Electromechanical response of reduced graphene oxide–polyvinylidene fluoride nanocomposites prepared through in-situ thermal reduction

Both ionic and electronic electroactive polymers (EAPs) have displayed great potential as actuators. Current ionic EAPs have limited practical implementation due to their slow response time and low force produced. Furthermore, the ion transport-based mechanism necessitates presence of an electrolyte, which further complicates issues of packaging and device lifetime. Similarly, despite the many advantages of electronic EAPs such as their low density, good energy densities, and high bandwidth, there are major obstacles facing their transition to application; notably they require high actuation voltages, have low blocked stress, and low operating temperatures. In order to overcome these limitations, electronic EAPs modified using conductive or semiconductive nanofillers have been explored. For example, carbon-based nanostructures such as carbon nanotubes (CNTs) were used as fillers to improve the electromechanical response of polymers.^1,2 In this study, the goal is to address the shortcomings of current electronic EAPs, notably polyvinylidene fluoride (PVDF), through the incorporation of 2D graphene, owing to its unique geometry and electrical properties. In particular, the high aspect ratio and high surface area of single layer graphene nanosheets (GNS) can yield simultaneous enhancement in mechanical, thermal, and electrical properties of polymers, which is of great promise for multifunctional applications.^3,4 The choice of PVDF as the polymer and graphene oxide (GO) as the filler is judicious and is addressed in the next paragraphs.

In our own work, and using single wall carbon nanotubes (SWNT) as fillers, we saw an order of magnitude increase in the electrostrictive coefficient compared to the highest performing polymer-based electrostrictors (such as polyurethane (PU), M₃₃ = 3.25 × 10⁻¹⁵ (m²/V²) (Ref. 5)) at much lower electric fields;^2,6 however, the SWNT-PVDF and SWNT-polyimide (PI) nanocomposites also exhibited low voltage breakdown and low energy density as a result of the increase in electrical conductivity. For that reason, based on promise gleaned from literature, we are now focusing on GO with the goal of achieving electromechanical coupling while minimizing increase in electrical conductivity and dielectric loss. Here, we synthesize and characterize PVDF-based nanocomposites with partially reduced GO as potential EAPs. The synthesis begins with GO as the filler, followed by thermal annealing to controllably reduce the GO such that a desirable level of electrical conductivity could be obtained in the nanocomposites. The use of a co-solvent and the reduction following drying of the polymer yielded nanocomposites with dispersed GO sheets in the PVDF polymer. The microstructure and electrical properties of these nanocomposites are thoroughly characterized to assess the effect of the partial reduction of GO. Actuation response is also analyzed and the mechanism driving the actuation response is discussed in detail.

Synthesis of well dispersed nanocomposites with graphene as the starting nanofiller is challenging due to the strong van der Walls forces between the graphitic sheets along with their weak interaction with polymers;⁷ whereas the use of GO is conducive to dispersion in a variety of polymers. Dispersion of GO is better due to its highly oxidized structure, which renders the sheets hydrophilic and soluble in certain organic solvents. This characteristic of GO has lead to its wide use as a starting material in the synthesis of polymer nanocomposites.⁸ GO can be readily dispersed in water while it usually forms less stable solutions when directly added to aprotic solvents such as N, N dimethylformamide (DMF).⁹ Moreover, it requires less sonication time to achieve good dispersion in water due to the hydrogen bonds that form between water and GO's functional groups. Park et al.¹⁰ dispersed the GO initially in water using mild sonication and then added 9:1 volume ratio of DMF to the solution and obtained a stable dispersion in the water/DMF co-solvent. Other researchers have dispersed GO directly in organic solvents alone and have seen enhancement in dielectric properties; however, this method requires more sonication time during processing with detrimentally affects the aspect ratio of the GO nanosheets;¹¹ consequently, because of the reduction in aspect ratio during processing, a higher weight fraction of GO nanosheets is required to obtain desirable properties.¹² 

Many polymer nanocomposites have been studied using GO and reduced GO as starting filler material with a focus on electrical properties. Ansari and Giannelis¹³ have studied functionalized GO-PVDF nanocomposites and found that the electrical percolation threshold for these nanocomposites is around 2 vol. %. Jinhong et al.¹² prepared graphene sheets-PVDF nanocomposites and achieved a dielectric permittivity higher than 300 at 1 kHz for a 7.5 vol. % reduced GO composite, although there is no mention of dielectric loss. Chen et al.¹⁴ synthesized GO-polyimide nanocomposites and found that GO was reduced during thermal treatment, which led to enhanced electrical properties of the nanocomposite. Even though many methods are explored for converting GO to reduced GO using chemical treatment, the in-situ thermal reduction method in the dried composite films results in more dramatic changes in properties. Tang et al.¹⁵ prepared reduced graphene-PVDF nanocomposites with the in-situ thermal reduction method and demonstrated existence of well dispersed reduced graphene sheets in PVDF matrix using TEM and x-ray diffraction analysis. The electrical conductivity of the reduced graphene-PVDF nanocomposites at 2.4% volume fraction was three orders of magnitude greater than the nanocomposites before in situ thermal reduction. The effective dielectric permittivity at 1 kHz was also high, around 662, compared to 7 for unmodified PVDF. Chemical reduction does not yield as high an increase in effective permittivity; Shang et al.¹⁶ prepared layered GNS/PVDF nanocomposite films by solution casting using in situ chemical reduction. Improvement in the electrical properties was limited with a dielectric permittivity of 63 at 100 Hz when the concentration of GNS was 1.27 vol. %. Other studies in the literature have focused on incorporating graphene instead of GO in the polymer; however, the difficulty in dispersing graphene lead to agglomeration in the polymer and therefore the need to use higher volume to obtain improvements in properties, which makes processing and dispersion more challenging. Cui et al.¹⁷ prepared high volume fraction, up to 12.5 vol. %, graphene-PVDF nanocomposites and obtained a dielectric permittivity of 2080 at 1 kHz with a percolation threshold of 4.08 vol. %. This very high percolation threshold is due to dispersion issues with graphene. To avoid dispersion challenges with graphene as a starting material, functionalizing the graphene is necessary for better interaction with polymers. Wu et al.¹⁸ functionalized the graphene sheets with hyper branched aromatic polyamide and increased the interaction between the polyurethane and graphene to achieve a dielectric permittivity of 900 at 1 kHz at only 3 vol. %. Fang et al.¹⁹ followed a covalent polymer functionalization of graphene in order to improve the dispersion and obtained a 0.9 wt. % (∼0.5 vol. %) graphene-polystyrene nanocomposite which enhanced the tensile strength and Young's modulus by around 70% and 57%, respectively. Thus, the above mentioned research confirms that dispersion of graphene nanosheets in polymer is challenging due to their strong interlayer cohesive energy and additional effort in functionalization is necessary to avoid aggregation. On the other hand reduced GO, which has functional groups, is conducive to good dispersion in polymers. Among the different routes to prepare reduced graphene oxide nanocomposites, the in-situ thermal reduction on the polymer films proves to be an efficient method to maintain the dispersion state of GOs in polymer.

While the electrical properties of reduced GO-based nanocomposites have been explored in depth, little work has focused on the characterization of their electromechanical properties. Seveyrat et al.²⁰ prepared nanocomposites based on PU and poly(vinylidene fluoride-trifluoroethylene-chlorofluoroethylene) (PVDF TrFE CFE) terpolymer filled with varying volume % of graphene nanoflakes. It was demonstrated that there was no interaction between the polymer and the fillers. Consequently, the electrostrictive coefficient showed only a marginal increase for very high volume fraction of the fillers such that M₃₃ of 3.25 × 10⁻¹⁵ (m²/V²) for unmodified PU was increased to 5.53 × 10⁻¹⁵ (m²/V²) for PU with 6.5 vol. % of graphene. Another research by Eswaraiah et al.²¹ focused on preparation of 2 vol. % functionalized graphene-PVDF nanocomposite films, which showed nearly 20 times better strain sensing performance than its counterparts based on carbon nanotube polymer composites. Javadi et al.²² prepared nanocomposites based on functionalized graphene nanosheets (FGNS)/P(VDF-TrFE-CFE) terpolymer. In order to achieve better interaction with the polymer, the graphene nanosheets were chemically modified using 3,5-di(trifluoromethyl)-aniline. Effort was made to improve the electromechanical properties by increasing the dielectric constant by the inclusion of conductive FGNS. At 1.09 vol. % FGNS, the out-of-plane strain (S₃₃) increased from 0.8% for unmodified terpolymer to 4.1% for the nanocomposite under the same electric field (23 MV/m). Eventhough the FGNS had a huge impact on the electromechanical properties of the terpolymer, a complex chemical modification route which uses oxalyl chloride (COCl)₂ was needed to achieve that. On the other hand, the reduced GO obtained from in-situ thermal reduction has some functional groups, which can help to achieve good dispersion and eliminates the need for a complex chemical modification route. As of now, and to the best of our knowledge, there is no study available in the literature which evaluates the effect of reduced GO on the electromechanical properties of the EAPs. More importantly, the potential of the reduced GO to show actuation when used in a polymer which is inactive before the modification, has to be explored.

Graphene oxide was prepared via the modified Hummer's method, as described in recent literature.^11,15,23 Graphite flake (20 g, Alfa Aesar, 325 mesh, 99.8%) was pre-oxidized by slowly adding it to a solution of K₂S₂O₈ (10 g, Alfa Aesar, 97%) and P₂O₅ (10 g, Alfa Aesar, 98%) in H₂SO₄ (60 ml, VWR, 95%–98%) at 80 °C. The mixture was stirred at 80 °C for 30 min and then cooled to room temperature at 10 °C/h. The pre-oxidized graphite was obtained through centrifugation and washed with ultrapure water (18 MΩ cm) until the pH value was neutral followed by drying at room temperature overnight. The resulting pre-oxidized graphite (4 g) and H₂SO₄ (92 ml) were mixed and stirred in an ice bath followed by the slow addition 12 g of KMnO₄ (Alfa Aesar, 99.0%). The mixture was stirred at room temperature for 2 h, then 184 ml of deionized water was added and the temperature was allowed to reach 90 °C for 15 min at which point 560 ml of ultrapure water was added, followed by 10 ml of 30% H₂O₂ (Mallinckrodt chemical, 30%), resulting in a yellow-brown colored solution. The reaction mixture was then washed with diluted hydrochloric (1:10) to remove the metal ions. The graphite oxide was purified through centrifugation at 200 rpm for 2 min to remove it from hydrochloric solution; then the sample was re-dispersed in deionized water twice and centrifuged at 5000 rpm for 5 min to remove the highly oxidized sheets and hydrate the graphite oxide. Finally, after dispersion once more in water, the materials were centrifuged at 11 000 rpm for 15 min, after which the pH of the supernatant was approximately neutral. The supernatant was then added into deionized water and sonicated (Branson 2510) for 15 min to exfoliate the GO layers, while the bottom layer of materials was discarded. After oxidation, purification, and exfoliation, the material was freeze dried (Labconco LYPH LOCK 4.5) to create GO powder. Final purification of the GO powder was accomplished with dialysis (Spectra/Por 2, 6.4 ml/cm, 12–14 000 MWCO) in ultrapure water (10 mg/ml) at 50 °C for 3 days. The GO was then freeze dried again to yield purified GO powder.

Kynar 701 PVDF was supplied by Arkema Inc. DMF was purchased from Sigma Aldrich. Neat PVDF is prepared by dissolving the PVDF powder in DMF and the solution casted films are dried at 60 °C. GO-PVDF nanocomposites are processed by a co-solvent approach involving water and DMF.²⁴ The method involves dispersion of GO in water and DMF co-solvents and subsequent removal of water. The GO is first dispersed in water using 5 min of bath sonication at 80 W power (Fisher Scientific FS-20) and 10 s of high power sonication at 150 W (Cole-palmer ultrasonic processor-750 W). Then, DMF is added to the GO/water with a weight ratio of 9:1 (DMF to water) followed by bath sonication for 2 min. The co-solvent solution is then heated to 65 °C in a silicone oil bath for 12 h to remove water. As the flash point of DMF is around 60 °C, some DMF also will evaporate from the solution while heating; hence, DMF is added at regular intervals to maintain the volume. PVDF (20 wt. %) is then separately dissolved in DMF solution. GO/DMF solution is added to this DMF/PVDF solution and dispersion of the GO in PVDF is achieved by bath sonication for 5 min and stirring in the hot plate at 100 rpm at 60 °C. After degassing the solution under vacuum, it is poured on a glass plate and cast using a solution casting knife. The film is then dried under vacuum at 60 °C for 3 h. The PVDF nanocomposites with well-dispersed GO is hot pressed at 150 °C for 2 h. Temperature is maintained at 150 °C for 2 h to keep it well below the melting point of PVDF to prevent re-aggregation of GO.

0.5 wt. % partially reduced GO-PVDF (0.5 wt. % PR-GO-PVDF) and 1 wt. % partially reduced GO-PVDF (1 wt. % PR-GO-PVDF) are prepared using the above said co-solvent method, and the properties are compared with the unmodified PVDF. Apart from that, 1 wt. % reduced GO-PVDF (1 wt. % R-GO-PVDF) is prepared with modified co-solvent method. In the modified process, the dispersion of GO in the DMF/water co-solvent solution is subjected to higher temperature during evaporation up to 110 °C. This was done in an effect to increase the conversion rate of GO to reduced GO in the nanocomposites as the functional groups are removed from the GO at higher temperature. The nanocomposite film before hotpress (0.5 wt. % GO-PVDF) is also characterized.

Chemical composition of the GO and reduced-GO was studied by high resolution X-ray photoelectron spectroscopy (XPS), Fourier transform infrared spectroscopy (FT-IR), and thermo gravimetric analysis (TGA). XPS studies were performed on samples drop cast (3 mg/ml in DMF) onto copper foils and reduced in a convection oven for the given amount of time consistent with the nanocomposite processing. Samples were analyzed in a Perkin Elmer 5100 XPS system with an Al kα x-ray source and a takeoff angle of 45°. FTIR was conducted on either a Thermo/Nicolet iS10 with a SmartiTR attenuated total reflectance (ATR) accessory or a FTIR-ATR-Brucker-FTIR spectrometer. TGA (TA Instruments Q50) was used to determine the amount of mass loss due to thermal reduction of the material, with a ramp rate of 1 °C/min to 150 °C followed by a 4 h hold time in air. The dispersion of GO in PVDF was studied with an optical microscope (Zeiss-Observer-D) and a transmission electron microscope (JEOL TEM 1200) at 80 kV. A Kurt J. Lesker Evaporator was used to deposit silver metal (100 nm) on top of the nanocomposite films for electrical characterization. Dielectric permittivity, loss tangent, and electrical conductivity were measured with a precision LCR meter (HP 4284A) from 20 Hz to 1 MHz. A TREK voltage amplifier and HP function generator were used to supply voltage to samples for the bending actuation measurements. The video of the bending actuation response was recorded using a Photron PCI-R2 high speed camera. Photron motion analysis software was used to accurately measure the tip displacement as a function of the applied electric field. The experimental setup for the bending actuation measurement is shown in Fig. 1.

It is well documented that thermal processing of GO can vastly change the physical and chemical properties of the material. Here, GO is chemically characterized with FTIR and XPS before and after the thermal treatment at 150 °C. The temperature of thermal treatment is selected to remain below the melting temperature of PVDF (∼165 °C) and thermal reduction temperature of GO (200 °C). At 150 °C, PVDF can be hot-pressed; at the same time, reduction and potential re-aggregation of GO is minimized. GO before and after thermal treatment were drop cast on glass microscope slides and were analyzed with FTIR as shown in Fig. 2. The reduction does not dramatically change the chemical structure; however, the thermal reduction does decrease the presence of adsorbed water in the films. The water peaks (2160 cm⁻¹, 2038 cm⁻¹, and 1978 cm⁻¹) are reduced in intensity with the 2 h and 4 h reduction. The other significant change in the chemical structure is the formation of C-O-C rings within the structure (1247 cm⁻¹), which are formed at the edges of defective areas in the basal plane of the graphene.²⁵ Perhaps the best indicator of the degree of reduction is the quantity of presence of hydroxyl groups, with a broad absorbance corresponding to O-H stretching from 3000 cm⁻¹ to 3500 cm⁻¹. The reduction at 150 °C does not substantially change the O-H stretching, which indicates that the reduction is not complete and the conjugated network is not fully restored. Previous studies of the reduction of GO indicate that only a mild reduction will be achieved^25,26 with a temperature of 150 °C.

XPS of the reduced GO films reinforces the evidence from the FTIR data that the reduction is not complete and not all the functional groups are removed. The data in Fig. 3 show the high resolution C1s data for the GO and reduced GO films. There is an increase in the relative fraction of lower binding energy carbon atoms (∼284.7 eV), which indicates that the reduction does increase the C–C bonding through the removal of oxygen functional groups. The intensity of both the carbonyl (∼287.5 eV) and carboxylic acid (∼289.0 eV) peaks decreases with the reduction because the high temperatures cause some of the functional groups to decompose into water and carbon dioxide. The restoration of the C–C bonding as illustrated in Table I restores some of the properties of sp² hybridized graphene and enables the material to function as an effective filler for the nanocomposite actuator as discussed in Secs. III B–III D. It should also be noted that while there are changes to the chemical structure of the material during the reduction, the reduction is not as significant as previous results at 200 °C (Ref. 15) and there remain significant functional groups on the surface. The results confirm that hot-pressing at 150 °C only minimally affects the presence of functional groups and would help maintain the initial good dispersion of GOs obtained in the PVDF.

TGA indicates that the materials undergo a mild loss of oxygen or intercalated water during the reduction; however, the bulk of the sample remains intact as shown in Fig. 4. The GO loses 5.5% of the initial mass during the ramp segment of the reduction and loses 3.5% during the 4 h holding period. There is not significant change in the mass loss from 2 h reduction to 4 h reduction, as very little additional oxygen is desorbed during that extra 2 h reduction. The reduction of the GO is not complete during this reduction as previous works have shown as much as 60% functional groups by weight.¹⁵ It is clear that the 4 h holding period continues to reduce the materials and eliminate the functional groups from the GO, even after the 4 h holding period. But, most of the mass loss and thermal reduction happens during the ramp up to 150 °C and 2 h of holding period. So, based on these characterization results on the thermal treatment on GO nanosheets, 2 h thermal reduction at 150 °C on the nanocomposites was used.

Our dispersion study revealed that the lateral dimension of GO nanosheets is reduced to approximately 2 μm when subjected to bath sonication of more than 10 min. This decrease in size reduces the aspect ratio of the sheets and may affect the properties of the nanocomposites. In contrast, the GO sheets dispersed in water are very thin, transparent, and have high aspect ratio with lateral dimension as large as 5 μm.²⁴ The co-solvent approach helps to disperse GO as individual nanosheets in water/DMF before the water is removed, allowing the dispersion to be maintained in DMF. The TEM image in Figs. 5(a) and 5(b) shows that the lateral dimensions of the nanosheets are around 2 μm for GO in DMF compared to 5 μm for GO in co-solvent. Optical microscope image (Fig. 5(c)) of the nanocomposite solution showed the good dispersion of GO nanosheets in PVDF is maintained. Finally, the TEM image of the dried nanocomposites films shows visibly thin layers of GO without agglomeration (Fig. 5(d)). The combination of co-solvent method of synthesis for GO-PVDF nanocomposites and further in situ reduction process to obtain partially reduced-GO-PVDF nanocomposites help to achieve well dispersed reduced GO in the PVDF matrix.

XRD spectra of GO dispersed in water and co-solvent show distinct peak at 2θ ∼ 10° for exfoliated graphene oxide with 8.2 Å interlayer spacing as shown in Fig. 6. Similarly, Park et al.¹⁰ observed the peak at 2θ = 11° corresponding to graphene oxide with interlamellar water trapped between hydrophilic graphene oxide sheets. In contrast, GO dispersed in DMF showed multiple peaks apart from the peak at 2θ = 10°; the peak at 2θ = 13° corresponds to graphite oxide (unexfoliated multi layer stacking) which was not able to disperse well in DMF solvent. Also, the peaks at 16°, 20°, 21°, 23° correspond to buckyball (C₆₀) fullerene as observed by Zhang et al.,²⁷ which might have been caused by breaking of GO sheets under sonication in DMF. Thus, XRD analysis also confirms the level of exfoliation is better in water and co-solvent when compared to using only DMF.

PVDF has many crystalline phases based on processing methods where predominantly α and γ phases are achieved by solution casting method, whereas β phase can be obtained by stretching α phase PVDF under controlled environment. Both β and γ phases are polar and may lead to electromechanical response in PVDF, while α is nonpolar. Hence, which type of crystalline phases exist in our nanocomposites and how the microstructure changes in the presence of nanofillers are important considerations when determining the mechanism of property enhancement. Fig. 7 shows that the PVDF made through our processing method yields a microstructure which is predominantly γ phase. The peaks at 510 cm⁻¹, 833 cm⁻¹, and 1233 cm⁻¹ all represent the dominance of γ phase peaks and it is consistent for all the nanocomposite films dried at 60 °C. FTIR result of the unmodified PVDF, which are dried at 180 °C, shows the dominance of α phase with peaks at 615 cm⁻¹ and 760 cm⁻¹. Also, the results confirm that microstructure is not altered by the addition of GO and additional hot-pressing.

Dielectric permittivity of the unmodified PVDF, 0.5 wt. % GO-PVDF before hotpress and 0.5 wt. % PR-GO-PVDF are shown in Fig. 8. The inclusion of GO did not impact the dielectric permittivity of PVDF. But, after the hotpress when the GO is converted to partially reduced GO, the dielectric permittivity increases throughout the frequency range (20 Hz–1 MHz). Dielectric permittivity, dielectric loss, and electrical conductivity of the unmodified PVDF and the reduced GO nanocomposites measured at room temperature at various frequencies are shown in Fig. 9. The partially reduced GO nanocomposites made using co-solvent approach show remarkable improvement in dielectric permittivity for the 0.5 wt. % case compared to unmodified PVDF. The dielectric permittivity increases by a factor of 4 at 1 kHz to reach 30 compared to 7 for PVDF. The electrical conductivity of this nanocomposite shows two orders of magnitude increase at 1 kHz. The properties of 1 wt. % PR-GO-PVDF are also similar to the 0.5 wt. % PR-GO-PVDF. Even though the GO content increases from 0.5 wt. % to 1 wt. %, the amount of conversion of GO to reduced GO is limited in this hotpress method (as evidenced by FTIR and XPS results). Hence, both these nanocomposites show similar amount of partially reduced GO. The 0.5 wt. % and 1 wt. % PR-GO-PVDF polymer films did not show conductive behavior. As seen in Fig. 9, the electrical conductivity shows a frequency dependent behavior for these two nanocomposites. Generally, when conductive nanofillers are added in a polymer matrix, the electrical conductivity shows a frequency dependent behavior below the percolation threshold of the wt. % of the nanofillers. In the case of the 1 wt. % R-GO-PVDF made using the modified co-solvent method, the dielectric permittivity at 1 kHz increases to 55 and electrical conductivity shows four orders of magnitude increase. And, the electrical conductivity shows a frequency independent behavior suggesting that the restoration of the sp² hybridized properties of graphene yielded a nanocomposite with percolated network of reduced GO. The dielectric permittivity and electrical conductivity of reduced GO film are high compared to the partially reduced GO films due to the increase in the conversion rate of GO to reduced GO.

The change in the dielectric permittivity and loss tangent with temperature for 0.5 wt. % PR-GO-PVDF films are shown in Fig. 10. There is an increase of dielectric permittivity due to the relaxation around glass transition temperature (T_g) of PVDF (−30 °C), and this relaxation temperature increases for higher frequencies. This behavior confirms that the increase in the dielectric permittivity is due to the β dielectric relaxation in PVDF.²⁸ Dielectric relaxation strength is the difference between the dielectric permittivity before and after the relaxation; a dielectric relaxation strength of 51 is seen in 0.5 wt. % PR-GO-PVDF compared to 4 for unmodified PVDF at 20 Hz as shown in Fig. 11. Therefore, the remnant polarization which is a function of the dielectric relaxation strength (as shown in Eq. (1)) increases for the nanocomposite compared to the unmodified PVDF.²⁹ This dielectric relaxation study suggests that the mechanism behind the improvement in dielectric properties for the nanocomposites is the increase in the dipolar polarization due to the presence of partially reduced GO

where P_r is the remnant polarization, $ε0$ is the dielectric permittivity of free space, $Δε$ is the dielectric relaxation strength, and $E$ is the applied electric field

The reduced GO-PVDF films made using the co-solvent approach shows bending actuation response under an applied DC electric field. Specifically, the films exhibit an intrinsic unimorph behavior such that the sample bends in the same direction for both positive and negative DC-electric field. It is noted that unmodified PVDF and GO-PVDF nanocomposites did not show any bending actuation response when tested up to 10 MV/m. Fig. 12 shows the variation of tip displacement with electric field for the reduced GO-PVDF nanocomposites. The 0.5 wt. % and 1 wt. % PR-GO-PVDF showed similar displacements in the 2–6 MV/m range. The 1 wt. % R-GO-PVDF films showed bending response with tip displacement up to 1.5 mm at very low electric field ∼0.5 MV/m. The tip displacement is used to calculate the S₁₁ strain, by assuming a cantilever beam configuration with a uniform bending moment acting due to the electric field.³⁰ S₁₁ is calculated along the length direction of the sample when the electric field is applied in the thickness direction. At constant curvature and using Hooke's law, the strain along the length can be calculated by the following equation:

where $d$ is the tip displacement, $t$ is the thickness of the film, and $L$ is the length of the part of the cantilever beam which are not constrained.

The mechanism of the observed bending actuation can be explained by the formation of resin-rich layer and GO-rich layer during the solution casting process due to wall depletion effect. Similar wall depletion effect was observed in solution cast polymer nanocomposite films by previous researchers.^2,31–34 When the nanocomposite solution is cast on a glass plate, formation of a resin-rich region happens because of the non-interaction of the nanofillers with the hard glass surface. Nanofillers appear at a smaller concentration in this resin-rich region whereas the rest of the film shows a well dispersed higher concentration of nanofillers as depicted in Figure 13(a). As a result, the nanocomposite film acts as inherent unimorph when electric field is applied; GO-rich region expands whereas the resin-rich region constrains the deformation, leading to bending of the thin film under applied electric field as shown in Figure 13(b). The measured electric field induced strain S₁₁ could be due to several factors like piezoelectricity, electrostriction, electrostatic effect, Joule heating, etc. However, there is no evidence of β phase in the PVDF films and moreover these films are not poled suggesting that there is no contribution from piezoelectric behavior. Strains resulting from Maxwell stress effect for these films are negligible due to low applied electric field and high Young's modulus of PVDF (∼1 GPa at 25 °C). Our calculation using an electric field magnitude of 0.5 MV/m, and a dielectric permittivity ε = 100 (25 °C, 20 Hz) yields a strain value of 2 × 10⁻⁷, which is an order of magnitude below values observed in our study. Also, Maxwell stress effect usually results in a contraction through the thickness direction, while in our study we observe an electric field-driven expansion which caused the film to bend, confirming that the mechanism is not Maxwell stress-driven. IR camera measurements when the sample is subjected to electric field record a maximum temperature increase of 6 °C; therefore, the strain contribution from Joule heating is also negligible. The increase in the dipolar relaxation strength to 51 for reduced GO-PVDF film from 4 for GO-PVDF film confirms that the strain observed in the active layer is due to the polarization-driven electrostriction response, and it shows quadratic response with the applied electric field. The coefficient of electrostriction calculated from the plot of S₁₁ versus square of the electric field for 0.5 wt. % and 1 wt. % PR-GO-PVDF by the co-solvent approach were found to be 1.5 × 10⁻¹⁸ (m²/V²) and 2.25 × 10⁻¹⁸ (m²/V²), respectively; whereas the 1 wt. % R-GO-PVDF showed a coefficient of electrostriction value that is two orders of magnitude higher, namely, 1.7 × 10⁻¹⁶ (m²/V²). This value of M₁₁₃₃ is higher than that of most existing electrostrictive polymers such as PU and P(VDF-TrFE CFE) whose values are 14 × 10⁻¹⁸ (m²/V²) and 8 × 10⁻¹⁸ (m²/V²), respectively.^35,36 The last two rows in Table II list the electrostrictive coefficient and the maximum applied field for two CNT modified PVDF composites. For the SWNT-PVDF case,⁶ the coefficient of electrostriction is very high, in the order of 10⁻¹² m²/V², however the presence of SWNTs reduces the dielectric breakdown substantially such as the maximum possible applied electric field is so low that it greatly limits application. In the case of the multi-walled carbon nanotube (MWNT)-PVDF based composite,³⁷ the best coefficient of electrostriction is lower than ones achieved in our current study. Hence, the GO modified PVDF studied here is a promising system for the development of better EAPs with tunable properties such as dielectric permittivity, electrical conductivity, and dielectric breakdown.

A co-solvent method for synthesis of nanocomposites involving water and DMF has been successfully developed to yield well-dispersed GO in a PVDF matrix. The co-solvent approach helps to disperse the GO as individual nanosheets in water and DMF, then as water is removed, the nanoscale dispersion is maintained in DMF. After the nanocomposite films are made by solution casting, the GO was partially reduced through hot pressing. The hot press temperature was controlled at 150 °C which is lower than the melting temperature of PVDF to avoid agglomeration of the dispersed GO nanosheets. The 1 wt. % R-GO-PVDF nanocomposites show tremendous improvement in dielectric permittivity and electrical conductivity. The dielectric permittivity at 1 kHz increased almost eight folds whereas electrical conductivity shows four orders of magnitude increase compared to unmodified PVDF. Dielectric relaxation spectroscopy results of 0.5 wt. % PR-GO-PVDF suggest that there is a strong dipole relaxation happening near the glass transition of the polymer with higher dielectric relaxation strength for the nanocomposite ∼51 compared to ∼4 for unmodified PVDF at 20 Hz. The reduced GO-PVDF polymer films showed bending actuation response with DC electric field and demonstrated the potential to be EAPs. Mechanism responsible for this bending actuation response is mainly electrostriction because the strain (S₁₁) shows quadratic response with the applied electric field. The coefficient of electrostriction value (M₁₁₃₃) of 1 wt. % R-GO-PVDF showed is 1.7 × 10⁻¹⁶ (m²/V²) which is higher than most of the existing electrostrictive polymers like PU and P(VDF-TrFE CFE) terpolymer, whose values lies in the range of 14 × 10⁻¹⁸ (m²/V²) to 8 × 10⁻¹⁸ (m²/V²). Therefore, using this method we can develop EAPs with tunable parameters like dielectric permittivity, electrical conductivity, and dielectric breakdown strength, which will lead to nanocomposite films with modest electrostrictive response with higher operating voltages.

This work was performed under the framework of the NSF International Institute for Multifunctional Materials for Energy Conversion (NSF IIMEC), Grant No. DMR-0844082.