Sandwich-structured poly(vinylidene fluoride-hexafluoropropylene) composite film containing a boron nitride nanosheet interlayer

High performance dielectric polymer materials are a key point for energy storage capacitors, especially film capacitors. In this paper, a sandwich-structured polymer film is constructed to achieve high energy density and high efficiency. High dielectric materials of poly(vinylidene fluoride-hexafluoropropylene) (P(VDF-HFP)) doped with barium titanate (BaTiO3) are used as the outer layer to achieve a high dielectric constant, and a boron nitride nanosheet (BNNS) layer is inserted between P(VDF-HFP)/BaTiO3 to obtain a high breakdown field strength of composite films. The results indicate that when the doping amount of the BaTiO3 nanoparticles reaches 10 wt% and the mass fraction of the BNNS layer is 0.75 wt%, a significant improvement of energy storage performance is obtained. The energy storage density of the P(VDF-HFP)/BaTiO3/BNNSs composite film can reach 8.37 J cm−3, which is higher than 6.65 J cm−3 of the pure P(VDF-HFP) film. Compared with the P(VDF-HFP) film doped with BaTiO3, significant improvement of the breakdown field strength (about 148.5%) is achieved and the energy storage density increases 235% accordingly, resulting from the inserted BNNSs layer blocking the growth of electrical branches and suppressing leakage current. This novel sandwich-structured film shows promising future applications for high performance dielectric capacitors.


Introduction
Dielectric polymer materials such as poly(vinylidene uoride-hexauoropropylene) (P(VDF-HFP)) are particularly widely used in energy storage, especially lm capacitors, because of their high dielectric constant, low loss, and ease of processing. Electrostatic lm capacitors have been widely used in electric vehicles, exible high-voltage DC power transmission and transformation systems, medical equipment, electromagnetic ejection systems and other elds. [1][2][3][4] However, the energy storage density of the most polymer materials is generally low, which limits the development of lm capacitors. For instance, the energy storage density of biaxially oriented polypropylene (BOPP) lm is only 1 J cm À3 . 5 Energy storage density is a key indicator for evaluating the performance of lm capacitors. As a key material for lm capacitors, dielectric lms can be divided into linear polymers and nonlinear polymers. 6 The storage density of linear polymers can calculate by the formula: and the energy storage density of a nonlinear polymer can be calculated according to the formula: where E b is the breakdown eld strength and D is the electrical displacement which can be expressed as D ¼ 3 0 3 0 r E b (3 0 is the vacuum dielectric constant, and 3 0 r is the relative dielectric constant of the material). Therefore, it is believed that the dielectric constant and breakdown eld strength of the lied material can effectively increase the energy storage density.
In general, compared with polymer materials, inorganic ceramic materials exhibit higher dielectric constant. In recent years, composite materials composed of polymer materials and inorganic ceramics have been widely studied. [7][8][9][10] In order to increase the dielectric constant of polymer materials, the introduction of inorganic ceramic particles, such as BaTiO 3 (BT), Ba x Sr 1Àx TiO 3 (BST), P(Mg 1/3 Nb 2/3 )O 3 -PbTiO 3 , etc., [11][12][13][14] into polymer matrix is considered an effective method. However, in order to achieve a higher dielectric constant of the composite material, it is generally necessary to add a higher volume fraction of the ceramic materials, which leads to problems with the compatibility and dispersion of ceramic particles with the matrix. Hence, several kinds of research have been carried out to solve this problem: surface modication of ceramic llers, 15-17 regulation of ller distribution, 18,19 the addition of third phase llers, etc. [20][21][22] Ceramic materials aer surface treatment and structural control have better compatibility and dispersion.
Boron nitride nanosheets (BNNSs) are a typical twodimension nanomaterial that have become an ideal ller for high performance polymer composites. It has a high forbidden bandwidth (z6 eV) and a high intrinsic breakdown voltage. 23,24 It is widely used in research to improve the energy storage characteristics of polymer materials. For example, BNNSs are directly added to the matrix, which resulted in a signicantly improved breakdown eld strength. 21,25,26 Most of these work is centered around the direct compounding of llers and polymer materials, as well as the uniform distribution of llers in polymer composite lms. However, the structural design of lm is rarely studied. Recently, Huang et al. used the structural design of BNNSs insulation layer to improve the breakdown eld strength and reduce the loss. 27 This structure has been proven very reliable for increasing energy storage density. In order to improve the energy storage characteristics of materials, BaTiO 3 nanoparticles were chosen as the ceramic ller to increase the dielectric constant of the matrix. At the same time, considering that BaTiO 3 will cause a decrease in breakdown eld strength, inspired by their work, 27 BNNSs were chosen as the third phase ller and designed as a separate mezzanine. While using the inorganic ceramic particles to increase the dielectric constant, the effect of suppressing the reduction of breakdown eld strength is achieved by adding BNNSs. Through the coordination between the materials of each phase, it should be an effective way to increase the dielectric constant while suppressing a signicant reduction in breakdown strength. As for details, P(VDF-HFP) is selected and a sandwich structure polymer lms are constructed. High dielectric materials of P(VDF-HFP) doped with BaTiO 3 are used as the outer layer, and a BNNSs layer is inserted between P(VDF-HFP)/BaTiO 3 . Different from the method of directly doping BNNSs into the matrix material, the separate mezzanine is prepared by solution casting. Each layer structure is sequentially prepared on the quartz substrate by a scraper from bottom to top, nally forming a sandwich composite lm. In addition, BaTiO 3 nanoparticles are surface-modied with dopamine to improve the dispersion of nanoparticles in the matrix. The sandwich-structured lm presents a signicant improvement in dielectric constant and energy storage density, the breakdown eld strength of the composite lm is not seriously reduced by the addition of BaTiO 3 . For instance, the sandwich-structured lm has a dielectric constant of 10.99, which is 1.446 times of pure P(VDF-HFP), and the energy storage density has also improved signicantly. This work provides a new way to obtain high performance energy storage media composites.
First of all, ultrasonic degradation method which is reported previously is applied here to prepare BNNSs. 21,26,28 2 g h-BN was added to 200 mL DMF solvent, continuously stripping by ultrasonic breaker for 20 h (340 W). Aer the completion of the ultrasonic, the solution was placed in a centrifuge and centrifuged at 3000 rpm for 45 min, then the supernatant was collected and centrifuged again at 9000 rpm for 10 min. In order to ensure the dispersion of BNNSs, freeze-drying method was used to dry the precipitate at the bottom of the centrifuge tube. As reported in the literature, BNNSs were prepared by ultrasonic degradation method with a yield of 2.5-3%. 27,29 The experimentally collected BNNSs were 60 mg, and the acquisition rate was within the normal range. The experimental results preliminary indicated that h-BN was stripped successfully.
In addition, barium titanate nanoparticles was coated by dopamine through water bath. 30 0.04 g of dopamine hydrochloride was dissolved in an aqueous ammonia solution with pH of 8.5 to form a mixed solution of 2 g L À1 . Appropriate amount of BaTiO 3 nanoparticles was added to the above solution, and stirred in a water bath at 60 C for 12 h. The product was collected by centrifugation. Finally, BT@DPA particles were vacuum dried at 60 C for 10 h.
The specic experimental process can be seen in Fig. 1. Pure P(VDF-HFP) was rst physically blended with 10 wt% of BT@DPA nanoparticles in DMF solvent and mechanically stirred for 12 h until P(VDF-HFP) was completely dissolved. The BNNSs were dispersed in isopropanol to form a solution with 0.75 wt%. The preparation process of the sandwich-structured lm was as follows. Firstly, a P(VDF-HFP)/BT@DPA bottom layer lm was casted on a quartz substrate with a scraper, dried at 70 C for 6 hours in a vacuum oven to remove solvents. Then, the interlayer of BNNSs and the top layer lm were casted through the same process. The sandwich-structured lm was dried overnight in a vacuum oven at 70 C to remove the solvents. Finally, the quartz substrate was placed on a heating table at 200 C for 10 min and quenched in ice water. Herein-aer, this sandwich-structured lm is named PBP/BT@DPA. In addition, a solution of only doped with BaTiO 3 nanoparticles and a solution of co-doped with BaTiO 3 nanoparticles and BNNSs (the doping amount of BNNSs and BaTiO 3 were 10 wt%) were prepared. The same preparation process was used to prepare four other lms: pure P(VDF-HFP), P(VDF-HFP)/BT, P(VDF-HFP)/BT@DPA, HFP/BT@DPA/BNNSs, all of which were single layer (the thickness were 12-13 mm).
A series of morphological characterization and electrical performance tests on the prepared materials and lms were performed. First of all, BNNSs were characterized by Atomic Force Microscope (AFM) and Transmission Electron Microscope (TEM) respectively. The cross section of the lms were observed by Scanning Electron Microscopy (SEM, Hitachi S-4800). In addition, the lms were characterized by Fouriertransform infrared spectroscopy (FTIR, 8400S, Shimadzu) and X-ray diffraction (XRD, Malvern Panalytical, Inc.). The thickness of lm was measured by a magnetic induction thickness-meter (Fischer DUALSCOPE MPO). Finally, dielectric spectrum characteristics were measured by using a precision impedance analyzer (4294A, Agilent Technologies, Inc.) from 40 Hz to 5 MHz. Breakdown eld strength was tested with a withstand voltage tester (TH9120, Changzhou Tonghui electronics co., LTD), and P-E hysteresis loops of lms were measured by using a ferroelectric tester (Radiant Technologies, Inc.).

Characterization of BNNSs and lms
The AFM image is shown in Fig. 2(a) and the degree of peeling and aggregation of BNNSs are expressed by TEM as shown in Fig. 2(b). The picture shows that h-BN aer stripping has become nanosheets of different sizes, and there is no signicant agglomeration between the BNNSs. Fig. 2(c) shows a crosssectional height map of a piece of nanosheet randomly taken, indicating that the thickness of BNNSs is around 2-3 nm. These results verify that h-BN is successfully stripped into BNNSs with very low thickness and relatively good dispersion state. The microscopic morphology of the cross-section of the lms was analyzed by SEM. In order to protect the cross-section morphology of lms, all characterization samples were prepared in liquid nitrogen and then sputter-coated with a homogeneous gold layer. At a magnication of 10000Â, the internal morphology of lms can be clearly observed. The mixed doped lm ller exhibits a poor distribution effect as shown in Fig. 3(c), while the dispersibility of the dopamine-coated BaTiO 3 particles was signicantly improved as shown in Fig. 3(a) and (b). When two nanomaterials are simultaneously doped into the matrix, not only the aggregation between the single particles but also the mutual adsorption between the two phases occur. Such a morphology will lead to a signicant increase in dielectric loss, as evidenced by subsequent testing of the lm. It can be seen from Fig. 3(d) that the upper and lower layers are matrix lms doped with BaTiO 3 particles, and the ller is evenly  This journal is © The Royal Society of Chemistry 2020 RSC Adv., 2020, 10, 2295-2302 | 2297 distributed among them. There is a thin intermediate layer in the middle, namely the cast BNNSs insulation layer, and the nanosheets are densely arranged in the matrix. Three layers can be clearly observed, which veries that this is indeed a lm of sandwich structure.
FTIR and XRD results are visible in the ESI. † FTIR technology is a productive method to identify the different crystal phases of PVDF-based polymers, and the spectra of lms are demonstrated in Fig. S1. † P(VDF-HFP) is a polar polymer with a long molecular chain extending from the crystalline region to the amorphous region. The absorption peaks at 879 cm À1 are the amorphous phase absorption peak of P(VDF-HFP), while the absorption peaks at 764 cm À1 and 796 cm À1 are the a phase absorption peaks of P(VDF-HFP) lm. 31,32 These lms mainly exhibited a phase, indicating that the preparation process used had no signicant effect on the crystalline phase of the matrix material. Analyzing the XRD results in Fig. S2, † the characteristic diffraction peaks of BaTiO 3 are very obvious, and the diffraction peaks of BaTiO 3 at (100), (110), (111) crystal planes can be clearly seen. 33,34 It is worth noting that a diffraction peak appears at 2q z 26.4 , corresponding to the BNNSs (002) crystal plane. 21,35 Because the amount of BNNSs added to the composite lm is small, the peak intensity is very weak due to the inuence of the BaTiO 3 diffraction peak.

Dielectric performances of P(VDF-HFP) lms
The dielectric spectrum of P(VDF-HFP) lms are shown in Fig. 4. The dielectric constant ð3 0 r Þ decreases with increasing frequency, due to the periodic transformation of the electric eld at high frequency resulting in the decrease of charge distribution in the direction of the electric eld. In other words, the role of the molecular dipole moment is gradually reduced. It can be seen from the gure that the dielectric constant of the composite lm doped with BaTiO 3 is much higher than that of the pure P(VDF-HFP) lm. The 3 0 r of PBP/BT@DPA at 1 kHz is 10.99, while the pure P(VDF-HFP) lm is only 7.6, which shows a very large enhancement rate of 3 0 r (144.6%). In contrast, the 3 0 r of HFP/BT@DPA/BNNSs is not increase signicantly, because the doping amount of BNNSs is 10 wt% and its dielectric constant is inherently low ($4). The result shows that the effect of BaTiO 3 on the improvement of dielectric constant is very obvious. The dielectric loss tan d of all lms shows a gradually rising trend. 36 Compared with pure P(VDF-HFP) lm, the composite lm doped with BaTiO 3 particles shows a higher loss at low frequency. On the one hand, Maxwell-Wagner-Sillars (MWS) polarization dominates the dielectric response at low frequencies. On the other hand, the addition of BaTiO 3 particles also increases the dielectric loss. 37,38 The loss increases rapidly at high frequencies, which can be attributed to the dynamic relaxation of the matrix material. 39 The dielectric loss of HFP/ BT@DPA/BNNSs lm is much larger than other lms, this data validates the results of the SEM. The dielectric loss of the polymer is not only related to the nature of the material itself, but also to the state of aggregation between the phases. 40,41 The agglomeration and local accumulation of ceramic particles directly increase the porosity of the mixture and increase the water absorption of the composite. When there are many pores in the matrix, the phase composition of each material in the actual matrix deviates from the theoretical value, and the dielectric loss increases by order of magnitude. The moisture absorption of pores will increase the moisture content of the dielectric material, which has a great inuence on the dielectric loss. The insulation layer of BNNSs are separately added into sandwich-structured lm, avoiding mutual aggregation between the phases, which reduces the porosity of the composite and achieves better interface conditions. Compared with the lm directly doped with two kinds of inorganic materials, this structure makes the density distribution of the lm more uniform, which is very helpful for improving the dispersion of the ller. However, the insulation layer of BNNSs doesn't signicantly reduce the dielectric loss at high frequency. This may be due to the large amount of BaTiO 3 added, and the effect of suppressing the loss of BNNSs is not conspicuous. Ceramic llers can effectively increase the dielectric constant, but the  content is oen high, which also weakens the exibility of the polymer lm. Compared with PVDF, P(VDF-HFP) has a lower glass transition temperature and a higher degree of amorphization, resulting in abetter exibility. This is the reason why P(VDF-HFP) is chosen as the matrix material. 42,43 In short, the composite lm with BNNSs interlayer shows a great improvement in dielectric properties. Not only does the dielectric constant increases signicantly, but also the dielectric loss is suppressed to some extent.

Breakdown characteristics of several lms
The Weibull distribution is used to evaluate the breakdown eld strength characteristics, the expression is as follows: where P(E) is the cumulative electrical breakdown probability, E is the experimental electrical breakdown strength, and E b is the characteristic breakdown strength when the cumulative probability reaches 63.2%. The shape factor b shows the degree of dispersion of the data, and the higher the value, the higher the reliability. Before conducting the comparative experiment, the BNNSs concentration test was performed to choose a suitable concentration. Using isopropyl alcohol as a solvent, BNNSs were formulated into a solution at a ratio of 0.25 wt%, 0.5 wt%, 0.75 wt% and 1 wt%. Four solutions were added as an insulating layer to the matrix material and tested for their breakdown eld strength. The results are visible in Fig. S3 of the ESI. † Aer 0.75 wt% of BNNSs were added into the matrix, the breakdown strength reached 414.76 kV mm À1 , the highest of all samples. Therefore, the 0.75 wt% of BNNSs was selected for subsequent experiments, and the test results of several lms are shown in Fig. 5(a) and (b). The E b and b of pure P(VDF-HFP) lm are 434.09 kV mm À1 and 8.88, respectively. Compared with the P(VDF-HFP) lm, the breakdown eld strength of the lm doped only with BaTiO 3 particles is seriously reduced. The breakdown eld strength of a three-phase mixed doped lm with a large amount of BNNSs added is not improved, which is at the same level as P(VDF-HFP)/BT@DPA. Aer three-phase mixed, there are many defects and holes inside the lm inevitably. As a kind of pre-breakdown phenomenon in dielectric lm, electric branches mainly occur in areas where defects, voids, and conductive llers are concentrated. The three-phase mixed lm, despite the large content of BNNSs, does not form a complete topological barrier, and the electric branches will grow freely in the voids, which is also the direct cause of the decrease in breakdown eld strength. On the contrary, during the growth of electric tree branches, the BNNSs interlayer acts as a barrier or scattering interface in sandwich-structured lm. 27 The breakdown eld strength of PBP/BT@DPA is 414.76 kV mm À1 , which is not much different from P(VDF-HFP). The shape factor b of the lms doped with BNNSs is higher than other samples, in particular, the shape factor b of PBP/BT@DPA is 1.82 times that of pure P(VDF-HFP). The results indicate that the introduction of BNNSs interlayer can signicantly improve the breakdown characteristics of composites compared to directly doped BaTiO 3 nanoparticles.
Energy storage characteristics of several P(VDF-HFP) lms The composite lm prepared by referring to the structural design of the BNNSs insulating layer is to increase the dielectric constant under the condition of maintaining the breakdown eld strength, thereby increasing the energy storage density of the lm. From the results, the purpose of experiment are also veried. As shown in Fig. 5(c), the energy storage density of modied lms are higher than of pure P(VDF-HFP) lm under the same breakdown eld strength. The energy storage density of P(VDF-HFP)/BT, P(VDF-HFP)/BT@DPA, and HFP/BT@DPA/ BNNSs is close to the limiting value at 300 kV mm À1 , however, the energy storage density of PBP/BT@DPA can continue to rise. The energy storage density of the lm under the maximum breakdown eld strength are calculated, and the results are shown in Fig. 5(d). The pure P(VDF-HFP) lm has a storage density of 6.65 J cm À3 under the maximum breakdown eld strength. In contrast, compared with P(VDF-HFP), the lms of P(VDF-HFP)/BT, P(VDF-HFP)/BT@DPA, and HFP/ BT@DPA/BNNSs, not only the breakdown eld strength is lower, but the energy storage density is also not as good as the former. The sandwich-structured lm exhibits a storage density of 8.37 J cm À3 under the maximum breakdown eld strength (414.76 kV mm À1 ), which is 125.9% higher than that of pure P(VDF-HFP) lm. The performance of the prepared PBP/ BT@DPA composite lm is compared with other polymer lms reported in the literature, as shown in Table 1. The table reveals that the sandwich-structured lm of this work has better comprehensive properties than other lms. In addition, there is still a lot of research space for this sandwich-structured lm, and how to further improve the breakdown eld strength is very worth exploring in subsequent studies. P-E hysteresis loops and leakage currents of several lms are visible in the ESI, † and the charge-discharge efficiency are shown in Fig. 6(a). The efficiency of the lm doped with BaTiO 3 decreases very sharply with the increase of electric eld, and the charge-discharge efficiency is less than 50% at 200 kV mm À1 . Compared with P(VDF-HFP), the efficiency of the sandwichstructured lm does not decrease signicantly. As the electric eld strength increases, the difference in electrical properties between the base material and the ller causes an increase in the charge carrier concentration of the lm, thereby causing a decrease in charge-discharge efficiency during the electric eld switching process. BNNSs have a high length height ratio and specic surface area, and the high energy barrier of the nanosheets can suppress the movement of charges under high electric elds. 51 The interlayer formed by solution casting has a dense network structure and plays a vital role in hindering the migration of charge carrier. 52 In Fig. 6(b), the insulation resistivity of the lm was calculated according to leakage currents. The leakage current of the dielectric lm can be expressed as: where U is an applied voltage and R I is an insulation resistance. The R I can be calculated by other parameters: where r I is insulation resistivity, the d is the dielectric lm thickness, and A is the area of the aluminum electrode evaporated on the dielectric lm. From these two equations, the calculation formula of the insulation resistivity can be derived: Under low electric eld, the conductive effect of BaTiO 3 is not obvious, so the HFP/BT@DPA/BNNSs lm with 10 wt%  BNNSs added shows the highest insulation resistivity. As the electric eld strength increases slowly, the conduction of BaTiO 3 is also gradually enhanced. Meanwhile, the role of the BNNSs interlayer is gradually becoming prominent, which hinders the conduction of leakage current. Under the same electric eld strength, sandwich structure lm has a higher insulation resistivity.

Conclusion
A novel composite material formed by adding high dielectric inorganic ceramic particles and BNNSs interlayer into a sandwich-structured lm. This method is indeed feasible by increasing the dielectric constant to achieve high storage density while ensuring a high breakdown eld strength. The doped BaTiO 3 particles are as high as 10 wt%, while the added BNNSs insulating layer is only 0.75 wt%. Compared with the direct doping of a large amount of BNNSs, this method greatly reduces the amount of BNNSs used. The dielectric constant of the composite lm is 10.99, which is 144.6% of that of the pure P(VDF-HFP) lm. The breakdown eld strength is not much different from pure P(VDF-HFP), furthermore, the shape factor is greatly improved. The sandwich-structured lm with BNNSs interlayer not only increases the energy storage density, but also reduce the leakage current conduction with the BNNSs interlayer as a topological barrier, keeping the charge-discharge efficiency of the lm at a normal level. In short, this is a reliable sandwich structure that provides new ideas for simultaneously increasing the dielectric constant and breakdown eld strength.
In future work, how to increase the breakdown eld strength while maintaining a high dielectric constant and chargedischarge efficiency may be the focus of research.

Conflicts of interest
There are no conicts to declare.