Bioinspired, compositionally graded cellulose-based dielectrics with Schottky-engineered interfaces for high-performance and sustainable energy storage

Abstract

Inspired by the hierarchical structure of the feathers of black swans at the Shaanxi University of Science and Technology (SUST), we developed compositionally graded cellulose-based composite films incorporating BCZT ceramic fillers with varying compositions into a cellulose/P(VDF–HFP) blend film (C8/PH2) for high-performance and sustainable dielectric capacitors. Three configurations—single-layer, down-graded trilayer (C8/PH2-BCZT-dg), and up-graded trilayer (C8/PH2-BCZT-ug)—were fabricated and systematically evaluated. The C8/PH2-BCZT-dg film achieved the highest recoverable energy storage density (W_rec = 38.73 J cm⁻³) and efficiency (η = 79.39%), attributed to the stable Schottky emission conduction across its interfaces, as revealed by current–voltage fitting and band-diagram analysis. In contrast, the C8/PH2-BCZT-ug film structure exhibited a conduction-mechanism transition to Ohmic contact, leading to reduced breakdown strength. Finite element simulations confirmed the experimental breakdown trends and highlighted the role of internal potential distribution. The C8/PH2-BCZT-dg film also demonstrated excellent frequency stability (10 Hz–10 kHz), cycling durability (10⁶ cycles), and high-power performance, with rapid energy release (t_0.9 = 41.97 ns) and a discharge energy density of 21.07 J cm⁻³ at 5.0 MV cm⁻¹. Furthermore, combustion testing revealed the superior fire resistance of the film, underscoring its safety for long-term operations. These results establish hydrogen-bond-engineered, compositionally graded cellulose composites as promising eco-friendly alternatives to petroleum-based dielectric materials for advanced energy-storage applications.

---

New concepts

We present a new concept of dual-mode cellulose-based piezoelectric nanogenerators (C-PNGs) that simultaneously harness d₃₃ and d₃₁ piezoelectric responses by integrating compositionally graded BCZT ceramic fillers within a cellulose/P(VDF–HFP) matrix. Unlike conventional nanogenerators that rely on a single piezoelectric mode or uniform filler distribution, our design leverages hierarchical filler gradients and multi-scale hydrogen-bonded interfaces to synergistically activate both out-of-plane and in-plane polarization pathways. This dual-mode operation not only enhances energy-conversion efficiency but also enables concurrent energy harvesting and real-time sensing functions, a capability rarely achieved in current systems. The introduction of compositional gradients further provides a material-level strategy to couple the piezoelectric response with dielectric band-structure regulation, offering new insights into how multi-scale structural design governs polarization dynamics in organic–inorganic composites. This concept bridges the gap between sustainable bio-based substrates and high-performance ferroelectric ceramics, advancing the field toward multifunctional and eco-friendly energy devices. By demonstrating that carefully engineered filler gradients can reprogram piezoelectric modes, our work establishes a generalizable design principle for next-generation smart materials and opens a pathway to integrated energy-sensing platforms beyond the limitations of traditional single-mode nanogenerators.

1. Introduction

Compared to rigid ceramics, dielectric capacitors based on flexible polymers offer significant advantages, including superior mechanical properties and broader application potential across diverse scenarios.^1–8 In recent decades, polymer dielectrics have primarily evolved along two major research directions: those based on polyvinylidene fluoride (PVDF) and those derived from polyimide (PI) or polyetherimide (PEI). PVDF-based dielectrics are particularly noted for their high polarization, which originates from their all-trans molecular conformations. In contrast, PI- or PEI-based materials exhibit excellent thermal stability, attributed to their rigid aromatic backbones and robust imide linkages. These distinct molecular features endow each class of polymers with unique advantages and research prospects for advanced dielectric applications. For PVDF-based dielectrics, ceramic fillers are commonly embedded to form composite structures, wherein the deliberate introduction of 0-3 type interfaces has been shown to significantly enhance the energy storage density (W_rec, as defined in SI) through interface engineering.^9–11 Beyond this, the incorporation of 2-2 type structural interfaces has proven effective in increasing W_rec. This is largely due to their ability to impede the propagation of electric currents, a phenomenon the authors previously demonstrated to enhance the dielectric breakdown strength (E_b),^12–16 thereby contributing further to energy-storage performance.^17,18 For PI-based dielectrics, Q. Wang et al. reported the use of in situ polymerization to incorporate ceramic fillers, which promoted more uniform filler dispersion and improved W_rec values.¹⁹ Additionally, H. Wang et al. demonstrated that in situ formation of oxide phases within PI-based composites can suppress charge injection at the electrode/dielectric interface, thereby enabling increased W_rec and improved thermal stability under elevated operating temperatures.²⁰

Although extensive studies have been reported, these two types of polymers still suffer from the drawback of non-biodegradability, posing new challenges for the development of next-generation flexible energy-storage capacitors. In the search for greener alternatives, cellulose—primarily derived from bamboo, cotton, and wood—has attracted increasing attention due to its intrinsic polarization. According to Eiichi Fukada's work, the spontaneous polarization of cellulose arises from symmetry breaking within the hydrogen-bond network, and cellulose exhibits significant piezoelectricity along the d₄₁ and d₂₅ orientations.^21,22 Building on this foundation, cellulose has been extensively explored over the past decade for applications in energy storage and conversion.^23–27 As discussed above, the 2-2 type structure facilitates the enhancement of E_b to achieve increased W_rec values. However, constructing a 2-2 structure on cellulose films is challenging because the self-assembly behavior of the hydrogen bonds during the tape-casting process causes the films to shrink.²⁸ This shrinkage makes it almost impossible to build a second layer on top of the first, becoming a major bottleneck in increasing the W_rec values of cellulose-based dielectric capacitors. In the authors’ previous work, by employing a hydrogen-bond replacement strategy through the addition of PVDF, the self-assembly behavior was effectively weakened due to the higher electronegativity of F⁻ compared to that of O.^2–30 As a result, 2-2-structured cellulose-based films were successfully fabricated on 80 wt% cellulose-20 wt% PVDF (C8/P2) blend films. Benefiting from the interfacial polarization-coupled Schottky barrier height and electric field redistribution, the W_rec of the cellulose-based composite films reached 27.20 J cm⁻³ and later 31.07 J cm⁻³, demonstrating the great potential of cellulose for use in energy-storage capacitors. Moreover, this hydrogen-bond replacement strategy has not only been applied to enhance the energy-storage performance of cellulose-based dielectric capacitors but has also proven effective in optimizing the output performance of piezoelectric nanogenerators (PENGs).³¹

To enhance the W_rec, various strategies, such as domain engineering, interface engineering, and entropy engineering, have been extensively explored in recent studies.^12,32,33 Although the W_rec values achieved with polymers still lag behind those of epitaxial thin films, the performance of cellulose-based capacitors appears to have reached a bottleneck. Based on the authors’ long-term research on energy-storage dielectric capacitors, it is noted that only the positive branch of the polarization–electric field loop (ferroelectric hysteresis loop or P–E) contributes effectively to W_rec, despite the symmetric nature of the loop about the origin. In most cases, the negative branch provides little or no contribution. Drawing on the authors’ previous work on ferroelectric memristors, which demonstrated the formation of distinct resistive states due to variations in the Schottky barrier height,^34,35 it is theoretically feasible to improve the positive W_rec by sacrificing the polarization response on the negative side. This approach requires films to exhibit different conduction mechanisms under positive and negative electric fields. Inspired by the unique structure of black swan feathers observed at the Shaanxi University of Science and Technology (SUST), as shown in Fig. 1(a1), a compositionally graded architecture was proposed to meet the above requirements. Compared to the case in the authors’ previous work, in this study, PVDF was replaced with P(VDF–HFP) due to its higher β-phase content, and an 80 wt% cellulose-20 wt% P(VDF–HFP) (C8/PH2) matrix was formulated. As ceramic fillers, (1 − x)BaZr_0.2Ti_0.8O₃–x0.7Ba_0.7Ca_0.3TiO₃ (BCZT) particles with different compositions of BCZT-37 (x = 0.3), BCZT-55 (x = 0.5), and BCZT-73 (x = 0.7) were employed. Utilizing a hydrogen-bond replacement strategy, both single-layered (C8/PH2-BCZT-37, C8/PH2-BCZT-55, C8/PH2-BCZT-73, and C8/PH2-BCZT-m) and multilayered films (C8/PH2-BCZT-dg and C8/PH2-BCZT-ug) were successfully fabricated. By analyzing the asymmetric electrical behavior in conjunction with band theory and conduction-mechanism transitions, the compositionally down-graded structure (C8/PH2-BCZT-dg) exhibited the highest W_rec. This work offers a promising new strategy for the design of high-performance, cellulose-based energy-storage capacitors.

Fig. 1 (a1) The photograph of a black swan residing on the lake at SUST; (a2) a closer look at the feather microstructure; (b) the sketch of the layer structure for all the films in this work; (c1) the XRD patterns in the 2θ range of 17.5°–77.5°; (c2) the enlarged-view around the (110) plane of (c1); (d) the FT-IR spectra of all the films; (e1)–(e3) the Rietveld refinement results of the XRD data for three different BCZT ceramic powders, with the corresponding ball-and-stick crystal structure models as insets in each figure; photographs of water-contact-angle measurements on the surface of the (f1) C8/PH2-BCZT-dg; (f2) C8/PH2-BCZT-ug films; and (g) the transmittance property of all the films.

2. Experimental section

Fig. S1 depicts the process routing of the C8/PH2-BCZT films. Cellulose was dissolved in DMAc/LiCl and blended with a P(VDF–HFP) solution (80/20 by mass). BCZT nanoparticles (0–10 vol%) were incorporated in the as-prepared blend solution, and the mixtures were homogenized by stirring for a long period, followed by ultrasonic-assisted agitation to obtain BCZT@cellulose/P(VDF–HFP) precursors. Multilayer films (∼15 μm) were prepared via tape casting, with each new layer deposited before the previous one dried, ensuring hydrogen-bond-assisted interfacial integration. Two structures were fabricated: uniform multilayers (an equal BCZT content in all sublayers) and sandwich multilayers (outer cellulose/P(VDF–HFP), inner BCZT-containing layer). The characterizations included XRD (crystalline structure), SEM/TEM (morphology), FTIR (functional groups), impedance analysis (dielectric properties), ferroelectric testing (J–E, P–E loops), charging–discharging tests (energy storage), and universal testing machine measurements (mechanical performance). The detailed description can be found in SI.

3. Results and discussions

3.1 Results and discussions of the phase and microstructure

Fig. 1(a1) shows a photograph of a black swan residing on the lake at SUST. These swans can be observed gliding freely across the water throughout all seasons, with their feathers exhibiting excellent water-repellent properties. A closer look at the feather microstructure, as illustrated in Fig. 1(a2), reveals a natural gradient architecture composed of three distinct layers from the exterior to the interior: flight feathers, contour feathers, and down feathers. Each layer possesses unique characteristics and functions. The down feathers are soft and lightweight, contributing to thermal insulation. The contour feathers are denser and play a crucial role in waterproofing and wind resistance. The outermost flight feathers are the most rigid and densely packed, providing aerodynamic lift and maneuverability during flight.^36,37 The hierarchical design of the swan feathers, refined through long-term evolution, offers valuable inspiration for the rational design of gradient-structured materials in advanced energy-storage applications. The long-term evolutionary adaptation of swan feathers highlights the advantages of gradient structural design, offering valuable inspiration for the development of advanced functional materials. Motivated by this biological model, as illustrated in Fig. 1(b), the authors designed and fabricated both single-layer and three-layer gradient-structured composite films using BCZT piezoelectric powders with varying compositions as fillers and a C8/PH2 blend film as the polymer matrix. Fig. 1(c1) presents the X-ray diffraction (XRD) patterns of all films in the 2θ range of 17.5° to 77.5°, while an enlarged view around the (110) plane is shown in Fig. 1(c2). Consistent with previous reports, the broad peak around 21° corresponds to the crystalline regions of cellulose and P(VDF–HFP), while the sharp, indexed peaks originate from the BCZT ceramic phase. Two notable phenomena are observed after transitioning from C8/PH2-BCZT-73 to C8/PH2-BCZT-37: (1) a shift of the (110) peak to lower diffraction angles and (2) an increase in the peak intensity and sharpness. The first observation can be attributed to the increase in the BCT content, which leads to an expansion of the lattice parameters. This expansion results from the substitution of Ba with Ca at the A-site, inducing greater lattice distortion and unit-cell enlargement.³⁸ The second observation is associated with the incorporation of Zr⁴⁺ ions, which help stabilize the perovskite structure by mitigating internal stresses and lattice mismatches. This stabilization enhances the tolerance factor and promotes improved crystallinity in the ceramic powders.³⁹Fig. 1(d) presents the Fourier transform infrared (FT-IR) spectra of all the composite films. The characteristic absorption peak at 840 cm⁻¹ is attributed to the electroactive β-phase of P(VDF–HFP), with no detectable signals corresponding to the α-phase. This indicates a complete α-to-β phase transition, which is induced by the incorporation of cellulose into the polymer matrix. The absorption bands in the range of 1000–1160 cm⁻¹ are assigned to the C–O–C and C–OH stretching vibrations in cellulose, confirming the successful blending and interaction between P(VDF–HFP) and cellulose. Additionally, the peaks observed around 1400 cm⁻¹ and 2890 cm⁻¹ correspond to CF₂/CF₃ clusters and C–H stretching vibrations, respectively. The broad absorption band spanning 3130 cm⁻¹ to 3560 cm⁻¹ is attributed to the O–H stretching vibration of hydroxyl groups within the cellulose matrix. This broad peak also signifies the formation of an extensive hydrogen-bonding network among cellulose, P(VDF–HFP), and the BCZT particles during the dissolution-regeneration process,²⁹ which plays a critical role in enhancing the interfacial compatibility and structural integrity of the composite films. Fig. 1(e1)–(e3) present the Rietveld refinement results of the XRD data for three different BCZT ceramic powders, with corresponding ball-and-stick crystal structure models provided as insets of each figure. The refinement analysis reveals that BCZT-37 and BCZT-73 exhibit single-phase structures, corresponding to the tetragonal (T) and cubic (C) phases, respectively, while BCZT-55 displays three coexisting phases: tetragonal (T), rhombohedral (R), and cubic (C). These findings are consistent with previous reports³⁹ and confirm the successful modulation of the crystal phases via compositional adjustment. Flexible capacitors are always applied under complex working conditions, including in damp environments. To confirm their hydrophobicity, a water-contact-angle test was thus implemented. Fig. 1(f1) and (f2) show photographs of the water-contact-angle measurements on the surface of the C8/PH2-BCZT-dg and C8/PH2-BCZT-ug films, respectively. Both samples exhibit good hydrophobicity, indicating their suitability for applications in humid environments. The water-contact-angle measurements of all the films are displayed in Fig. S2. For optical characterization, Fig. 1(g) summarizes the transmittance data of the films. All samples exhibit comparable transparency, with an average transmittance of approximately 40% at a wavelength of 800 nm, demonstrating the potential of these composite films for applications requiring moderate optical transparency.

To investigate the microstructure of the composite film, a cross-sectional scanning electron microscopy (SEM) image of the C8/PH2-BCZT-dg sample was obtained and is shown in Fig. 2(a1). Consistent with our previous results, the film exhibits a dense structure with minimal porosity. The corresponding energy-dispersive X-ray spectroscopy (EDS) elemental mappings for carbon (C), oxygen (O), barium (Ba), titanium (Ti), and calcium (Ca) within the dashed region of Fig. 2(a1) are displayed in Fig. 2(a2)–(a6). The dark regions circled by white dash-dotted lines in Fig. 2(a2) coincide with the bright regions in the O, Ba, and Ti maps, confirming the presence of BCZT ceramic fillers. It is particularly noteworthy that in the EDS maps of Ba and Ti (Fig. 2(a4) and (a5)), the brightness appears relatively uniform across the filler regions. In contrast, the Ca map in Fig. 2(a6) shows noticeably higher intensity in the lower part of the film. This observation indicates a compositional gradient: the Ca content is low in the upper layers and high in the lower layers of the film. These findings confirm the successful fabrication of a gradient three-layer structure in the C8/PH2-BCZT-dg film, in which the distribution of Ca is intentionally varied to create the desired gradient architecture.

Fig. 2 (a1) A cross-sectional SEM image of the C8/PH2-BCZT-dg film; (a2)–(a6) the corresponding EDS mappings of C, O, Ba, Ti, and Ca elements of (a1); (b1) a TEM image of the sliced C8/PH2-BCZT-dg film; (b2) a magnified view focusing on an individual ceramic particle in (b1); (b3) a magnified view focusing on the particle in (b2) and the corresponding EDP is inserted; (c1) a general TEM view of the C8/PH2-BCZT-dg film, and the corresponding EDS mappings of C, F, O, Ba, Ca, Ti and Zr are displayed from (c2)–(c8).

Fig. 2(b1) shows a transmission electron microscopy (TEM) image of the sliced C8/PH2-BCZT-dg film, where the C8/PH2 blend matrix appears as gray regions, and the embedded BCZT ceramic fillers are visible as dark dots. A magnified view focusing on an individual ceramic particle is presented in Fig. 2(b2), in which ferroelectric domains can be clearly identified within the particle. Further magnification of the particle, as shown in Fig. 2(b3), reveals well-defined lattice fringes, indicating a high degree of crystallinity. The corresponding electron diffraction pattern (EDP), inserted in the lower right corner of Fig. 2(b3), displays discrete bright spots, confirming the single-crystalline nature of the BCZT particle. Fig. 2(c1) presents a general TEM view of the composite film, with corresponding energy-dispersive X-ray spectroscopy (EDS) elemental mappings for C, F, O, Ba, Ca, Ti, and Zr shown in Fig. 2(c2)–(c8). Since both cellulose and P(VDF–HFP) contain carbon, the C element is uniformly distributed throughout the region (Fig. 2(c2)). The distribution of fluorine (Fig. 2(c3)) suggests that P(VDF–HFP) chains are relatively shorter and more localized compared to the more widespread cellulose. In the elemental mappings for the BCZT components (Fig. 2(c4)–(c8)), the bright, localized regions correspond to the ceramic particles. As expected, the distributions of O, Ba, and Ti are relatively uniform, while noticeable variations in brightness are observed for Ca and Zr (Fig. 2(c6) and (c8)), reflecting compositional differences among the BCZT particles. Moreover, a comparison between the distributions of F and Ba suggests that the BCZT particles tend to aggregate around P(VDF–HFP) domains. This behavior is attributed to strong hydrogen-bonding interactions between the fluorine atoms (F⁻) in P(VDF–HFP) and the hydroxyl groups (OH⁻) on the surfaces of the ceramic particles.^40,41

3.2 Results of the energy-storage performance and mechanical property

Before evaluating the energy-storage performance of the films, their E_b values were first determined. For each sample, measurements were taken from 10 electrodes, and the results were analyzed using the Weibull distribution method, as shown in Fig. 3(a). Interestingly, not all multilayer-structured films exhibited higher E_b values than their single-layer counterparts, in contrast with previous reports.^29,30 This suggests that structural design plays a critical role in regulating the voltage endurance of the C8/PH2-BCZT films. The unipolar P–E loops measured near the breakdown field for all films are presented in Fig. 3(b), along with the corresponding W_rec and energy-storage efficiency (η), which are annotated above the loops. Notably, the C8/PH2-BCZT-dg film achieved a W_rec of 38.73 J cm⁻³, with an efficiency of 79.39%. The P–E loops at electric fields ranging from 1.0 MV cm⁻¹ up to the breakdown field are displayed in Fig. S3, and the associated energy-storage performance for each loop is summarized in Fig. 3(c). Fig. 3(d) compares the W_rec and η values of the C8/PH2-BCZT-dg film with those of representative cellulose-based dielectric capacitors reported in the literature.^{29,30,42–55} Clearly, the film developed in this study demonstrates superior overall energy-storage performance. Given their flexibility, mechanical properties are also crucial. The stress–strain curves for all films are shown in Fig. 3(e1), with the elongation at break and tensile strength extracted from these curves summarized in Fig. S4. Compared to the single-layered films, the multilayer-structured ones exhibit significantly enhanced mechanical properties, which can be attributed to the formation of 3D hydrogen bonds between the sublayers, as previously reported.²⁹ The corresponding Young's moduli, derived from the stress–strain curves, are plotted in Fig. 3(e2). No significant variation was observed, indicating that the stiffness of the films remains largely unaffected by the structural design.

Fig. 3 (a) The Weibull distribution of all the films; (b) the unipolar P–E loops measured near the E_b for all the films as well as the corresponding W_rec and η; (c) the W_rec and η of all the films at different fields; (d) the comparison of the W_rec and η values of the C8/PH2-BCZT-dg film with those of representative cellulose-based dielectric capacitors reported in recent literature; (e1) the stress–strain curves for all the films; (e2) the Young's modulus calculated from (e1).

3.3 Discussion of the dielectric breakdown strength

As discussed earlier, an unexpected phenomenon was identified: multilayer-structured films display lower E_b values than their single-layer counterparts. To investigate this observation in more detail, a comparative analysis of the E_b values across all film types was carried out, as illustrated in Fig. 4(a). The dashed line labeled “Average” denotes the mean E_b of the three single-layer films—BCZT-73, BCZT-55, and BCZT-37. Interestingly, the BCZT-m and C8/PH2-BCZT-dg films exhibit E_b values above this average, whereas the value for the C8/PH2-BCZT-ug film falls below it. This contrast suggests that the introduction of multilayer architectures can influence the dielectric strength in both positive and negative ways, depending on the structural design. To further elucidate the role of film structure in electrical behavior, bipolar P–E hysteresis loops were recorded for the three multilayer films—C8/PH2-BCZT-m, C8/PH2-BCZT-dg, and C8/PH2-BCZT-ug—under a fixed electric field of 3.0 MV cm⁻¹, as shown in Fig. 4(b1)–(b3). Interestingly, the C8/PH2-BCZT-m film exhibits a nearly linear and symmetric P–E loop, whereas the other two multilayered films display significant asymmetry. In particular, C8/PH2-BCZT-dg shows an open-loop feature on the negative side, while C8/PH2-BCZT-ug exhibits a similar opening on the positive side. These asymmetric polarization behaviors prompted further investigation through current–voltage (I–V) measurements, as shown in Fig. 4(c1) and (c2) for C8/PH2-BCZT-dg and C8/PH2-BCZT-ug, respectively. To gain a broader understanding of the conduction behavior across different architectures, the I–V characteristics of all samples—including single-layer and gradient trilayer structures—are provided in Fig. S5. As expected, both C8/PH2-BCZT-dg and C8/PH2-BCZT-ug demonstrate distinctly asymmetric I–V profiles. Specifically, the C8/PH2-BCZT-dg film exhibits a high current density under negative bias and a low current density under positive bias, while the opposite trend is observed for C8/PH2-BCZT-ug. To elucidate the underlying conduction mechanisms responsible for these behaviors, two widely accepted models were considered: Schottky emission and Ohmic conduction (or space-charge-limited current (SCLC)). These mechanisms are typically expressed by the following respective equations:and

I = σE (2)

Fig. 4 (a1) A comparative analysis of E_b values across all films; (b1)–(b3) bipolar P–E loops of C8/PH2-BCZT-m, C8/PH2-BCZT-dg, and C8/PH2-BCZT-ug under 3.0 MV cm⁻¹; (c1), (c2) the I–V characters of C8/PH2-BCZT-dg and C8/PH2-BCZT-ug; (c3), (c4) the fitting result of C8/PH2-BCZT-dg and C8/PH2-BCZT-ug according to the Schottky emission mechanism; (c5), (c6) the fitting result of C8/PH2-BCZT-dg and C8/PH2-BCZT-ug according to the SCLC mechanism; (d1), (d2) the schematic of capacitor configurations for the C8/PH2-BCZT-m and C8/PH2-BCZT-dg/ug films; (d3)–(d5) the band diagram of C8/PH2-BCZT-m, C8/PH2-BCZT-dg, and C8/PH2-BCZT-ug with V_a = 0; (d6)–(d8) the band diagram of C8/PH2-BCZT-m, C8/PH2-BCZT-dg, and C8/PH2-BCZT-ug with V_a > 0.

In these formulas, J represents the current density, Φ_S the Schottky barrier potential, E the electric field across the material, and T the absolute temperature. Constants A*, k, q, ε₀, ε_r and σ denote the effective Richardson constant, Boltzmann constant, electronic charge, vacuum dielectric constant, relative dielectric constant, and electrical conductivity, respectively.⁵⁶ These conduction models were applied to both the C8/PH2-BCZT-dg and C8/PH2-BCZT-ug films, with the corresponding fitting results presented in Fig. 4(c3)–(c6). On the low-current sides, i.e., under positive bias for C8/PH2-BCZT-dg and negative bias for C8/PH2-BCZT-ug, as shown in Fig. 4(c3) and (c4), the I–V curves exhibit clear linearity, indicating that Schottky emission is the dominant conduction mechanism in these regions. In contrast, the high-current sides—negative bias for C8/PH2-BCZT-dg and positive bias for C8/PH2-BCZT-ug—display more complex characteristics: while Schottky emission dominates at low electric fields, space-charge-limited current (SCLC) conduction becomes increasingly significant at higher fields. These I–V results are also in strong agreement with the observed dielectric breakdown behavior. Notably, the C8/PH2-BCZT-ug film shows higher leakage current under positive bias compared to C8/PH2-BCZT-dg, which aligns with its lower breakdown strength, as increased leakage is typically associated with reduced voltage endurance. To further explore the underlying cause of the voltage endurance differences between C8/PH2-BCZT-dg and C8/PH2-BCZT-ug, energy-band diagrams were constructed for analysis. Fig. 4(d1) and (d2) present schematic illustrations of the capacitor configurations for the C8/PH2-BCZT-m and C8/PH2-BCZT-dg/ug films, respectively, each with gold (Au) electrodes sputtered on both sides. As previously reported, composite films comprising cellulose, P(VDF–HFP), and BCZT-based ceramic fillers exhibit characteristics of n-type semiconductors. When interfaced with Au, a Schottky barrier forms at the junction between the film and the negatively biased electrode. The definitions of the physical quantities and symbols used in the band diagrams are provided in SI.^31,57

For the C8/PH2-BCZT-m film, due to its symmetric structure, it is sufficient to analyze only one side of the capacitor to understand its electrical behavior. At zero applied voltage (V_a = 0), band bending occurs at the Au/film interface, driven by the difference in work function between the Au electrode and the film. When a non-zero V_a is applied, the Fermi level of the film (E_F-m) near the negatively biased Au electrode shifts downward, resulting in increased band bending, as illustrated in Fig. 4(d3) and (d6). As the applied voltage continues to rise, electrons accumulate at the Schottky interface, and the width of the depletion region (R_d-m) increases to a new value, , until dielectric breakdown occurs. Due to the low hysteresis characteristics of both the BCZT filler and the C8P2 blend matrix, along with the film's symmetric architecture, the P–E response of the C8/PH2-BCZT-m film remains linear and symmetric, as shown in Fig. 4(b1).

For the C8/PH2-BCZT-dg and C8/PH2-BCZT-ug films, when connected to Au electrodes in the absence of an external electric field, band bending also occurs at the interfaces, as illustrated in Fig. 4(d4) and (d5), due to the same work function mismatch described earlier. Notably, the top sublayer C8/PH2-BCZT-73 possesses a higher work function (Φ₇₃) than the bottom sublayer C8/PH2-BCZT-37 (Φ₃₇), i.e., Φ₇₃ > Φ₃₇. As a result, the built-in potential within the C8/PH2-BCZT-ug structure is larger than that in C8/PH2-BCZT-dg. Consequently, more electrons tend to accumulate at the Au/C8/PH2-BCZT-dg interface than at the Au/C8/PH2-BCZT-ug interface. Upon applying a non-zero V_a, the two capacitors—C8/PH2-BCZT-dg and C8/PH2-BCZT-ug—can be viewed as having structurally identical multilayer stacks but connected in opposite directions. Their distinct energy-storage behaviors can thus be rationalized using their corresponding band diagrams. Under increasing V_a, both Φ₇₃ and Φ₃₇ shift downward, leading to increased electron accumulation at their respective Schottky interfaces and the expansion of the depletion regions from R_d-dg and R_d-ug to and , as shown in Fig. 4(d7) and 4(d8). For C8/PH2-BCZT-dg, the Schottky interfaces between adjacent sublayers behave uniformly. As the V_a increases, the valence bands of all sublayers shift downward in parallel, and the dominant conduction mechanism at each interface remains unchanged. In contrast, for C8/PH2-BCZT-ug, the built-in potential between adjacent sublayers—for example, between C8/PH2-BCZT-73 and C8/PH2-BCZT-55—is lower than that in the C8/PH2-BCZT-dg configuration. As a result, when V_a increases, the valence band of the middle layer (e.g., C8/PH2-BCZT-55) may shift lower than that of the top layer (C8/PH2-BCZT-73), potentially altering the interfacial conduction mechanism. This change, supported by the I–V characterization results, is likely indicative of a transition from Schottky emission to Ohmic conduction.

Based on the above analysis, the conduction mechanisms at the interfaces between all sublayers in C8/PH2-BCZT-dg remain unchanged before and after applying the V_a. In contrast, the C8/PH2-BCZT-ug film undergoes a transition in conduction mechanism from Schottky emission to Ohmic contact. Compared to Schottky emission, Ohmic contact facilitates electron transport, thereby increasing leakage currents. Additionally, the electron accumulation at the Au//C8/PH2-BCZT-73 interface in C8/PH2-BCZT-ug is greater than that at the Au//C8/PH2-BCZT-37 interface in C8/PH2-BCZT-dg, further contributing to the reduced voltage endurance observed in the former. These factors collectively explain why C8/PH2-BCZT-ug exhibits a lower E_b compared to C8/PH2-BCZT-dg. Like the black swans inhabiting the lake of SUST, where different types of feathers fulfill distinct functions and their optimal arrangement allows the swans to thrive across the seasons, the structural organization in energy-storage films plays a similarly decisive role. If the feather components were misplaced—for instance, with flight feathers positioned close to the skin and down feathers exposed on the outside—the plumage would lose both its waterproofing and thermal-insulation functions. In an analogous manner, the strategic arrangement of sublayers is essential in the design of energy-storage films. Even when the constituent materials are identical, such as C8/PH2-BCZT-dg and C8/PH2-BCZT-ug, films with carefully engineered structural configurations can achieve significantly enhanced energy-storage performance compared to those with suboptimal layer arrangements.

To further validate the voltage endurance results and visualize the dielectric breakdown process, the finite element method (FEM) was employed to simulate the breakdown behavior of the C8/PH2-BCZT-m, C8/PH2-BCZT-dg, and C8/PH2-BCZT-ug films. The detailed physical model and simulation parameters are provided in SI Prior to the simulation, the ε_r and electrical conductivity σ of each sublayer were assigned based on experimental measurements. Fig. 5(a1)–(a3) present the simulated current distribution at an applied electric field of 5.75 MV cm⁻¹ for the three films. As expected, the C8/PH2-BCZT-dg film withstood the field without breakdown, while both C8/PH2-BCZT-m and C8/PH2-BCZT-ug broke down, consistent with the experimental observations. The corresponding simulated electric potential distributions are shown in Fig. S6. It is evident that, in the multilayered C8/PH2-BCZT-dg and C8/PH2-BCZT-ug films, the interfaces between adjacent sublayers cause fluctuations in the potential distribution, which in turn influence the overall dielectric breakdown behavior of the films.

Fig. 5 (a1)–(a3) The simulated current distribution at an applied electric field of 5.75 MV cm⁻¹ for the C8/PH2-BCZT-m, C8/PH2-BCZT-dg, and C8/PH2-BCZT-ug films; (b1), (b2) the contour maps showing the variation in P_max with frequency and electric field and another showing that with the cycle number and electric field of the C8/PH2-BCZT-dg film under 5.0 MV cm⁻¹; (b3), (b4) the calculated W_rec and η values of the C8/PH2-BCZT-dg film from Fig. S7(a) and (b); (c1), (c2); the image of the C8/PH2-BCZT-dg film (d1) before; (d2) after the combustion experiment.

3.4 Results and discussions of the reliability

In addition to excellent short-term or single-frequency energy-storage performance, long-term operational and frequency stability is, in many cases, of great practical importance. To evaluate the reliability of the C8/PH2-BCZT-dg film, its P–E loops were measured at 5.0 MV cm⁻¹ across frequencies of 10 Hz to 10 kHz and over 10⁶ polarization cycles, as shown in Fig. S7(a) and (b), respectively. To better understand how loop evolution, which strongly influences W_rec, varies with frequency and cycling, two contour maps were constructed: one showing the variation in the maximum polarization (P_max) with frequency and electric field (Fig. 5(b1)) and the other with the cycle number and electric field (Fig. 5(b2)). In both maps, P_max remains largely stable at low fields; however, under higher fields, it increases from the top-left to the bottom-right corner in the frequency map and from the bottom-left to the top-right corner in the cycling map. To assess the impact of these changes on the energy-storage performance, W_rec and η were calculated for both processes, with results presented in Fig. 5(b3) and (b4). From 10 Hz to 10 kHz, W_rec and η vary by only 14.87% and 12.09%, respectively; from the 1st to the 106th cycle, the variations are 8.14% and 19.88%, respectively. These results demonstrate that the C8/PH2-BCZT-dg film maintains a stable, high energy-storage performance under a broad range of working conditions and after prolonged use.

To evaluate the figure-of-merit of the C8/PH2-BCZT-dg film for high-power applications, both undamped and overdamped discharge currents were measured using an R–L–C circuit, as illustrated in Fig. S8(a), under various applied electric fields. The corresponding undamped and overdamped waveforms are shown in Fig. S8(b) and Fig. 5(c1), respectively. In the undamped mode, similar discharge behaviors were observed across all fields, with the first current peak increasing in magnitude with increasing field strength. As shown in Fig. S8(c), the maximum current (I_max), current density (C_D), and power density (P_D) all increase monotonically with the electric field, consistent with eqn (S16) and (S17). In the overdamped mode, the current peak occurs at approximately 42 ns and remains essentially unchanged with increasing field strength. From these data, the discharge energy density (W_d)–time profiles were calculated using eqn (S18) (detailed in SI) and are presented in Fig. 5(c2). The discharge time (t_0.9), defined as the time required to release 90% of the stored energy, was determined to be ∼41.97 ns, and it shows negligible dependence on the applied field, underscoring the potential of this material to replace fully petroleum-based flexible dielectric capacitors. At an applied field of 5 MV cm⁻¹, W_d reaches ∼21.07 J cm⁻³, which is lower than the value obtained from static P–E loop measurements, a difference attributed to the clamping effect on polar regions under high-field, high-frequency conditions.⁵⁸Fig. 5(d1) and (d2) show photographs of the C8/PH2-BCZT-dg film before and after ignition in the combustion test. Notably, the film exhibits almost no visible flame upon ignition, indicating excellent fire resistance and enhanced safety for long-term operation. The excellent fire resistance of the cellulose/P(VDF–HFP)-based composite can be attributed to several synergistic mechanisms. Cellulose contributes to condensed-phase char formation, generating a dense carbonaceous layer that acts as a thermal and oxygen barrier during combustion. The fluoropolymer P(VDF–HFP), with its strong C–F bonds, enhances thermal stability and suppresses the release of combustible volatiles, while inorganic ceramic fillers (e.g., BCZT or BT) further absorb heat and form stable inorganic residues that physically shield the underlying matrix. The strong interfacial interactions between cellulose and P(VDF–HFP) improve the film's structural integrity and delay its decomposition, thereby reducing the emission of flammable gases. In some cases, additional gas-phase and condensed-phase synergies may occur, such as the dilution of combustible species and simultaneous formation of protective residues. Collectively, these effects endow the composite with superior flame-retardant performance compared to those of conventional polymer systems.^59,60

4. Conclusions

In this work, inspired by the hierarchical architecture of swan feathers, we designed and fabricated cellulose-based composite films with compositionally graded structures using BCZT ceramic fillers of varying compositions embedded in a C8/PH2 matrix. Through systematic structural, electrical, and mechanical characterizations, we demonstrated that the down-graded trilayer configuration (C8/PH2-BCZT-dg) delivers the best overall performance, achieving a recoverable energy storage density of 38.73 J cm⁻³ with an efficiency of 79.39%. Band-diagram analysis and conduction-mechanism fitting revealed that the superior voltage endurance of the C8/PH2-BCZT-dg film originates from its stable Schottky emission behavior across interfaces, in contrast to the conduction-mechanism transition observed in the up-graded structure. Finite element simulations confirmed the experimentally observed breakdown resistance, highlighting the role of internal potential distribution in determining dielectric reliability. Beyond its high W_rec, the C8/PH2-BCZT-dg film exhibited outstanding frequency stability and cycling durability, with minimal variation in its energy-storage performance over 10 Hz–10 kHz and after 10⁶ polarization cycles. Due to its high-power capability, it shows rapid energy release (t_0.9 = 41.97 ns) and a discharge energy density of 21.07 J cm⁻³ at 5 MV cm⁻¹. These results demonstrate the potential of compositionally graded, hydrogen-bond-engineered cellulose composites to replace petroleum-based polymer dielectrics in advanced, high-performance, and sustainable energy-storage capacitors.

Conflicts of interest

The authors have no conflicts to disclose.

Data availability

The data that support the findings of this study can be made available by the corresponding author, Zixiong Sun, upon reasonable request.

Detailed description on the experimental section and the finite element simulation, as well as some supporting figures mentioned in the main manuscript are provided in the supplementary information (SI). See DOI: https://doi.org/10.1039/d5mh01558h.

Acknowledgements

All feathers used in this study were naturally shed and collected from swans’ habitat without harm to the animals. The feathers we use in this work were obtained from the habitat of the swans, rather than unplugging directly from the body of the swans. There is no harm to the swans in the whole experimental process. This work was financed by the National Natural Science Foundation of China (52472132) and the Opening Project of State Key Laboratory of Polymer Materials Engineering (Sichuan University) (grant no. sklpme2024-1-24).

References

1. L. Cao, R. Xi, C. Zhou, G. He, F. Yang, L. Xu and H. Li, Coatings, 2024, 14, 1193.
2. M. Singh, I. E. Apata, S. Samant, W. Wu, B. V. Tawade, N. Pradhan, D. Raghavan and A. Karim, Polym. Rev., 2022, 62, 211–260.
3. Z. Sun, Z. Wang, Y. Tian, G. Wang, W. Wang, M. Yang, X. Wang, F. Zhang and Y. Pu, Adv. Electron. Mater., 2019, 6, 1900698.
4. P. Yin, L. Lei, Q. Tang, D. Dastan, Y. Liu, H. Wang and Z. Shi, Energy Storage Mater., 2025, 77, 104213.
5. J.-Y. Pei, L.-J. Yin, S.-L. Zhong and Z.-M. Dang, Adv. Mater., 2023, 35, 2203623.
6. G. Liu, L. Zhu, Z. Chang, D. Dastan, Y. Liu, R. Fan and Z. Shi, Adv. Funct. Mater., 2025, 35, 2425001.
7. Z. Sun, Y. Bai, J. Liu, G. Jian, C. Guo, L. Zhang and Y. Pu, J. Alloys Compd., 2022, 909, 164735.
8. Z. Sun, L. Diwu, R. Gao, P. Sun, H. Jing, S. Huang, Y. Tian, Z. Wang, L. Jin and D. Q. Tan, Adv. Funct. Mater., 2025, e16412.
9. X. Zhang, Y. Shen, Q. Zhang, L. Gu, Y. Hu, J. Du, Y. Lin and C.-W. Nan, Adv. Mater., 2015, 27, 819–824.
10. K. Bi, M. Bi, Y. Hao, W. Luo, Z. Cai, X. Wang and Y. Huang, Nano Energy, 2018, 51, 513–523.
11. J. Zhao, G. Ji, X. Zhang, R. Hu and J. Zheng, Polymer, 2021, 214, 123372.
12. Z. Sun, C. Ma, M. Liu, J. Cui, L. Lu, J. Lu, X. Lou, L. Jin, H. Wang and C.-L. Jia, Adv. Mater., 2016, 29, 1604427.
13. S. Gong, Y. Shi, Y. Su, Y. Li, L. Ding, J. Lin, G. Yang, B. Li, X. Wu, J. Zhang, H. Xie and H. Sun, ACS Sustainable Chem. Eng., 2021, 9, 13385–13394.
14. Z. Sun, Y. Bai, H. Jing, T. Hu, K. Du, Q. Guo, P. Gao, Y. Tian, C. Ma, M. Liu and Y. Pu, Mater. Horiz., 2024, 14, 3330–3344.
15. Z. Sun, H. Xin, L. Diwu, Z. Wang, Y. Tian, H. Jing, X. Wang, W. Hu, Y. Hu and Z. Wang, Mater. Horiz., 2025, 7, 2328–2340.
16. Z. Ren, L. Lei, L. Zhu, S. Xia, R. Fan, D. Dastan, H. Cui and Z. Shi, Adv. Funct. Mater., 2025, 35, 2417156.
17. S. Luo, J. Yu, S. Yu, R. Sun, L. Cao, W.-H. Liao and C.-P. Wong, Nano Energy, 2020, 78, 105247.
18. L. Sun, Z. Shi, B. He, H. Wang, S. Liu, M. Huang, J. Shi, D. Dastan and H. Wang, Adv. Funct. Mater., 2021, 31, 2100280.
19. D. Ai, H. Li, Y. Zhou, L. Ren, Z. Han, B. Yao, W. Zhou, L. Zhao, J. Xu and Q. Wang, Adv. Energy Mater., 2020, 10, 1903881.
20. J. Dong, R. Hu, X. Xu, J. Chen, Y. Niu, F. Wang, J. Hao, K. Wu, Q. Wang and H. Wang, Adv. Energy Mater., 2021, 31, 2102644.
21. E. Fukada, J. Phys. Soc. Jpn., 1955, 10, 149–154.
22. E. Fukada, Wood Sci. Technol., 1968, 2, 299–307.
23. D. Zhao, Y. Zhu, W. Cheng, W. Chen, Y. Wu and H. Yu, Adv. Mater., 2021, 33, 2000619.
24. G. Du, J. Wang, Y. Liu, J. Yuan, T. Liu, C. Cai, B. Luo, S. Zhu, Z. Wei, S. Wang and S. Nie, Adv. Sci., 2023, 10, 2206243.
25. Y. Song, Z. Shi, G.-H. Hu, C. Xiong, A. Isogai and Q. Yang, J. Mater. Chem. A, 2021, 9, 1910–1937.
26. Q. Lv, X. Ma, C. Zhang, J. Han, S. He, K. Liu and S. Jiang, Int. J. Biol. Macromol., 2024, 259, 129268.
27. J. Sun, Y. Dong, X. Wang, J. Cao, M. Gong and C. Li, Bull. Chem. Soc. Ethiop., 2021, 35, 669–675.
28. M. Ichwan, J. Appl. Polym. Sci., 2012, 124, 35104.
29. Z. Sun, H. Wei, S. Zhao, Q. Guo, Y. Bai, S. Wang, P. Sun, K. Du, Y. Ning, Y. Tian, X. Zhang, H. Jing, Y. Pu and S. Zhang, J. Mater. Chem. A, 2024, 12, 128–143.
30. Z. Sun, J. Liu, H. Wei, Q. Guo, Y. Bai, S. Zhao, S. Wang, L. Li, Y. Zhang, Y. Tian, X. Zhang, H. Jing, Y. Pu and S. Zhang, J. Mater. Chem. A, 2023, 11, 20089–20101.
31. Z. Sun, S. Wang, S. Zhao, H. Wei, G. Shen, Y. Pu and S. Zhang, J. Mater. Chem. C, 2024, 12, 859–867.
32. H. Pan, F. Li, Y. Liu, Q. Zhang, M. Wang, S. Lan, Y. Zheng, J. Ma, L. Gu, Y. Shen, P. Yu, S. Zhang, L.-Q. Chen, Y.-H. Lin and C.-W. Nan, Science, 2019, 365, 578–582.
33. B. Yang, Q. Zhang, H. Huang, H. Pan, W. Zhu, F. Meng, S. Lan, Y. Liu, B. Wei, Y. Liu, L. Yang, L. Gu, L.-Q. Chen, C.-W. Nan and Y.-H. Lin, Nat. Energy, 2023, 8, 956–964.
34. Z. Sun, S. Huang, W. Zhu, Y. A. Birkhölzer, X. Gao, R. A. Avila, H. Huang, X. Lou, E. P. Houwman, M. D. Nguyen, G. Koster and G. Rijnders, APL Mater., 2023, 11, 101129.
35. Z. Sun, E. P. Houwman, S. Wang, M. D. Nguyen, G. Koster and G. Rijnders, J. Alloys Compd., 2024, 981, 173758.
36. S. Hendrickx-Rodriguez and D. Lentink, J. R. Soc., Interface, 2025, 22, 20240776.
37. R. S. Terrill and A. J. Shultz, Biol. Rev. Cambridge Philos. Soc., 2023, 98, 540–566.
38. J. Du, L. Qiu, C. Yang, Y. Chen, K. Zhu and L. Wang, Crystals, 2022, 12, 896.
39. W. Liu and X. Ren, Phys. Rev. Lett., 2009, 103, 257602.
40. C. Ehrhardt, C. Fettkenhauer, J. Glenneberg, W. Münchgesang, H. S. Leipner, G. Wagner, M. Diestelhorst, C. Pientschke, H. Beigec and S. G. Ebbinghaus, RSC Adv., 2014, 4, 40321–40329.
41. M.-F. Lin, V. K. Thakur, E. J. Tan and P. S. Lee, RSC Adv., 2011, 1, 576–578.
42. X. Zeng, L. Deng, Y. Yao, R. Sun, J. Xu and C.-P. Wong, J. Mater. Chem. C, 2016, 4, 6037–6044.
43. P. Wang, Y. Yin, L. Fang, J. He, Y. Wang, H. Cai, Q. Yang, Z. Shi and C. Xiong, Composites, Part A, 2023, 164, 107325.
44. C. Zhang, Y. Yin, Q. Yang, Z. Shi, G.-H. Hu and C. Xiong, ACS Sustainable Chem. Eng., 2019, 7, 10641–10648.
45. J. He, Y. Yin, M. Xu, P. Wang, Z. Yang, Q. Yang, Z. Shi and C. Xiong, ACS Appl. Energy Mater., 2021, 4, 8150–8157.
46. X. Zheng, Y. Yin, P. Wang, C. Sun, Q. Yang, Z. Shi and C. Xiong, Int. J. Biol. Macromol., 2023, 243, 125220.
47. Y. Yin, C. Zhang, W. Yu, G. Kang, Q. Yang, Z. Shi and C. Xiong, Energy Storage Mater., 2020, 26, 105–111.
48. L. Wu, J. Zhao, Z. Li, Y. Zhai, Y. Zhang, Q. Zhen, Y. Cheng, X. Ding, P. Li, J. Liu and Z. Pan, J. Mater. Chem. C, 2022, 10, 15416–15423.
49. Y. Yin, C. Zhang, J. Chen, W. Yu, Z. Shi, C. Xiong and Q. Yang, Carbohydr. Polym., 2020, 249, 116883.
50. S. Zhang, J. Liu, Q. Guo, N. Wei, Y. Ning, Y. Bai, Y. Tian, T. Wang, Z. Sun and Y. Pu, Composites, Part A, 2022, 165, 107329.
51. S. Goodman, J. Che, W. Neri, J. Yuan and A. Dichiara, Energy Storage Mater., 2022, 48, 497–506.
52. J. Bao, J. Lao, Y. Hu, Y. Song, M. Xu, F. Niu, Q. Yang, C. Xiong and Z. Shi, Cellulose, 2023, 30, 5259–5271.
53. J. Lao, H. Xie, Z. Shi, G. Li, B. Li, G.-H. Hu, Q. Yang and C. Xiong, ACS Sustainable Chem. Eng., 2018, 6, 7151–7158.
54. H. Luo, D. Zhang, L. Wang, C. Chen, J. Zhou and K. Zhou, RSC Adv., 2015, 5, 52809–52816.
55. S. Liu, S. Xue, S. Xiu, B. Shen and J. Zhai, Sci. Rep., 2016, 6, 26198.
56. F.-C. Chiu, Adv. Mater. Sci. Eng., 2014, 1, 578168.
57. P. Wang, H. Xin, X. Wang, J. Luo, R. Xu, Z. Wang, Z. Wang, S. Zhang, Z. Sun and J. H. Jung, J. Mater. Chem. C, 2025, 13, 11938–11949.
58. L. Yang, X. Kong, F. Li, H. Hao, Z. Cheng, H. Liu, J. Li and S. Zhang, Prog. Mater. Sci., 2019, 102, 72–108.
59. X. Wei, Y. Deng, X. Hu, Z. Yang, G. Han, H. Xu and Z. Zhang, Chem. Eng. J., 2024, 15, 157725.
60. Y. Chen, L. Qiu, X. Ma, L. Dong, Z. Jin, G. Xia, P. Du and J. Xiong, Carbohydr. Polym., 2020, 15, 115907.

---