strong influence on the material's properties, as is the case with ultrathin films. We demonstrated that the magnetocrystalline anisotropy of ordered iron-platinum (FePt) and iron-palladium (FePd) intermetallic compounds can be reversibly modified by an applied electric field when immersed in an electrolyte. A voltage change of –0.6 volts on 2-nanometer-thick films altered the coercivity by –4.5 and +1% in FePt and FePd, respectively. The modification of the magnetic parameters was attributed to a change in the number of unpaired d electrons in response to the applied electric field. Our device structure is general and should be applicable for characterization of other thin-film magnetic systems.

Spin electronics has recently emerged as a field of science and technology in which the electron's spin is directly exploited. Giant magnetoresistance and tunnel magneto-resistance sensors, used in read heads of hard disks, are both examples of the application of spin-electronic devices. In contrast, there are very few active (i.e., non-sensor) systems that exploit spin electronics. Such systems are operated through electrical current actuation, which requires much larger energy consumption than electric field (E) actuation used in usual electronic circuits and/or electromechanical systems. E actuation of magnetic properties has been demonstrated in the semiconducting (Ir,Mn) As system, in which the Curie temperature TC was varied by 1 K under a voltage of 125 V, corresponding to an electric field of approximately 1.6 × 108 V/m (1, 2). However, the possible applications of this phenomenon are limited as a result of the low TC of magnetic semiconductors (≪ room temperature). The recent renewed interest in multiferroic materials in which piezoelectricity and ferromagnetism may coexist is also driven by the same objective of practical application (3, 4). In this case, the deformation of the material's structure under E results in the modification of the magnetic properties and vice versa. Again, this effect is limited to low temperatures in single-phase multiferroics.

Because the TC values of the 3d metals—Fe, Co, and Ni—and of some of their alloys are above room temperature, the possible use of an electric field to modify and control the intrinsic magnetic properties [e.g., magnetization or magnetocrystalline anisotropy (KU)] of such metallic systems is attractive. However, as a result of screening by the E-induced surface charge, the field does not penetrate the bulk of the material and is confined to a depth on the order of atomic dimensions. It may be expected that substantial E-induced effects may be found in nanosystems where the surface-to-volume ratio is high, with a large E obtained by the application of a high voltage at both sides of an insulating layer (e.g., a thin dielectric deposited on the surface of the metal). However, the preparation of thin and homogeneous dielectric layers free of pinholes is challenging. The immersion of metal particles in an electrolyte was suggested as an elegant approach to overcome this problem (5). In a liquid electrolyte, an insulating ionic layer—the electrolytic double layer—forms naturally in front of a conducting surface, and most of the potential drop between the electrodes occurs across this layer (6). Subsequently, it was shown (7) that a macroscopic strain of 0.15% is created in Pt nanoparticles under a voltage difference of 0.6 V, and this was attributed to the modification in metal bonding resulting from the E-induced change in the surface electron density at the Fermi level EF.

In metallic systems that show itinerant electron magnetism, the material's intrinsic magnetic properties are primarily determined by unpaired d electrons with energies close to EF. It isexpected that these properties will be affected by changing the electron density at EF under E. We selected L10-ordered FePt and FePd ultrathin films, because these compounds combine high TC (750 K), saturation magnetization (1.4 T), and KU (6.6 MJ/m3 in FePt and 1.8 MJ/m3 in FePd) values and exhibit high coercivity. In addition, these compounds are chemically stable.

Epitaxial films of FePt and FePd (2 and 4 nm thick) were grown by means of Pt and Pd buffer layers, respectively, on MgO(001) substrates (8). The superstructure lines as seen by x-ray diffraction indicate that the alloy crystallizes in the L10 phase, with the Fe and Pt (or Feand Pd) atoms being highly ordered on their two respective crystallographic sites. All films exhibit the magnetic easy axis perpendicular to the layer plane with anisotropy fields in excess of 8 and 2.5 T, as expected for the tetragonal L10 FePt and FePd compounds, respectively (8).

The effect of E on the alloy magnetic properties was observed at room temperature by measurements of the magnitude of the signal and coercive field (HC) via the polar Kerr effect. For these measurements, the electrolyte propylene carbonate (PC) was chosen to prevent hydrogen formation and its diffusion into the film when a negative potential was applied to the sample. Traces of water in the electrolyte were removed by introducing small Na pieces, which at the same time provided the Na+ ions necessary for the formation of the electrolytic double layer at the sample surface. A large sheet of inert Pt was used as the counter electrode (Fig. 1).

The HC variation as a function of the applied voltage U was derived from the measurements of the easy-axis hysteresis loops (Fig. 2). A sharp switching of the magnetization occurs at HC. To avoid possible dissolution of Fe from the alloy, we maintained the sample potential at or lower than –400 mV. The potential was also restricted to values above –1 V in order to avoid other electrochemical reactions at the film surface. Small but detectable irreversible film degradation occurred above –400 mV and below –1 V. Figure 3 shows the coercivity change δHC (relative to the value at –400 mV) versus U for FePt and FePd. A change of voltage ΔU = –600 mV results in a δHC by –4.5% for FePt and +1% for FePd films with a thickness of 2 nm. The 4-nm-thick films show a lower variation of –1% for FePt and no variation (within experimental resolution) for FePd. The observed change in HC as a function of U was essentially reversible, and the Kerr signal variation with magnetic field was the same upon increasing E as upon decreasing it. Both results imply that the observed effect is intrinsic and not due to film contamination or degradation that would be associated with the occurrence of irreversible phenomena (8). Note that δHC is stronger than a linear response in all cases.

In addition to the effect of E on the coercivity, an increase in the Kerr rotation by 3% is observed upon changing U from –400 to –1000 mV for the 2-nm-thick FePt film. Within experimental accuracy, no such effect was found in the other samples.

The results reveal that the magnetic properties of itinerant electron magnetic systems can be changed in a controlled way under E. During the measurements, the film microstructure is expected to remain essentially unaltered (to the extent that the very slow irreversible film degradation can be neglected), and, to first-order approximation, the coercivity can be assumed to be directly proportional to the KU from which it originates. The observed δHC values were evaluated with respect to KU energies (MAE) derived from electronic structure calculations (9). A capacitance value C per surface area A of C/A = 30 μF/cm2 was used, because this is a typical value for a clean metal surface in an electrolyte such as PC that is characterized by a large dielectric constant k = 66 (10). Under a voltage U that acts almost exclusively at the FePt/PC interface because of the 30 times larger surface area of the Pt counter electrode as compared to the sample surface, a charge CU accumulates in the whole film volume Ad, where d is the film thickness. Considering that the volume of the L10 FePt or FePd primitive cell is Vcell = a2c/2 = 0.027 nm3, where a is 0.385 nm and c is 0.371 nm, the charge variation per unit cell becomes $Math$(1)

In particular, for U = 600 mV and d = 2 nm, this amounts to 0.015 electrons per unit cell. From electronic structure calculations (9), a decrease of the MAE by 200% per electron is predicted for FePt and an increase of the MAE by 70% per electron is predicted for FePd. This result corresponds to expected changes of –3% and +1% for 2-nm-thick FePt and FePd, respectively. The induced excess charge in such metallic films is concentrated close to the surface and not distributed homogeneously. However, to first-order approximation, the anisotropy variation depends solely on the total film excess charge. This is a direct consequence of the linearity of all operations (i.e., it does not make a difference whether an anisotropy distribution is calculated first and then averaged over the whole film or whether the electron distribution is averaged right away).

The sign and magnitude of δHC that was measured agree well with the calculations for both alloys. It is quite notable that this simple concept is sufficient to explain the experimental results. Considering that we are dealing with ultrathin films, surface anisotropy should be included in the comparison between the experiment and calculation. However, neither experimental nor calculated data are available on the surface anisotropy of FePt. As a very qualitative argument, it may be argued that surface anisotropy in FePt is not expected to be dominant to the same extent as it is in cubic systems where surface symmetry breaking is the only source for the occurrence of a second-order anisotropy term. It has been shown (11) that there is only a weak thickness dependence of the nucleation field in FePt down to 2 nm, which supports this assumption.

The fact that the voltage dependence of the coercivity is stronger than a linear response can be attributed to a change in the thickness of the electrolytic double layer with U: When the charge is increased at the sample surface, the double layer is compressed and, in turn, C and E acting on thesampleare increased.

Because the Kerr angle is another intrinsic material property, its variation (detected in FePt only) upon the application of a voltage is another proof of the influence of E on the magnetic properties. However, in the absence of any calculated data, it is not possible to discuss the importance of the observed effect any further.

Our results have shown that the magnetic properties of thin-film ferromagnetic systems can be varied in a controlled way by an applied electric field. It must be stressed that the majority of modern magnetic materials belong to this category of materials. Beyond this proof of principle, this approach could offer an alternative and attractive generic actuation mechanism for electronic and electromechanical systems.

Supporting Online Material

Materials and Methods

Figs. S1 to S3

References