Supercritical CO₂ Foaming of Molecular Chain-Modified Polyimide for High-Performance Electromagnetic Wave Absorption

Abstract

The microwave absorption performance depends on both impedance matching and electromagnetic energy dissipation. Conventional filler-functionalized polyimide (PI) composites can provide strong electromagnetic attenuation, but their limited foamability under supercritical carbon dioxide (scCO₂) restricts porous structure formation and thus impairs impedance matching. To address these limitations, this study proposes an intrinsic molecular modification strategy based on a chain segment design. Furthermore, by combining synergistic dielectric-magnetic loss from nanofillers with a double-layer heterogeneous architecture, a PI-based electromagnetic shielding nanocomposite with strong microwave absorption performance and low-reflection characteristics was constructed. Bisphenol A dianhydride (BPADA) was used as the dianhydride monomer, and a series of PI copolymers with tunable chain segment compositions were prepared through the reaction of BPADA with mixed aromatic diisocyanate (MDI) and aliphatic diisocyanate (HDI), followed by thermal imidization/cyclization. The changes in molecular chain mobility, scCO₂ foaming behavior, and cellular structure associated with the nominal HDI feed fraction were systematically investigated. The results showed that increasing the nominal HDI feed fraction was associated with an improved rigid-flexible balance of the PI backbone, enhanced segmental mobility, a broader foaming window, and the formation of PI foams with uniform and compact cellular structures. CNT@Fe₃O₄ nanofillers with both dielectric and magnetic loss capabilities were further introduced to improve the intrinsic electromagnetic loss of the composite foams. The calculated microwave absorption results showed that, after pore structure regulation, the PI/CNT@Fe₃O₄ nanocomposite foam exhibited markedly enhanced microwave absorption in the X-band. At a porosity of 46 vol %, the minimum reflection loss reached −38 dB at a thickness of 2 mm. A double-layer heterogeneous structure comprising a CNT@Fe₃O₄-containing porous absorbing layer and a CNT-containing dense reflecting layer was then constructed to exploit impedance gradient transition, dielectric-magnetic synergistic loss, and multiple reflection/reabsorption processes. The optimized sample showed an average absorption coefficient of 0.92 in the X-band and a maximum A/R ratio of 332, indicating pronounced high absorption and low reflection characteristics. This study provides a new strategy for designing and fabricating lightweight, high-performance PI-based electromagnetic protection materials with an absorption-dominated shielding behavior.

1. Introduction

With the increasing use of wireless communication and high-frequency electronic systems, the electromagnetic environment has become increasingly complex. Electromagnetic interference (EMI) and electromagnetic pollution have, therefore, become important factors affecting the stable operation, information security, and electromagnetic compatibility of electronic devices. Developing efficient, lightweight, and designable electromagnetic shielding materials has become a major research direction in electromagnetic functional materials. , At present, shielding materials dominated by electromagnetic wave absorption, namely, absorption-dominated EMI shielding materials, are attracting increasing attention − because they can effectively reduce secondary electromagnetic pollution caused by reflection.

The central objective of microwave absorbers is to guide incident waves into the material as much as possible and convert them to thermal energy for dissipation. The dissipation mechanisms of electromagnetic waves in media are mainly divided into dielectric and magnetic losses. Dielectric loss includes conduction loss, dipolar polarization, and interfacial polarization, whereas magnetic loss is mainly associated with natural resonance, exchange resonance, eddy-current loss, and magnetic-moment relaxation. , In multiphase composites, introducing magnetic components can effectively compensate for the insufficient attenuation capacity of purely dielectric systems in the high-frequency regions. Therefore, the design of high-performance microwave absorbers is no longer limited to a single dielectric loss or magnetic loss mechanism. Greater emphasis is now placed on dielectric-magnetic synergistic loss, interface regulation, and multiscale structural design. , Recent studies have shown that a single homogeneous dense material usually struggles to achieve high impedance matching, strong attenuation, and low reflection simultaneously. Regulating built-in electric fields at heterogeneous interfaces can redistribute local charges and thereby tune electromagnetic response and impedance matching. Porous, layered, gradient, and Janus structures are, therefore, emerging as important design directions. Porous structures lower the effective dielectric constant by introducing an air phase, thereby optimizing the impedance matching and enhancing internal multiple reflection. Layered and gradient structures create stepwise variations in conductivity and electromagnetic parameters across spatial scales, enabling electromagnetic waves to undergo a continuous process of entry, propagation, and dissipation between different layers. − Such structures can improve impedance matching, strengthen multiple scattering inside the material, promote interfacial polarization and layer-by-layer dissipation, and help achieve the combined goals of high shielding, high absorption, and low reflection. At the molecular scale, facet modulation in ferroelectric polymers can tailor chain packing and polarization behavior, providing another route to regulate dielectric response. They also meet the key requirement of absorption-dominated EMI-shielding materials for extremely low surface reflection, thereby suppressing secondary electromagnetic pollution that is readily induced by traditional highly conductive substrates.

Polyimide (PI) has excellent heat resistance, flame retardancy, chemical corrosion resistance, and mechanical stability, making it a representative matrix for electromagnetic protection materials. However, improving the shielding performance of PI-based materials solely by functional fillers often causes an excessive surface reflection. Although functional fillers can enhance EMI shielding effectiveness, their high conductivity leads to severe impedance mismatch at the material-free space interface. A large fraction of electromagnetic waves is therefore reflected at the surface, which not only limits their entry into the material for effective dissipation but may also cause secondary electromagnetic pollution in the surrounding environment. The core objective of absorption-dominated shielding is to maximize the contribution of absorption loss while maintaining high shielding effectiveness. From the electromagnetic theory, good impedance matching and efficient internal attenuation cannot be achieved by composition regulation alone. Macro- and mesostructural design must also be used to optimize impedance matching and internal dissipation pathways simultaneously. , Mn-doping-induced lattice distortion provides an atomic-scale route for modulating permittivity through defect- and lattice-related polarization.

Introducing a three-dimensional porous structure into the PI matrix is an effective strategy for simultaneously addressing impedance matching and electromagnetic-wave attenuation. On the one hand, the large amount of air phase within a foam structure can significantly reduce the effective dielectric constant of the system, thereby improving interfacial impedance matching and weakening direct surface reflection. On the other hand, the three-dimensional microporous framework provides more complex propagation pathways for electromagnetic waves entering the material, increasing multiple reflection, electromagnetic scattering, and interfacial polarization at the cell walls and thereby enhancing loss. Ultralight hollow SiC/C nanofibers likewise demonstrate that hollow architectures can extend propagation pathways while reducing density, thereby strengthening electromagnetic-wave attenuation. PI foams therefore provide a suitable structural platform for lightweight, absorption-dominated EMI shielding materials. −

Supercritical fluid foaming, represented by supercritical CO₂ (scCO₂) foaming, has become an important direction in functionalized foaming research. Compared with conventional chemical or physical foaming routes, scCO₂ foaming offers advantages such as no solvent residue, environmental friendliness, tunable pore structure, and potential recyclability. − However, PI systems with rigid molecular chains often suffer from restricted segmental motion, difficult nucleation, and insufficient cell stability. These factors hinder the precise control of cellular structure and limit PI foamability. To address these problems, Li et al. synthesized a fluorinated PI material in which the −CF₃ groups enhanced the CO₂ uptake and improved cell nucleation. Nevertheless, although −CF₃ introduction improves pore nucleation in PI, the rigid PI backbone still shows high resistance to cell growth, resulting in a foam porosity below 40%.

In addition, the neat PI framework is electromagnetically wave-transparent. Porosity mainly optimizes impedance matching at the material/electromagnetic wave interface, whereas substantial electromagnetic energy conversion still requires functional fillers. Carbon-based components such as graphene, , MXene, − and carbon nanotubes can construct three-dimensional conductive networks and increase dielectric loss. Magnetic particles such as Fe₃O₄ and Ni can supplement magnetic loss mechanisms and suppress the abrupt increase in dielectric constant caused by highly conductive networks. Studies have shown that the synergy of multiple mechanisms, including conduction loss, magnetic loss, and polarization relaxation, is essential for constructing high-performance shielding materials. , Across a broader spectral range, near-perfect light-capture materials further illustrate how multiscale structural coupling can coordinate electronic transitions, molecular motion, and spatial scattering to strengthen wave-matter interactions. However, adding these particles to a PI matrix further increases the difficulty of the scCO₂ foaming.

On the basis of these considerations, this study proposes an intrinsic PI molecular modification strategy to significantly enhance the electromagnetic absorption performance. Bisphenol A dianhydride (BPADA) was used as the dianhydride monomer. A series of PI copolymers were prepared by reacting BPADA with mixed aromatic diisocyanate (MDI) and aliphatic diisocyanate (HDI), followed by thermal imidization and cyclization. This strategy regulated the proportions of rigid aromatic ring and flexible aliphatic chain units in the PI backbone and improved its scCO₂ foamability. PI copolymer microcellular foams were then obtained, and CNT@Fe₃O₄ nanofillers were introduced to enhance synergistic dielectric-magnetic loss. Finally, the high foamability of the high-HDI-content PI and the high-reflectivity behavior of the low-HDI-content PI were combined to construct a CNT@Fe₃O₄-containing porous absorbing layer/CNT-containing dense reflecting layer structure, thereby achieving absorption-dominated, low-reflection EMI shielding.

2. Experimental Section

2.1. Materials

Bisphenol A dianhydride (BPADA) was purchased from Shanghai Macklin Biochemical Co., Ltd., China. Hexamethylene diisocyanate (HDI) and diphenylmethane diisocyanate (MDI) were purchased from Tokyo Chemical Industry Co., Ltd. N-methyl-2-pyrrolidone (NMP) was purchased from Adamas-beta. Sodium hydroxide (NaOH, small pellets, AR grade, ≥98.0%) was purchased from Chengdu Kelong Chemical Co., Ltd. Deionized water was prepared in the laboratory. Multiwalled carbon nanotubes (MWCNTs, NC 7000, average length = 1.5 μm, average diameter = 10 nm) were purchased from Nanocyl. FeCl₃·6H₂O, FeCl₂·4H₂O, and NH₃·H₂O were purchased from Aladdin Chemistry Co., Ltd., China. Carbon dioxide (purity ≥ 99.99%, Linde Gas) was used as the foaming agent.

2.2. Sample Preparation

2.2.1. Polymer Synthesis

The PI copolymer precursor solution was prepared through the reaction of BPADA with mixed diisocyanates under alkaline conditions, with the diisocyanate solution added dropwise through a constant-pressure dropping funnel. First, the diisocyanates (a mixture of MDI and HDI) and 50 wt.% aqueous sodium hydroxide solution were separately dissolved in sealed containers. BPADA and 9.44 mL NMP were then added to a three-neck round-bottom flask equipped with a nitrogen inlet and a condenser. After sodium hydroxide was completely dissolved, it was also added to the flask. Once the mixture became clear and transparent, the diisocyanate solution was slowly added through the constant-pressure dropping funnel and reacted at 85 °C for 8 h to ensure full monomer conversion. By regulating the BPADA-HDI-MDI molar ratio, a series of copolymers were successfully prepared. After the reaction, the copolymer solution was precipitated into deionized water under stirring. The flocculent product was collected by filtration, soaked overnight, washed repeatedly, and dried at 80 °C to a constant weight.

The feed amounts and nominal HDI feed fractions of the copolymers are provided in Table . By adjusting the molar fraction of HDI in the total diisocyanates from 0% to 40%, a series of copolymers with different nominal HDI feed fractions were prepared and denoted as Rx, where R represents the PI copolymer and x represents the molar percentage of HDI in the total diisocyanate feed.

1. Synthetic Feed Composition .

	BPADA		MDI		HDI		HDI feed fraction
Sample	mmol	g	mmol	g	mmol	g	mol %
R0	15	7.81	15.75	3.94	-	-	0
R10	15	7.81	14.18	3.55	1.58	0.26	10
R20	15	7.81	12.6	3.15	3.15	0.53	20
R30	15	7.81	11.03	2.76	4.73	0.79	30
R40	15	7.81	9.45	2.36	6.30	1.06	40

^aHDI feed fraction denotes the nominal molar percentage of HDI in the total diisocyanate feed, calculated as 100 × n­(HDI)/[n­(HDI) + n­(MDI)].

2.2.2. Synthesis of CNT@Fe₃O₄ Nanofillers

CNT@Fe₃O₄ nanofillers were prepared from MWCNTs using a modified one-step coprecipitation method. First, 0.2 g of MWCNTs were dispersed in 100 mL of distilled water under ultrasonic stirring for 2 h. Then, 100 mL of a mixed salt solution containing 0.5 g FeCl₂·4H₂O and 1.0 g FeCl₃·6H₂O was added to the dispersion, followed by stirring for another 30 min. Finally, 25–28 wt.% aqueous NH₃·H₂O (approximately 4.76 mL) was added dropwise over approximately 5 min (average addition rate, approximately 0.95 mL min^–1) until the final pH reached at least 9.2. The mixture was stirred at 50 °C for 1 h under ambient air. The obtained product was filtered, washed three times each with deionized water and absolute ethanol, and freeze-dried for approximately 72 h using a BIOCOOL FD-1A-80+ freeze-dryer. The resulting MWCNT-supported Fe₃O₄ nanofiller is denoted as CNT@Fe₃O₄ throughout this work.

2.2.3. Preparation of PI/CNT@Fe₃O₄ Nanocomposites and Foams

The synthesized CNT@Fe₃O₄ nanofillers were added to the PI copolymer solution, and the nanocomposites were prepared by ultrasonic dispersion, mechanical blending, and thermal imidization. The scCO₂ foaming processes for the PI copolymer and its nanocomposite were similar. First, the bulk sample was placed in a high-pressure autoclave, heated to the preset temperature, and then supplied with CO₂. After the pressure reached 15 MPa, the system was maintained for 60 min to ensure a sufficient level of CO₂ dissolution in the matrix. Foaming was then induced by rapid depressurization, producing microcellular foams. For ease of description, the sample obtained by compounding R30 with 3 wt % CNT@Fe₃O₄ is denoted as R30/CNT@Fe₃O₄-3%.

For layered heterogeneous foaming, layers with high (30%) and low (0%) nominal HDI feed fractions were prepared and combined by hot pressing. During subsequent foaming, the high-HDI-feed-fraction layer formed the CNT@Fe₃O₄-containing porous absorbing layer, whereas the low-HDI-feed-fraction layer remained unfoamed as the CNT-containing dense reflecting layer.

2.3. Characterization

Proton nuclear magnetic resonance (¹H NMR) and Fourier-transform infrared (FTIR) spectra of samples were recorded using a Bruker 400 MHz NMR spectrometer and a Nicolet 6700 FTIR spectrometer, respectively. The microstructure of microcellular foams was characterized by scanning electron microscopy (SEM, Phenom Pro). Thermal properties were measured using differential scanning calorimetry (DSC, TA 250) and thermogravimetric analysis (TGA, Mettler Toledo TGA2). Compressive stress–strain tests on the representative R30 PI foam were performed using an electronic universal testing machine (Instron 68TM-10, Instron, USA). CNT@Fe₃O₄ nanofillers were characterized by transmission electron microscopy (TEM, FEI Talos F200x), vibrating sample magnetometry (VSM, LakeShore 7404), and X-ray photoelectron spectroscopy (XPS). Electromagnetic parameters and electromagnetic interference (EMI) shielding performance were characterized in the X-band (8.2–12.4 GHz) using a PNA vector network analyzer (Keysight N5224B). For microwave-absorption evaluation, the complex permittivity and permeability were measured by the waveguide method, and the reflection loss (RL) and normalized input impedance (Z_in/Z₀) at specified thicknesses were calculated using transmission-line theory. For EMI-shielding evaluation, S11 and S21 were measured directly for the double-layer samples; the total shielding effectiveness (SE _T ), reflection shielding effectiveness (SE _R ), absorption shielding effectiveness (SE _A ), reflection coefficient (R), transmission coefficient (T), and absorption coefficient (A) were calculated from these scattering parameters. The corresponding equations are provided in the Supporting Information. The two-port waveguide system was equipped with an X-band WR-90 standard section (Keysight 00896-60008) and calibrated by the thru-reflect-line (TRL) method. All specimens for the electromagnetic measurements were prepared using a rectangular mold with a fixed in-plane dimension of 22.68 × 10.16 mm². The three porous specimens containing 3 wt.% CNT@Fe₃O₄, arranged from left to right in Figure S2 at void fractions of 37%, 42%, and 55%, had measured postfoaming thicknesses of 2.68, 2.77, and 3.07 mm, respectively. The three porous specimens containing 4 wt.% CNT@Fe₃O₄, arranged from left to right in Figure , had measured postfoaming thicknesses of 2.84, 2.76, and 3.09 mm, respectively. The three porous specimens containing 5 wt.% CNT@Fe₃O₄, arranged from left to right in Figure S3 at void fractions of 36%, 45%, and 52%, had measured postfoaming thicknesses of 2.84, 2.95, and 2.77 mm, respectively. The reported RL curves were calculated from these measured electromagnetic parameters at the specified design thicknesses using an ideal metal-backed single-layer absorber model; no physical metal backing was used during parameter measurement. For the direct EMI-shielding measurements, the double-layer samples comprised a 3.25–3.65 mm porous absorbing layer and a 1.00 mm dense reflecting layer (total thickness, 4.25–4.65 mm), with the incident wave entering from the porous absorbing-layer side.

6.

Reflection loss of PI-based nanocomposite foams (4 wt.% nanofiller) at various thicknesses with different porosities: (a) 29 vol.%; (b) 46 vol.%; (c) 57 vol.%. Impedance-matching of the corresponding foams at various thicknesses with different porosities: (d) 29 vol.%; (e) 46 vol.%; (f) 57 vol.%.

3. Results and Discussion

3.1. Regulation of PI Molecular-Chain Structure and Foaming Behavior

The rigid-flexible characteristics of the PI molecular chain are among the key intrinsic factors governing its scCO₂ foaming behavior. A series of PI systems with different molecular chain structures were constructed by adjusting the ratio of aromatic diisocyanate (MDI) to aliphatic diisocyanate (HDI), enabling systematic investigation of the resulting foaming behavior. Mixed diisocyanate feeds with nominal HDI fractions of 0–40% were reacted with BPADA to obtain PI copolymers, denoted as R0-R40.

As shown in Figure a, the upper part of the scheme corresponds to the condensation reaction between the raw materials, in which the combined MDI/HDI isocyanate system was introduced. The resulting copolymer structure contains BPADA-MDI-derived aromatic rigid imide units and BPADA-HDI-derived aliphatic flexible imide units, whose relative contributions are nominally regulated through the MDI/HDI feed ratio and schematically represented by m and n in Figure a. Figure b identifies the characteristic peaks corresponding to the MDI, HDI, and BPADA blocks by ¹H NMR. No ¹H NMR signal corresponding to amide groups, which would be expected near 9.44 ppm, was observed. FTIR was also used to characterize PI systems with different diisocyanate ratios. As shown in Figure c, the peaks at 1853 cm^–1 (asymmetric stretching vibration of CO), 1801 cm^–1 (symmetric stretching vibration of CO), 1459 cm^–1 (symmetric stretching vibration of C–N), and 849 cm^–1 (imide-ring vibration) confirm the presence of imide bonds in the main chain. TGA was then used to characterize the thermogravimetric behavior of PI copolymers prepared with different nominal HDI feed fractions, as shown in Figure d and e. The DTGA peak of the PI without HDI segments appeared at approximately 540 °C. The DTGA peak changed only slightly with increasing nominal HDI feed fraction, indicating that the chain-structure regulation applied in this study had little effect on PI thermal stability. The decomposition temperatures of all five samples remained above 500 °C, as shown in Figure f, demonstrating good thermal stability.

1.

(a) Schematic illustration of MDI/HDI/BPADA copolymerization; (b) ¹H NMR spectrum of the R30 sample; (c) FTIR spectra; (d) TGA curves; (e) DTGA curves; and (f) decomposition temperatures of PI copolymers with different HDI contents.

MDI contains aromatic rings, which markedly increase PI intermolecular π–π stacking, leading to dense chain packing and restricted segmental motion. By contrast, HDI contains aliphatic methylene segments. Its incorporation can disrupt dense PI-chain packing and increase both chain mobility and free volume. , As shown in Figure a, DSC measurements clearly show that the glass-transition temperature (T _g ) of PI decreased from 225.6 to 164.3 °C with increasing nominal HDI feed fraction. The PI molecular chain gradually shifted from an aromatic-rigid-structure-dominated state to a rigid-flexible synergistic structure, allowing CO₂ molecules to penetrate, dissolve, and diffuse more uniformly and efficiently.

2.

(a) DSC curves of PI copolymers with different HDI contents; (b) porosity variation of PI copolymers with different HDI contents at various foaming temperatures; (c) foaming behavior map of PI copolymers at different HDI contents and foaming temperatures; (d) representative compressive stress–strain curve of R30 PI foam; (e, f) SEM images showing the porous foam layer and magnified cell-wall morphology of the PI/CNT@Fe₃O₄ composite foam; (g–i) corresponding SEM image and EDS elemental mapping images of Fe and O for the PI/CNT@Fe₃O₄ composite foam.

As the nominal HDI feed fraction increased, the PI system exhibited markedly enhanced foaming activity under the same foaming conditions. In PI systems prepared with different nominal HDI feed fractions, foaming performance improved clearly with increasing nominal HDI feed fraction, and the foaming temperature and optimal foaming window shifted toward lower temperatures. As shown in Figure b, PI foams with porosity above 20% could be prepared when the nominal HDI feed fraction exceeded 10%. For construction of the foaming-window map in Figure c, a void fraction of 20% was used as an operational threshold: samples with VF ≥ 20% were classified as foamable, whereas those with VF < 20% were classified as nonfoamable. This threshold was used to distinguish conditions producing appreciable bulk cellular expansion from those producing only limited expansion and is not intended as a universal criterion for all polymer foams. At a nominal HDI feed fraction of 40% and a foaming temperature of 200 °C, PI copolymer foam with a porosity of approximately 68% was obtained. Notably, the porosity results showed that R30-R40 PI copolymer foams all achieved high and consistent porosity when the nominal HDI feed fraction reached approximately 30%. At this point, CO₂ dissolution and nucleation capacity in the PI system approached an upper limit, whereas further molecular-chain regulation weakened the viscoelastic support of the cell walls and limited cell-growth stability. − Thus, the results demonstrate a clear but nonexclusive association between nominal HDI feed fraction and PI foaming behavior under the investigated processing conditions, as shown in Figure c.

The compressive stress–strain curve further verifies the compressive load-bearing capability of the representative R30 PI foam. As shown in Figure d, the R30 PI foam exhibits a progressive increase in compressive stress with increasing strain, reaching approximately 0.90 MPa at 10% strain, 17.44 MPa at 50% strain, and 26.28 MPa at 60% strain. This result indicates that the optimized foam structure retains sufficient load-bearing capability under large compressive deformations, providing a mechanical basis for subsequent structural-functional integration.

Furthermore, a moderate increase in the nominal HDI feed fraction was associated with a reduction in the melt strength of the PI matrix sufficient to facilitate cell growth while maintaining structural integrity, thereby promoting a uniform and stable cellular structure. In the SEM analysis, Figure shows the cellular structures and relative pore sizes of PI foams prepared with different nominal HDI feed fractions foamed at 190 and 220 °C. At 190 °C, the average cell size of the PI foams clearly increased with increasing nominal HDI feed fraction, which is related to the lower melt strength of the PI matrix. When the nominal HDI feed fraction was too high, excessive chain softening could reduce cell-wall strength and lead to cell coalescence or collapse. − The prepared PI foams showed compact and uniform pores, with average pore sizes far smaller than those of foams prepared by chemical foaming. ,

3.

Cellular morphology and pore-structure statistics of R10, R20, R30, and R40 at the foaming temperatures of 190 °C (a–d) and 220 °C (a′–d′), together with the corresponding cell size (e) and cell density (f).

The morphology of the CNT@Fe₃O₄-containing composite foam was also examined to evaluate whether nanofiller incorporation affected foaming behavior. As shown in Figure e,f, the composite foam maintains a continuous porous structure, and the magnified cell-wall SEM image shows that nanofillers are distributed along/within the foam cell walls. The corresponding SEM/EDS mapping images in Figure g–i further reveal the spatial distributions of Fe and O elements in the selected foam region, confirming that Fe₃O₄-related components are present and distributed within the foam skeleton. These observations indicate that CNT@Fe₃O₄ incorporation does not destroy the scCO₂-induced cellular morphology and provides direct morphological and elemental evidence of filler distribution in the composite foam.

In summary, adjusting the MDI/HDI ratio enables precise regulation of the PI molecular chains, thereby effectively controlling the scCO₂ foaming behavior. This intrinsic relationship between molecular structure and foaming behavior provides an important basis for subsequent pore-structure optimization and performance regulation. Considering foamability, cellular structure integrity, and thermal properties, R30 provided the best overall balance.

3.2. CNT@Fe₃O₄ Nanofillers

CNTs have excellent electrical conductivity and high aspect ratios, allowing them to form electron-transport pathways in nanocomposites and thereby enhance conduction loss and polarization response. Fe₃O₄, as a typical magnetic component, can contribute magnetic-loss mechanisms, including natural resonance, exchange resonance, and eddy-current loss. To introduce CNT@Fe₃O₄ nanofillers with both dielectric and magnetic loss capabilities, nanoscale Fe₃O₄ was attached to the CNTs during preparation, as shown in Figure a. To confirm the crystal structure of the magnetic component, CNT@Fe₃O₄ nanofillers were characterized by XRD. As shown in Figure b, the CNT@Fe₃O₄ sample exhibited a series of distinct diffraction peaks at 2θ = 30.2°, 35.8°, 43.4°, 53.6°, 57.3°, and 63.0°. These peaks can be assigned to the (220), (311), (400), (422), (511), and (440) crystal planes of the inverse spinel structure of Fe₃O₄, indicating that the introduced magnetic component was well-crystallized Fe₃O₄. Within the detection limit of XRD, no obvious diffraction peaks from other impurity phases were observed, supporting Fe₃O₄ as the predominant crystalline phase.

4.

Structural and chemical characterization of CNT@Fe₃O₄ nanofillers: (a) schematic illustration; (b) XRD pattern; (c) XPS Fe 2p spectra; (d) TEM images; (d′) Fe element mapping; (d″) O element mapping.

XPS was further used to verify the surface chemical composition of CNT@Fe₃O₄, as shown in Figure c. Compared with pristine CNT, CNT@Fe₃O₄ showed distinct Fe 2p­(1/2) and Fe 2p­(3/2) signals, confirming the presence of Fe-containing species after coprecipitation and providing complementary chemical evidence for the introduction of Fe₃O₄ onto the CNTs. Air-atmosphere TGA further estimated the Fe₃O₄ content in CNT@Fe₃O₄ to be 32.2 wt.% (approximately 32 wt.%; Figure S6). The magnetic response of CNT@Fe₃O₄ is provided in Figure S5, where the VSM curve shows an S-shaped magnetic response at room temperature, supporting its magnetic-loss contribution. The microstructure and elemental distribution of the CNT@Fe₃O₄ nanofiller were further characterized by TEM-EDS elemental mapping, as shown in Figure d. The results demonstrate that Fe₃O₄ nanoparticles were successfully anchored on the surface of CNTs with good overall dispersion and only a minor agglomeration. The Fe₃O₄ component may contribute natural and exchange resonance in the microwave frequency range, with a possible eddy-current contribution under appropriate conditions, thereby providing magnetic loss for the nanocomposite. This magnetic loss acts synergistically with the dielectric loss contributed by CNTs and may enhance both microwave absorption and EMI shielding. Thus, CNT@Fe₃O₄ nanofillers can simultaneously provide conduction loss, interfacial polarization loss, and magnetic loss, offering multiple loss pathways for subsequently improving the microwave-absorption performance of PI-based nanocomposite foams.

3.3. Microwave-Absorption Performance of PI Copolymer Nanocomposites and Corresponding Foams

To investigate the influence of the pore structure on the electromagnetic response, the complex permittivity and complex permeability of PI foams with different porosities were measured in the X-band, and reflection loss was calculated. The results show that changes in the porosity significantly affect the dielectric parameters and final microwave absorption performance of the materials.

As shown in Figure , the unfoamed samples with nanofiller contents of 3, 4, and 5 wt.% failed to reach the effective absorption threshold at various thicknesses. As nanofiller content increased, the microwave-absorption performance of the unfoamed samples further deteriorated, which can be attributed to excessive dielectric response and impedance mismatch caused by high conductive-filler loading.

5.

Reflection loss (a–c) and impedance-matching (a′–c′) performance of unfoamed PI-based nanocomposites containing 3, 4, and 5 wt.% CNT@Fe₃O₄ at various thicknesses.

Compared with the unfoamed PI-based nanocomposites, the nanocomposite foams showed substantially improved microwave absorption performance. As shown in Figure , the reflection-loss results further indicate that PI-based nanocomposite foams with different porosities all exhibited measurable microwave-absorption performance, and the absorption peaks shifted overall toward lower frequencies as sample thickness increased. Samples with different porosities showed clear differences in absorption peak depth and effective absorption bandwidth, indicating that higher porosity is not always better and that an optimal balance exists between impedance matching and loss capability. As shown in Figure b, the nanocomposite foam with 46 vol.% porosity exhibited good microwave-absorption performance within the tested frequency range, and its reflection-loss curve showed a pronounced downward shift. At a thickness of 2 mm, the nanocomposite foam achieved a minimum reflection loss (RL _min ) of −38 dB, indicating that incident electromagnetic waves could be effectively attenuated inside the material. In summary, this nanocomposite foam improved impedance matching through its porous structure and combined CNT conduction loss, Fe₃O₄ magnetic loss, and interfacial polarization effects to achieve a strong electromagnetic-wave absorption performance.

3.4. Absorption-Dominated, Low-Reflection EMI Shielding of Double-Layer Composites

Traditional highly conductive shielding materials often rely predominantly on reflection, resulting in high reflection shielding effectiveness (SER). However, severe impedance mismatch causes a large fraction of incident electromagnetic waves to be directly reflected at the air/material interface, leading to secondary electromagnetic pollution. Based on the combined dielectric and magnetic loss advantages of the CNT@Fe₃O₄ nanofillers discussed above, this study further proposes a double-layer heterogeneous structure. The upper layer is the CNT@Fe₃O₄-containing porous absorbing layer (PI copolymer-based nanocomposite foam), and the lower layer is the CNT-containing dense reflecting layer (PI copolymer-based nanocomposite film), as shown in Figure b. This design enables a synergistic shielding mechanism involving initial low-reflection entry, strong internal dissipation, and final blocking by the reflecting layer, combining efficient shielding with reduced secondary reflection. The electromagnetic protection mechanism of this double-layer structure can be summarized as a multistage “absorption-reflection-reabsorption” dissipation process. When electromagnetic waves are incident on the material surface, they first pass through the upper porous absorbing layer. Owing to its low effective dielectric constant and favorable impedance matching, this layer substantially reduces primary interfacial reflection and allows more electromagnetic waves to enter the material. Meanwhile, CNT@Fe₃O₄ nanofillers provide conductive loss, interfacial polarization loss, and magnetic loss, while the porous structure further extends the internal propagation path and enhances multiple scattering, thereby achieving efficient first-round dissipation. The portion of electromagnetic waves that is not fully absorbed continues to propagate to the CNT-containing dense reflecting layer. Because this layer has high conductivity and strong interfacial reflection ability, it reflects the remaining electromagnetic waves back into the upper foam region. The reflected waves then undergo polarization, conduction, and magnetic-loss processes again in the upper layer and are further dissipated.

7.

Structural design and low-reflection EMI shielding performance of double-layer heterogeneous PI copolymer-based nanocomposites with different absorbing-foam-layer thicknesses: (a–d) structural schematics; (a′–d′) EMI shielding effectiveness; (a″–d″) reflection, transmission, and absorption coefficients; (a‴–d‴) A/R ratios. Source: Original artwork and experimental data created by the authors for this work.

To further clarify the influence of absorbing-layer thickness on the shielding mechanism of the double-layer structure, the upper foam thickness was adjusted to t = 3.25–3.65 mm, and EMI shielding effectiveness was measured in the X-band (8.2–12.4 GHz). Figure a–a‴ shows the shielding results at different absorbing-layer thicknesses. At an absorbing-layer thickness of 0 mm, the sample consisted only of the CNT-containing dense reflecting layer. Because this layer had high conductivity, its shielding behavior was dominated by reflection, as indicated by SER, and the average interfacial reflection coefficient R was high (average R = 0.84), indicating that a large fraction of incident electromagnetic waves was directly reflected at the air/material interface. As shown in Figure b–b‴, when the absorbing-layer thickness increased to 3.25 mm, the material showed the optimal “low-reflection absorption-type” shielding characteristic. SE _R decreased to its lowest level, whereas SE _A became dominant, indicating that nearly all incident electromagnetic waves were attenuated through absorption pathways. The corresponding average reflection coefficient was as low as 0.05, demonstrating excellent low-reflection characteristics.

In this work, the reflection coefficient (R), absorption coefficient (A), and transmission coefficient (T) were calculated, and the A/R ratio was further analyzed. As shown in Figure a–a‴, when no absorbing foam layer was present, the material showed high R and low A, indicating that shielding mainly relied on reflection. As the absorbing-layer thickness increased, R decreased rapidly, A increased significantly, and T remained extremely low. In particular, at the optimal thickness, the average absorption coefficient A reached 0.92, and the A/R ratio increased markedly (average A/R = 77.29, maximum A/R = 332.41). These results indicate that the double-layer structure achieved a shielding mechanism in which absorption greatly exceeded reflection, meaning that incident electromagnetic waves were attenuated mainly by absorptive dissipation rather than reflection. These low-reflection and high-absorption characteristics are favorable for reducing secondary electromagnetic pollution and are consistent with absorption-dominated EMI shielding strategies.

In summary, the performance of the double-layer PI copolymer-based nanocomposite arises from a synergistic regulation at three levels. At the molecular level, HDI incorporation improves the intrinsic foamability of PI; at the mesoscopic level, the porous structure optimizes impedance matching in the high nanofiller system; and at the macroscopic level, the double-layer heterogeneous structure establishes gradient propagation and cyclic dissipation pathways for electromagnetic waves.

The electromagnetic measurements in this study were limited to the X-band (8.2–12.4 GHz). Therefore, applicability to other frequency bands used in 5G/6G communications, radar, and high-frequency electronic systems requires a further evaluation.

4. Conclusions

In this study, a series of PI copolymers with tunable molecular-chain mobility were successfully synthesized by regulating the nominal MDI/HDI feed ratio, and the effects of molecular structure regulation on PI scCO₂ foaming behavior were systematically investigated. The results showed that optimizing the nominal feed ratio between aromatic and aliphatic diisocyanates was associated with improved PI molecular chain mobility, a broader foaming window, and more controllable cellular structure, yielding PI copolymer foams with high porosity and uniform cells. CNT@Fe₃O₄ nanofillers with both dielectric and magnetic loss characteristics were then introduced to enhance the electromagnetic attenuation of the nanocomposite system. The conductive network constructed by CNTs and the magnetic-loss mechanism introduced by Fe₃O₄ acted synergistically through multiple loss pathways, effectively improving the electromagnetic-wave absorption performance. The PI copolymer-based nanocomposite foam exhibited excellent microwave absorption in the X-band with a minimum reflection loss of −38 dB at a thickness of 2 mm. Finally, a double-layer EMI-shielding material with absorption-dominated and low-reflection characteristics was constructed by combining a CNT@Fe₃O₄-containing porous absorbing layer with a CNT-containing dense reflecting layer. Benefiting from the gradient attenuation effect of the layered heterogeneous structure and the synergistic effects of dielectric loss, magnetic loss, and multiple scattering, the double-layer system substantially reduced interfacial reflection while maintaining a high total shielding effectiveness. The optimized structure achieved an average absorption coefficient of 0.92 in the X-band, a maximum A/R ratio of 332.41, and an average A/R ratio of 77.29, indicating that the electromagnetic shielding process was dominated by absorption and had pronounced low-reflection characteristics.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.6c06706.

• Calculation methods for void fraction, cell size, cell density, microwave absorption, impedance matching, and EMI shielding; supplementary dielectric and magnetic parameters; reflection-loss and impedance-matching results; representative ¹H NMR spectra before and after thermal treatment; VSM and air-atmosphere TGA results; and a quantitative comparison with reported PI foams and bilayer microwave-absorbing/EMI-shielding materials (PDF)

#.

J.W. and J.Z. contributed equally to this work.

The authors declare no competing financial interest.