Enhancement of the piezoelectric coefficient in PVDF-TrFe/CoFe 2 O 4 nanocomposites through DC magnetic poling

In the last years flexible, low-cost, wearable, and innovative piezoelectric nanomaterials have attracted considerable interest regarding the development of energy harvesters and sensors. Among the piezoelectric materials, special attention has been paid to electroactive polymers such as poly(vinylidene fluoride) (PVDF) and its copolymer poly(vinylidene fluoride- co -trifluoroethylene) (PVDF-TrFe), which is one of the most extensively investigated piezoelectric polymers, due to the high β phase content resulting from specific curing or processing conditions. However, to obtain a high piezoelectric


Introduction
In the last years, innovative energy harvesting systems based on the piezoelectric effect, able to convert vibrational mechanical energy, which is one of the most present accessible and alternative form of energy, into electric energy, have attracted the interest of the scientific community. The most common piezoelectric materials are ceramics, such as barium titanate, lead zirconate titanate, and lithium niobate, which have a high piezoelectric constant and, therefore, a high conversion efficiency. However, these materials are toxic, brittle, and not environmentally friendly. In order to overcome these disadvantages, several studies have investigated polymeric piezoelectric materials and nanostructured materials with high dielectric constant and good piezoelectric properties suitable for the fabrication of flexible piezoelectric nanogenerators [1][2][3][4][5][6][7].
One of the most extensively investigated piezoelectric polymers is poly(vinylidene fluoride) (PVDF). This polymer has attracted a lot of interest due to its excellent piezoelectric and ferroelectric properties, high chemical resistance, high thermal stability, large polarization, short switching time, and mechanical flexibility. All these combined characteristics make it suitable for a wide range of advanced applications, from sensing to energy harvesting [8][9][10][11].
It is well known that PVDF is characterized by four main polymorph phases, namely α, β, γ, and δ [12]. Due to its high thermodynamic stability at room temperature the α phase is the predominant one. However, the α phase does not contribute to the piezoelectricity of the polymer, whereas the γ and δ phases, the so-called TGTG′ (trans-gauche-trans-gauche) chain conformation, show a weak piezoelectric response after electric poling. The β phase, the so-called TTT (all trans) planar zigzag chain conformation, is the one responsible for the electroactive response of the polymer. It is a polar phase that shows the strongest ferro-, piezo-, and pyroelectric behavior [13], due to the largest spontaneous polarization (7 × 10 −30 C·m) [14]. In order to obtain a high content of the β phase, PVDF is generally polymerized with trifluoroethylene (TrFe), producing a copolymer of PVDF, namely poly(vinylidene fluoride-co-trifluoroethylene) (PVDF-TrFE) [15][16][17]. This copolymer shows a higher piezoelectric coefficient than PVDF, while still maintaining a high flexibility, sensitivity, and mechanical durability. It is also, like PVDF, a biocompatible polymer.
In order to obtain the best piezoelectric response, the dipoles in PVDF/PVDF-TrFe must be oriented along a preferential direction. The dipole orientation is generally obtained by electrical poling, that is, the application of a strong DC electric field (ca. 10 6 V·cm −1 ) at elevated temperature (ca. 120 °C) through top and bottom electrodes [18]. However, this technique is not cost-effective and is not easy in the practical implementation, since it requires the use of high-voltage apparatus. Recently, alternative techniques with respect to electric poling and different strategies have been investigated in order to enhance the β phase content in PVDF. These included mechanical stretching [19,20], spin coating [21], quenching [22], a combination of the latter two techniques [2], and the addition of external additives to the PVDF matrix, such as metal nanocomposites [14,23], ceramic filler [24], and graphene nanoplatelets (GNPs) and their combination with zinc oxide nanostructures [1,3,5,6]. Recently, it was shown that the β phase content of PVDF can be improved introducing CoFe 2 O 4 nanoparticles into the polymer and applying a DC magnetic field [25]. This effect has been ascribed to the strong tensile stress at the CoFe 2 O 4 /PVDF interfaces originating from the interaction between the magnetic field and the CoFe 2 O 4 nanoparticles. As a result, a very high β phase content (up to 95%) has been measured in the PVDF nanocomposite with 5 wt % CoFe 2 O 4 . However, the authors limited the analysis to the β phase content and it was not clear if the β phase increase corresponded to an increase of the piezoelectric coefficient as well. In fact, high values of d 33 cannot be achieved, in general, with an enhancement of the β phase content only, but they are associated to an alignment of the piezoelectric domains within the material [18].
In this work, we developed and characterized, in terms of piezoelectric coefficient, a novel piezoelectric nanocomposite, based on PVDF-TrFe and CoFe 2 O 4 nanoparticles. We adopted PVDF-TrFe instead of PVDF, as in [25], due to its superior β phase content, as discussed above. The PVDF-TrFe/CoFe 2 O 4 nanocomposite was poled under different strengths of a DC magnetic field and at different temperatures (room temperature (RT) and 65 °C), in order to enhance the alignment of piezoelectric domains. We demonstrate that this approach is an effective and simple method to improve the piezoelectric response of PVDF-TrFe. The piezoelectric coefficient was evaluated through piezoresponse force microscopy (PFM), which is a very powerful technique to characterize the micro-and nanoscale piezoelectric response of piezoelectric materials [1,2,4,26,27].

Results and Discussion
In order to understand whether the β phase fraction was influenced by magnetic poling, we performed FTIR analyses of six different samples. Figure 1 shows the FTIR spectra of the nanocomposite specimen produced with a CoFe 2 O 4 nanoparticle content of 5 wt %, poled for increasing time (from 60 to 120 min) at increasing strengths of the DC magnetic field (50 and 110 mT). As already reported in [1][2][3]5,6,28], it is possible to determine the relative fraction of the β phase, F(β), through the analysis of the peaks at 763 and 840 cm −1 , which are char- acteristic of the α phase and the β phase, respectively. To be more specific, F(β) can be estimated by the using following equation [28]: (1) where A α and A β are the absorbance at 763 and 840 cm −1 , respectively, for the α and β phases, and K α and K β are the absorption coefficients at the corresponding wavenumbers (7.7 × 10 4 and 6.1 × 10 4 cm 2 ·mol −1 , respectively). As it can be seen from Figure 1, the spectra of neat PVDF-TrFE presents two clear features at 763 cm and 840 cm −1 and, by using Equation 1, we obtained an estimation of F(β) = 81.06%. This is higher than the value of F(β) of pure PVDF [3] and is in agreement with values reported for PVDF-TrFe in [29]. However, when PVDF-TrFe is filled with CoFe 2 O 4 nanoparticles, the measured FTIR spectra show a broadband shoulder close to the α phase peak located at 763 cm −1 (probably due to the interaction between polymer and nanoparticles), making the evaluation of the relative fraction of β phase by using Equation 1 difficult. By deconvolving the shoulder from the absorption band related to the α phase centered at 763 cm −1 , we estimated for the different samples a value of F(β) = (82 ± 3)%. The relative-ly large error is due to the uncertainty in the deconvolution procedure. From this estimation, and considering that the intensity of the β phase peaks remains unaffected by the introduction of CoFe 2 O 4 nanoparticles as well as by the successive magnetic poling processes, we conclude that the relative fraction of the β phase remains nearly constant upon the introduction of the nanoparticles and it is practically unaffected by the magnetic poling. This is in contrast to what has been reported in [25], where an increase of F(β) was observed, with a maximum increase for magnetic fields around 400 Oe.
Furthermore, analyzing the FTIR data in the range of 1600-1900 cm −1 , as reported in Figure 1, we observed that a broad peak at 1740 cm −1 appears in the spectra of all PVDF/ TrFe/CoFe 2 O 4 nanocomposite samples. We attributed this absorption band to carbonyl groups (C=O), in agreement with what has been reported in [30], likely originating from the formation of bonds between O atoms of the CoFe 2 O 4 nanoparticles and C atoms of PVDF-TrFe chains.
In order to quantify the piezoelectric response of the PVDF-TrFe/CoFe 2 O 4 nanocomposites we performed PFM. Three different regions of each sample were measured to determine an average piezoelectric coefficient (d 33 ), through a procedure developed to quantitatively evaluate d 33 [2, 3,31]. For neat PVDF-TrFe we obtained d 33 = 4.32 ± 1.07 pm/V, a much lower value than those obtained after electrical poling (d 33 ≈ 30 pm/V) [32,33].
First, we analyzed the piezoelectric response of nanocomposites with different content of CoFe 2 O 4 nanoparticles. As shown in Figure 2, d 33 increases when CoFe 2 O 4 nanoparticles are added, showing a maximum of 6.34 ± 1.1 pm/V at 5 wt %, in agreement to what has been reported in [25], where the maximum β phase content was found for a content of nanoparticles of 5 wt %. After that, we focused our investigation on the effect of DC magnetic poling on nanocomposites with CoFe 2 O 4 nanoparticles at 5 wt %. After curing, the spin coated nanocomposite thin films were exposed to a DC magnetic field for 1 h at 65 °C. In Figure 3 the d 33 values are reported as function of the applied DC magnetic field strength. As can be seen, the piezoelectric coefficient increases when the magnetic field is increased and reaches a maximum value for a magnetic field strength of 50 mT. Also shown in Figure 3 is the d 33 value of a sample magnetically poled with a field of 50 mT at RT. The value, although higher than that of the un-poled sample, is substantially lower than that of the sample poled with the same magnetic field at 65 °C.
We assume that the d 33 increase induced by DC magnetic poling was due to the increased alignment of the β phase domains along the magnetic field direction, rather than due to an increase of the β phase content, as proposed in [25]. In particular, we speculate that chemical bonding can take place between PVDF-TrFe and CoFe 2 O 4 nanoparticles, with the forma- tion of a carbonyl group (C=O), as suggested by the FTIR spectra reported in Figure 1. When the DC magnetic field is applied, the ferromagnetic nanoparticles orient themselves along the direction of the applied field and then drag the polymeric chains of PVDF-TrFe to align in the same direction. At relatively high magnetic field strengths (B > 50 mT) we found that the d 33 increase is lower, in agreement with what has been observed in [25]. This behaviour has been attributed to the formation of CoFe 2 O 4 nanoparticle aggregates [25]. The presence of these aggregates will be investigated in the following part dealing with the morphological analysis.
Also, temperature plays a relevant role in the orientation of the β phase. As shown in Figure 3, when the DC magnetic poling is performed at RT the value of d 33 does not increase as much as in the case in which the temperature is kept at 65 °C. Increasing the temperature makes the polymeric nanocomposite softer, thus facilitating the orientation of β phase domains along the magnetic field direction.
We have also investigated the effect of the application time of the DC magnetic field on d 33 . Figure 4 shows the measured d 33 values of samples exposed to the magnetic field for increasing periods of time. The d 33 values tend to saturate after 90 min of magnetic field poling, reaching, in the case of the samples poled with a magnetic field of 50 mT, an average value as high as 34 pm/V. We note that such a value is comparable to or even higher than those reported for electrically poled PVDF-TrFe samples [32,33]. This renders the process of magnetic poling of PVDF-TrFe/CoFe 2 O 4 nanocomposites particularly attractive for applications where electrical poling may have limitations.  In order to verify the presence of the aggregates of the CoFe 2 O 4 we investigated the morphology of the CoFe 2 O 4 nanoparticles ( Figure 5a) and of the PVDF-TrFe/CoFe 2 O 4 nanocomposites (Figure 5b) using field-emission scanning electron microscopy (FESEM). As shown in Figure 5b, the morphology of the produced nanocomposites is homogeneous and it is characterized by spherulitic structures with an average diameter of ca. 1 μm.
The samples were also measured after magnetic poling under different magnetic field strengths and we did not observe appreciable changes in the spherulitic structure ( Figure 6). However, Figure 6b,c shows the aggregation of CoFe 2 O 4 nanoparticles in some areas, in analogy with what has been reported in [25]. The formation of aggregates was confirmed by energy-dispersive X-ray spectroscopy (EDX). In correspondence with the agglomeration we observed a higher intensity of the O, Fe, and Co signals, as reported in the chemical maps shown in Figure 7, indicating the presence of clusters of CoFe 2 O 4 nanoparticles.
Despite the presence of these agglomerates Fe and Co appear to be uniformly distributed in the nanocomposite film, confirming that the nanoparticles are well dispersed in the polymeric matrix. It should be pointed out that we also observed some smaller aggregates in the un-poled samples, but with a much lower density then in the magnetically poled samples.

Conclusion
In this work we developed a simple and cost-effective method to fabricate a polymeric nanocomposite with high piezoelectric coefficient by adding ferromagnetic CoFe 2 O 4 nanoparticles to PVDF-TrFe and performing magnetic poling. The highest d 33 value, up to 39 pm/V, was obtained after adding 5 wt % of CoFe 2 O 4 nanoparticles and applying a DC magnetic field of 50 mT. Application of the magnetic field produced an alignment of the β phase along the direction of the applied magnetic field rather than an increase of the β phase content itself.
We believe that the increase of the d 33 is due to the strong interaction between the molecular chains of the polymer nanocomposite and the magnetic field, thanks to the drag effect produced by the CoFe 2 O 4 nanoparticles, which are strongly anchored to PVDF-TrFe. In fact, based on the FTIR analysis of the produced samples, showing the appearance of an absorption band attributed to carbonyl groups (C=O) when the CoFe 2 O 4 nanoparticles are introduced, we propose that chemical bonds are formed between O atoms of the CoFe 2 O 4 nanoparticles and C atoms of PVDF-TrFe chains. When the DC magnetic field is applied, the orientation of the CoFe 2 O 4 nanoparticles drags the polymeric chains along the direction of the applied field. We observed that the maximum value of d 33 was reached when a DC magnetic field of 50 mT was applied, while for higher values of magnetic fields the d 33 value tends to decrease. This could be due to the aggregation of the CoFe 2 O 4 nanoparticles forced by intense magnetic fields, resulting in a reduced alignment efficiency of the polymeric chains, as suggested in [25].   The FTIR measurements of PVDF-TrFe/CoFe 2 O 4 nanocomposites showed a broadband shoulder close to the α phase peak located at 763 cm −1 , making the evaluation of the β phase difficult. Nevertheless, deconvolving the shoulder from the absorption band related to the α phase we estimated for the different samples a relative fraction of the β phase of F(β) = (82 ± 3)%. From this estimation, and considering that the intensity of the β phase peaks remains nearly unaffected by the introduction of CoFe 2 O 4 nanoparticles as well as by the successive magnetic poling processes, we conclude that the relative fraction of the β phase remains nearly constant. This implies that the primary effect of the DC magnetic poling is to increase the orientation of the β phase domains along the magnetic field direction, rather than to increase the β phase content, as proposed in [25].
The developed combination of polymeric nanocomposite, including ferromagnetic CoFe 2 O 4 nanoparticles, and the application of a DC magnetic poling provides an alternative route to obtain highly efficient piezoelectric materials with excellent d 33 values, avoiding electrical poling or mechanical stretching. This process may be particularly attractive in the fabrication of energy harvesting devices or wearable sensors for flexible electronics applications.

Experimental
The PVDF-TrFe/CoFe 2 O 4 nanocomposite thin films were produced through spin coating. The CoFe 2 O 4 nanoparticles (Sigma-Aldrich, 99%) were dispersed in N,N-dimethylformamide (DMF, Sigma-Aldrich, ≥99%) with a concentration of 1 wt % and tip-sonicated using an ultrasonic processor in pulsed mode for 20 min at 70% of the power amplitude. Subsequently, the CoFe 2 O 4 suspension was added to a PVDF-TrFE solution (Piezotech FC Ink L, 7 wt %) at the desired ratio. The mixture was bath-sonicated for 30 min and then spin coated (30 s with a spin speed of 2000 rpm) onto polyethylene terephthalate/indium tin oxide (PET/ITO, Sigma-Aldrich, R s = 60 Ω/sq).
The obtained melt was cured in the oven at 120 °C for 2 h. After curing, the nanostructured thin film was placed over a heated mat reaching 65 °C and a DC magnetic field was applied through a ferromagnetic core (Figure 8).
The morphology of the PVDF-TrFe/CoFe 2 O 4 nanocomposite thin films was investigated using a field-emission scanning electron microscope (FESEM, Auriga, Carl Zeiss) operated with an accelerating voltage of 5 kV. Chemical composition analysis was performed through EDX using the FESEM, equipped with a Bruker Quantax EDX setup capable of an energy resolution of 123 eV at Mn Kα. In order to prevent charging the samples were metalized with 20 nm of Cr using a Quorum Technologies Q150T ES sputter coater.
All FTIR absorption spectra were recorded at RT using a FTIR spectrometer (Vertex 70 by Bruker Optics GmbH) equipped with a single reflection diamond ATR cell. The analyses were performed in the typical mid-IR spectral range (4000-400 cm −1 ) at a resolution of 2 cm −1 and with 256 scans.
The piezoelectric coefficients were measured through PFM [1,5,27] using a commercial Bruker-Veeco Dimension Icon AFM with a Co/Cr-coated-tip silicon cantilever (MESP-RC-V2, Bruker). Following the procedure described in [2,3], we scanned three different areas, 5 × 5 μm 2 in size, of each sample, with 256 × 256 acquisition points per scanning area using a scan rate of 0.5 Hz. An alternating voltage with a frequency of 15 kHz was applied to the tip.
In order to obtain a quantitative evaluation of d 33 , the amplitude of the PFM signal is measured as function of the applied voltage V ac in the selected range of 1-5 V. First, a calibration sample of periodically poled lithium niobate (PPLN), with a known value of the piezoelectric coefficient (d 33PPLN = 7.5 pm/V), is measured. Then, the sample under investigation is tested and the PPLN is measured again. The two piezoelectric calibration signals of PPLN are compared for each value of the applied voltage V ac . If the difference between them is less than 20%, the measurement of the sample under test is considered reliable [2,3].
The calibration factor ξ = m PPLN /d 33PPLN (in which m PPLN is the slope of the straight line of the amplitude of the PFM signal vs V ac , and d 33PPLN is the known piezoelectric coefficient of PPLN) is evaluated, averaging over the two calibration measurements. The d 33 value of the sample under test is finally calculated using the following expression: (2) in which V sample is the PFM signal of the sample under test.
PFM measurements were performed on three different areas for each sample and two samples for each condition were tested. We determined the average PFM response and the corresponding standard deviations, which are reported as error bars in Figures 2-4.