Hysteretic Magnetoelectric Behavior of CoFe2O4–BaTiO3 Composites Prepared by Reductive Sintering and Reoxidation

Magnetoelectric composites (CoFe2O4)x–(BaTiO3)1-x with x = 0.1, 0.2, 0.3, 0.4 and 0.5 were prepared by a polyol mediated synthesis route. The densification step was performed in a reducing forming gas atmosphere at 1250 1C. Under these conditions Co2 and Fe3 are reduced to the corresponding metals and thus a reaction of these elements with the BaTiO3 matrix is prevented. Reoxidation in air to CoFe2O4 at a rather low temperature of 800 1C leads to samples free of secondary phases. While the course of the synthesis was monitored by TGA, XRD and photometric analysis, the final composites were characterized via SEM, EDX, impedance spectroscopy and magnetic and magnetoelectric (ME) measurements. All samples show pronounced ME hystereses and aME maxima dependent on the Hdc field sweep direction. The sample with x = 0.4 exhibits the highest maximum aME of 420 mV Oe-1 cm-1.


Introduction
Multiferroic composites consist of a combination of compounds, which exhibit at least one ferroic order phenomenon like ferroelectricity and ferro-or ferrimagnetism. The coupling between magnetostrictive and piezoelectric phases allows manipulating the electric polarization by a magnetic field or the magnetization by an electric field. These so-called direct and indirect magnetoelectric (ME) effects promise new applications and devices, such as spintronics and MERAMS. [1][2][3][4] Several composites have been identified as magnetoelectric materials. 5 The first discovered magnetoelectric composite -and still one of the most prominent combinations -is the system consisting of CoFe 2 O 4 (CFO) and BaTiO 3 (BTO). 6 The hard ferrimagnetic CoFe 2 O 4 shows large magnetostriction while BaTiO 3 is ferroelectric with high piezoelectricity. Furthermore, this system is free of resource-critical elements and is predicted to exhibit high ME voltages. 7-10 CoFe 2 O 4 and BaTiO 3 have already been combined in all three prominent connectivities for composite multiferroics, namely 0-3, 1-3 and 2-2. 11,12 The reported a ME values for 0-3 composites are typically ranging from a few mV Oe À1 cm À1 to 10 mV Oe À1 cm À1 . [13][14][15] However, the theoretically predicted much higher a ME values have not yet been achieved. 16 This is most often explained by an insufficient interface between the piezoelectric and magnetostrictive phase that prevents mechanical transfers and thus good coupling. One approach to tailor the interface is to build up composites from nanoparticles because of their large surface to volume ratio. One disadvantage of this strategy is that impurities which originate from the reaction of BaTiO 3 with CoFe 2 O 4 at the interface, like substituted barium hexaferrite Ba(Co 0.5 Ti 0.5 ) x Fe 12Àx O 19 , are favored by large interface areas. While pure BaFe 12 O 19 is multiferroic on its own 17,18 and even composites with perovskites have been investigated, 19,20 the incorporation of Co 2+ and Ti 4+ in BaFe 12 O 19 has a significant impact on its magnetic properties. 21,22 To overcome the formation of impurity phases, we present in this article a polyol mediated synthesis route for magnetoelectric CoFe 2 O 4 -BaTiO 3 composite ceramics. Using a reductive sintering step followed by reoxidation under mild conditions, the formation of Ba(Co 0.5 Ti 0.5 ) x Fe 12Àx O 19 was successfully suppressed and phase pure samples with relative densities 480% were obtained. Measurements of the magnetoelectric coupling show frequency independent values of the ME coefficient above 300 Hz. All samples exhibit a pronounced ME hysteresis with a maximum of a ME at H dc E AE2500 Oe. Furthermore, we for the first time describe a dependence of the maximum a ME values on the direction of the DC-field sweep.
composites were synthesized with x CFO = 0.1, 0.2, 0.3, 0.4 and 0.5. A modified polyol mediated process was used to prepare the precursor powders for the ceramics. 23,24 The preparation scheme is shown in Fig. 1 and is described as follows.
Fe(NO 3 ) 3 Á9H 2 O (20 mmol, Sigma Aldrich) and Co(NO 3 ) 2 Á6H 2 O (10 mmol, Sigma Aldrich) were dissolved in deionized water (6 ml). NaOH (80 mmol, Grüssing) and diethylene glycol (250 ml, Carl Roth) were added to the red solution, which was then heated within 45 min to its boiling point (B160 1C) and refluxed for 1 h. After cooling to room temperature, acetone (250 ml, Overlack) was added, resulting in a brown precipitate. Subsequent centrifugation and washing with acetone led to a brown CoFe 2 O 4 precursor powder. The amount of ferrite was determined by thermogravimetric analysis.
For sample 0.5CFO-0.5BTO, 5 mmol of the presynthesized CoFe 2 O 4 precursor and 5 mmol Ba(OH) 2 Á8H 2 O (FLUKA) were mixed with diethylene glycol (250 ml, Carl Roth) in an argon flushed flask. Distilled Ti(i-OPr) 4 (5 mmol, Alfa Aesar) was added and the reaction mixture was heated within 45 min to the boiling point (B160 1C) and kept under reflux for 1 h. After cooling to room temperature, 250 ml of acetone (Overlack) were added to the gray suspension and subsequent centrifugation and washing with acetone led to a gray composite precursor. The precursor powders for the other samples were prepared accordingly, using the respective stoichiometric quantities.
The precursor powders were first calcined in static air at 700 1C for 1 h with a heating rate of 10 K min À1 . Afterwards, a first reduction step in flowing forming gas (80 ml min À1 , 10% H 2 ) at 950 1C for 1 h (heating rate 5 K min À1 ) was performed and the resulting light gray powders were pressed into disks (100 mg, + = 6 mm) that were sintered in flowing forming gas (80 ml min À1 , 10% H 2 ) at 1250 1C for 1 h (heating rate 5 K min À1 ). Finally, reoxidation in static air at 800 1C for 6 h (heating rate 10 K min À1 ) led to black ceramic bodies of CoFe 2 O 4 -BaTiO 3 composites.

Characterization
Thermogravimetric measurements in flowing synthetic air or forming gas (10% H 2 ) (flow rate 40 ml min À1 , heat rate 10 K min À1 ) were performed using a Netzsch STA 409 system. X-ray diffraction patterns were recorded at room temperature on a Bruker D8 Advance diffractometer operating with CuKa radiation. For the quantitative cobalt ferrite determination, small aliquots of the samples were dissolved in a mixture of hydrochloric acid (Z37%, Sigma Aldrich) and hydrogen peroxide solution (30%, Overlack). After dissolution, residual peroxides were decomposed by heating. The solutions were diluted to an approximated Fe concentration of 1 mg l À1 and Spectroquant Iron Test solution (Merck Millipore) was added. For absorbance measurements at l = 560 nm a VWR UV-3100PC Spectrophotometer was used. The Fe concentrations and corresponding CoFe 2 O 4 contents were determined by a calibration series. Scanning electron microscopy images in the backscattered electron (BSE) mode and EDX spectra were recorded using a Philips ESEM XL 30 FEG. For impedance measurements an eutectic Ga-In alloy was coated as electrodes on top and bottom surfaces of the ceramic bodies. The temperature and frequency dependent impedance spectra (0 to 180 1C; 100 Hz to 13 MHz) were recorded using a Hewlett-Packard 4192A impedance analyzer. Magnetic measurements were carried out using the ACMS option of a Quantum Design PPMS 9. Hysteresis loops were measured at 300 K with a cycling of the magnetic field between +90 and À90 kOe. For the magnetoelectric investigations, 100 nm thick gold electrodes were sputtered onto the sample surfaces using a Cressington Sputter Coater 108auto. Electric poling was done applying an electric DC field of 4 kV cm À1 to the samples at room temperature. Then the samples were heated to 200 1C for 1 h (heating rate 10 K min À1 ) and the electric field was dynamically adjusted setting the current limit to 0.1 mA. Due to the increasing conductivity of the samples the field decreased to a few V cm À1 at 200 1C but again reached 4 kV cm À1 during cooling to room temperature. After poling, the samples were short circuited for 10 min. Immediately afterwards, the ME measurements were performed at 300 K in a Quantum Design PPMS 9 using a custom made setup based on the AC-Transport measurement option. A magnetic AC field of H ac = 10 Oe with different frequencies was applied by a solenoid with 1160 loops of copper wire and the ME voltage was measured in dependence of the magnetic DC field upon cycling between +10 and À10 kOe. Magnetic AC and DC fields were aligned parallel to the electric polarization. Raw data were corrected for eddy currents measured on an empty sample holder. The magnetoelectric coefficient was calculated as a ME = U/(H ac Áh) with h being the sample height.

Thermal analysis
Simultaneous thermogravimetric and differential thermal analyses were carried out on the samples to investigate calcination, sintering and reoxidation behavior. The results are discussed for 0.3CFO-0.7BTO exemplarily as follows. Calcination of the dried composite precursor in air at 700 1C leads to a total weight loss of 45%. This weight change comprises the evaporation of residual solvent and adsorbed water, an exothermal combustion of organic residues and the endothermal decomposition of intermediately formed BaO x (CO 3 ) 1Àx . [25][26][27] Following the path of preparation and representing the calcination and sintering steps under reducing conditions, an air-calcined sample was heated in the thermobalance in flowing forming gas to 1250 1C and kept at this temperature for 1 h. The corresponding weight change and DTA curves are shown in Fig. 2. The main weight loss of 9.0% occurs between 350 and 1000 1C and is accompanied by two distinct exothermal signals in the DTA curve. During this step, the reduction of CoFe 2 O 4 to an alloy of Co and Fe (expected weight loss: 7.7%), the decomposition of residual BaCO 3 and the final formation of BaTiO 3 take place. During the dwell time of 1 h at 1250 1C a further slight weight loss of about 2% was observed. This weight loss could be caused by a gradual generation of oxygen defects and the accompanied partial reduction of Ti 4+ in the BaTiO 3 matrix. Overall a weight loss of 11.0% occurred.
Afterwards, the resulting CoFe 2 -BaTiO 3 sample was heated in synthetic air up to 800 1C and kept at that temperature for 6 h to investigate the reoxidation behavior. As shown in Fig. 3, a weight gain accompanied by an exothermic DTA signal starts at 430 1C and is finished about 1 h after reaching 800 1C. This process leads to a total weight change of +8.2%. Longer heating at 800 1C did not lead to any significant additional weight change. For the complete oxidation of the CoFe 2 alloy in the 0.3CFO-0.7BTO sample, a theoretical weight gain of 8.4% is expected. Thus, the observed change of +8.2% can be assigned completely to the reoxidation of the alloy to CoFe 2 O 4 as additionally supported by the XRD results discussed in the next paragraph.

X-ray diffraction
The course of the synthesis was monitored by X-ray powder diffraction as exemplarily shown for x CFO = 0.5 in Fig. 4. The diffractogram of the precursor obtained by precipitation without further temperature treatment (Fig. 4a) mainly shows crystalline BaCO 3 and CoFe 2 O 4 implying that at least one Ti-containing amorphous phase is present. Calcination in air at 700 1C for 1 h leads to the formation of BaTiO 3 although some BaCO 3 is still detectable in the powder (Fig. 4b). Calcination or sintering in air at temperatures higher than 700 1C leads to reactions between the BaTiO 3 matrix and the ferrite, i.e. the formation of impurities such as Ti 4+ -and Co 2+ -doped BaFe 12 O 19 , hexagonal BaTiO 3 as well as increased Co 2+ and Fe 3+ doping of BaTiO 3 . To avoid this, an additional calcination step in forming gas (10% H 2 ) at 950 1C for 1 h was performed. Under these conditions residual BaCO 3 decomposes and the X-ray diffraction pattern (Fig. 4c) indicates the formation of (pseudo-) cubic modification for BaTiO 3 . In addition, CoFe 2 O 4 is reduced to an alloy of Co and Fe that does not react with BaTiO 3 as the ferrite would. The resulting reduced powder was pressed into disks (+ = 6 mm, m = 100 mg) and sintered at 1250 1C for 1 h in forming gas. The corresponding X-ray diffraction pattern shows a mixture of tetragonal BaTiO 3 and the CoFe 2 alloy (Fig. 4d). Afterwards, the ceramic body was reoxidized in air at 800 1C for 6 h resulting in an oxidation of the CoFe 2 alloy to CoFe 2 O 4 (Fig. 4e).
As shown in Fig. 5, phase pure CoFe 2 O 4 -BaTiO 3 composites were obtained for all investigated compositions with x CFO = 0.1, 0.2, 0.3, 0.4, and 0.5. By Rietveld refinement the cell parameters   It was found that if the disks are too dense after the reductive sintering, the oxidation leads to cracks in the samples due to an increasing volume during the reaction of CoFe 2 to CoFe 2 O 4 . The optimal relative densities for reoxidizing (65 to 80%) are dependent on x CFO and decrease with higher CoFe 2 O 4 content. For the final samples relative densities, with respect to the (weighted) single crystal values, 28 between 80% and 90% were achieved.

Photometric CoFe 2 O 4 assay
In Table 1 the results of the photometric CoFe 2 O 4 quantifications are listed. The uniform deficiency in the CoFe 2 O 4 content compared to the nominal values is most likely caused by a certain solubility of the Co-and Fe-containing precursors during precipitation and washing of the composite powder. On the other hand the deviation between the expected and measured CoFe 2 O 4 content is below 2 mol% in most cases and for simplification the nominal percentages are used throughout the text for describing the compositions. Nevertheless, for all calculations with respect to x CFO , the experimentally determined contents were used.

SEM and EDX investigations
Scanning electron micrographs were taken from the polished surfaces of sintered samples as shown exemplarily in Fig. 6 for x CFO = 0.1, 0.3 and 0.5. A minor porosity was observed for all samples in accordance with the measured densities of 80% to 90%. The pores are distributed randomly throughout the samples. The BaTiO 3 phase was found to consist of grains with sizes of 1-5 mm independent of x CFO . From Fig. 6a it can be seen that for 0.1CFO-0.9BTO isolated CoFe 2 O 4 grains (dark gray) are embedded in the BaTiO 3 matrix (light gray) but most of the CoFe 2 O 4 is assembled in centers of partially interconnected grains separated and surrounded by BaTiO 3 . With higher x CFO values the amount of these CoFe 2 O 4 clusters increases. In 0.3CFO-0.7BTO only a small portion of the ferrite occurs as isolated particles completely surrounded by BaTiO 3 (Fig. 6b and  d). For x CFO = 0.4 and 0.5 (Fig. 6c), the CoFe 2 O 4 particles form irregular connected structures that permeate through wide areas, from several 10 to 100 mm in diameter.
To assign the two distinguishable phases in the BSE mode, EDX measurements were carried out at representative sample areas (Fig. 6d). The EDX spectra confirm the formation of two phases, namely CoFe 2 O 4 and BaTiO 3 , as can be seen in Fig. 7. The traces of barium and titanium, visible in the CoFe 2 O 4 spectrum, are due to the large interaction volume of the electron beam compared to the grain size.    In 0.5CFO-0.5BTO a third phase can be identified in the BSE images from its medium-gray contrast, although no additional reflexes show up in the corresponding X-ray diffraction pattern. EDX-line scans revealed this phase to be slightly oxygendeficient BaTiO 3Àd .

Impedance spectroscopy
In Fig. 8  According to Mitoseriu et al. the shift to 150 1C is due to the convolution of extrinsic defect-related dielectric relaxation with the intrinsic ferroelectric component. 33 The presence of the phase transition in the impedance data proves the formation of the tetragonal (ferroelectric) BaTiO 3 modification in accordance with the XRD results discussed above. The consistent phase transition temperatures hint at a uniform composition of the BaTiO 3 , indicating that no x CFO -dependent doping of Co 2+ or Fe 3+ occurred for any of the samples. The loss tangents at room temperature are below 0.5 with the exception of 0.5CFO-0.5BTO and only slightly increase with temperature below the tetragonal-cubic phase transition of BaTiO 3 .
Concerning the frequency dependence at room temperature, e 0 decreases rather strongly in the range of roughly 100 Hz-10 kHz and remains quite stable for higher frequencies for all samples, as can be seen in Fig. 10. This phenomenon is typical for electrically conducting particles embedded in an insulating matrix and is called Maxwell Wagner polarization. 34 A similar behavior has already been reported for various CoFe 2 O 4 -BaTiO 3 composites. 13,35,36

Magnetic properties
The field dependent magnetization curves of the composites are depicted in Fig. 11. All samples show clear hystereses in accordance with the ferrimagnetism of the spinel component. The saturation magnetization values were determined by a linear extrapolation of the magnetization in the high field ranges (50-90 kOe) to H = 0 Oe. As expected, saturation and remanent magnetization increase with CoFe 2 O 4 content.
Normalizing these magnetization values with respect to the ferrite content (i.e. emu g À1 of CoFe 2 O 4 ), as shown in Fig. 12, reveals a small maximum of M R with a value of 27.0 emu g À1 for x CFO = 0.3, while the M S data show a general slight increase with higher CoFe 2 O 4 contents, reflecting the increasing ferrite particle sizes. Coercivity fields increase from 1060 Oe (0.1CFO-0.9BTO) to a maximum of 1230 Oe for x CFO = 0.2 and decreases with higher CoFe 2 O 4 contents down to a minimum of 970 Oe for x CFO = 0.5. It is commonly known that coercivity values for CoFe 2 O 4 are dependent on its grain shape and structure. Therefore, the change in coercivity is probably due   to the formation of CoFe 2 O 4 clusters and interconnected structures with increasing x CFO values (see SEM part).

Magnetoelectric coupling
An example of the magnetic DC field dependence of the magnetoelectric coefficient a ME is shown in Fig. 13 for the composite with x CFO = 0.4. For all samples, the a ME values show maxima/minima around AE2.5 kOe and clear hysteretic behavior with coercive fields between AE400 and AE600 Oe as well as distinct remanent a ME values of AE7 to AE150 mV Oe À1 cm À1 . Thus, the sign of a ME is switchable with the direction of the applied magnetic DC field.
It is remarkable that the maximum values of a ME at AE2.5 kOe are dependent on the history of the magnetic DC field. Upon increasing the DC field strength (both in positive and negative direction), the a ME values are higher than for decreasing field strengths. For example, in the case of x CFO = 0.4 the maximum a ME value is 420 mV Oe À1 cm À1 when the magnetic field is increased up to 10 kOe (red circles, Fig. 13), whereas a ME(max) amounts to only 337 mV Oe À1 cm À1 when the field strength is lowered down from 10 kOe (blue diamonds, Fig. 13). To the best of our knowledge, such an effect has not yet been described for a ME and its origin needs to be investigated in further experiments.
A comparison between the magnetic and ME hystereses (Fig. 14) reveals that the ME coercivities are about half the size of the magnetic coercivites, i.e. the ME voltage diminishes although magnetization still exists. It is noteworthy that the DC field at which the maximum a ME is ovserved (AE2.5 kOe) shows no obvious relation to the characteristic points of the magnetic hysteresis. In particular, it is much smaller than the saturation field (roughly AE20 kOe) and also significantly different from the inflection points of magnetization which occur at about AE1 kOe. It seems that the ME effect is more affected by magnetostriction. For example, van Run et al. found the highest ME values at the maximum of the piezomagnetic strain per Oe. 37 In laminated PZT-Terfenol-D systems, ME maxima were found at the saturation point of the magnetostrictive strain 38 and at the maximum of the piezomagnetic coefficient. 39 Our paper is one of the few examples in which a hysteretic behavior of the ME coefficient is reported for CoFe 2 O 4 -BaTiO 3 composites. 37,[40][41][42][43] In the majority of articles (e.g. ref. 13 and 44-47), a ME values were found to be zero when the magnetic DC field was switched off, or only initial curves, i.e. without cycling of H dc , were published.     Fig. 15 shows the maximum a ME values for the investigated samples depending on x CFO and the frequency of H ac . The a ME values increase with higher CoFe 2 O 4 contents up to a maximum at x CFO = 0.4 with 420 mV Oe À1 cm À1 at 500 Hz and decrease again for x CFO = 0.5 at all investigated frequencies. The magnetoelectric remanences follow this trend and the sample 0.4CFO-0.6BTO shows the highest remanent a ME value with 150 mV Oe À1 cm À1 . The maximum a ME values are within the typical range of a few mV Oe À1 cm À1 to some mV Oe À1 cm À1 reported for CoFe 2 O 4 -BaTiO 3 composites with 0-3 connectivity measured under similar conditions. [13][14][15]41,43 ME measurements at five different H ac frequencies showed that a ME is significantly lower for f = 100 Hz than for higher frequencies, while for frequencies from 300 to 900 Hz the obtained a ME values show only little deviations. This is in contrast to previously reported frequency dependencies where a ME values increase more or less linearly up to 1 kHz. 44,45 Other articles show a giant increase of a ME (up to 2.5 V Oe À1 cm À1 ) under resonance conditions in the range between 15 and 430 kHz. [46][47][48] Measurements at such higher frequencies are therefore planned for the future.

Conclusions
In multiferroic composites, the strength of the magnetoelectric coupling depends on the quality of the interface and therefore an intimate contact between the two components is mandatory. Because of the low sintering activity of BaTiO 3 , such highquality interfaces require high sintering temperatures, which usually lead to the formation of secondary phases like BaFe 12 O 19 and/or ionic exchanges like the incorporation of iron or cobalt ions in BaTiO 3 . To avoid these problems we describe a new approach starting with the polyol-assisted soft-chemistry preparation of a CoFe 2 O 4 -BaTiO 3 composite precursor consisting of submicrometer-sized particles. This composite powder is sintered under a reducing atmosphere, resulting in CoFe 2 -BaTiO 3 composites. In a final reaction step these ceramics are reoxidized under mild conditions leading to the final dense CoFe 2 O 4 -BaTiO 3 composites (Z80% of the crystallographic density), which according to XRD are completely free of secondary phases.
The presence of tetragonal BaTiO 3 with lattice parameters independent of the ferrite fraction further indicates that no considerable amounts of Co 2+ or Fe 3+ are incorporated in BaTiO 3 . This finding is additionally corroborated by dielectric measurements showing the occurrence of the ferroelectric-paraelectric phase transition, which is shifted to 150 1C probably due to defect-related dielectric relaxations. EDX investigations gave no hints for secondary phases or detectable Ti/Fe/Co exchange between the two phases. All composites show ferrimagnetic behavior. Their (normalized) saturation magnetizations slightly increase with the CoFe 2 O 4 content, reflecting the increasing grain sizes of the ferrite. Field-dependent measurements of the magnetoelectric coupling show a hysteresis of a ME with maxima at DC magnetic fields of AE2.5 kOe and remanent a ME values in the order of 10-150 mV Oe À1 cm À1 . We observed distinct deviations between the coercivities of the magnetic moment and the ME effect, i.e. the magnetoelectric coupling diminishes already at roughly AE500 Oe while the required field to extinct the magnetization is about twice as high. Concerning the effect of the H ac frequency, we found almost constant a ME values from 300 Hz to 900 Hz while the magnetoelectric coupling is much smaller at 100 Hz. In addition, an increase of the ME effect with increasing CoFe 2 O 4 content was observed leading to a maximum in the order of 400 mV Oe À1 cm À1 for x = 0.4. With a higher ferrite content, a ME decreases again. Most interestingly, for all samples we observed strong differences for the maximum a ME values depending on the direction of the magnetic dc-field sweep. When the field strength is increased (both in positive and negative directions) the maximum ME values are up to 20% larger than upon decreasing H dc . Thus, the measured ME voltages of the CoFe 2 O 4 -BaTiO 3 composites not only depend on the H dc field strength but also on its history. This 'memory effect' may give rise to additional future applications.