Pronounced and reversible modulation of the piezoelectric coefficients by a low magnetic field in a magnetoelectric PZT-5%Fe3O4 system

Composite magnetoelectric compounds that combine ferroelectricity/piezoelectricity and ferromagnetism/magnetostriction are investigated intensively for room-temperature applications. Here, we studied bulk composites of a magnetostrictive constituent, ferromagnetic Fe3O4 nanoparticles, homogeneously embedded in a ferroelectric/piezoelectric matrix, Pb(Zr0.52Ti0.48)O3 (PZT). Specifically, we focused on PZT-5%Fe3O4 samples which are strongly insulating and thus sustain a relatively high out-of-plane external electric field, Eex,z. The in-plane strain-electric field curve (S(Eex,z)) was carefully recorded upon successive application and removal of an out-of-plane external magnetic field, Hex,z. The obtained S(Eex,z) data exhibited two main features. First, the respective in-plane piezoelectric coefficients, d(Eex,z) = 200–250 pm/V, show a dramatic decrease, 50–60%, upon application of a relatively low Hex,z = 1 kOe. Second, the process is completely reversible since the initial value of d(Eex,z) is recovered upon removal of Hex,z. Polarization data, P(Eex,z), evidenced that the Fe3O4 nanoparticles introduced static structural disorder that made PZT harder. Taken together, these results prove that the Fe3O4 nanoparticles, except for static structural disorder, introduce reconfigurable magnetic disorder that modifies the in-plane S(Eex,z) curve and the accompanying d(Eex,z) of PZT when an external magnetic field is applied at will. The room-temperature feasibility of these findings renders the PZT-x%Fe3O4 system a solid basis for the development of magnetic-field-controlled PE devices.

. The above mentioned studies [22][23][24][25][26][27] , as well as the majority of studies concerning ME composites [28][29][30][31][32] , examine the electric-field control of the magnetic properties of the composites. The inverse, i.e. the control of FE or PE properties by an external magnetic field, H ex , has been less investigated [33][34][35][36] , focusing mostly on the electric polarization. Corresponding studies, performed by means of piezoresponse force microscopy, demonstrate the modulation of FE domains, upon application of an H ex on the order of 2-9 kOe. Among them, Evans et al. 35 and Xie et al. 36 based on Pb(Zr 1−x ,Ti x )O 3 as PE component, studied the changes in polarization 35 and piezoresponse 36 upon H ex , for the cases of the ME composites PbZr 0.53 Ti 0.47 O 3 -PbFe 0.5 Ta 0.5 O 3 (PZTFT) single crystals and PbZrTiO-TbDyFe (PZT-TDF) bilayer, respectively. These changes, were significant in magnitude, although partially reversible 35 or irreversible at all 36 .
Apart from the polarization and piezoresponse, another basic PE parameter, the PE coefficients d ij , have not been investigated upon application of an H ex , until now. Here, we study the bulk composite system consisting of PE Pb(Zr 0.52 Ti 0.48 )O 3 (with composition at the so-called morphotropic phase boundary (MPB); called simply PZT for the rest of the paper) and FM Fe 3 O 4 nanoparticles (NPs) that have noticeable magnetostrictive nature. Specifically, the FM NPs are embedded in the PE matrix in the desired weight percentage PZT-5%Fe 3 O 4 . We focus on the variation of the in-plane PE coefficients upon application of an E ex,z (ranging within ± 10 kV/cm) and a relatively low H ex,z (1 kOe), that are both applied out-of-plane at room temperature. By means of a method based on optical microscopy 37,38 , we showed in a direct way that the application of such a low H ex = 1 kOe, causes a pronounced decrease of the in-plane PE coefficients on the order of 50-60%, that intriguingly is restored upon successive removal of H ex . Polarization data showed that with the addition of the Fe 3 O 4 NPs, PZT becomes harder when compared to its plain form. This is a proof that the FM NPs act as structural disorder that pins the FE domain walls efficiently, a fact that can be attributed to the similar size (50-100 nm) of the FM Fe 3 O 4 NPs (see 'Sample preparation' in 'METHODS') and the FE domains of PZT (see 17,[39][40][41]. These combined results prove that the FM Fe 3 O 4 NPs serve not only as a simple static structural disorder but, most important, as a reconfigurable magnetic disorder that can be applied at will by means of an external magnetic field to control the PE properties of PZT. Based on the strong PE character of PZT and the noticeable magnetostrictive nature of Fe 3 O 4 , these findings can be ascribed to a strain transfer mechanism realized at the interface of the FM Fe 3 O 4 NPs and the surrounding FE PZT matrix. Consequently, the quantitatively significant and qualitatively reversible modulation of the in-plane PE coefficients render PZT-5%Fe 3 O 4 a candidate ME material for applications in functional devices with high ME performance.

Results and Discussion
A complete series of PZT-xFe 3 O 4 samples (x = 0-50% w∕w), that is Fe 3 O 4 NPs homogeneously embedded in the PZT matrix as schematically shown in Fig. 1(a), was preliminary investigated to choose the appropriate insulating sample for the subsequent study. Obviously, a basic criterion is that the sample should remain strongly insulating for a reasonable range of electric field values. Figure 1(b), presents the leakage current as function of an electric field applied out-of-plane, E ex,z , for two specific samples with x = 5% and x = 20%. We clearly see that the sample with x = 5% remains insulating, while the sample with x = 20% exhibits a noticeable leakage current for a comparatively much lower value of E ex,z . Accordingly, all the experimental results presented below refer to PZT-5%Fe 3 O 4 . The constitutive Strain-Electric field curve, S(E ex,z ), of samples PZT-5%Fe 3 O 4 was estimated experimentally by using an OM-based technique already discussed in 37,38 for out-of-plane E ex,z up to 10 kV/cm. Here, we made another modification in the home-made aluminum platform hosting the sample, so that a constant, also out-of-plane magnetic field, H ex,z , is applied at will, by a NdFeB permanent magnet, (disc-shaped with diameter 20 mm and thickness 3 mm). The magnet is placed just below the sample as shown in Fig. 2(a-c) and explained in detail at its caption. By means of this experimental setup, the in-plane strain curves, S zx (E ex,z ) and S zy (E ex,z ), can be recorded even under the presence of an H ex,z (see 'Section I' in Supplementary Information). We note that all materials used in this construction are non magnetic, else the magnetic field produced by the NdFeB permanent magnet would be non-homogeneous. Since our study aims to investigate the PE response upon the absence/presence of an external magnetic field, the homogeneity of H ex,z is crucial. In this context, accurate mapping of H ex,z was performed along the z coordinate, as shown in Fig. 2(d), and along the ρ coordinate, as shown in Fig. 2(e) (vertical dotted lines denote the boundary of the permanent magnet). The sample is placed at z s = 5 mm, with its center aligned with that of the magnet. Accordingly, the data of Fig. 2(d,e) prove that the external magnetic field of value H ex,z (z s ) = 1 kOe is fairly homogeneous over the entire volume of the PZT-5%Fe 3 O 4 sample. This magnetic field brings the FM Fe 3 O 4 NPs close to saturation (see magnetization data, below).
Detailed XRD data is shown in Fig. 3(a-c) on a comparative basis, with assignment of the main peaks, for non-sintered PZT-5%Fe 3   field, E ex,z , produced by a DC-voltage supply and a constant magnetic field, H ex,z , produced by a disc-shape NdFeB permanent magnet. E ex,z can be varied continuously within −10 kV/cm to +10 kV/cm during a set of measurements, while H ex,z is constant, 1 kOe, and exhibits azimuthal symmetry. (c) Detail of (a) in the sample area. The top and bottom pins hold the sample at its center so that the entire sample is left free to deform. The black dashed rectangle denotes the NdFeB magnet that is embedded in the aluminum platform, with its center, O, aligned with that of the sample (vertical dotted lines denote the boundary of the permanent magnet). The coordinate axes z and ρ are shown with white dotted lines. (d) Variation of H ex,z along the z coordinate (ρ = 0 mm). (e) Variation of H ex,z along the ρ coordinate (z s = 5 mm). Figure 4(a-f) present detailed SEM data referring to, topography ( Fig. 4(a)), EDS spectrum for elemental analysis ( Fig. 4(b)), and EDS elemental mapping ( Fig. 4(c-f)) for a sintered PZT-5%Fe 3 O 4 composite (sintering conditions: T = 1000 °C, for t = 2 h in air) in its as-prepared form (without polishing). These results prove that the  . The deduced magnetization values for a magnetic field of 5 kOe are m = 63 emu/g, 0.4 emu/g, 6 × 10 −3 emu/g, and 24 emu/g, respectively. We see that the non-sintered Fe 3 O 4 NPs have high magnetization value, 63 emu/g, however lower than that of bulk magnetite due to their reduced size. When the Fe 3 O 4 NPs are sintered in free form (not included in a PZT matrix) they completely oxidize to α-Fe 2 O 3 as evidenced by the extremely lower magnetization value, 0.4 emu/g, obtained after sintering that is the magnetic fingerprint of hematite 44 . Referring to sintered PZT, it exhibits negligible magnetization, 6 × 10 −3 emu/g, as expected. Accordingly, we can conclude that for the sintered PZT-5%Fe 3 O 4 composite the deduced magnetization value, 24 emu/g, originates from Fe 3 O 4 NPs that have not oxidized to α-Fe 2 O 3 since they are protected by the PZT matrix. Using these magnetization values we can estimate that 38% of the NPs remain in the Fe 3 O 4 form after sintering. In agreement to the XRD results the magnetization data show that the former PZT-5%Fe 3 O 4 system evolves to a PZT-5%(Fe 3 O 4 /α-Fe 2 O 3 ) composite (for clarity, simply denoted PZT-5%Fe 3 O 4 in the rest of the paper). Figure 6 shows detailed data of the in-plane PE coefficients obtained upon application and removal of a constant H ex,z for a square-shaped sample, (a.i)-(d.i), and for a disc-shaped sample, (a.ii)-(d.ii). The insets of the upper panels, (a.i) and (a.ii), show photos of the specific samples together with an arbitrarily chosen coordinate system that assists us to define the symmetry axes x (SA x ) and y (SA y ). As it should be, the investigated area was at the sample edge, in the specific case onto SA y , as shown with a white dot for both samples (for details see 'Section III' of Supplementary Information and 37,38 ). Concerning the experimental procedure, six consecutive E ex,z loops were recorded designated by a loop index; first, without the application of H ex,z (loop index: 1 and 2), next, under the presence of H ex,z = 1 kOe (loop index: 3 and 4) and finally, when removing H ex,z (loop index: 5 and 6). Figure 6(a.i,a.ii) present the mean absolute value of the in-plane PE coefficient in the y-direction, <|d zy |>, for the complete sequence of measurements (the other in-plane PE coefficient in the x-direction, d zx , is not shown since it is obviously zero 37,38 ). Specifically, these panels present explicit data on < |d zy | > before applying (loop index: 1 and 2), upon application (loop index: 3 and 4) and after removal (loop index: 5 and 6) of a constant H ex,z = 1 kOe. For each loop the <|d zy |> is calculated from the slopes |dS zy /dE ex,z | of the linear parts of the detailed S zy (E ex,z ) data. Representative sets of the latter are shown in Fig. 6(b.i-d.i,b.ii-d.ii) for the square-shaped and disc-shaped sample, respectively, for loop index: 2 (without H ex,z , (b.i) and (b.ii)), loop index: 4 (with H ex,z , (c.i) and (c.ii)) and loop index: 6 (without H ex,z , (d.i) and (d.ii)). The raw data of panels (c.i) and (d.i) are accessible in 'Section III' of Supplementary Information and the accompanying Supplementary Videos. Please note that, as expected 37,38 , the respective S zx (E ex,z ) loops are practically horizontal lines, since the investigated area was onto SA y (see insets of Fig. 6(b.i-b.ii)). With the help of the raw data presented in Fig. 6(b.i-d.i)/(b.ii-d.ii), the main results of the present work become evident: first, the S zy (E ex,z ) curves become narrower upon H ex,z application ( Fig. 6( Fig. 6(c.i)/(c.ii)), and second they recover their original form, almost completely, upon H ex,z removal ( Fig. 6( Fig. 6(d.i)/(d.ii)). Thus, this data evidences that the application of an external magnetic field, H ex,z = 1 kOe, causes drastic decrease to the respective in-plane PE coefficients on the order of 50-60% which is restored upon H ex,z removal, as summarized in the deduced <|d zy |> data that are presented in Fig. 6(a.i)/(a.ii). To clarify the S zy (E ex,z ) curves obtained with our local OM-based method we performed standard measurements of the polarization in both sintered PZT and PZT-5%Fe 3 O 4 (sintering conditions: T = 1000 °C, for t = 2 h in air). Figure 7 shows raw P(E ex,z ) data, while its inset presents the derivative, dP/dE ex,z , for the case of the PZT-5%Fe 3 O 4 composite. From this data becomes apparent that the addition of the FM Fe 3 O 4 NPs to PZT makes it harder; the P(E ex,z ) loop is noticeably broadened since the coercive field, E ex,z coer , increases from 6.7 kV/cm to 10.7 kV/cm. This fact can be ascribed to the efficient pinning of FE domain walls by the structural disorder introduced by the FM NPs; the similar size (50-100 nm) of the FM Fe 3 O 4 NPs (see 'Sample preparation' in 'METHODS') and the FE domains of PZT (see 17,[39][40][41] ) can probably motivate and or promote this feature. On the other hand, the nucleation field, E ex,z nuc , is practically unaltered at approximately 4-5 kV/cm. Notably, the P(E ex,z ) data clearly show that the peaks observed in the S zy (E ex,z ) curves of Fig. 6 ii) correspond to the nucleation fields, E ex,z nuc,+ and E ex,z nuc,− , where FE domains appear and start to move/rotate, ultimately dictating complete reversal of the bulk polarization. Most important, the comparison of the S zy (E ex,z ) results with the P(E ex,z ) ones clearly evidences that the FM Fe 3 O 4 /Fe 2 O 3 NPs do not simply serve as static structural disorder; they introduce reconfigurable magnetic disorder that modifies the in-plane strain-electric field curves and the accompanying PE coefficients when an external magnetic field is applied at will.
Referring to the underlying mechanism motivating the observed ME coupling between the PZT matrix and the Fe 3 O 4 NPs we recall that three mechanisms are generally considered: first charge modulation, second exchange interaction modulation and third strain transfer 45 . Since the first two mechanisms are active at a short range by a FE/FM interface, we suggest that the strain transfer mechanism is dominant in the case of the PE/FM PZT-5%Fe 3 O 4 samples studied here (noticeably, the latter is active at length scales on the order of 100-200 nm 46,47 ). Recently, A. Kumar et al. 48 employed the strain transfer mechanism to explain the results obtained in a relevant   48 . In our case the same mechanism could be at play due to the strong PE character of PZT and the noticeable magnetostrictive nature of Fe 3 O 4 and Fe 2 O 3 . Specifically, the magnetostriction coefficient, λ, of bulk magnetite ranges within 20 × 10 −6 to 90 × 10 −6 for single crystals and 30 × 10 −6 to 50 × 10 −6 for polycrystalline samples 49,50 . In addition, the magnetostriction coefficient of single crystals of bulk hematite is somehow lower, 10 × 10 −6 at maximum 51  ones) can effectively alter the PE strain of the PZT matrix when an external magnetic field is applied. Indeed, the experimental fact that the PE coefficients are completely restored upon removal of the magnetic field (reversible procedure) hints that the underlying cause is a linear magnetostrictive effect of magnetoelastic nature. Irrespectively of the underlying mechanism, here we report for the first time experimental data on the magnetic-field control of a ME parameter, specifically that of the PE coefficients d zi (i = x,y), of three significant characteristics: first, it is impressive in magnitude, 60%, second, it is completely reversible, while third, and probably most important, these characteristics are feasible at room temperature upon using a relatively low magnetic field, H ex = 1 kOe. At this point, it is worth mentioning that recent studies 35,36 concerning composite ME systems report variations of similar magnitude in other relevant parameters such as the polarization and piezoresponse, but not in the PE coefficients reported here. In 35 D. M. Evans et al. report an average change of 60% in the polarization of PZTFT single crystals for a relatively high H ex variation of 6 kOe, between the +3 kOe and −3 kOe states. Unfortunately, as already discussed by the authors 35 , this noticeable percentage variation is partially reversible upon reversing once again the magnetic field back to +3 kOe, since the polarization attains a 40% higher value in comparison to that of the original state. In 36 S. H. Xie et al. observe a change of 40% in the piezoresponse of PZT-TDF bilayer upon application of +2 kOe. Notably, in contrast to the partially reversible behavior observed in 35 , in 36 the authors report a further increase of the piezoresponse upon reversing the magnetic field to −2 kOe, a clear proof that in this case the overall process is entirely irreversible. The above comparative discussion with the data reported in 35,36 shout that our results could be important for applications at room temperature.

Conclusions
Here, we investigated the modulation of PE coefficients for a bulk PZT-5%Fe 3 O 4 system, which is a strongly insulating hybrid of FM Fe 3 O 4 NPs (noticeably magnetostrictive) homogeneously embedded in a FE PZT matrix (highly piezoelectric). By means of an OM-based method we recorded the in-plane strain loops for electric fields applied out-of-plane in the range −10 kV/cm ≤ E ex,z ≤+ 10 kV/cm, upon successive application and removal of an also out-of-plane external magnetic field of low value, H ex,z = 1 kOe. The respective in-plane PE coefficients, 200-250 pm/V, display a dramatic decrease on the order of 50-60% upon H ex,z application, that is completely restored upon H ex,z removal. Polarization data performed in a wider range −20 kV/cm ≤ E ex,z ≤ + 20 kV/cm showed that  techniques. Since the present study aims to attain information that could be of interest for practical applications it is exclusively focused at room temperature. As a consequence, all experimental techniques described below were employed at room temperature.  [3][4][5][6][7][8]. Information on the topography for the evaluation of the microstructure was obtained with secondary electron imaging (SEI), while elemental analysis to obtain both energy-dispersive x-ray spectroscopy (EDS) information and compositional mapping images was recorded with backscattered electron imaging (BSE). Au-coated samples were used for SEI, while both Au-coated and non-coated samples were used for BSE to cross-check the results due to the overlapping of the M-spectral line of Au (2.123 keV) with the Lα-spectral line of Zr (2.044 keV) and the M-spectral line of Pb (2.342 keV).

X-ray diffraction (XRD
Current-voltage (I-V) characteristics. I-V characteristics were taken for the composite Pb(Zr 0.52 Ti 0.48 ) O 3 -x%Fe 3 O 4 samples, x = 0-50%, using a DC-voltage supply (model IP-32, Healthkit Co., USA) to apply the voltage across each sample, while I was monitored with the digital multimeter (MY-67, V&A). This information is important since we want to apply the maximum possible electric field to the sample without loss of its insulating properties.
Optical Microscopy for piezoelectric characterization. The constitutive Strain-Electric field curve, S(E ex,z ), of the samples was estimated experimentally by using an OM-based technique introduced in 37 . An ORTHOLUX (Leitz, Wetzlar, Germany) OM was used, equipped with a linear xy translation stage on which a home-made aluminum platform was mounted. The magnification used was x100-x150 (objective lens x10). The calibration in the length scale of the OM images was accomplished by using the standard grating test TGZ3 (NT-MDT Co, Moscow, Russia). Using a DC-voltage supply (model IP-32, Healthkit Co., USA) we applied E ex,z up to 10 kV/cm along the sample thickness, i.e. out-of-plane, during the measurements. More details can be found in 38 . Here, we made another modification in the home-made aluminum platform hosting the sample, so that a constant magnetic field, H ex,z , is applied at will, also out-of-plane (z axis), by a NdFeB permanent magnet, (disc-shaped with diameter 20 mm and thickness 3 mm). Mapping of H ex,z was performed with a Gaussmeter Model 410 (Lake Shore Cryotronics Inc, Ohio, USA), with its probe placed on a linear xyz stage with micrometer resolution. Magnetization measurements. The constitutive Magnetization-Magnetic field curve, M(H ex,z ), of the samples was measured experimentally by using a SQUID magnetometer MPMS 5.5 Tesla (Quantum Design, San Diego, CA, USA).
Polarization measurements. The constitutive Polarization-Electric field curve, P(E ex,z ), of the samples was measured experimentally by using a TF Analyzer 2000 (aixACCT Systems GmbH, Aachen, Germany) ferroelectric analyzer connected with a high-voltage source Trek Model 610E (TREK INC, Lockport, NY, USA). The samples were immersed in silicone oil during the measurements to prevent arcing. The waveform and frequency of the applied electric field are triangle and 10 Hz, respectively.