Aqueous graphene oxide (GO) dispersion, with an average lateral size of approximately 30 μm and an initial concentration of 10 mg·mL^-1, was purchased from Hangzhou Gaoxi Technology Co., Ltd. (Hangzhou, China). Polycarbosilane (PCS) was obtained from Suzhou Saifen Technology Co., Ltd. (Suzhou, China). Sodium polyacrylate (PAS), polyvinylpyrrolidone (PVP), and Triton X-100 were purchased from Macklin Biochemical Co., Ltd. (Shanghai, China). Hydroiodic acid, N,N-dimethylformamide (DMF), toluene, and absolute ethanol (analytical grade) were purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). All reagents were used as received without further purification. Deionized water was used throughout the experiments.

The commercial GO dispersion and PAS aqueous solution were concentrated to 20 mg·mL^-1 before use. The GO5 core-spinning solution was prepared by mixing the GO dispersion, PAS solution, and Triton X-100 at a volume ratio of 1:1:0.1, yielding a GO concentration of approximately 9.52 mg·mL^-1 (0.95 wt%). GO6 was prepared by changing the volume ratio of GO dispersion to PAS solution to 6:4 while maintaining the total volume of the two solutions and the Triton X-100 content; the corresponding mixing ratio was 1.2:0.8:0.1, yielding a GO concentration of approximately 11.43 mg·mL^-1 (1.14 wt%). Both dispersions were filtered before use. Unless otherwise stated, GO5 was used as the baseline core solution for fabricating G@SiC NFFs.

The shell spinning solution was prepared by dissolving PCS and PVP in a toluene/DMF mixed solvent at a volume ratio of 7:3. The mixture was stirred at 25 °C and 120 rpm for 12 h to obtain a homogeneous solution. Unless otherwise specified, the baseline shell solution contained 20 wt% PCS and 6 wt% PVP, with the weight percentages calculated relative to the total mass of the shell spinning solution.

To investigate the influence of shell composition on coaxial electrospinning behavior and core-shell fiber formation, two shell-composition series were prepared. In the PCS-content series, the PVP content was fixed at 6 wt%, whereas the PCS content was varied from 10 to 40 wt%; the corresponding samples were denoted as GPCS10, GPCS15, GPCS20, GPCS25, GPCS30, GPCS35, and GPCS40, respectively. In the PVP-content series, the PCS content was fixed at 20 wt%, whereas the PVP content was varied from 2 to 8 wt%; the corresponding samples were denoted as GPVP2, GPVP4, GPVP6, and GPVP8, respectively. Additionally, formulations containing PVP at concentrations of 0, 2, 4, 6, 8, 10, and 12 wt% were prepared for the filament-drawing tests using the same procedure, with only the PVP concentration varied.

For clarity, pristine graphene nanofiber fabrics (GNFFs) without SiC encapsulation are denoted as GNFF, while SiC nanofiber fabrics without a graphene core are denoted as SiCNFF. The core-shell graphene@SiC nanofiber fabrics are denoted as G@SiC NFFs. Unless otherwise stated, G@SiC NFF refers to the baseline sample prepared using GO5 as the core solution and the shell formulation containing 20 wt% PCS and 6 wt% PVP. This baseline sample corresponds to GPCS20 in the PCS-content series and GPVP6 in the PVP-content series.

G@SiC NFFs were fabricated by coaxial electrospinning followed by sequential chemical reduction, oxidative crosslinking, and high-temperature ceramization^[16,18]. The core and shell solutions were loaded into separate syringes and delivered through a coaxial spinneret with inner and outer needle diameters of 0.5 and 1.2 mm, respectively. During flow-rate optimization, the shell flow rate was maintained at 0.032 mL·min^-1, whereas the core flow rate was varied between 0.014 and 0.028 mL·min^-1, corresponding to shell-to-core flow-rate ratios of 1.14-2.28. For all subsequent experiments, the optimized shell and core flow rates were fixed at 0.032 and 0.018 mL·min^-1, respectively, corresponding to a shell-to-core flow-rate ratio of 1.78. Electrospinning was performed at 45-55 °C and 5%-15% relative humidity. The applied voltage and tip-to-collector distance were adjusted within 17-23 kV and 10-15 cm, respectively, to maintain a stable coaxial jet. The resulting nanofibers were collected on a high-speed rotating drum to form uniform fabrics.

The as-collected graphene oxide@polycarbosilane (GO@PCS) precursor fabrics were converted into G@SiC NFFs through a three-step post-treatment process. First, the precursor fabrics were reduced by hydroiodic acid vapor at 90 °C for 12 h. Second, the reduced fabrics were heated in air to 200 °C at 2 °C·min^-1 and held for 2 h to induce oxidative crosslinking of PCS. Third, the crosslinked fabrics were transferred to a tube furnace and heated to 1,400 °C at 2 °C·min^-1 under flowing Ar, then held for 2 h. After ceramization, the furnace was naturally cooled to room temperature under Ar.

Control samples were prepared for comparison. GNFFs were prepared by single-nozzle electrospinning of the GO/PAS solution, followed by HI vapor reduction and Ar annealing under the same conditions used for G@SiC NFFs. SiC nanofiber fabrics (SiCNFF) were prepared by single-nozzle electrospinning of the PCS/PVP shell solution, followed by air crosslinking and Ar ceramization under the same thermal conditions used for G@SiC NFFs.

The morphology and microstructure of the NFFs were examined by field-emission scanning electron microscopy (FE-SEM, S400, Hitachi, Japan) at an accelerating voltage of 5 kV. The core-shell structure was characterized by transmission electron microscopy (TEM, JEM-2100F, JEOL, Japan) operated at 200 kV. Elemental distributions were analyzed by energy-dispersive X-ray spectroscopy (EDS) coupled with SEM or TEM. Cross sections of G@SiC NFFs were prepared and examined using a focused ion beam dual-beam microscope (FIB, DB550, CIQTEK, China).

The crystal structure and phase composition were analyzed by X-ray diffraction (XRD, D8 Advance, Bruker, Germany) using Cu Kα radiation. Raman spectroscopy (inVia Reflex, Renishaw, UK) was performed with a 532 nm laser to characterize the graphitic structure. The surface chemical composition and bonding states were investigated by X-ray photoelectron spectroscopy (XPS, K-Alpha+, Thermo Scientific, USA). All binding energies were calibrated using the C 1s peak at 284.8 eV as the reference. Fourier-transform infrared spectroscopy (FTIR, Nicolet iS50, Thermo Fisher Scientific, USA) was conducted in attenuated total reflectance mode to identify functional groups. Thermogravimetric analysis (TGA, Q500, TA Instruments, USA) was carried out in air at a heating rate of 10 °C·min^-1 to evaluate thermal stability.

The electrical conductivity (σ) of the NFFs was measured at room temperature using an HPS2526 precision sheet resistance tester:

(1) $$ \sigma =\frac{L}{R\cdot w\cdot t} $$

where σ is the electrical conductivity (S·cm^-1), R is the measured resistance (Ω), and L, w, and t represent the length (cm), width (cm), and thickness (cm) of the sample, respectively.

The Joule heating characteristics were monitored using a source meter (2450, Keithley Instruments, USA) and an infrared thermal imaging camera (T630sc, FLIR, USA).

EMI shielding measurements were conducted over the X-band frequency range of 8.2-12.4 GHz using a vector network analyzer (3672C, 10 MHz - 140 GHz, China Electronics Technology Group Corporation, China). The thickness of each sample was measured using a digital micrometer at three locations, including one central position and two positions near opposite edges. The standard deviation of the thickness measurements was 0.8 μm, suggesting limited thickness variation across the measured locations. The scattering parameters (S₁₁ and S₂₁) were recorded, where S₁₁ represents the reflection coefficient and S₂₁ represents the transmission coefficient. The power coefficients of reflection (R), transmission (T), and absorption (A) were calculated as follows^[8,19]:

(2) $$ R=|S_{11}|^2 $$

(3) $$ T=|S_{21}|^2 $$

(4) $$ A=1-R-T $$

The total EMI shielding effectiveness (SE_T), reflection shielding effectiveness (SE_R), and absorption shielding effectiveness (SE_A) were calculated using the following equations^[19]:

(5) $$ SE_T=-10\mathrm{log}T=-10 \mathrm{log}|S_{21}|^2 $$

(6) $$ SE_R=-10\mathrm{log}(1-R)=-10 \mathrm{log}(1-|S_{11}|^2) $$

(7) $$ SE_A=SE_T-SE_R $$

(8) $$ SE/t=\frac{SE_T}{t} $$

Constructing a ceramic-confined graphene fibrous architecture requires coordinated control over fiber formation, radial precursor distribution, and phase conversion^[15,16,20]. As illustrated in Figure 1A and B, G@SiC NFFs were fabricated through sequential coaxial electrospinning, chemical reduction, oxidative curing, and thermal annealing^[21]. During coaxial electrospinning, the GO/PAS core solution and PCS/PVP shell solution were spatially confined within the inner and outer channels, respectively, thereby defining the radial precursor arrangement^[3,18]. Subsequent chemical and thermal conversion transformed this programmed architecture into a graphene fiber core encapsulated by a dense SiC ceramic shell while preserving the fibrous framework.

As shown in Figure 1C, the annealed fabric consists of abundant interconnected fibers, forming a continuous nonwoven network. The retention of the fibrous morphology after multistep conversion indicates that the electrospun framework remains structurally intact during chemical and thermal treatments^[15]. Cross-sectional SEM observation of an individual fiber further reveals two regions with distinct contrast, where the outer region continuously encloses the inner domain [Figure 1D]. This contrast feature suggests that the coaxially electrospun core-shell architecture remains intact after multistep conversion.

TEM and elemental mapping reveal the nanoscale compositional distribution of the G@SiC nanofibers. As shown in Figure 1E, carbon is mainly concentrated in the inner region, whereas silicon is preferentially enriched near the periphery, confirming the formation of a carbon-rich core and a silicon-containing outer shell. High-resolution TEM (HRTEM) imaging of the G@SiC interface resolves distinct lattice fringes from both domains [Figure 1F]. The shell region exhibits a lattice spacing of approximately 0.25 nm, characteristic of the SiC (111) plane, whereas the core region shows a spacing of approximately 0.35 nm, corresponding to the (002) plane of graphitic carbon^[15,22,23]. These lattice features confirm the successful encapsulation of a graphitized carbon core within a dense SiC ceramic shell.

Stable coaxial electrospinning requires synchronized deformation of the core and shell jets during fluid delivery, compound Taylor cone formation, and subsequent electrohydrodynamic stretching^[24]. As illustrated in Figure 2A, the outer shell fluid must confine the internal stream while maintaining sufficient deformability under the applied electric field. Insufficient shell viscosity weakens confinement of the core stream, causing incomplete encapsulation, non-uniform shell coverage, or partial core exposure. Conversely, excessive shell viscosity or viscoelastic mismatch suppresses jet elongation and destabilizes the compound jet, resulting in interrupted spinning, bead formation, or irregular fiber deposition. Because carbon filler spatial distribution governs conductive-network formation and electrical conductivity^[25], continuous core-shell precursor fibers require not only intrinsically spinnable solutions but also rheological compatibility between the GO/PAS core and PCS/PVP shell solutions^[26-29].

Rheological assessments and spinnability evaluations delineate the precise operational window required for stable coaxial fiber formation. The shear-rate-dependent viscosity curves reveal that all PCS/PVP shell solutions exhibit typical shear-thinning behavior [Figure 2B], which is favorable for electrospinning because it facilitates smooth fluid extrusion through the spinneret while preserving sufficient interchain cohesion before jet initiation. Furthermore, the viscosity varies markedly with the PVP concentration, confirming its critical role as a rheological modifier for the PCS precursor. A delicate rheological balance is therefore important. In this study, rheological compatibility was evaluated based on the combined consideration of the viscosity level within the processing-relevant shear-rate range, the shear-thinning response, and the frequency-dependent storage modulus (G′), loss modulus (G″), and loss tangent (tanδ). A favorable formulation was defined as one that balanced sufficient elasticity for core confinement with adequate fluidity for extrusion and electrohydrodynamic stretching. Inadequate shell viscosity failed to robustly encapsulate the core stream, whereas excessive viscosity severely impedes stretching and triggers unstable deposition. Consequently, the optimal shell formulation must precisely reconcile fluidic processability with the structural integrity necessary for continuous encapsulation.

Frequency-dependent storage and loss moduli provide deeper insights into the viscoelastic compatibility between the core and shell solutions [Figure 2C and Supplementary Figure 1]. Solution rheology is known to strongly influence electrospun jet stability and fiber formation^[29]. Among the evaluated shell formulations, PVP6 demonstrates relatively stable moduli and optimal viscoelastic matching with the GO5 core, thereby ensuring synchronized deformation during coaxial jet formation and stretching. The loss tangent (tanδ) offers complementary perspectives on the dynamic balance between viscous dissipation and elastic recovery [Figure 2D]. A moderate and frequency-stable tanδ is desirable because the shell fluid must effectively accommodate extensional stress while retaining adequate elasticity to preserve interfacial continuity. Subsequent extensional and spinnability evaluations further validate the rheological screenings [Figure 2E and Supplementary Figure 2]. Formulations lacking sufficient interchain entanglement undergo rapid capillary breakup during extension. Conversely, the optimized core-shell system sustains continuous and stable fluidic jets, manifesting superior resistance to capillary-driven droplet fragmentation.

Having established the optimal PVP content, we systematically investigated the interplay between the PCS concentration and the shell-to-core flow rate ratio. The spinnability map in Figure 2F summarizes the coefficient of variation of fiber diameter under different PCS concentrations and flow rate ratios, with the full diameter distributions shown in Supplementary Figure 3. During flow-rate optimization, the shell flow rate was maintained at 0.032 mL·min^-1, whereas the core flow rate was varied between 0.014 and 0.028 mL·min^-1 to obtain shell-to-core flow-rate ratios of 1.14-2.28. Among all tested conditions, the combination of 20 wt% PCS and a shell-to-core flow-rate ratio of 1.78 yielded the lowest coefficient of variation in fiber diameter, indicating a narrower diameter distribution. Consistently, SEM observations showed relatively uniform fibers, suggesting enhanced jet stability under these conditions. SEM characterization revealed that an insufficient PCS concentration (10 wt%) fails to supply adequate precursor for continuous shell encapsulation, while excessively high concentrations (30 and 40 wt%) led to structural artifacts, including severe fiber thickening, localized agglomeration, and bead defects [Supplementary Figure 4]. Accordingly, PVP6, a PCS concentration of 20 wt%, and a shell-to-core flow-rate ratio of 1.78 were selected as the baseline formulation and processing conditions for fabricating G@SiC precursor fibers. Samples with varied PCS and PVP contents were prepared according to the compositions specified for each comparison.

Under the optimized conditions, a large-area and freestanding fiber network with homogeneous fiber morphology was successfully realized. After sequential chemical reduction, oxidative curing, and high-temperature annealing, the color of the fabric changed from brownish yellow to brownish black, indicating the formation of a SiC shell [Figure 2G]. Cross-sectional SEM images revealed a continuous fibrous network and a relatively uniform SiC shell surrounding the individual G@SiC nanofibers [Supplementary Figure 5A and B]. The good flexibility of the G@SiC NFF is shown in Supplementary Figure 6. Tensile and cyclic deformation measurements further confirmed its mechanical flexibility [Supplementary Figure 7]. The tensile strain of the fabric could reach ~15%, while the resistance variation is highly recoverable during repeated stretching and bending cycles. Figure 2H shows the XRD pattern of the resultant fabric. A broad diffraction peak centered at ~26° is attributed to the (002) plane of graphitized carbon, whereas distinct diffraction peaks at 36.0°, 41.6°, 60.0°, and 72.0° can be assigned to crystalline SiC^[21,23]. Comparative XRD patterns of G@SiC NFF, GNFF, and SiCNFF further support these phase assignments [Supplementary Figure 8]. Collectively, these results confirm the complete ceramization of the PCS-derived shell into a robust SiC phase during thermal conversion, as well as the formation of a graphene fiber core.

The characteristic D and G bands in the Raman spectra [Figure 2I] at approximately 1,350 and 1,580 cm^-1 are assigned to defective carbon structures and the in-plane vibration of sp²-bonded graphitic carbon, respectively^[30]. Comparative Raman spectra of GNFF, GO@PCS, G@SiC NFF, SiC, and SiO₂ further clarify the carbon- and ceramic-related vibrational features [Supplementary Figure 9]. G@SiC NFF retains distinct D and G bands after the thermochemical conversion. Moreover, the I_D/I_G intensity ratio undergoes a sharp reduction from 0.761 in the pristine GNFF to 0.483 in G@SiC NFF. Since both samples were annealed at 1,400 °C, the lower I_D/I_G ratio of G@SiC NFF is unlikely to arise solely from the annealing temperature. It may instead be associated with the formation and reorganization of carbon-rich domains during PCS-to-SiC conversion and the development of the SiC/C interface^[31]. Previous studies have reported the formation of graphitic carbon layers and interface-related structures on SiC surfaces, together with changes in their morphology and Raman response under different growth conditions^[32]. These findings provide a relevant context for considering the role of SiC-C interfacial processes in the local carbon structure of G@SiC NFF. Because the Raman response is sensitive to near-surface and interfacial regions, the lower I_D/I_G ratio primarily indicates a difference in the local carbon structure probed by Raman spectroscopy.

The surface chemistry and bonding states were further investigated by XPS and FTIR analyses. The XPS survey spectra show that the pristine GNFF is mainly composed of C and O, whereas the G@SiC NFF exhibits a prominent Si signal alongside these elements [Figure 2J]. Given the inherent surface sensitivity of XPS, the pronounced Si signal in G@SiC NFF confirms the successful formation of the SiC shell. High-resolution deconvolution of the C 1s spectrum distinguishes specific bonding environments corresponding to Si–C, C–C/C=C, and C–O [Figure 2K]^[33]. The Si 2p spectrum resolves a dominant Si–C peak accompanied by a minor Si–O component [Figure 2L]. The synchronized emergence of Si–C features in both the C 1s and Si 2p spectra provides direct evidence of covalent silicon-carbon bonding, indicating that the PCS-derived shell underwent ceramic conversion during thermal treatment. The minor Si–O contribution can be attributed to superficial oxidation or trace residual oxygen-containing species.

In the FTIR spectra displayed in Figure 2M, GNFF exhibits only weak infrared absorption, whereas G@SiC NFF shows distinct Si–C stretching vibrations accompanied by a weak Si–O band. Together with the XPS results, the FTIR spectra support the formation of Si–C bonds within the outer shell^[10].

The oxidation resistance of G@SiC NFFs was first evaluated by TGA in an air atmosphere. As depicted in Figure 3A, the pristine GNFF undergoes rapid mass loss upon heating, reflecting oxidative ablation of the exposed carbon framework^[5,6]. In contrast, the G@SiC NFF exhibits a drastically retarded degradation profile, retaining ~90.6% of its initial mass even at 650 °C. The PCS-concentration-dependent TGA curves further reveal the critical role of sufficient ceramic-shell formation in suppressing oxidative mass loss [Supplementary Figure 10]. The SiCNFF exhibits relatively high thermo-oxidative stability, with a slight mass loss at intermediate temperatures followed by a gradual mass gain above approximately 1,000 °C, consistent with the oxidation behavior of SiC and the formation of a nonvolatile SiO₂ layer^[16,17]. This exceptional mass retention confirms that the SiC shell functions as an impenetrable physical barrier, effectively shielding the highly graphitized carbon core from direct oxygen attack^[34]. Quantitative analysis of the initial mass loss and high-temperature mass gain further supports the oxidation-retarding effect of the SiC shell [Supplementary Figure 11]. The slight mass gain observed above approximately 1,000 °C is attributed to the passive oxidation of SiC and the formation of nonvolatile SiO₂ according to SiC + 3/2O₂ → SiO₂ + CO↑. The resulting SiO₂ layer impedes oxygen diffusion and retards further oxidation. In air, initial mass loss is markedly suppressed in the G@SiC NFF relative to bare GNFF and generally decreases with increasing PCS-derived SiC content. This protective encapsulation is also evidenced by the derivative thermogravimetric (DTG) profiles [Figure 3B]. Specifically, the temperature corresponding to the maximum degradation rate undergoes a remarkable shift from ~425 °C for the bare GNFF to ~685 °C for the G@SiC NFF. These results demonstrate that the core-shell encapsulation paradigm endows the conductive carbon network with unprecedented oxidative resilience in extreme, high-temperature environments.

This improved oxidative resistance guarantees stable electrical properties of the G@SiC NFFs under extreme conditions^[1,35-37]. As shown in Figure 3C, during continuous exposure to an alcohol-lamp flame, pristine GNFF burns rapidly while the G@SiC NFF preserves its structural integrity with only moderate surface color change. Figure 3D and Supplementary Movie 1 demonstrate the normalized resistance evolution of both GNFF and G@SiC NFF during prolonged exposure to an alcohol-lamp flame at approximately 650 °C for 120 min. The GNFF rapidly loses electrical conductivity, which is a direct consequence of the oxidative ablation of its exposed carbon network. By contrast, the G@SiC NFF maintains stable resistance throughout the flame exposure, with the resistance fluctuation (|ΔR/R₀|) remaining below 13%. This minimal deviation indicates that the conductive pathways within the graphene fiber core remain intact, shielded from the thermo-oxidative stress^[36,38].

After annealing in ambient air at 650 °C for 1 h [Figure 3E], the unprotected GNFF was completely decomposed, while the G@SiC NFF remained an intact freestanding membrane. Under localized exposure to an alcohol-torch flame (~1,100 °C) for 180 s, G@SiC NFF retained its macroscopic integrity [Figure 3F]. SEM observations before and after the 180 s torch exposure further showed that the fibrous network morphology was largely preserved [Supplementary Figure 12]. In a separate time-dependent experiment, G@SiC NFF retained more than 87% of its initial mass after torch exposure for up to 200 s [Supplementary Figure 13].

To visually demonstrate this robust electrical continuity, the G@SiC NFF was integrated as the primary conductive interconnect in a series light-emitting diode (LED) circuit operating directly within the open flame [Figure 3G]. The sustained LED illumination confirms the uninterrupted charge-transport capability endowed by this core-shell architecture under extreme thermal conditions, as further visualized in Supplementary Movie 2. A separate 30 s flame-exposure demonstration is provided in Supplementary Movie 3.

Collectively, these results show that the dense SiC shell functions as a ceramic barrier that effectively shields the highly conductive graphene fiber core from severe oxidative degradation, preserving both the structural integrity and the interconnected charge-transport pathways under high-temperature oxidative conditions. These structural and electrical advantages provide a critical basis for durable EMI shielding performance under thermo-oxidative conditions^[33].

The EMI shielding performance of G@SiC NFF was evaluated across the X-band using scattering parameters obtained by vector network analysis. The total shielding effectiveness (SE_T), absorption contribution (SE_A), and reflection contribution (SE_R) were derived from the measured S-parameters based on the power-balance principle^[8,19]. As summarized in Figure 4A, both electrical conductivity and post-oxidation residual mass depend on PCS content. Increasing the PCS loading increases the ceramic fraction and thereby improves thermo-oxidative mass retention. However, excessive PCS loading simultaneously reduces the bulk electrical conductivity because the SiC phase partially interrupts the conductive graphene nanofiber network. This trade-off decides the optimum design of the core-shell architecture: the SiC shell must provide sufficient conformal continuity to protect the graphene nanofiber core against oxidative ablation, while the internal conductive fiber network must retain electrical percolation to sustain charge transport and electromagnetic attenuation. Therefore, the ultimate EMI shielding response is determined not only by the content of the conductive phase, but also by the balance among electrical percolation, ceramic protection, and graphene/SiC interfacial polarization^[39].

The EMI shielding performance of the selected G@SiC NFFs is shown in Figure 4B. All samples were tested at a fixed thickness of 40 μm. Under identical thickness conditions, all samples tested in Figure 4B exhibit SE_T values above 20 dB in the X-band, meeting the commonly accepted threshold for practical EMI shielding^[40]. GPCS15 shows the highest SE_T of approximately 37 dB, suggesting an optimal balance among electrical conduction, impedance matching, and ceramic protection. Further increasing the PCS content improves ceramic coverage but disrupts charge transport through the graphene nanofiber network, leading to reduced shielding effectiveness. For all samples, SE_A is substantially higher than SE_R, indicating that electromagnetic waves are mainly dissipated within the hierarchical fibrous network rather than reflected at the surface. This absorption-dominated shielding behavior is desirable for mitigating secondary electromagnetic radiation^[11,19].

The thickness-dependent EMI shielding performance is consistent with enhanced electromagnetic attenuation through extended internal propagation and dissipation pathways. As shown in Figure 4C, the frequency-averaged X-band SE_T of the G@SiC NFFs increases from approximately 12 dB at a thickness of 25 μm to over 80 dB at thicknesses of approximately 85-95 μm. At a nominal thickness of 85 μm, the G@SiC NFFs exhibit a frequency-averaged SE_T of 81.03 ± 5.14 dB (85.5 dB for a representative 80 μm-thick specimen). Meanwhile, the thickness-normalized shielding effectiveness reaches approximately 10³ dB·mm^-1, despite a thickness below 0.1 mm. This value compares favorably with those reported for ultrathin flexible EMI shielding materials^[41,42]. Notably, the increase in SE_T mainly originated from the enhanced SE_A, whereas SE_R remained nearly constant. Because reflection is primarily governed by the impedance mismatch at the air-material interface, the identical composition and comparable surface characteristics of the specimens resulted in only minor changes in SE_R with increasing thickness. In contrast, a greater thickness increased the effective propagation distance within the porous fibrous network, providing more opportunities for conductive loss, interfacial polarization, and internal scattering/multiple reflections at the fiber surfaces, pore interfaces, and G@SiC core-shell interfaces^[9]. Thus, the thickness-dependent enhancement in SE_T was predominantly absorption-driven. The EMI shielding retention characteristics of 80 μm-thick G@SiC NFFs prepared using a precursor formulation containing 20 wt% PCS were subsequently investigated after prolonged thermo-oxidative exposure. Figure 4D shows the frequency-dependent SE_T before and after isothermal oxidation in ambient air at elevated temperatures. After high-temperature treatment at 400 °C for 1 h, SE_T of the SiC-protected fabric decreases slightly and remains close to 80 dB across the measured frequency range. Even after oxidation at 650 °C for 1 h, the average SE_T remains above 76.7 dB. The temperature-dependent summary in Figure 4E further confirms that the SE_T shows strong resistance to oxidative degradation, with absorption (SE_A) governing the overall attenuation across the X-band [Figure 4F]. Frequency-resolved decomposition of SE_T after 650 °C oxidation further supports this absorption-dominated shielding mechanism across the X-band [Supplementary Figure 14]. The high-temperature EMI shielding performance of G@SiC NFF was further benchmarked against representative EMI shielding materials designed for high-temperature environments [Figure 4G]^{[25,36,37,43-53]}. Detailed comparison data are provided in Supplementary Table 1. Compared with previously reported metal-based, carbon-based, polymer-based, and ceramic-containing shields, G@SiC NFF combines high shielding effectiveness, small thickness, and oxidation tolerance, exhibiting excellent comprehensive performance.

This advantage stems from the hierarchical G@SiC core-shell architecture. As schematically illustrated in Figure 4H, a small portion of incident electromagnetic waves is initially reflected when encountering the fabric because of the impedance mismatch between air and the conductive network (I). The primary attenuation occurs after the electromagnetic waves penetrate the porous three-dimensional network. The interconnected nanofibers induce multiple internal reflections and scattering, thereby extending the propagation pathways of the incident waves (II). The SiC shell and G@SiC heterointerfaces may contribute to dielectric relaxation and interfacial polarization under the alternating electromagnetic field, as reported for related carbon/SiC and carbon-based heterostructures^{[23,33,54,55]} (III). Meanwhile, the wrinkled rGO core provides a conductive scaffold that facilitates carrier migration and ohmic dissipation^[1,2] (IV). Furthermore, the surface roughness of the ceramic shell increases local field heterogeneity and expands the effective interfacial area, promoting additional scattering and polarization relaxation. The synergy of these coupled electromagnetic processes accounts for the absorption-dominated EMI shielding behavior and its retention after thermal oxidation. The practical shielding durability of the G@SiC NFF was qualitatively demonstrated under direct exposure to an alcohol-torch flame with a nominal temperature of approximately 1,100 °C [Figure 4I]. When the G@SiC NFF was placed between the electromagnetic source and the indicator lamp, the lamp remained off during flame exposure, indicating that the film retained its electromagnetic-shielding function under flame-induced thermal shock.

Beyond serving as a robust thin film for passive electromagnetic protection, the preserved electrical networks and outstanding thermo-oxidative resilience of the G@SiC NFFs suggest their potential for active electrothermal heating in oxidative atmospheres, where carbon-based conductive networks commonly suffer from rapid oxidation and electrical failure at elevated temperatures.

The electrothermal behavior of core-shell G@SiC NFF was first investigated under different driving voltages^[56]. As shown in Figure 5A, the G@SiC NFF exhibits a rapid and reversible Joule heating response under applied voltages from 5 to 35 V. Upon electrical bias, the surface temperature increases rapidly and reaches a stable plateau within 0.14-0.35 s. Once the voltage is switched off, the fabric rapidly cools back to the ambient state. Detailed transient heating-rate profiles and enlarged cooling curves further confirm the rapid thermal activation and efficient heat dissipation of the G@SiC NFF across the examined voltage range [Supplementary Figure 15]. The steady-state temperature can be continuously regulated by the applied voltage, increasing from 52.4 °C at 5 V to 651.7 °C at 35 V. Furthermore, the approximately linear relationships between heating rate and power density and between steady-state temperature and U² confirm the Joule-heating-dominated electrothermal behavior of the G@SiC NFF [Supplementary Figure 16]. This broad voltage-programmable temperature window demonstrates that the G@SiC NFF can deliver tunable thermal outputs, spanning mild heating to high-temperature operation.

Infrared thermal images further reveal the spatial uniformity of Joule heating. As shown in Figure 5B, the fabric displays a homogeneous thermal distribution across all examined bias voltages, indicating a uniform conductive network and stable heat generation. The thermal response kinetics were further quantified in Figure 5C. Both the heating and cooling response times were in the sub-second regime, with t₉₀ values of 0.14-0.35 s and t₁₀ values of 0.19-0.45 s, respectively. Because these measurements were conducted under different applied voltages, the shorter t₉₀ at higher steady-state temperatures reflects the greater Joule-heating power, whereas the shorter t₁₀ at higher initial temperatures is attributed to enhanced heat dissipation, particularly thermal radiation. Under stepwise voltage loading, the surface temperature follows the applied voltage in a clear stepwise manner and stabilizes at each thermal plateau [Figure 5D], verifying voltage-dependent thermal regulation. During cyclic on/off tests under repeated electrical stimulation, the fabric exhibits highly reproducible heating and cooling profiles at different voltages [Figure 5E]. No obvious temperature decay, response delay, or unstable fluctuation is observed during repeated stimulation, demonstrating the good reversibility and short-cycle electrothermal stability.

The protective role of the SiC shell becomes particularly important during long-term high-temperature Joule heating in air. As shown in Figure 5F, pristine GNFF can initially be electrically heated to 350 °C for a short period; however, its surface temperature rapidly decreases after sudden electrical failure. This degradation is attributed to severe thermo-oxidative erosion of the exposed carbon network during high-temperature operation in ambient air. In contrast, the G@SiC NFF maintains stable Joule heating at approximately 600 °C over a prolonged duration, with minor temperature fluctuations. This comparison confirms that the SiC shell effectively suppresses oxidative damage to the encapsulated conductive framework, thereby preserving electrical continuity and sustained heat output under harsh oxidative conditions^[15,17,36].

Taking advantage of its favorable electrothermal properties, the G@SiC NFF was employed as a localized heating platform for the rapid conversion of an In/Ga/Zn-containing metal-salt precursor. This pulse-assisted electrothermal conversion is conceptually related to pulse Joule-heating synthesis, in which transient high temperatures enable rapid material transformation^[57]. The G@SiC NFF used for the electrothermal treatment had dimensions of approximately 2.0 cm × 1.5 cm × 1.0 mm (length × width × apparent thickness). The precursor was subjected to ten electrothermal pulses, each consisting of 30 s of heating followed by a 10 s interval. As shown in Figure 5G, the untreated precursor exhibits a broad amorphous-like diffraction background, whereas the electrothermally treated product displays several weak diffraction features, indicating structural evolution and increased ordering during electrothermal treatment. Given the compositional and structural complexity of In-Ga-Zn-O materials, these diffraction features are not assigned to specific crystalline phases. Semi-quantitative SEM-EDS analysis further detected In, Ga, Zn, and O in the treated product [Supplementary Figure 17 and Supplementary Table 2]. Collectively, these results are consistent with the rapid formation of an In-Ga-Zn-O multicomponent oxide product through pulse-assisted electrothermal conversion. The optical photographs in Figure 5H further show the visible color change of the precursor droplets after treatment, consistent with the electrothermal conversion of metal-salt precursors into oxide products. This proof-of-concept demonstration highlights the capability of G@SiC NFF to generate a localized high-temperature field under oxygen-containing conditions, offering a rapid route for oxide synthesis.