4,4′-Diaminodiphenyl ether (ODA, ≥ 98%) and pyromellitic dianhydride (PMDA, ≥ 99%) were purchased from Shanghai Aladdin Biochemical Technology Co., Ltd. Triethylamine (TEA, 99%), N,N-dimethylacetamide (DMAc, 99.8%) and TA (99%) were obtained from Shanghai Macklin Biochemical Technology Co., Ltd. Acetone (analytical reagent grade) and anhydrous ethanol (AR) were supplied by Hangzhou Gaojing Fine Chemical Co., Ltd. GO (≥ 99%) was purchased from Guangzhou Dazhan Nano Materials Co., Ltd. AgNW (10 mg/mL) suspension was purchased from Shanghai Ouyi Organic Optoelectronic Materials Co., Ltd.

A 15 wt% poly(amic acid) (PAA) solution was synthesized by dissolving ODA (7.4627 g) in DMAc (91 mL) under sonication, followed by the incremental addition of PMDA (7.5373 g) under continuous mechanical stirring (200 rpm) in an ice bath for 3 h. TEA (7.5373 g) was then added to yield the poly(amic acid) salt (PAS) solution. To remove residual solvents (acetone and DMAc), the product was repeatedly washed with ethanol and vacuum-dried at 40 °C for 24 h. Finally, the intermediate was mechanically fragmented into granules and subjected to a second vacuum-drying cycle at 40 °C for 24 h to ensure complete solvent removal and structural stabilization, yielding solvent-free solid PAS particles.

The solid PAS particles (synthesized in Section “Preparation of water-soluble PAS”) were dispersed in deionized water via magnetic stirring at room temperature (approximately 25 °C) at 500 rpm for 12 h to form a uniform solution, followed by the addition of GO powder (0.21 g) and a predetermined volume of AgNW suspension. After 30 min of ultrasonication and continuous stirring at room temperature (approximately 25 °C) at 500 rpm for 24 h, the resulting homogeneous PAS/GO/AgNW dispersion was cast into a mold and bidirectionally frozen using liquid nitrogen.

The frozen monoliths were then lyophilized at -50 °C for 72 h to yield the aerogel precursors. Subsequently, thermal imidization was conducted in a tube furnace (CVD-6-12TFC) under a nitrogen atmosphere. A stepwise heating program was used, with dwells at 100, 200, and 300 °C for 1 h each, with a constant ramp rate of 2 °C/min. The final aerogels were designated PGAA-0, PGAA-1, PGAA-2, and PGAA-3, corresponding to solid-phase AgNW mass fractions of 0%, 0.5%, 1%, and 1.5%, respectively.

TA (5 g) was dissolved in anhydrous ethanol (5 mL) and ultrasonicated for 30 min to initiate pre-polymerization. The as-prepared PGAA aerogel was subsequently immersed in this precursor solution and incubated at 40 °C for 4 h. This thermal treatment drove the in situ polymerization of TA across the 3D skeleton, yielding the conformal poly(thioctic acid)-coated composite aerogel (PGAA-PTA, Figure 1).

Morphological and structural characterizations were performed using scanning electron microscopy (SEM, Hitachi S-4800, Japan) at an accelerating voltage of 5 kV. Surface chemical structures were elucidated via attenuated total reflectance Fourier transform infrared (ATR-FTIR) spectroscopy (Thermo Scientific Nicolet iS50, America) from 500 to 4,000 cm^-1 over 64 scans at ambient temperature. Thermal stability was evaluated by thermogravimetric analysis (TGA, Netzsch TG 209 F1, Germany) in an air atmosphere from 25 to 800 °C at a heating rate of 10 °C min^-1.

The compressive mechanical properties of the aerogels were measured using a universal testing machine (Instron 3367, America) equipped with a 500 N load cell, applying a constant strain rate of 5 mm/min. Infrared (IR) thermal imaging (InfraTec VarioCAM HD head 620, Germany) was employed to evaluate the thermal insulation performance of 10 mm thick samples at a working distance of 30 cm.

Complex electromagnetic parameters were acquired via the waveguide method using a vector network analyzer (VNA, Keysight N5222A, America) across the X-band (8.2-12.4 GHz) and Ku-band (12.4-18.0 GHz). Samples were precisely machined to match waveguide flange specifications (23 mm × 10 mm for the X-band and 6 mm × 8 mm for the Ku-band). For strain-dependent microwave absorption assessments, the aerogels were encapsulated in paraffin wax to physically lock the applied compressive strain. For structural analysis, the ice crystal growth direction is defined as the z-axis, with the perpendicular plane designated the x-y plane. Both compression and electromagnetic testing were conducted along the x-y plane, perpendicular to the z-axis.

Shape-memory behaviors were systematically quantified via dynamic mechanical analysis (DMA, TA Instruments Q850, America) using a compression clamp. In a standard one-way shape-memory cycle, the aerogel was isothermally heated to 200 °C and equilibrated for 10 min, compressed to a target maximum strain (ε_m) of 60%, and subsequently quenched to -50 °C under a constant applied stress to fix the temporary shape. After equilibration at -50 °C, the external load was removed and the strain retained in the unloaded state was recorded as the fixed strain (ε_f). The sample was subsequently reheated from -50 to 200 °C under zero load, while the strain was continuously monitored. The strain remaining after completion of the recovery process was recorded as the residual strain (ε_r).

The DMA measurement was conducted at a maximum compressive strain of 60%. Separate isothermal shape-memory tests were conducted at 60 °C using programmed maximum strains (ε_m) of 20%, 40%, and 60%. The samples were maintained at 60 °C for 10 min before being compressed to the prescribed strain and cooled to room temperature under the applied load. After unloading, the fixed strain (ε_f) was recorded. The samples were subsequently reheated to 60 °C for 10 min under zero load, and the residual strain (ε_r) after recovery was measured. The shape-fixity ratio (R_f) and shape-recovery ratio (R_r) were calculated as R_f = (ε_f/ε_m) × 100% and R_r = [(ε_m - ε_r)/ε_m] × 100%, where ε_m is the maximum programmed compressive strain, ε_f is the strain fixed after unloading, and ε_r is the residual strain after recovery.

For the THz measurements, terahertz (THz) imaging was performed at 0.34 THz using a THz focal plane array (THz-FPA) detector (Moruixin Terahertz Technology Co., Ltd., Suzhou, China).

The morphology and structural evolution of the PGAA and PGAA-PTA aerogels were examined to clarify the roles of bidirectional freezing and in situ polymerization. As shown in Figure 2A, PGAA-0 displays a highly aligned lamellar structure along the ice-growth direction, indicating that bidirectional freezing effectively creates an anisotropic porous framework. This lamellar architecture is preserved after the incorporation of AgNWs [Figure 2B]. After in situ polymerization, PGAA-PTA still retains the aligned lamellar scaffold and interlayer bridges, while the skeleton surface is uniformly covered by a conformal PTA layer [Figure 2C]. This coating keeps the pore channels open while introducing abundant nanoscale heterointerfaces, thereby increasing interfacial heterogeneity and favoring interfacial polarization and multiple scattering of EMWs. Meanwhile, bridge-like connections gradually develop between adjacent lamellae as the AgNW content increases [Figure 2D and Supplementary Figure 1]. These features likely arise from the alignment of AgNWs during directional freezing, which reinforces coupling between neighboring lamellae and helps build continuous conductive pathways throughout the framework. Elemental mapping [Figure 2E and F] further confirms that the AgNWs remain homogeneously distributed and that the conductive network is preserved after PTA deposition. Overall, PGAA-PTA retains the anisotropic conductive scaffold of PGAA while acquiring an interface-rich PTA coating.

The structural composition and chemical evolution of the aerogels were further examined by XRD, FTIR and TGA. As shown in Figure 2G, the PI/rGO aerogel exhibits a broad diffraction feature, consistent with the predominantly amorphous nature of the PI-based framework. After incorporation of AgNWs, distinct reflections assigned to face-centered-cubic silver appear in PGAA, confirming the successful introduction of AgNWs while retaining the largely amorphous character of the framework. In the FTIR spectra [Figure 2H], PGAA-PTA exhibits distinct additional absorption bands in the 500-600 cm^-1 region, which are assigned to stretching vibrations of disulfide (-S-S-) bonds formed during in situ polymerization. Meanwhile, the intensity of the carbonyl (C=O) stretching vibration near ~1,700 cm^-1 is significantly enhanced compared with that of pristine PGAA, indicating an increased content of polar functional groups. These spectral changes provide direct evidence for the successful incorporation of PTA onto the aerogel skeleton. Moreover, the introduction of disulfide-containing polymer chains not only modifies the chemical composition of the framework but also enriches the interfacial heterogeneity and surface polarity, which are expected to facilitate interfacial polarization and thereby contribute to enhanced dielectric loss in subsequent EMW absorption performance.

The TGA curves further support this interpretation [Figure 2I]. Both PGAA and PGAA-PTA retain high thermal stability because of the PI framework, whereas PGAA-PTA exhibits an additional weight-loss stage near 300 °C, which is attributed to the cleavage of disulfide-containing moieties in PTA. These results confirm that PGAA-PTA retains an anisotropic PI/rGO/AgNW scaffold while introducing a PTA network into the aerogel framework.

The anisotropic conductive framework of PGAA is closely related to its microwave absorption behavior. Among the PGAA aerogels, PGAA-2 showed excellent overall absorption performance [Supplementary Figure 2], with a minimum reflection loss (RL_min) of -49.77 dB at 9.88 GHz and an effective absorption bandwidth (EAB) of 4.95 GHz at a matching thickness of 3.20 mm. Therefore, PGAA-2 was selected for further investigations under compressive loading [Supplementary Figure 3]. The compression results [Figure 3A-C and Supplementary Figure 4] further show that PGAA-2 maintains structural integrity under repeated deformation, indicating that the aerogel can serve as a stable platform for strain-regulated microwave absorption.

As the compressive strain increased from 0% to 60%, the electromagnetic response of PGAA-2 changed accordingly. Both the real (ε′) and imaginary (ε′′) parts of the complex permittivity increased progressively [Figure 3D], indicating enhanced charge-storage capability and dielectric dissipation in the densified conductive framework. The Cole-Cole plots in Figure 3E also reveal the presence of multiple polarization relaxation processes within the heterogeneous framework. The attenuation constant (α) increased similarly [Figure 3F], indicating enhanced electromagnetic attenuation under compression. Notably, the increase from 40% to 60% was much smaller, suggesting a diminishing contribution from further densification at high strain. At the same time, the |Z_in/Z₀| values of PGAA-2 remained within 0.5-1.5 over a broad frequency range at compressive strains of 0%, 20%, 40% and 60% [Figure 3G], indicating favorable impedance matching under different deformation states. To quantitatively evaluate impedance matching, the proportion of frequency points with |Z_in/Z₀| in the optimal range (0.8-1.2) was calculated. PGAA-2-40% exhibited the highest proportion (90.33%), significantly exceeding those of PGAA-2-0 (49.19%), PGAA-2-20% (51.32%), and PGAA-2-60% (74.44%). This result confirms that moderate compression (40%) achieves the best impedance matching. Accordingly, PGAA-2-40% achieved an RL_min of -54.34 dB at a matching thickness of 1.80 mm and an EAB of 7.84 GHz, nearly covering the entire X and Ku bands [Figure 3H-J]. These results demonstrate that appropriate compressive deformation can effectively optimize the microwave absorption performance of PGAA-2. However, the deformation-defined absorption state still requires an applied external force and cannot be retained after unloading. This limitation can be addressed by introducing thermally triggered shape-memory behavior into the aerogel framework.

The introduction of a PTA network capable of reversible disulfide exchange endowed PGAA-PTA with thermally triggered shape-memory behavior, enabling fixation of the compressed configuration and recovery of the original architecture. Figure 4A and B shows photographs and SEM images of PGAA-PTA at different compressive strains of 20%, 40% and 60%. It can be observed that the increase in compressive strain causes a decrease in the height of the PGAA-PTA samples. The corresponding SEM images show that the initially separated PGAA-PTA lamellae progressively contract with increasing locked strain, leading to a denser lamellar arrangement at higher strain. These compressed microstructures remained stable after unloading and recovered their original aligned architecture after thermal activation. To further analyze the shape memory behavior of PGAA-PTA at different compressive strains, the thermomechanical shape memory cycles were investigated, as displayed in Figure 4C and D. At programmed strains of 20%, 40% and 60%, PGAA-PTA exhibited shape-fixity ratios (R_f) of 100%, 92% and 90%, together with recovery ratios (R_r) of 98.1%, 97.5% and 93.6%, respectively. As expected, the shape recovery ratios decreased with increasing strain percentage. This can be attributed to the lamination of adjacent PTA-coated sheets. Figure 4E illustrates the shape memory route based on the dynamic covalent bonds. The presence of reversible disulfide bonds and carboxylic hydrogen bonds on the PTA-coated PGAA surfaces allows structural rearrangement via thermal dynamic bond exchanges during thermomechanical programming, locking the deformed state during cooling and reactivating recovery upon heating. The shape memory of PGAA-PTA can be realized in the following sequence. Upon exerting compressive strain on the PGAA-PTA structure, the adjacent PTA-coated surface moves and aligns to promote contact between the surfaces. When heat is applied, dynamic bond exchange reactions take place between the surfaces, and cooling deactivates the dynamic reaction, thus forming new covalent bonds and fixing the new shape under compression force. This new shape can be held for an extended period of time. Finally, heating the unloaded sample causes the new covalent bonds to break and separate the adjacent sheets due to internal stresses, causing macroscopic shape recovery.

This ability to retain strain-defined microstructures after unloading at different strains is directly reflected in the microwave absorption behavior of PGAA-PTA as shown in Figure 4F, Supplementary Figures 5 and 6. In the uncompressed state, PGAA-PTA exhibited an RL_min of -19.17 dB at a matching thickness of 5.10 mm, with an EAB of 2.33 GHz. Fixing the compressed configuration progressively improved the absorption performance. At 20% locked strain, the RL_min and EAB of PGAA-PTA-20 improved to -23.34 dB and 3.19 GHz, respectively. The best overall performance was obtained in the compressed configuration locked at 40% strain, where PGAA-PTA-40 achieved an RL_min of -72.95 dB at a reduced matching thickness of 3.20 mm and an EAB of 7.11 GHz, surpassing many previously reported absorbers [Supplementary Table 1]. Further increasing the locked strain to 60% resulted in a less negative RL_min value of −37.95 dB and reduced the EAB to 4.71 GHz, indicating that excessive densification became unfavorable for absorption.

This strain-dependent evolution in microwave absorption is governed by the balance between attenuation capability and impedance matching in the locked state. As illustrated in Figure 5, the aligned lamellar architecture, together with the AgNW bridges, prolongs the propagation path of incident EMWs by promoting multiple internal reflections and scattering within the framework. Locking the compressed configuration further densifies the interconnected PGAA-PTA network, thereby strengthening conduction loss through more efficient charge transport. Meanwhile, the abundant heterogeneous interfaces among PTA, rGO, AgNWs and PI promote interfacial polarization, while residual polar groups and structural defects contribute additional dipolar relaxation. These processes together enhance dielectric dissipation in the locked state [Supplementary Figure 7]. With further densification, attenuation is strengthened, but once the conductive framework becomes overly compact, the resulting impedance mismatch begins to offset this gain. Consequently, the best overall absorption performance is obtained in the compressed configuration locked at 40% strain, where attenuation is sufficiently enhanced without causing severe mismatch. These results demonstrate that the PTA network stabilizes the compression-defined state and enables its reversible microwave absorption response when exposed to heat.

The electromagnetic response of PGAA-PTA was further examined in the THz regime to evaluate its potential for high-frequency electromagnetic protection. As shown in Figure 6A, PGAA-PTA effectively suppresses the transmitted THz signal, whereas the uncoated substrate shows a distinct propagation pattern. This contrast indicates that PGAA-PTA can effectively attenuate incident EMWs even at higher frequencies. Such high-frequency attenuation capability also implies the potential for regulating electromagnetic scattering.

To assess this possibility, radar cross-section (RCS) simulations were carried out at the optimal absorption frequencies corresponding to the 0% and 40% strain states, representing the X-band and Ku-band conditions, respectively. As shown in Figure 6B and C, both PGAA-PTA-0 and PGAA-PTA-40 exhibit lower RCS values than perfect electric conductor (PEC) across most observation angles in the X-band and Ku-band ranges, confirming their ability to suppress electromagnetic scattering. Compared with PGAA-PTA-0, PGAA-PTA-40 shows a further reduction in scattering intensity, indicating that the locked compressed state is more effective in lowering radar signatures. The 3D scattering patterns in Figure 6D₁-D₄ further support this result. In contrast to the relatively broad and intense lobes observed for PEC in Figure 6E₁-E₄, PGAA-PTA shows more compact scattering patterns with reduced peak intensity. This difference becomes more pronounced in the locked state, consistent with its stronger electromagnetic attenuation and improved scattering suppression.

The oblique-incidence response of the PGAA-PTA aerogels was further evaluated using CST simulations at incidence angles of 15° and 45° in the X and Ku bands [Supplementary Figures 8 and 9]. The PGAA-PTA aerogels under compression strains of 0%, 20%, 40%, and 60% were individually benchmarked against a PEC plate with identical dimensions. At both investigated incidence angles, PEC produced a pronounced specular-scattering lobe and a high RCS near the dominant scattering direction. In contrast, each PGAA-PTA sample exhibited a weaker principal scattering lobe and a lower RCS than PEC. This reduction is associated with electromagnetic dissipation within the conductive and polarizable networks, as well as repeated reflection and scattering within the porous structure, which decrease the energy returned to the far field. These results demonstrate that PGAA-PTA maintains effective scattering suppression relative to PEC under the investigated compression states and at the selected incidence angles of 15° and 45°.

Beyond electromagnetic attenuation, the anisotropic porous architecture of the aerogel also contributes to thermal insulation. As shown in Figure 7A, the surface temperatures of both PGAA-x and PGAA-z increased with heating time and gradually stabilized, while PGAA-z consistently exhibited a higher surface temperature than PGAA-x. Correspondingly, Figure 7B shows that PGAA-x maintained a larger ΔT over the entire test period, indicating a stronger thermal-barrier effect. The IR thermal images and temperature profiles in Figure 7C further support this result. This anisotropic thermal-insulation behavior arises from the orientation of the aligned porous framework [Figure 7D]. In the PGAA-x mode, the layered solid/gas structure impedes heat transport across the thickness, whereas in the PGAA-z mode, the aligned channels facilitate heat propagation along the heating direction. These results show that the aerogel combines directional thermal insulation with tunable electromagnetic functionality, which is beneficial for multifunctional protection applications.