100-nm-sized magnetic domain reversal by the magneto-electric effect in self-assembled BiFeO3/CoFe2O4 bilayer films

A (001)-epitaxial-BiFeO3/CoFe2O4 bilayer was grown by self-assembly on SrTiO3 (100) substrates by just coating a mixture precursor solution. The thickness ratio of the bilayer could be controlled by adjusting the composition ratio. For example, a BiFeOx:CoFe2Ox = 4:1 (namely Bi4CoFe6Ox) mixture solution could make a total thickness of 110nm divided into 85-nm-thick BiFeO3 and 25-nm-thick CoFe2O4. Self-assembly of the bilayer occurred because the perovskite BiFeO3 better matched the lattice constant (misfit approximately 1%) and crystal symmetry of the perovskite SrTiO3 than the spinel CoFe2O4 (misfit approximately 7%). The magnetic domains of the hard magnet CoFe2O4 were switched by the polarization change of BiFeO3 due to an applied vertical voltage, and the switched magnetic domain size was approximately 100nm in diameter. These results suggest that self-assembled BiFeO3/CoFe2O4 bilayers are interesting in voltage driven nonvolatile memory with a low manufacturing cost.

As for the process for fabricating solid-state memory devices, most are fabricated using an ultra-high-vacuum process; therefore, expensive fabrication equipment is necessary. If these devices could be fabricated using a wet chemical process, the low cost of investment in facilities at the start of research would be an advantage. In addition, using spray coating or chemical solution deposition (CSD) would lower the fabrication cost per unit. In a previous report, when a Bi-rich BiFeO 3 target was used, BiFeO 3 was epitaxially grown on the SrTiO 3 (001) substrate, and excess Bi was grown on the BiFeO 3 as Bi 2 O 3 9 . Due to the large lattice mismatch between Bi 2 O 3 and SrTiO 3 compared with that of BiFeO 3 and SrTiO 3 , Bi 2 O 3 formed at the surface of the film. This result suggested that by adjusting the lattice misfit between a SrTiO 3 substrate and a ferromagnet or BiFeO 3 , a BiFeO 3 /ferromagnet bilayer can be prepared even by a one-timeonly liquid phase process. CoFe 2 O 4 is a candidate material for a ferromagnetic layer because the lattice mismatch between CoFe 2 O 4 and SrTiO 3 is 7.3%, which is much larger than that between BiFeO 3 and SrTiO 3 (1.4%). These material combinations are expected to enable fabrication of a BiFeO 3 /CoFe 2 O 4 bilayer on a SrTiO 3 substrate by utilizing the differences in lattice mismatch. Whether the magnetization can be reversed by the ME effect of BiFeO 3 when using materials with high-magnetic-anisotropy has yet to be verified in the case of a bilayer system. Without that verification, it is not possible to apply BiFeO 3 to high-density memories. From the viewpoint of the magnetocrystalline anisotropy energy (K u ), CoFe 2 O 4 is one of the candidate hard magnetic materials (K 1 of approximately 3 3 10 6 erg/ cm 3 in bulk). Self-assembled CoFe 2 O 4 -BiFeO 3 (or BaTiO 3 ) epitaxially grown on SrTiO 3 substrates using pulsed laser deposition (PLD) has been reported [10][11][12][13][14][15][16] . In many of these reports, the CoFe 2 O 4 nanopillars were embedded in the BiFeO 3 matrix; namely, it is a nanocomposite structured film. The CoFe 2 O 4 nanopillars were complexly influenced by the exchange bias from the BiFeO 3 matrix; moreover, they might be influenced by a strain effect 13 from the BiFeO 3 because a nanopillar has a degree of freedom to move in the vertical direction of the film. In the case of a bilayer structure, the ME effect becomes simpler than that in the case of a nanopillar structure. However, polarization reversal using a BiFeO 3 /CoFe 2 O 4 bilayer has not been phenomenologically investigated much in terms of magnetization switching. In the meantime, the minimum magnetic domain size in the case of a layered structure is not clear. It can be considered that the growth of the PLD process is not strongly influenced by the differences between the lattice misfits of CoFe 2 O 4 and SrTiO 3 substrates because, for example, PLD can prepare a non-equilibrium phase. A wet chemical process of thermal equilibration (such as CSD) is expected to more effectively expose the influence of lattice misfit of materials.
In this study, we demonstrated a novel chemical solution method; namely, introducing lattice misfit in relation to a single crystal SrTiO 3 substrate using a BiFeO 3 /CoFe 2 O 4 bilayer film, is proposed. The BiFeO 3 /CoFe 2 O 4 bilayer was microfabricated as a vertical device structure to reduce the operation voltage based on the ME effect, and it was experimentally verified that CoFe 2 O 4 magnetic domains (of 100-nm-diameter scale) with relatively high magnetocrystalline anisotropy could be switched.
Enhanced metalorganic decomposition (EMOD) solutions (Kojundo Chemical Laboratory Co., Ltd.) were used in this study. Two metal-diethylhexanoate compositions were used as the starting precursor solutions. As for the first, the ratio of Bi 31 and Fe 31 precursors was even in terms of atomic percent (P BFO ) (i.e., a typical condition for preparing BiFeO 3 films). For the second, the ratio of Co 21 and Fe 31 precursors was 152 in terms of atomic percent (P CFO ) (i.e., a typical condition for preparing CoFe 2 O 4 films). Each precursor solution was mixed to form an atomic percent ratio (P BFO 5P CFO ) of 451. The mixed precursor solution ratio (%), i.e., Bi 31 5Fe 31 5Co 21 , was 3655559. The solution was spin-coated at 6000 rpm for 50 sec on a 5-at.% La-doped SrTiO 3 (La-SrTiO 3 ) (001) conductive single crystal substrate. The spin-coated films were dried at 150uC for 1 min and calcined at 350uC for 5 min using hot plates in air. This process was repeated six times. Sintering for crystallization was carried out at 650uC for 10 min in air using an infrared lamp heating system. Circular Pt top electrodes (with a diameter of 100 mm and a thickness of 60 nm) were deposited on the film surface by DC magnetron sputtering. The crystal structures of the films were evaluated by X-ray diffraction (XRD; Philips, X9pert-Pro MRD). The cross sections of the films were observed by transmission electron microscopy (TEM; Hitachi HF-2000, FEI Company Tecnai G 2 F20). The ferroelectric properties of the films were evaluated at 90 K using a ferroelectric tester (TOYO Corporation, FCE-1A). The temperatures of the films were measured using a thermocouple contacted with silver paste. The ferroelectric domains were observed by piezoelectric force microscopy (PFM; Asylum Technology Cypher). Various poling methods were carried out before ferroelectric measurement. The magnetic hysteresis loops were measured using a superconducting quantum interference device (SQUID; Quantum Design MPMS). The magnetic domains were observed by magnetic force-microscopy (MFM; Bruker AXS Digital instruments, NanoScope IVa, Dimension 3100 stage AFM system).
The cross-sectional TEM observations of the film on the La-SrTiO 3 (100) substrate are shown in Figs. 1(a) to 1(g). A bright-field TEM image is shown in Fig. 1(a). A slight bumpy contrast between the film surface (bright) and film body (dark) was obvious. The nanobeam electron diffraction patterns of the regions indicated by circles (i) to (iv) in Fig. 1(a) are shown in Figs. 1(b)-1(e), respectively. At the surface area, randomly oriented CoFe 2 O 4 diffraction spots were observed in Fig. 1(b). BiFeO 3 diffraction spots were observed in Fig. 1(d) and these diffraction spots corresponded to those of the La-SrTiO 3 (100) substrates [ Fig. 1(e)]. This correspondence indicated that BiFeO 3 was epitaxially grown on the La-SrTiO 3 substrates in a cube-on-cube crystal relationship (under the assumption of a pseudo-cubic-perovskite structure in BiFeO 3 ). Both epitaxial BiFeO 3 and randomly oriented CoFe 2 O 4 diffraction spots coexisted at the interface [ Fig. 1(c)]. The high-resolution TEM image in Fig. 1(f) revealed that the interface between CoFe 2 O 4 and BiFeO 3 was clear and no intermixing interfacial layer of these two materials occurred. It should be noted that although a mixture precursor solution was used and the film was deposited only once, the film was separated into two layers, namely, an epitaxial BiFeO 3 layer and a polycrystalline CoFe 2 O 4 layer. The lattice misfit between BiFeO 3 and the SrTiO 3 substrate (1.4%) is smaller than that between CoFe 2 O 4 and the SrTiO 3 substrate (7.3%). It is therefore considered that the small lattice misfit of BiFeO 3 was the reason for its precedential growth in regard to CoFe 2 O 4 , which meant that bilayer samples can be produced by utilizing the lattice mismatch with respect to the substrate materials. We believe that bilayer films formed instead of nanopillar composite films 9-10,14 because crystal growth by CSD is in a thermal equilibrium state in comparison with that by PLD; as a result, the influence of the differences between lattice misfits of the substrates and epitaxial films was strong in the case of CSD. The total film thickness was estimated to be 110 nm, and the thicknesses of the BiFeO 3 and CoFe 2 O 4 layers were estimated to be 85 nm (t BFO ) and 25 nm (t CFO ), respectively. It is noteworthy that the thickness ratio (t BFO 5t CFO 5 3.451.0) approximately corresponded to the precursor-solution ratio under the assumption that the ratio of BiFeO 3 and CoFe 2 O 4 was given as P BFO 5P CFO 5 4.051.0. This result indicated that the thickness ratio of BiFeO 3 and CoFe 2 O 4 can be controlled by adjusting the composition ratio of P BFO and P CFO . The h -2h X-ray diffraction (XRD) pattern is shown in Fig. 1(h), and grazing incident XRD (GI-XRD) patterns are shown in Figs. 1(i) and 1(j). The incident angle was fixed at 1.5u for the GI-XRD measurements. The diffractions by the (311) and (440)

www.nature.com/scientificreports
The positions of the (311) and (440) diffraction patterns corresponded to the bulk CoFe 2 O 4 diffraction angles, indicating that CoFe 2 O 4 in the BiFeO 3 /CoFe 2 O 4 bilayer film was not strained in the as-grown state. BiFeO 3 showed an ME effect based on the switching of three different polarizations due to its rhombohedral distorted BiFeO 3 7,8 . In the case of tetragonal symmetry, only polarization of the (001) plane switched, and symmetry of the antiferromagnetic plane does not change; therefore, the ME effect does not occur. The crystal symmetry of BiFeO 3 is strongly influenced by the type of substrate, and the conditions of the sputtering; therefore, determine the crystal symmetry in film form is a key factor in using the ME effect. X-ray reciprocal space maps (RSMs) around the 004 and 204 spots are shown in Figs. 1(k) and 1(l), respectively. When BiFeO 3 has a tetragonal symmetry, only one spot is observed in the 204 RSM regions, and two spots are observed along the Q z axis if BiFeO 3 has a rhombohedral symmetry. The split of diffraction spots related to the rhombohedral symmetry were clearly observed; thus, the BiFeO 3 layer had a rhombohedral crystal symmetry with a space group of R3c. The lattice parameters of BiFeO 3 (estimated from the RSMs) were a 5 0.396 nm and b 5 89.5u, which corresponded to those of bulk BiFeO 3 17 . The polarization-electric field (P-E) hysteresis loops of the (001)epitaxial-BiFeO 3 /CoFe 2 O 4 bilayer film is shown in Figs. 2(a). The P-E loop with a rounded shape measured at 2 kHz became sharp as frequency increased above 5 kHz. The leakage current was linear as a function of time; however, ferroelectric switching occurred within a few tens of nanoseconds. Sharp P-E loops were therefore obtained at high frequency due to the reduction of the leakage current. The polarization value of the BiFeO 3 /CoFe 2 O 4 bilayer film estimated from the 20 kHz loop was 91 mC/cm 2 . To confirm the polarization of the BiFeO 3 /CoFe 2 O 4 bilayer film, the electrical displacement was measured by the positive-up-negative-down (PUND) method. [Fig. 2(c)] Schematic illustrations of the PUND responses for positive and up pulses are divided into four components: spontaneous polarization (SPC), initial increment (IC), paraelectric (PC), and leakage (LC) components.
[ Fig. 2(d)] For comparison of time scale, the wave forms of the applied voltage to the samples for the P-E (2 kHz) and PUND are illustrated in Fig. 2(b). SPC and PC increased when the electric field applied, and LC linearly increased. SPC remains; however, PC and LC disappear when the electric field was removed. SPC can be calculated by subtracting PC from IC. It is considered that the PUND method may express spontaneous polarization more accurately than with the P-E loops measured by a ferroelectric tester. The PUND measurement was discussed in detail elsewhere 18 . The spontaneous polarization evaluated by PUND was 84 mC/cm 2 , which is almost same as that evaluated by P-E loops. The ferroelectric switching characteristics of the local area and the influence of the surface potential on the magnetic stray field were evaluated by scanning probe microscopy (SPM). A PFM phase loop and amplitude loops taken by switching spectroscopy (SS) are shown in Fig. 3(a) and Fig. 3 Fig. 3(i)] 21 The SS-PFM phase loops showed that the ferroelectric domain showed a small coercive electric field compared with that observed in the P-E hysteresis loops. In the case of the P-E measurement, the diameter of the Pt electrode was 100 mm which included the various types of ferroelectric domains. In contrast, the SS-PFM measurement selected the ferroelectric domains having active responses to the vertical direction, which might be the reason for small coercive electric field in SS-PFM to the vertical direction compared with those obtained from the P-E measurement. Before the ME effect was measured, images of the BiFeO 3 /CoFe 2 O 4 bilayer film were taken using MFM and Kelvin force microscopy (KFM). [Figs. 3(e) and 3(f)] MFM is detected in the tapping mode of the phase shift of the resonance frequency of the cantilever due to attractive (or repulsive) forces between the magnetized cantilever and the magnetic moment at the film surface. AFM and KFM are also used in tapping mode to observe the surface morphology and surface electrical potential, respectively. To clarify the influence of the surface electrostatic potential and the surface morphology on the MFM measurement, first obtained the surface morphology image by AFM, and then the cantilever distant from surface, KFM measurements were performed using an electrical feedback circuit. Finally, MFM measurements were carried out. The difference in the contrasts of the MFM image and the KFM image revealed that the surface electrostatic potential was not necessary taken into consideration for the ME effect measurement.
To reduce the switching voltage and understand the phenomenology of the ME effect in the vertical direction, a two-step electrode was designed. A thick top electrode is necessary to prevent penetration of an electrode by the detection needle used for measuring the ferroelectric polarization reversal, where a thin electrode is necessary for MFM to detect the stray-magnetic-field signal. To solve these two contradictory matters, the needles for detecting the ferroelectric polarization reversal were connected to 60-nm-thick electrodes, and the MFM tips were connected to 5-nm-thick electrodes. A two-step Pt electrode was prepared by manually shifting the shadow mask slightly when the Pt electrode was sputtered for the first and second times. A schematic diagram of the setup for evaluating the ferroelectric and magnetic properties is shown in Fig. 4(a). The reversal of the ferroelectric domains in the 60-nm-thick electrode area was confirmed by investigating the relationship between the electrode size and the polarization reversal charge. For theME measurements, the polarization of BiFeO 3 was switched in the upward direction by applying a voltage of 20 V (1.8 MV/cm), and then the magnetization reversal was observed by MFM. As described in Fig. 2(a), 20 V (1.8 MV/cm) could reverse the ferroelectric polarization. To confirm the polarization reversibility, the polarization of BiFeO 3 was switched in the downward direction, and the magnetization reversal in the same area was again observed by MFM. The switched polarization in BiFeO 3 was stable for several weeks at RT in air. MFM observation was carried out in taping mode with a CoCr coated cantilever, and the magnetization direction (magnetic north) of the cantilever was upward to the film plane. The space resolution deduced from the digital data points (256 points/ 1.0 mm) is 4 nm. In fact, the space resolution of MFM observation depends on the distance between the cantilever and the film surface; therefore, the actual space resolution was a few tens of nanometers. In order to reduce the influence of the atomic force in tapping mode, the distance of the cantilever was slightly increased during MFM observation compared with AFM observation. The gap between the cantilever and the film surface was kept at approximately 20 nm during the MFM observation; that is, the actual distance between the CoFe 2 O 4 and the cantilever was approximately 25nm.  Fig. 4(d), which showed many dotlike contrast changes. Here, the influence of piezoelectric strain due to the polarization reversal in BiFeO 3 domains on magnetization in CoFe 2 O 4 was explained as follows. Piezoelectric intrinsically cannot conserve their strain without an electric field; therefore, a large piezoelectric constant is necessary for the polarization reversal. The piezoelectric constant (d 33 ) estimated from the slope of the SS-PFM amplitude in the case of the BiFeO 3 and CoFe 2 O 4 /BiFeO 3 bilayer films was approximately 15 pm/V and 10 pm/V, respectively were smaller than the reported range of 50 to 100 pm/V 22 . The piezoelectric strain under an applied electric field of 1.0 MV/cm was less than 1%, which meant that the piezoelectric strain from the BiFeO 3 layer could not reverse the magnetization of CoFe 2 O 4 . It may be noted that interface of the CoFe 2 O 4 and BiFeO 3 layers was slightly wavy, [ Fig. 1(a)] which might make the strain influence on the ME www.nature.com/scientificreports SCIENTIFIC REPORTS | 5 : 9348 | DOI: 10.1038/srep09348 effect enhanced at the interface edge. It can thus be considered that the magnetization switching of CoFe 2 O 4 was basically derived from polarization switching of BiFeO 3 and exchange coupling between antiferromagnetic BiFeO 3 and ferromagnetic CoFe 2 O 4 ; moreover, the strain effect at the edge of interfaces possibility enhanced the ME effect. In Fig. 4(d), the polarization was partially reversed; www.nature.com/scientificreports however, the contrast did not change across the whole area because the BiFeO 3 epitaxial film has three ferroelectric domains and only the 180u ferroelectric domains can not switch the magnetization by the ME effect. Therefore, the 71u and 109u ferroelectric domains might have respond to the ME effect 23,24 . Another possible reason is that even though the magnetization was switched by the 71u and 109u domains, the magnetization change of CoFe 2 O 4 was not always in a state that was not directly detected as a change in the cantilever direction of MFM; Enlarged MFM images with typical upward and downward polarizations area, indicated as green squares (i) in Figs. 4 (b) and 4(c), are shown in Figs. 4(e) and 4(f). The gradation of brown contrast differed in the cases of upward and downward polarization. Line profiles taken from the areas of the MFM images indicated as green squares (i), (ii), and (iii) are shown in Fig. 4(g). These line profiles showed that the magnetic domains were reversed by the electric field in three areas; in particular, the smaller magnetic domains indicated by areas (i) and (ii) were approximately 100 nm in diameter. As mentioned above, the actual space resolution of MFM was a few tens of nanometers; therefore, a magnetizationswitching signal from an area with diameter of approximately  100 nm was within the measurement range. As for the size of a magnetic domain [ Fig. 4(d)], the minimum size of a switched magnetic domain was around a few dozen nano meters in diameter. These reversed magnetic domain sizes were roughly consistent with ferroelectric domain size of BiFeO 3 . [Figs. 3(a) and 3(c)] The ferromagnetic and ferroelectric domains seemed to couple in a one-toone relationship 8 . However, the MFM resolution is lower than that of the PFM contact measurement; accordingly, small magnetic domains are necessary for further investigation. In this study, the relatively large magnetocrystalline anisotropy of CoFe 2 O 4 could be reversed by applying an electric field through exchange bias, and this result indicated that materials with higher magnetocrystallinity (such as an L1 0 -ordered alloy) can be expected to be applied to solid-state memories in the near future.
BiFeO 3 /CoFe 2 O 4 bilayer films were formed on La-SrTiO 3 (001) substrates by a one-time-only liquid phase process that involved spin coating of a mixed precursor solution. Cross-sectional TEM analysis confirmed that the BiFeO 3 was epitaxially grown on the La-SrTiO 3 substrates and that polycrystalline CoFe 2 O 4 was grown on the BiFeO 3 layer. The bilayer could be formed by an ''all-at-once chemical process'' in which BiFeO 3 (i.e., not CoFe 2 O 4 ) preferentially grew on the La-SrTiO 3 (001) substrate because the lattice mismatch between BiFeO 3 and La-SrTiO 3 is much smaller than that between CoFe 2 O 4 and La-SrTiO 3 . Two-step top electrodes were used to evaluat the ME effect generated by applying a vertical electric field. The orientation of the small magnetic domains of CoFe 2 O 4 changed when the polarization of BiFeO 3 was switched to the opposite direction by applying a voltage. The key points regarding the ME effect are twofold: a material such as CoFe 2 O 4 with relatively large magnetocrystalline anisotropy could be switched through ME coupling with BiFeO 3 , and the orientation of the magnetic domains of CoFe 2 O 4 (namely, 100 nm in diameter) could be reversed. These results suggest that a novel, low power, high-density MRAM and HDDs, which can be written by applying a voltage, can be created on the basis of the two points described above.