Introduction

With the rapid advancement of next-generation electronic communication systems, radar warning, satellite detection, and deep space exploration play a crucial role in the fields of aerospace, information technology, and daily life1,2,3,4,5. To guarantee the normal operation of these systems, it is both a goal and a challenge to develop electronic device casings that integrate outstanding electromagnetic interference (EMI) shielding performance with mechanical stability and lightweight characteristics6,7,8.

Nowadays, three-dimensional (3D) structured graphene has shown excellent EMI shielding properties by virtue of good conductivity and porous microstructures9,10,11. There are still significant obstacles for it to bear loads as a rigid structural material. This issue restricts the application of 3D structured graphene in electronic device casings that require both excellent EMI shielding performance and robust load-bearing capability. It is an innovative strategy to introduce a high-stiffness component (molten Mg alloy) into the pores of 3D structured graphene, thereby enabling the Mg alloy to act as a rigid load-bearing component. By leveraging the advantages of continuous connectivity inherent in the 3D network structure of graphene embedded within the Mg alloy matrix, efficient electron and load transfer can be achieved. The construction of 3D structured graphene within the Mg alloy and the design of heterointerfaces are key factors affecting the performance of the composites.

There are many processing routes, including liquid-state fabrication methods (casting12, selective laser melting13, etc.) and solid-state fabrication techniques (powder metallurgy14, friction stir processing15, etc.) for the preparation of graphene reinforced Mg matrix (G/Mg) composites to solve the problem of graphene aggregation16,17,18,19. Graphene can uniformly disperse at grain boundaries and impede grain growth of Mg, thereby enhancing the strength-ductility compatibility of G/Mg composites via the above methods. Nevertheless, the current preparation methods of G/Mg composites mentioned above are insufficient for constructing 3D network-structured graphene within the Mg matrix. An innovative strategy of 3D skeleton preconstruction-infiltration filling based on a liquid-solid infiltration extrusion method presents an intriguing solution. This approach first constructs a 3D graphene skeleton, which is subsequently infiltrated with molten Mg alloy, effectively maintaining the 3D network structure of the graphene. The liquid-solid infiltration extrusion method, including low-pressure infiltration followed by high-pressure densification, was independently developed by our research group20.

The preservation of the 3D structure of graphene, along with the design of heterogeneous interfaces between graphene and Mg alloy, is crucial for optimizing the overall properties of the composites. On one hand, among many derivatives of graphene, pristine graphene (PG) exhibits fewer defects and a lower content of oxygen-containing functional groups, allowing it to fully leverage its intrinsic properties (lightweight, high conductivity, large specific surface area, and superior mechanical performance)7,21,22. However, constructing a 3D network-structured PG preform with substantial compressive strength poses a challenge due to its weak gelation ability23. On the other hand, the design and construction of heterogeneous interfaces between PG and Mg are of great significance for multi-component composites to achieve excellent EMI shielding and enhanced mechanical performance24. However, the poor interface of PG/Mg composites would be detrimental to the attenuation of electromagnetic waves (EMWs) within the composites and load transfer efficiency. It is crucial to introduce gradient interface layers to improve both the compressive strength of the PG preform and the interface bonding of the composites.

Herein, we present an innovative synthesis of PG@pyrocarbon (PG@PyC) reinforced AZ91D matrix (PG@PyC/AZ91D) composites, achieved through the preconstruction of a 3D PG@PyC skeleton followed by molten AZ91D infiltration filling. The prepared composites show exceptional EMI shielding performance and mechanical properties. Our design leverages the construction of a 3D connected PG@PyC network within the AZ91D matrix through this method, which can not only provide multiple reflections and scatterings of the EMWs and enhance interfacial polarization but also achieve a better load transfer strengthening effect. This innovative concept paves a way to obtain excellent lightweight structural-functional materials and shows promise for applications in next-generation electronic communication systems.

Results

Morphologies and structures of PG@PyC preforms

The synthesis process of preforms and composites is shown in Supplementary Fig. 1. The initial step in the proposed strategy (3D skeleton preconstruction-infiltration filling) is to fabricate a multifunctional 3D PG@PyC preform that exhibits both EMI shielding and mechanical performance. Methane (CH4) was introduced as a carbon source at 1070–1080 °C for 60, 90, and 120 min on the PG preform, and the obtained preforms were named PG@PyC-60, PG@PyC-90, and PG@PyC-120, respectively. When the deposition time of PyC is 60–120 min, PyC would significantly affect the compressive strength and porosity of the PG@PyC preform, thereby affecting the overall properties of the composite. Figure 1 and Supplementary Fig. 2 show the morphologies of the PG and PG@PyC preforms, where PG sheets interconnect to generate a 3D network. In Fig. 1a1-d4, the 3D honeycomb structure can be retained even after deposition, and the uniform pore diameter is 15–80 μm. The PG sheets are uniformly coated with PyC, consisting of granular particles (0.2–1.5 μm). As the deposition time increases, the granular PyC gradually grows. The deposition of PyC would thicken the PG sheets (Fig. 1e1, f1, g1, h1, i1, i2, j1, and j2). PyC is grown on the PG sheets with an interlayer spacing of ~0.34 nm (Fig. 1e2, f2, g2, and h2). The selected area electron diffraction (SAED) pattern of PG changes from a hexagonal structure to a circular ring composed of symmetrical diffraction patterns on both sides after depositing PyC, indicating that the crystallinity of PyC is inferior to that of PG. There are more and more continuous carbon layers arranged on the PG sheets with the increase in deposition time.

Fig. 1: SEM, TEM, and AFM images of preforms.
Fig. 1: SEM, TEM, and AFM images of preforms.
Full size image

a1–a4 SEM images of PG preform. b1–b4 SEM images of PG@PyC-60 preform. c1–c4 SEM images of PG@PyC-90 preform. d1–d4 SEM images of PG@PyC-120 preform. e1, e2 TEM images of PG sheets. f1, f2 TEM images of PG@PyC-60 sheets. g1, g2 TEM images of PG@PyC-90 sheets. h1, h2 TEM images of PG@PyC-120 sheets. i1, i2 AFM images of PG sheets. j1, j2 AFM images of PG@PyC-60 sheets. The AFM image height (nm) is shown by the color bar. Pristine graphene and pyrocarbon are defined as PG and PyC, respectively. Inverse fast fourier transform is abbreviated as IFFT.

The entire process of precursor pyrolysis and PyC deposition is analyzed through elementary reactions and morphologies of the PG@PyC preform, as shown in Fig. 2 and Supplementary Table 1. The rate constants are fitted according to the modified Arrhenius expression25 (Supplementary Equation (1)). The precursor pyrolysis of PyC progresses from CH4 to C2-species, followed by C4-species and C6-species. To elucidate the reactions that play significant roles in the conversion of CH4 to C6H6, the reaction enthalpy (\(\Delta\)H) and Gibbs free energy change (\(\Delta\)G) of each elementary reaction are calculated by HSC software, and the results are shown in Fig. 2a–f. Notable differences in \(\Delta\)H and \(\Delta\)G values are observed across different reactions. Reactions with higher rates and lower \(\Delta\)G values are likely to be predominant during the precursor pyrolysis process. The first aromatic ring (C6H6) would be condensed to polycyclic aromatic hydrocarbons (PAHs). In the initial stage of deposition, C2H2 and C2H4 primarily react at the edge carbon atoms of PG, supplemented by the nucleation of PAHs on the PG or newly formed aromatic carbon planes. Figure 2g-j show the morphologies of the PG@PyC preforms when the deposition time of PyC is 20 min and 40 min, respectively. Some small granular PyC is grown on the PG sheets. A cross-linked structure between PyC and PG can be formed (Fig. 2j). As the deposition time increases, granular PyC is deposited and gradually merges until the PG substrate is completely covered, which can be further confirmed by Fig. 1b4, c4, and d4. The above schematic diagram of the PyC deposition process is displayed in Fig. 2k.

Fig. 2: The entire process of precursor pyrolysis and PyC deposition.
Fig. 2: The entire process of precursor pyrolysis and PyC deposition.
Full size image

a–c Reaction enthalpy (\({{\boldsymbol{\triangle }}}\)H) and d–f Gibbs free energy change (\({{\boldsymbol{\triangle }}}\)G) of each elementary reaction among the reactions from CH4 to C6H6 calculated by HSC software. g SEM image of PG@PyC-20 preform. h–j SEM and TEM images of PG@PyC-40 preform. Pristine graphene and pyrocarbon are defined as PG and PyC, respectively. k Schematic diagram of the PyC deposition process.

XRD, Raman spectroscopy, and XPS are employed to investigate the structure evolution of PG and PG@PyC. A primary peak located at 25.5-26.5° is identified as the (002) plane of graphite in Fig. 3a. Moreover, the interlayer spacing (\({d}_{002}\)), mean microcrystalline height (\({L}_{c}\)(002)), and graphitization degree (g) are calculated by Supplementary Equations (2)-(4)26 and the results are listed in Supplementary Table 2. It is evident that with increasing deposition time, an increase in \({d}_{002}\) and a decrease in both \({L}_{c}\)(002) and g for PG@PyC preforms are obtained. The PG@PyC-60 preform exhibits a lower graphitization degree, a more disordered carbon layer structure, and a smaller graphene microcrystalline size than the PG preform. The defects in carbon materials can be detected by Raman spectroscopy, and the results are given in Fig. 3b and c. There are three typical vibration peaks (D, G, and 2D bands) in PG preform, while the PG@PyC preforms show two obvious characteristic peaks (D and G bands). The Raman spectra of PG@PyC are fitted with five Lorenz curves located at 1200 cm-1 (D4), 1350 cm-1 (D), 1500 cm-1 (D3), 1580 cm-1 (G), and 1620 cm-1 (D’)27. After depositing PyC, highly defective D3 and D4 bands appear, while the 2D band of the graphene lattice disappears, indicating a significant difference in graphitization degree between PG and PyC. The G band reflects the graphitic degree related to the in-plane motion of sp2 carbon atoms in the carbon materials4,28. Since the graphitization degree of PyC is lower than that of the PG, the G peak of PG@PyC would decrease sharply after depositing PyC on the PG in Fig. 3b. The width of graphene microcrystals (\({L}_{a}\)) can be calculated from the intensity ratio of D band to G band (ID/IG, Supplementary Equation (5)29), as shown in Supplementary Table 3. The ID/IG value increases from 0.15 to 1.13 after depositing PyC, accompanied by a decrease in \({L}_{a}\) from 29.33 nm to 3.89 nm. The chemical compositions of PG@PyC preforms are elucidated through XPS spectra in Fig. 3d, e. The high-resolution C 1 s spectrum is deconvoluted into four sub-peaks at 284.5 eV, 285.5 eV, 286.5 eV, and 289.0 eV, which correspond to C-C (sp2), defect peak (sp3), C-O, and O-C = O bonds, respectively. The area ratio of sp3 to sp2 (sp3/sp2) serves as an indicator of the defect level within carbon materials. Notably, the sp3/sp2 ratio of PG@PyC preforms shows an increase with higher PyC content. It demonstrates that the introduction of PyC leads to an increase in defects and structural disorder inside the materials, which can enhance dipole polarization loss and attenuation of EMWs28.

Fig. 3: Characterization of PG, PG@PyC, and pure PyC preforms.
Fig. 3: Characterization of PG, PG@PyC, and pure PyC preforms.
Full size image

a XRD patterns. b, c Raman spectra. d Wide-scan XPS spectra. e High-resolution XPS of C 1 s. f Surface roughness of preform and contact angles between molten AZ91D droplet and preform. The height of the surface (μm) is shown by the color bar. g, h Compressive stress-strain curves. i Infrared images and thermal conductivity. The temperature (°C) is shown by the color bar. Data are presented as mean values ± standard deviations (SD) from n = 3 impendent experiments.

The surface roughness of the preforms and the contact angles between molten AZ91D droplets and preforms are displayed in Fig. 3f. The roughness parameters Sa and Sq are calculated by Supplementary Equations (6)-(7)30,31. Compared with PG preform, the Sa values of PG@PyC-60, PG@PyC-90, and PG@PyC-120 are 31.2 μm, 44.1 μm, and 51.9 μm, respectively, showing increases of 100.00%, 182.69%, and 232.69%. The wetting angle test is detailed in the Supplementary Methods. The contact angles between AZ91D droplet and PG, PG@PyC-60, PG@PyC-90, and PG@PyC-120 preforms are 111°, 123°, 131°, and 128°, respectively. An extended deposition time of PyC leads to an increase in both surface roughness and contact angle. Although the interface between PyC and AZ91D is in a non-wetting state, the enhanced surface roughness of PG endowed by PyC would form a tight interlocking effect with the matrix, making it difficult for PG and Mg to slide along the interface and strengthening the interfacial bonds32,33. The compressive stress-strain curves of PG, PG@PyC, and pure PyC preforms are shown in Fig. 3g, h. The preparation process of pure PyC preform is detailed in the Supplementary Methods. The compressive strengths of PG, PyC-60, PyC-90, and PyC-120 preforms are 4 kPa, 34 kPa, 66 kPa, and 113 kPa under 50.0% deformation, respectively. In contrast, the compressive strengths of PG@PyC-60, PG@PyC-90, and PG@PyC-120 preforms are 35 kPa, 69 kPa, and 120 kPa, significantly increasing by 8.75, 17.25, and 30.00 times, respectively, compared with the PG preform. The deposition of PyC increases the thickness of the cell walls of the PG preform, significantly improving its compressive strength. The PG@PyC preforms demonstrate remarkable mechanical stability, which is essential for withstanding the infiltration pressure of the molten AZ91D during the preparation of the composites.

The thermal management capability of preforms can be estimated by the infrared thermal imaging method and measurements of thermal conductivity (Fig. 3i and Supplementary Fig. 3). The thermal conductivities of the PyC-60, PyC-90, and PyC-120 preforms are 0.039 W∙m-1∙K-1, 0.050 W∙m-1∙K-1, and 0.065 W∙m-1∙K-1, respectively. The thermal conductivities of PG preform increase from 0.052 W∙m-1∙K-1 to 0.066 W∙m-1∙K-1, 0.072 W∙m-1∙K-1, and 0.075 W∙m-1∙K-1, respectively, after depositing different contents of PyC. The thermal conductivity of both pure PyC and PG@PyC preforms rises with increasing PyC content. The PG and PG@PyC preforms are placed on a hot stage at 180 °C, and then an infrared thermal camera is used to record the temperature during heating. It is found that the heating rates of all PG@PyC preforms exceed that of the PG preform. Thermal conduction in the 3D structured PG preform comprises solid heat conduction, gas heat conduction, and radiative heat conduction34. An increase in carbon content coupled with a reduction in porosity within the preform enhances solid heat conduction, thereby improving thermal conductivity.

The EMI shielding performance, calculated by Supplementary Equations (8)–(14), is related to the electrical conductivity (Supplementary Fig. 4) of the preform, which plays a crucial role in that of the composites. The EMI shielding effectiveness (SE) of the preforms in the X band is presented in Fig. 4a and b. The PG preform exhibits a low EMI SE value of 25.62 dB, whereas the EMI SE values of PG@PyC increase significantly to 35.28 dB, 80.50 dB, and 77.77 dB, respectively, with rising PyC content. The enhancement in EMI SET with more PyC content aligns with the corresponding increase in electrical conductivity of the preform. In Fig. 4c, the reflection coefficient (R) gradually improves with extended growth time, surpassing the absorption coefficient (A), demonstrating that the primary shielding mechanism of the PG@PyC preform is the reflection process.

Fig. 4: EMI shielding performance.
Fig. 4: EMI shielding performance.
Full size image

a, b EMI SE values in the X band. c Average R, T, and A values. Reflection coefficient (R), transmission coefficient (T), and absorption coefficient (A) represent the ability of reflection, transmission, and absorption of the EMWs, respectively. d Schematic diagram of the EMI shielding mechanism of PG@PyC preform. When incident EMWs reach the surface of the PG@PyC preform, most are shielded. The remaining absorption EMWs are dissipated by heterogeneous interface polarization loss, multiple reflections and scatterings, multistage conduction loss, and dipole polarization loss.

The EMI shielding mechanism is illustrated in Fig. 4d. When incident EMWs reach the surface of the PG@PyC preform, most are shielded due to the impedance mismatch between the preform and free space. Electrically conductive networks with numerous charge carriers enhance EMI SE through reflection. The remaining EMWs penetrate into the preform and are dissipated by the following processes. Firstly, PG shows a high G peak while PyC exhibits a low G peak due to their different graphitization degrees, resulting in abundant heterogeneous interfaces. The mismatch of electrical conductivity between PG and PyC results in significant interfacial polarization. When subjected to an alternating EM field, variations in charge carrier mobility and polarization response rates across different dielectric materials cause free charges to accumulate at the PG-PyC interface. The network structure with many interfaces and conductivity mismatches would enhance charge storage capacity and amplify interfacial polarization35. Secondly, the 3D network structure of the PG@PyC preform provides pathways for electron motion and EMWs transmission through multiple reflections and scatterings. With the increase of PyC content, more cross-linking among PG sheets and more electron transfer pathways can be obtained. Thirdly, Raman analysis indicates that PyC contains various lattice defects, inducing dipole polarization under an alternating EM field36. Consequently, the EMI SE increases with the PyC content. Compared with PG@PyC-90, the EMI shielding performance of PG@PyC-120 is reduced. This reduction can be attributed to the fact that the increased deposition of PyC tends to cover some small pores within the preform. Consequently, this leads to a diminished capacity for electron and phonon transport within the 3D structure. In summary, these findings demonstrate a multifunctional PG@PyC-90 preform, which achieves excellent EMI shielding performance, thermal transfer efficiency, and mechanical stability.

Microstructures of composites

Through the liquid-solid infiltration extrusion method, the obtained composites were designated as PG/AZ91D, PG@PyC-60/AZ91D, PG@PyC-90/AZ91D, and PG@PyC-120/AZ91D composites, respectively, corresponding to different preforms. Based on the extrusion direction during the preparation of the composites (Supplementary Fig. 5), the composites both perpendicular and parallel to the extrusion direction are selected for SEM characterization. The infiltration temperature of molten AZ91D (Supplementary Fig. 6), the concentration of PG (Supplementary Fig. 7), and the introduction of PyC significantly influence the microstructures of the composites. Fig. 5a1–d4 illustrates the microstructures of the composites with varying PyC content grafted at their interfaces. In these composites, the black areas pointed by yellow arrows represent PG and PG@PyC, while lighter areas correspond to the Mg matrix. For both PG/AZ91D and PG@PyC-60/AZ91D composites, aggregated PG sheets (marked with red boxes), small pores within those sheets, and the disrupted network can be observed in Fig. 5a1–b4. As indicated in Fig. 5a3, a4, b3, and b4, it is evident that the distribution of PG sheets tends to align perpendicularly to the extrusion direction, which proves that the preforms undergo flattening during the infiltration process, resulting in the agglomeration of PG. It can be attributed to the lower compressive strength of both PG and PG@PyC-60 preforms, which are unable to withstand the infiltration pressure. In contrast, when the growth time of PyC is prolonged to 90 min, the arrangement of PG becomes more homogenous, and the continuous 3D network of PG@PyC remained intact in the composites (Fig. 5c1-c3), verified by 3D reconstruction results (Fig. 5e and Supplementary Movies 1-3). The PG@PyC-90 preform with enhanced compressive strength can resist the infiltration pressure, which helps to maintain the uniform network structure of PG@PyC in the AZ91D matrix and prevent re-agglomeration. Additionally, some small pores in the PG layers are also well infiltrated in Fig. 5c4. Regarding the PG@PyC-120/AZ91D composites (Fig. 5d1–d4), it is found that a lamellar structure with partially interconnected bridges of PG@PyC is dispersed in the AZ91D matrix. A pressure exceeding 30 kPa is needed to overcome capillary infiltration resistance and achieve complete pore filling according to the Washburn equation37 (Supplementary Equation (15)). This requires the preform to have adequate compressive strength. During the infiltration process, molten AZ91D is flown into the PG@PyC preform under an external pressure of 30-40 kPa, without damaging the 3D structure of PG@PyC-90 and PG@PyC-120 preforms. The molten AZ91D moves from the top to the bottom and from the edges to the core regions within the preform38. This infiltration process remains stable, with pores gradually being filled by molten AZ91D. When the infiltration temperature is appropriate (~660 °C) and the compressive strength of the preform is sufficiently high (more than 40 kPa), uniform and adequate infiltration of molten AZ91D can be achieved without any non-infiltration regions. To decrease or eliminate possible existing pores, various strategies drawn on the techniques associated with the squeeze casting process or vacuum casting process can be employed, including controlling extrusion temperature39, applying high pressure during solidification40, lowering cavity air pressure41, combining vacuum with the casting process42, and so on.

Fig. 5: Characterization of the composites.
Fig. 5: Characterization of the composites.
Full size image

a1–a4 SEM images of PG/AZ91D composites. b1–b4 SEM images of PG@PyC-60/AZ91D composites. c1–c4 SEM images of PG@PyC-90/AZ91D composites. d1–d4 SEM images of PG@PyC-120/AZ91D composites. The composites both perpendicular (x-y plane) and parallel (x-z plane) to the extrusion direction are selected for SEM characterization. e 3D reconstruction result of 3D PG@PyC in the PG@PyC-90/AZ91D composites. The composites were characterized using an X-ray microscope (ZEISS, Xradia 515 Versa), resulting in a series of 2D section images. The 3D reconstruction result can be obtained from this collection of 2D images through image segmentation, surface reconstruction, and visualization utilizing Dragonfly 2021 software. f EDS spectra of Spot A-C in (a2), (b2), and (d2), respectively. g XRD diffractions of AZ91D and composites. h Raman spectra of the composites.

There are white particle precipitations near the PG in the composites, marked as Spot A-C and confirmed by EDS in Fig. 5f and Supplementary Table 4. The white particle precipitations contain Al and Mn elements with an atomic ratio close to 8:5. Fig. 5g shows the XRD patterns of AZ91D and composites, revealing that AZ91D and all composites display typical diffraction peaks of α-Mg. The primary constituents of AZ91D are α-Mg and Mg17Al12 phases43. The Mg17Al12 phase can be characterized in the PG@PyC/AZ91D composites. However, the Mg17Al12 phase is not detected in the PG/AZ91D composites due to its low content. The presence of Mg17Al12 phase in the PG/AZ91D composites can be proven by the TEM images (Supplementary Fig. 8). The peaks of Al8Mn5 appear in the PG/AZ91D and PG@PyC/AZ91D composites. Mn is almost insoluble in α-Mg and would react with Al solute to form Al8Mn5 rather than forming intermetallic compounds (IMCs) with Mg44. Al8Mn5 can be formed at various stages during handling, solidification, and heat treatment of AZ91D melt45,46,47, leading to a reduction in the content of the Mg17Al12 phase. The diffraction peak of PG is only detected in the PG/AZ91D composites due to the abundance of PG clusters. No aluminum carbides are found in the composites, as highly purified carbon materials (PG and PyC) would not react with Al, even at relatively high temperatures48. Figure 5h and Supplementary Fig. 9 present the Raman spectra of the composites. The increase in the ID/IG value of the PG/AZ91D is greater than that of the other composites after the preparation of the composites. This means that even though the preparation process of the composites may induce structural damage to the PG, PyC would protect PG and effectively enhance its performance.

Interfacial structures of composites

The interfacial structures between PG and AZ91D matrix are crucial in determining the EMI shielding and mechanical properties of the composites. TEM images of the composites are shown in Fig. 6. There are no interfacial products in the PG/AZ91D composites (Fig. 6a1 and a2), displaying mechanical bonding between PG and Mg. A disordered area and mutual diffusion (PG-Mg) are evident at the interface in Fig. 6a3 (yellow boxes). Furthermore, some dislocations are presented near the interface. HRTEM image of Area C (Fig. 6a4) reveals that PG comprises multiple layers of graphene. After introducing PyC, PyC-MgO rich zones are tightly embedded within the PG and matrix in Fig. 6b1–b3, c1, c2, c5, d1-d3, and d5. As the deposition time of PyC is prolonged, the thickness of the PyC-MgO transition layer increases, with the values of ~8.6 nm (PG@PyC-60/AZ91D), ~18.4 nm (PG@PyC-90/AZ91D), and ~314.5 nm (PG@PyC-120/AZ91D), respectively. From Fig. 6b3, c2, c5, and d5, dispersed nanoparticles (marked with yellow circles) accompanied by PyC at the interface can be confirmed as MgO phase, which exhibits an interlayer spacing of 0.21 nm and polycrystalline rings (Fig. 6c4). The MgO nanoparticles are formed in situ by the reaction between molten Mg and oxygen-containing groups on the PyC, as supported by \(\Delta\)H and \(\Delta\)G (Supplementary Equation (16) and Supplementary Fig. 10). The molten Mg would not react with C from either PG or PyC, which can be confirmed by Supplementary Fig. 11. It can be attributed to the following reasons. There is no wetting between C and Mg, which makes it challenging for the two elements to engage in chemical reactions. Moreover, magnesium carbides such as MgC2 and Mg2C3 are thermodynamically unstable. When the temperature is below 450 °C and 670 °C, they will decompose, respectively49,50. Unlike magnesium carbides, MgO is a thermodynamically stable phase49,50. Therefore, MgO can be obtained stably while the magnesium carbides cannot be formed. After depositing PyC on PG, PG@PyC comprises multiple disordered carbon layers (Fig. 6b3, c2, d2, and d3). The SAED patterns of PG and PyC are both circular rings (Fig. 6c3 and d4) due to the diffusion of PyC onto the PG. Mg can be determined by the HRTEM images (an inter-planar spacing of 0.28 nm in Fig. 6a5) and SAED pattern (Fig. 6b4). It discloses that the Mg (\(01\bar{11}\)) plane is along the Mg [\(\bar{2}110\)] zone axis. Fig. 6b5 shows an atomic-resolution HAADF-STEM image of Mg, with an inset displaying a repeatable stacking unit of the Mg lattice along <1\(\bar{2}\)10> . It indicates that Mg has a hexagonal close-packed (hcp) structure with an ABAB stacking arrangement. The PG is well-bonded to Mg through the PyC-MgO transition layer. The lattice distortion of the PG/PyC-MgO/Mg heterogeneous interface would induce more lattice defects and sufficient carriers to strengthen interfacial polarization and consume EMWs. The in-situ grown PyC, accompanied by MgO nanoparticles, would improve interfacial bonding and enhance load transfer efficiency51.

Fig. 6: Interface structures of a1-a5 PG/AZ91D, b1-b5 PG@PyC-60/AZ91D, c1-c5 PG@PyC-90/AZ91D, and d1-d5 PG@PyC-120/AZ91D composites.
Fig. 6: Interface structures of a1-a5 PG/AZ91D, b1-b5 PG@PyC-60/AZ91D, c1-c5 PG@PyC-90/AZ91D, and d1-d5 PG@PyC-120/AZ91D composites.
Full size image

a1, b1, c1, d1 TEM images. a2 HRTEM image of Area A in (a1). a3 IFFT image of Area B in (a2). a4 HRTEM image of Area C in (a1). a5 HRTEM image of Area D in (a1). b2 HRTEM image of Area E in (b1). b3 HRTEM image of Area F in (b1). b4 SAED pattern of Area G in (b2). b5 Atomic-resolution HAADF-STEM image of Area H in (b3) (inset image: schematic diagram of HCP Mg). c2 HRTEM image of Area I in (c1). c3 SAED pattern of Area J in (c2). c4 FFT pattern of Area K in (c2). c5 HRTEM image of Area L in (c1). d2, d3 HRTEM images of Area M in (d1). d4 SAED pattern of Area N in (d3). d5 HRTEM image of Area O in (d1). Pristine graphene and pyrocarbon are defined as PG and PyC, respectively.

EMI shielding and mechanical performances of composites

Figure 7a, b and Supplementary Fig. 12 show the EMI shielding performance of the AZ91D and composites in the X band. All samples were prepared with a thickness of 3 mm for EMI shielding evaluation, with the exception of the PG@PyC-90/AZ91D composite, which was also assessed at 2 mm. After introducing PG, the average SET value of the PG/AZ91D composites increases from 21.81 dB to 26.13 dB. As the deposition time of PyC is extended to 60 min, 90 min, and 120 min, the corresponding SET values are 36.71 dB, 76.70 dB, and 57.54 dB, respectively. The average SET value of the PG@PyC-90/AZ91D composites with a thickness of 2 mm is 69.59 dB. When the thickness of the material is much greater than the skin depth, increasing the thickness of the material has little effect on the EMI shielding performance52. The skin depth, which is related to electrical conductivity (Supplementary Fig. 13), can be calculated by Supplementary Equations (17)–(18)53, as shown in Supplementary Fig. 14. Compared with AZ91D, the enhancement in SET values of the composites can be attributed to the improved SEA values with no obvious variation in SER values. The SER of the composites can also be expressed as Supplementary Equation (14)54. The minimal change in SER is attributed to slight variations in electrical conductivity. The designed 3D network structure of PG@PyC in the AZ91D matrix promotes higher SEA compared to that of the lamellar structure, resulting in an increased SET value of PG@PyC-90/AZ91D in contrast to the PG@PyC-120/AZ91D composites. In Fig. 7a, a peak between 9.2 GHz and 9.5 GHz is found in the PG@PyC-90/AZ91D composites, while the peak has not been found in AZ91D and other composites. It suggests the presence of resonance behavior55 in the PG@PyC-90/AZ91D composites. The special 3D network structure of PG@PyC-90 in the AZ91D matrix induces eddy current effects and resonances that contribute to dielectric loss, resulting in high-performance EMI shielding and resonance behavior56. Existing studies indicate that the EMI SE values reaching near 100 dB in Mg alloy or Mg matrix composites are mainly based on the Mg-Li system with the unique α-Mg/β-Li dual-phase54,57. The EMI SE value of Mg-Li matrix composites increases from 77-96 dB to 89-100 dB due to the introduction of reinforcements (MWCNT or Ni0.4Zn0.4Co0.2Fe2O4), representing an improvement of 4-30%54,57. To highlight the enhancement effect of PG@PyC in this work, comparisons of shielding performance (enhancement efficiency (\(\Delta\)SE) of the materials with various EMI SE values) of the PG@PyC-90/AZ91D composites with other graphene-based, Mg matrix composites, and Mg alloys are shown in Fig. 7c and Supplementary Table 554,58,59,60,61,62,63. The \(\Delta\)SE value can be evaluated by Supplementary Equation (19)58. A significant \(\Delta\)SE value of 457.6 can be obtained, which is much higher than other existing studies. The remarkable strengthening efficiency and EMI SE value highlight the distinct advantage of constructing 3D network-structured PG@PyC within the AZ91D matrix.

Fig. 7: EMI shielding properties of AZ91D and composites.
Fig. 7: EMI shielding properties of AZ91D and composites.
Full size image

a, b EMI SE values. c Comparison of EMI shielding performance (\({{\boldsymbol{\triangle }}}\)SE values with various EMI SE values (X band)) between existing studies and this work. d Verification test of PG@PyC-90/AZ91D composite as an EMI shielding material. e Schematic diagram of the EMI shielding mechanism of PG@PyC-90/AZ91D composites. The EMWs are reflected at the surface of the composites when they reach the surface of the composites. The absorption attenuation within the composites can mainly come from conductive loss, multiple internal reflections, interfacial polarization, and dipole polarization.

To validate the EMI shielding performance of the PG@PyC-90/AZ91D composites, we employed a Tesla coil wireless power transmission system as the EM emission source53. As indicated in Fig. 7d and Supplementary Movie 4, when the PG@PyC-90/AZ91D composite is placed between the Tesla coil and bulb, the light is immediately off, demonstrating that the EMW signal can be effectively shielded by the PG@PyC-90/AZ91D composite. The primary factors for the enhancement of the EMI shielding property of the PG@PyC-90/AZ91D composites are summarized in Fig. 7e. Firstly, EMWs are reflected at the surface of the composites due to impedance mismatch. Secondly, the rest of the EMWs enter into the composites. The absorption attenuation within the composites can mainly come from dielectric loss and multi-level reflections. Dielectric loss is related to the dielectric conductivity and dielectric polarization54. On one hand, the highly connected 3D-PG@PyC establishes extensive conductive pathways and provides conductive loss. On the other hand, the PG/PyC-MgO/Mg heterointerfaces would generate more lattice defects, which in turn increase carrier density. This enhancement promotes an uneven distribution and accumulation of spatial charges at the heterojunction contact, thereby promoting interfacial polarization and facilitating the absorption of EMWs59. The numerous heterointerfaces, dislocations, grain boundaries, and phase boundaries further contribute to the interfacial polarization. Moreover, defects and oxygen-containing functional groups on PyC are conducive to the dipole polarization, which results in energy consumption.

The nano-indentation technique can be used to evaluate the enhancement effects of PG and PyC. The suffixes -xy and -xz refer to the application of a 5 mN load on the x-y and x-z planes of the composites, respectively. As shown in Fig. 8a–d, the introduction of PG leads to a reduction in penetration depth compared to the AZ91D, demonstrating the strengthening effect of PG on the AZ91D matrix. As the deposition time of PyC increases, the indentation depth initially becomes shallower, then deeper, followed by shallower again. This trend is also observed in both hardness and elastic modulus. In comparison with AZ91D, the PG@PyC-90/AZ91D composites are endowed with increments of 34.00% in hardness (1.34 GPa) and enhancements of 5.82% and 4.96% in elastic modulus (49.30 GPa and 48.90 GPa), respectively. The mechanical properties are further investigated through compressive and tensile strength experiments (Fig. 8e–h, Supplementary Fig. 15, Supplementary Fig. 16, Supplementary Fig. 17, and Supplementary Fig. 18). The ultimate compressive strength (UCS) and failure strain of AZ91D (remolten AZ91D by liquid-solid infiltration extrusion method) are 247 MPa and 15.6%, respectively. Due to the agglomeration of PG in the AZ91D matrix, the UCS values of the PG/AZ91D and PG@PyC-60/AZ91D composites are lower than that of AZ91D, making it unable to exert the strengthening effect of PG. The UCS and failure strain of the PG@PyC-90/AZ91D composites are enhanced gradually and exhibit the best values of 276 MPa and 21.0%, getting 58.62% and 40.00% improvements, respectively, compared with PG/AZ91D composites. Compared to AZ91D, the ultimate tensile strengths (UTS) of the composites firstly decrease to 98 MPa and 127 MPa, and then increase to 231 MPa and 216 MPa with the addition of PG and PyC. The UTS is increased by 26.92% for PG@PyC-90/AZ91D composites and by 18.68% for PG@PyC-120/AZ91D composites, respectively, in contrast to AZ91D. The elongation of the PG@PyC-90/AZ91D composites is 4.9%. It indicates that constructing 3D network-structured PG@PyC in the AZ91D matrix contributes to simultaneously enhancing the UCS and UTS of the composites. The strengthening mechanism can be explained by the grain size of α-Mg in the composites, as presented in Supplementary Fig. 19, Supplementary Fig. 20, and Fig. 8i. The PG is distributed along the grain boundaries of α-Mg, which serves to restrict the growth of α-Mg grains and limit the migration of grain boundaries64,65. The PG/PyC-MgO/Mg interfaces facilitate load transfer. Both dislocation and fine grain size significantly enhance the reinforcing effect of PG, thereby facilitating superior nanomechanical strengthening.

Fig. 8: Mechanical performances of AZ91D and composites.
Fig. 8: Mechanical performances of AZ91D and composites.
Full size image

a, b Nano-indentation curves. c, d Histogram of hardness and elastic modulus. Data are presented as mean values ± SD from n = 3 impendent experiments. e Compressive stress-strain curves. f Histogram of UCS and failure strain. g Tensile stress-strain curves. h Histogram of UTS and elongation. i Schematic diagram of the nanomechanical response for PG@PyC/AZ91D composites.

Discussion

This work presents an innovative method of 3D skeleton preconstruction-infiltration filling. A continuous 3D PG@PyC network in the AZ91D matrix is fabricated using the liquid-solid infiltration extrusion method, aimed at producing lightweight structural-functional materials that achieve excellent EMI shielding performance and mechanical stability. Benefiting from the special conductive interconnected structure of the 3D PG@PyC in the AZ91D matrix, an EMI SE value of 76.70 dB for the PG@PyC-90/AZ91D composites can be obtained. The PG@PyC-90/AZ91D composite exhibits optimal mechanical properties, including a UCS of 276 MPa, UTS of 231 MPa, elastic modulus of 49.30 GPa, and hardness of 1.34 GPa. The approach of constructing a 3D multifunctional PG@PyC network within the AZ91D matrix has been proven to improve EMI shielding and mechanical properties. Structural-functional Mg matrix composites hold significant potential for applications across various fields, including the aerospace industry, information technology, and daily life.

Methods

Materials

PG sheets were synthesized from the liquid phase exfoliation method. AZ91D alloy (Mg-8.41 wt.% Al-0.65 wt.% Zn-0.22 wt.% Mn-0.05 wt.% Si-0.0025 wt.% Ni) was purchased from Dongguan Gelan Metal Materials Co., Ltd. All gases were sourced from Xi’an Weiguang Gas Co., Ltd. Poly(vinyl alcohol) (PVA, alcoholysis: 72.5-74.5 mol.%, viscosity: 4.2-5.0 mPa∙s), riboflavin sodium phosphate (RP, USP), and sodium dodecyl sulfate (SDS, GR) were bought from Shanghai Aladdin Biochemical Technology Co., Ltd. and Shanghai Macklin Biochemical Co., Ltd., respectively. All reagents were used without further purification.

Fabrication of PG and PG@PyC preforms

PG preform was prepared by self-assembly combined with freeze-drying methods. PG sheets (10.0 mg∙mL-1, 50 mg), PVA (2.0 mg∙mL-1, 10 mg), RP (0.4 mg∙mL-1, 2 mg), and SDS (15.0 mg∙mL-1, 75 mg) were first mixed in aqueous solution using bath ultrasound and then stirred for 5 min at 300 rpm and 3 min at 1200 rpm. Then, the mixture was frozen at −20 °C for 24-72 h and dried to obtain the PG preform. Finally, the PG preform was heat-treated at 900 °C for 90 min, protected by Ar. The PG@PyC preforms were prepared by depositing PyC onto the PG sheets in a chemical vapor deposition furnace using a mixture of CH4 and Ar.

Preparation of PG/AZ91D and PG@PyC/AZ91D composites

AZ91D was heated to 600–700 °C, then infiltrated into the PG preform and PG@PyC preforms via the liquid-solid infiltration extrusion method. The liquid-solid infiltration extrusion method includes low-pressure infiltration followed by high-pressure densification. A low pressure of 30-40 kPa was applied to facilitate the infiltration of molten AZ91D into the pores of the 3D structured PG@PyC preform, ensuring that the 3D structure remained intact during processing at elevated temperatures. Subsequently, a high pressure of 20–60 MPa was applied to the AZ91D in its liquid-solid state (520–560 °C) to realize densification of the composites without compromising the integrity of the 3D skeleton of the PG@PyC.

Characterization

Field emission scanning electron microscope (SEM, Zeiss Sigma 300) with energy-dispersive spectroscopy (EDS) and double Cs corrector transmission electron microscope (TEM, FEI, Themis Z) were used to investigate the morphologies of the PG preform, PG@PyC preforms, and composites. Optical microscope (OM, Nikon ECLIPSE LV150N) was applied to observe the microstructures of the AZ91D and composites. Atomic force microscope (AFM, Bruker, Dimension Icon) was utilized to analyze the thickness of the PG and PG@PyC sheets. X-ray diffraction (XRD, X’Pert PRO, Cu K\(\alpha\) radiation), Raman spectroscopy (Renishaw, inVia Reflex) with a 532 nm laser, and X-ray photoelectron spectroscopy (XPS, ESCALAB Xi + ) were employed to characterize the phase compositions. 3D confocal laser microscopy (NANOVEA, ST400) was carried out to measure the average surface roughness of the PG and PG@PyC preforms. X-ray microscope (ZEISS, Xradia 515 Versa, 3D CT) was used to characterize the distribution of PG in the composites.

Measurements of mechanical properties

A universal machine (Edburg, HP-20 and HP-100) was used to evaluate the compressive strength of the PG, pure PyC, and PG@PyC preforms. The dimensions of the samples for compression tests were 10 mm × 10 mm × 10 mm. When the deformation of the preform was 0.0-50.0% (i.e., the thickness of the preform was compressed from 10 mm to 5 mm), stress measurements were recorded. A nanomechanical testing system (Hysitron, TI980) was operated to measure the hardness and elastic modulus of the composites under a load of 5 mN. The tensile strength and compressive strength of the AZ91D and composites were determined by an electromechanical universal testing machine (CMT5504). The tensile test was conducted using specimens with a gauge length of 4.0 mm, a gauge width of 2.0 mm, and a gauge thickness of 1.5 mm. The dimensions of the compressed samples were 4.0 mm in diameter and 6.0 mm in height. The tensile and compressive tests were conducted with a loading speed of 0.5 mm/min at room temperature.

Measurements of electrical conductivity

The electrical conductivity of the preforms was measured using a two-probe method with a digital multimeter (DMM6500). The electrical conductivity of the AZ91D and composites was measured by an eddy current conductivity meter (FIRST, FD101).

Measurements of thermal conductivity

The thermal conductivity of the preforms was evaluated by a thermal constant analyzer (Hot Disk TPS2500S). An infrared camera (InfraTec, VarioCAM hr 384) was used to record the temperature changes of the preforms placed on a heating stage (180 °C).

Measurements of EMI SE values

Vector network analyzer (Anritsu, MS4644A) was applied to test S-parameters (S11, S12, S21, and S22) of the PG preform, PG@PyC preforms, and composites from 8.2 to 12.4 GHz (X band). The sample was machined to 22.86 mm\(\times\) 10.16 mm\(\times\) 3.00 mm.

Statistics and Reproducibility

The microscope experiments for the preforms and composites (including AFM, SEM, OM, and TEM) were repeated three times independently with similar results. The representative findings are presented in this paper. For data that includes error bars, the experimental tests were also independently repeated three times. The data are presented as mean values ± standard deviations (SD).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.