Introduction

With the fast-growing of advanced electronics and electric power system, soft film capacitors with both high-power and high-energy density at the same time have drawn more and more attention (Dang 2018). This film capacitors (TFCs) usually request high capacity density, large breakdown strength [typically exceeds 300 MV m−1 (Paniagua et al. 2014)], high-frequency response, fast charge/discharge speed, low dissipation, and long lifetime. Compared with electrochemical capacitors and batteries, the charge/discharge speed of capacitors depends on their polarizations, which are contributed by the orientation of dipoles and deformations of atoms and molecules and are much faster and less influenced by high frequency than charge carriers transport. With higher dielectrics, the capacitors with same capacitance can be smaller in size and lighter in weight, thus have higher capacity density. Polymer-based dielectrics surpass ceramics by its higher breakdown strength, lower dissipation, easier preparation and longer lifetime.

The most commonly used dielectric polymers for TFCs mainly are non-biodegradable and nonrenewable thermoplastic polymers, such as polypropylene (PP) (Kumari and Ghosh 2018; Lay et al. 2018), polyethylene terephthalate (PET) (Tang et al. 2018; Topala et al. 2007), polyethylene naphthalate (PEN), polyphenylene sulfide (PPS), polytetrafluoroethylene (PTFE) (Thomas et al. 2008; Wang and Dang 2018; Yuan et al. 2014, 2018) and polystyrene (PS). Besides, other polymers include polyimide (PI) (Chang et al. 2009; Feng et al. 2013, 2014; Ishmael et al. 2014; Kizilkaya et al. 2012; Koytepe et al. 2008; Lay et al. 2018; Lee et al. 2009, 2012; Lee and Wang 2010; Lin et al. 2017; Lu et al. 2017; Meena et al. 2012; Olariu et al. 2017; Wang et al. 2018, 2010b; Wu et al. 2017; Yin et al. 2014; Zha et al. 2010a, b), polyamide (PA) (Novac et al. 2017; Qi et al. 2017, 2018), and polyvinylidene fluoride (PVDF) (Al-Saygh et al. 2017; Alam et al. 2017; Deshmukh et al. 2017; Dou et al. 2017; Gan and Abd Majid 2014; Park et al. 2013; Prabakaran et al. 2014; Rekik et al. 2013; Ribeiro et al. 2018; Su et al. 2016; Wang et al. 2010b; Yang et al. 2016) are also studied. PI-based nanocomposites are reported to have enhanced corona aging performance (Lin et al. 2017; Lu et al. 2017; Yin et al. 2014; Zha et al. 2010a), but their relative dielectric constant is generally below 6 (1 kHz) (Feng et al. 2013, 2014; Lay et al. 2018; Wang et al. 2018). Qi et al. prepared polyamide11 (PA11)/BaTiO3/carbon nanotube (CNT) ternary nanocomposites with 3D segregated percolation routes, the relative dielectric constant is 16.2 with a dielectric loss of ~ 0.08 (1 kHz) (Qi et al. 2018). Alam et al. (2017) put titanium dioxide (TiO2) nanoparticles into \( \gamma \)-phase containing PVDF, and the dielectric constant of the nanocomposite film reaches 32 with a dielectric loss of 0.25 (1 kHz). Recently, low-cost and eco-friendly cellulose nanofibrils are being increasingly explored as a candidate to replace some conventional dielectric materials (Abdel-karim et al. 2018; Al-Saygh et al. 2017; Bonardd et al. 2018; Chiang and Popielarz 2002; Gaspar et al. 2014; Inui et al. 2015; Jayaramudu et al. 2018b; Le Bras et al. 2015; Madusanka et al. 2016, 2017; Milinskii et al. 2018; Milovidova et al. 2014; Poyraz 2018; Poyraz et al. 2017b; Rajala et al. 2016; Shi et al. 2018; Yagyu et al. 2017; Zeng et al. 2016; Zhou et al. 2018).

Cellulose nanofibril (CNF) has a low density and coefficient of thermal expansion (CTE) (12–28.5 ppm K−1), high mechanical strength (200–400 MPa) and Young’s modulus (7.4–14 GPa), excellent thermal stability (> 180 °C) and chemical durability, and it is an almost inexhaustible green material (Du et al. 2017; Fujisaki et al. 2014). The relative dielectric constant (εr) of traditional paper prepared from micro-sized cellulose is in the low range of 1.3–4.0, resulting from the porous microstructure (Inui et al. 2014; Inui et al. 2015). With a densely packed nanostructure, the εr of nanocellulose paper reaches 5.3 (at 1.1 GHz) (Inui et al. 2014, 2015) with a breakdown strength of 613.8 kV cm−1 (Zeng et al. 2016), making it a promising candidate for the dielectric matrix. Comparing to regenerated cellulose films that usually need toxic and expensive solvent, nanocelluloses can be made from pure mechanical grinding so toxic solvent is no longer needed. Furthermore, the dispersibility of TiO2 nanoparticles in regenerated cellulose/solvent is poor, and phase separation between nanoparticles and cellulose occurs during film preparation. However, CNF is a nanofibrils rather than soluble molecules so the phase separation between TiO2 particles and nanofibril network could be effectively prevented. However, the hydroxyl-richen cellulose shows strong hydrophilicity, which inevitably results in high electric leakage, high dielectric loss, low breakdown strength and low energy densities in humid environments (Shimizu et al. 2016; Yang et al. 2018a, b). Many studies have tended to focus on further improving the dielectric constant of cellulose nanopapers by introducing conductive fillers (Inui et al. 2015; Ji et al. 2017; Kafy et al. 2015b; Milovidova et al. 2014) but little attention has been paid on reducing dielectric loss.

Herein, we prepared a high dielectric composite film based on TiO2 and CNF by a solution casting method. TiO2 has a high dielectric constant (εr = 63.7 at 1 MHz), low dielectric loss (\( \tan\updelta < 0.051 \)) (Wypych et al. 2014) and is stable in a broad temperature range (< 1000 °C). Besides, the hydrophilic property of TiO2 offers a way out for homogeneous mixing with CNF suspension. Thus, TiO2 is a promising candidate for CNF based dielectrics. Homogenous composite films were made by mechanically mixing TiO2 nanoparticles with CNF. The relative dielectric constant (εr) and dielectric loss (\( \tan\updelta \)) were studied as the function of frequency and filler content. The effects of hot-press treatment on dielectric properties, microscopy, thermal stability, dynamic mechanical properties, and hydrophilicity of composite films were also studied.

Experimental

Materials

Both CNF slurry and 2,2,6,6-tetramethylpiperidinooxy (TEMPO)-oxidized CNF (TCNF) slurry were purchased from the University of Maine, with solid content of 3.4 wt% CNF and 1.1 wt% TCNF, respectively (Fukuzumi et al. 2010; Isogai et al. 2011; Kumar et al. 2014; Osong et al. 2016; Postek et al. 2013; Sacui et al. 2014; Saito et al. 2009; Stelte and Sanadi 2009). TiO2 nanoparticles of diameter ~ 21 nm (P25) were purchased from Nippon Aerosil Co. Ltd.

Preparation of CNF/TiO2 composite film

The purchased CNF and TCNF were diluted into 0.34 wt% and 0.30 wt% respectively by distilled water. According to the dry weight percentage, a certain amount of TiO2 was added, then the solution was stirred homogeneously by a homogenizer for 10 min before poured into a petri dish. The dried film was obtained after being put in a fume for 2–4 days. After being hot-pressed under 80 °C and 1.1 MPa for 3 h, light yellow, flat and thin films were obtained. At least three samples were prepared for each composition. The thickness of the sample films was in the range of 30–100 µm. Figure 1 shows the flow diagram of the preparation of the CNF/TiO2 composite film.

Fig. 1
figure 1

The flow diagram illustrates the preparation of CNF/TiO2 composite film

Full size image

Characterizations

Morphologies of both surface and cross-section of the composite films were analyzed by thermally assisted field emission scanning electron microscope (TFE-SEM, LEO 1530, Germany) at an accelerating voltage of 10 kV. The composite films were sputtered with gold in vacuum to avoid accumulation of charge before observation.

Thermogravimetric analysis and differential scanning calorimetry (TGA–DTA) was performed on a simultaneous thermal analyzer (PerkinElmer STA6000, USA), ranging from 35 to 700 °C with a heating rate of 10 °C min−1 under nitrogen atmosphere and hold at 105 °C for 10 min before heading to a higher temperature.

The dielectric properties of sample films were measured by an LCR meter (Keysight E4980 with a 16451B fixture, USA) in the frequency range of 20 Hz–2 MHz. The test for each sample film was repeated at least five times. The thickness of the sample film was measured by a micrometer and was averaged over seven measurements on each sample. Unless otherwise stated, the dielectric constant of a material refers to the relative dielectric constant, which is the ratio of its absolute dielectric constant to the dielectric constant of vacuum. The relative dielectric constants (εr) of the sample films were calculated by Eq. (1)

$$ C{ = }\varepsilon_{ 0} \varepsilon_{\text{r}} \frac{A}{d} $$
(1)

where C is the capacitance; \( \varepsilon_{ 0} \) is the absolute dielectric constant of vacuum, \( \varepsilon_{ 0} { = 8} . 8 5 4\times 1 0^{ - 1 2} \;{\text{F/m}} \); A is the electrode area, \( A = 1.963 \times 10^{ - 12} \;{\text{m}}^{2} \); d is the thickness of the sample film.

Tensile strength and ultimate elongation were studied with a dynamic mechanical analyzer DMA (Q800, TA Instruments, New Castle, DE, USA) with a test rate of 10% min−1 at room temperature. The film specimens were 5 mm wide and 20 mm long. At least four specimens were tested for each sample.

The densities of the sample films were calculated by Eq. (2)

$$ d = \frac{m}{A \times t} $$
(2)

where d is the density, m is the weight, A is the surface area, and t is the thickness.

Results and discussion

Microscopy

Figure 2 shows the surface and cross-section morphologies of pure CNF and TiO2 (50 wt%)/CNF composite film. The CNF was typically dozens of micrometers long with a diameter lower than 0.3 µm, and part of CNFs aggregated with each other (as shown in red circle). However, unlike normal paper which has a porous structure, no obvious pore was observed in pure CNF film. Figure 2b shows some level laminated structures with obvious layer gaps that might result from the peeling during the sample preparation. A dense layer and much smoother surface were observed, and it might result from the hot-press treatment. Figure 2d illustrates the distribution of TiO2 in CNF. Because of the poor capability between the inorganic filler and organic matrix, there were some small pores, which had great influence on the properties of the composite film.

Fig. 2
figure 2

Cross-section and surface SEM images of pure CNF (a, c) and hot-pressed CNF/TiO2 (50 wt%) (b, d)

Full size image

Dielectric properties

Prior to study the dielectric properties, a series of dielectric tests with different levels of oscillation signal (OSC level) was conducted. As shown in Fig. 3, the OSC level was adjusted in the range of 0.1–2 V. At lower OSC level of 0.1–1.5 V, the dielectric constant increased with the increase of OSC level, and the growth rate was decreased. After OSC level reached 2.0 V, the dielectric constant started to decrease. The same results were obtained in other composite films with different TiO2 content. It illustrated that the testing electric field starts to overpass the breakdown strength of the films. Higher OSC level brings a higher risk of breakdown. Thus, the following dielectric tests were conducted with OSC level at 1.0 V.

Fig. 3
figure 3

The influence of OSC level on the dielectric test (50 wt% TiO2/TCNF)

Full size image

Figure 4 shows the influence of frequency and filler content on relative dielectric constant (εr) and dielectric loss (\( \tan\updelta \)). With the increase of frequency, both εr and tanδ decreased. In the lower range of 20–100 kHz, sharp reductions were observed. It was caused by the electrode polarization which took place at the interface between metal electrode and samples, and Maxwell–Wagner–Sillars interfacial polarization which can be observed at the interface between CNF and TiO2 in inhomogeneous materials (Emmert et al. 2011; Lu and Zhang 2006; Samet et al. 2015; Anju and Narayanankutty 2016; Mohiuddin et al. 2015). Resulted from the interfacial polarization, the test results of εr and tanδ at the point of 20 Hz were abnormally high (several hundred or even higher) and were not showed in Fig. 4. In the higher frequency range of above 500 kHz, the reduction of both εr and tanδ tended to be flatter and showed much less dependence on frequency and filler content, which indicated that electronic and atomic polarization and orientation polarization started to play a predominant role.

Fig. 4
figure 4

The influence of frequency and filler content on the dielectric constant and dielectric loss

Full size image

The dielectric properties of the composite films can be explained by the multi-layered core model (Tanaka et al. 2005). According to this model, the interface of filler particle is chemically consisting of a bonded layer, a bound layer, a loose layer, and an electric double layer overlapping the above three layers. The addition of nano-sized TiO2 has contradicted effects on the dielectric properties. On the one hand, nano-sized TiO2 introduced large surface areas for interfacial polarization compared with micro-sized TiO2 (Kafy et al. 2015a), resulting in an extraordinary increase of both εr and tanδ in the low frequency range. On the other hand, the bonded layer and bound layer of the nanoparticles impair the motion of dipoles, leading to a reduction of εr and tanδ. And the dipoles and ionic carriers in the loose layer may act inversely. The far-field effect caused by the double electric layer makes the neighbored nano-particles collaborate with each other. The imperfection of heterogeneous structures can increase εr and tanδ. With the interfacial polarization becoming weak in the high frequency range, TiO2 content showed less effect. In the case of micro-sized TiO2, the increased εr is usually explained in term of the Lichtenecker–Rother logarithmic law of mixing (Tanaka 2005).

The εr of CNF/TiO2 increased with the increasing of TiO2 content in the range of 0–50 wt%. The maximum dielectric constant was 19.51 (at 1 kHz) at 50 wt%. As the TiO2 content continued to increase, the dielectric constant decreased. It was caused by the aggregation of TiO2 and the appearance of pores, as shown in SEM images. In the range of 0–70 wt%, the dielectric loss showed less dependency on TiO2 content and fluctuated in the range of 0.51–0.81 (at 1 kHz). The εr of TCNF/TiO2 increased with TiO2 content in the range of 10–70 wt% and reached the maximum value of 47.15 (at 1 kHz) at 70 wt%. The dielectric loss of TCNF/TiO2 composite films also showed less dependency on TiO2 content and was fluctuated in the range of 2.57–3.32 (at 1 kHz). Generally, the dielectric loss of TCNF/TiO2 composite films was three times higher than CNF/TiO2 composite films, which was caused by the residual ions after TEMPO oxidation treatment of CNF. Thus, from the perspective of reliability and energy saving, CNF/TiO2 was better than TCNF/TiO2. Compared with other reported CNF based dielectrics, our CNF/TiO2 showed a much lower dielectric loss.

Thermal properties

In order to study the thermal stability of the composite films, TGA–DTA measurements were conducted on a series of CNF/TiO2 composite films with the TiO2 content in the range of 10–50 wt%. Figure 5a, b shows the typical TGA–DTA curves for pure CNF and CNF/TiO2 (50 wt%) composite film after hot-press treatment, respectively. In the low temperature region below 105 °C, weight losses of 5.66–2.95% on the TGA curve and an endothermic peak at 41 °C on the DTA curve were observed. It was caused by water evaporation (Chenampulli et al. 2019; Hassan et al. 2019; Jayaramudu et al. 2018a; Lizundia et al. 2016; Poyraz 2018; Poyraz et al. 2017a; Raghunathan et al. 2017; Zeng et al. 2016). The temperature was held at 105 °C for 10 min, and it was showed obviously on the DTA curve as a marked drop. The 5% decomposition temperature of the sample films was 291–302 °C, which indicated that the TiO2/CNF has a low water absorption and good thermostability. The fluctuation of the DTA curve and sharply loss of weight around 400 °C showed the decomposition of CNF. At 700 °C, for the composite films, the residual weight percentage of sample films was per the TiO2 content, which indicated that CNF had been completely decomposed at this temperature. However, for pure CNF, the weight percentage was 15%. It indicated that the addition of TiO2 accelerates the thermal degradation of CNF.

Fig. 5
figure 5

a TGA–DTA curves of pure CNF film. b TGA–DTA curves of CNF/TiO2 (50 wt%) composite film after hot-press treatment. c Dynamic mechanical properties of hot-pressed (HP) and untreated CNF/TiO2 composite films with different TiO2 content. d Densities of untreated and hot-pressed (HP) CNF films

Full size image

Dynamic mechanical property

Figure 5c shows the DMA test results of CNF/TiO2 composite films. All the stress–strain curves show “S” shape, suggesting that with the increase of strain, the stress first increased slowly, then sharply, and turned to be slow again, and finally the sample broke down, which is a typical property of flexible films. For pure CNF, hot-pressing treatment improved the strain at break from 7.65 to 10.92% with the stress slightly decreased from 76.91 to 76.59 MPa, which was resulted from better bounding between CNFs after hot-pressing treatment. However, CNF/TiO2 (50 wt%) composite film has a contract result. After hot-pressing, the strain at break decreased from 5.45 to 2.63%, and the stress increased from 12.68 to 17.71 MPa, which may be caused by the weak bonding between TiO2 particles and CNFs.

Density

Figure 5d shows the densities of untreated and hot-pressed (HP) CNF films. Pure CNF films have higher densities than CNF/TiO2 (50 wt%) although the density of TiO2 is much higher than cellulose. Because the poor bonding between CNF and TiO2, a looser structure of the films was obtained by adding TiO2 particles, which is confirmed by SEM images shown in Fig. 2. Hot-pressing could improve the density of CNF films, and the rise of pure CNF is much higher than that of CNF/TiO2 (50 wt%), suggesting that the pores in the CNF/TiO2 films are hardly be removed by physical treatment.

Conclusions

In conclusion, a high dielectric film was prepared under facile condition. The relative dielectric constant of CNF/TiO2 composite film reached 19.51 (at 1 kHz). Compared with pure CNF films (εr = 6.92 at 1 kHz), the εr of composite films was improved about three times. It was also illustrated that hot-pressed CNF/TiO2 had good flexibility and thermal stability. The addition of TiO2 particles reduces the cellulose–cellulose bonding so generates more pores in the films, which has significant impacts on both dielectric and physical strength properties.