Bio-inspired cellular aerogel fibers integrating high mechanical strength and softness for thermal insulation textiles

Abstract

Integrating high mechanical strength yet softness and effective thermal insulation into the same aerogel materials presents a significant challenge. Inspired by penguin feathers, here we assemble aramid nanofibers (ANFs) into aerogel fibers of hierarchical structures through delicate control of covalent and non-covalent interactions during a wet-spinning process. The process involves initial cross-linking of deprotonated ANF sol to form a cellular structure, followed by acid-induced gelation that produces a rigid shell through hydrogen bonding. The shell imparts high tensile strength of up to 74.6 MPa, while the cellular core enables good softness with ultralow bending and compression stresses of 33.8 and 39.8 kPa, respectively. The process is scalable, and allows fabrication of large fabrics with dyeability, hydrophobicity, flame retardancy, moisture and chemical resistances. Notably, the fabrics exhibit good thermal insulation, with a 0.9 mm-thick sample outperforming much thicker commercial counterparts, including a 2.5 mm sweater and a 15 mm jacket.

Aerogels are promising for thermal insulation materials, though often have unfavorable mechanical properties. Here the authors design an aramid nanofiber aerogel to optimize mechanical and insulative properties.

Introduction

Since the dawn of civilization, humans have recognized the importance of thermal insulation, as reflected in the development of warm clothing and insulated dwellings^1–3. Over time, the advancement of society has driven continuous innovation in insulation materials—from natural resources such as animal hides, wool, and straw, to synthetic foams in the late 20th century, and more recently to advanced materials like vacuum insulation panels and aerogels^4–6. Among these, aerogels stand out for their ultralow density and high porosity, which confer exceptionally low thermal conductivity^7–9. However, these structural features unfortunately compromise the mechanical strength of the aerogels, posing significant challenges to their practical implementation^4,10,11.

Many are the strategies that try to address this fundamental challenge, including chemical cross-linking^12–14, surface coating^4,15, blending with mechanically robust materials^16,17, and incorporating phase-separated or gradient nanostructures^10,18. Several notable studies have successfully demonstrated the potential to overcome the intrinsic trade-off between mechanical strength and thermal insulation^4,10,11. Nevertheless, improvements in tensile strength are often achieved at the cost of softness, which remains a critical limitation for applications requiring high conformity and comfort. Clearly, softness—alongside thermal protection and tensile strength—constitutes a key performance parameter across a wide range of fields, such as textiles, vehicle interiors, sports equipment, transportation, among others^19,20. Of particular importance, therefore, are the approaches that can simultaneously enhance all three properties, enabling aerogels to be more effectively deployed in those fields.

With this goal in mind, penguin feathers come to our sight due to their critical roles in protecting penguins from the cold, harsh environment through resisting external impact, maintaining warmth and comfort. Examination of their microstructure reveals a cellular core with multi-scale pore sizes surrounded by a rigid sheath composed of densely packed keratins (Fig. 1a and Supplementary Fig. 1)^21,22. The sheath provides high tensile strength, whereas the cellular core leads to uneven stress distribution and renders the structure susceptible to bending and compression^23,24. Inspired by this natural design, we employed aramid nanofibers (ANFs) as the building blocks to replicate such an architecture. The choice of ANFs is rooted in their special structural characteristics; They exhibit highly ordered and aligned arrangements induced by π–π stacking and hydrogen bonding interactions among the repeat units²⁵. Upon deprotonation with alkali, the amide nitrogens are converted into amide anions, introducing electrostatic repulsion that disrupts the ordered structure^26,27. Interestingly, reprotonation with a strong acid rapidly restores interchain hydrogen bonding, enabling ANFs to realign into stacked structures, whereas water stabilizes the disordered state into porous networks^28,29. The divergent reprotonation behaviors offer a platform to construct core‒shell structures as in penguin feathers. However, the resulting pores are typically in a normal distribution of sizes, due to their thermodynamically favored nature in the water-induced reprotonation process, which potentially diminishes the softness of aerogels.

Fig. 1. Design and fabrication of bioinspired cellular aerogel fibers.

a Schematic illustration of hierarchical structures present in penguin feathers and our bioinspired cellular aerogel fibers (CAFs). b Schematic illustrations of a wet-spinning process to produce CAFs and the associated self-assemblies of the mixture of ANFs and dibromo cross-linkers in both acid and water.

To address this issue, we sought to re-establish the thermodynamic equilibrium that governs the self-assembly of deprotonated ANFs in water. Specifically, small amounts of dibromo compounds were introduced to react with amide anions, generating cross-links that applied pulling forces on reacted regions, while unreacted segments continued to repel each other. To our delight, this cross-linking process alters the assembly behavior of deprotonated ANFs and produces a cellular structure, with the cell wall consisting of nanoscale pores (Fig. 1b). This finding soon sparks our interest of leveraging the divergent yet controllable assembly of ANFs for the fabrication of technologically relevant materials—specifically, aerogel fibers that demand high strength, softness, and good thermal insulation (Fig. 1a). The preparation relies on an adapted wet-spinning process: Kevlar fibers were first deprotonated to form an ANF sol, into which dibromo cross-linkers were incorporated. The spinning sol was then passed through sequential coagulation baths—initially in acid to form a rigid shell, followed by water treatment to generate a cellular core (Fig. 1b). The hierarchical structure endows the cellular aerogel fibers (CAFs) with high tensile strength and ultralow bending and compression stresses. Moreover, we show that the sweaters woven from these fibers exhibit superior thermal insulation performance compared to a thicker commercial sweater and even a jacket.

Results

Controlled self-assembly of ANFs

We set out to investigate the self-assembly behavior of ANFs. First, following a well-established protocol^25,30, commercial Kevlar fibers were exfoliated through the deprotonation using an alkali mixture of potassium hydroxide (KOH), DMSO, and H₂O at a ratio of 3.75/100/4 (w/v/v). Scanning electron microscopy (SEM) analysis reveals a clear morphological evolution during the deprotonation process. After 5 minutes, fluff-like fibrils begin to appear on the surface of the staple fibers (ca. 62.7 nm in diameter), gradually covering the entire surface within 10 min (Supplementary Fig. 2a‒c). Upon an additional 15 min of treatment, the original staple fibers completely disappear and transform into a fibrous network consisting of ANFs with diameters of 4.4–31.8 nm and lengths of up to 5–10 μm (Supplementary Fig. 2d, e)³⁰. This morphology is retained when extending the treatment time to 35 min (Supplementary Fig. 2f). Concurrently, the surface charge of the fibers shifts from weakly positive to negative, as determined by Zeta Potential measurements (Supplementary Fig. 3a). These results indicate that deprotonation introduces negative charges onto the fiber surface, which trigger exfoliation through electrostatic repulsion^27,28,31. Upon reprotonation in water, the negatively charged ANFs assemble into a homogeneous porous structure (HS) with a normal pore size distribution, as confirmed by SEM (Supplementary Fig. 3b, c), a finding that mirrors the results reported in most works ^29,32,33.

The negative charges of deprotonated ANFs arise from amide anions, which are reactive toward electrophiles such as halogenated compounds^34,35. To modulate their assembly behavior in water, a small amount of 1,6-dibromohexane (DBH) in 10 mol% relative to the aramid repeat unit was introduced as a cross-linker. The reaction progress was monitored by in situ Fourier transform infrared spectroscopy (FTIR). Upon addition of DBH, two new vibration bands appear at 2924 and 1246 cm⁻¹ in 30 s, ascribed to the stretching of ‒CH₂‒ and C‒N, respectively. The intensities of these bands increase steadily and plateau after 120 s (Fig. 2a), confirming the rapid formation of alkyl–nitrogen linkages. This fast reaction kinetics is further reflected in the sol–gel transition of the ANFs/DBH mixture, which occurs within 4 min (Supplementary Fig. 4). After different reaction times, the cross-linked ANFs were reprotonated by water, followed by freeze drying to capture their structures, with the morphologies analyzed by SEM technique. It reveals that ANFs initially exhibit a nanoscale HS as in the case without the addition of cross-linker (Supplementary Fig. 5a). After 30 s of reaction, sporadically distributed micropores emerge within the nanofibrous networks, which eventually develop into a cellular structure (CS) composed of micropore-dominated networks with nanoporous walls (Fig. 2b and Supplementary Fig. 5). The morphology change is accompanied by the decrease in surface charge over time, determined by Zeta Potential measurements (Fig. 2c), suggesting the consumption of amide anions in the reaction with DBH during the assembly process (Fig. 1b).

Fig. 2. Dynamic cross-linking and self-assembly structure evolution of ANFs.

a In situ FTIR spectra, b SEM images, and c zeta potential of the mixture of ANFs and DBH after different reaction times (as indicated). d Coarse-grained molecular dynamics simulation on the assembly behavior between ANFs and DBH, with the snapshots taken in the initial (left), 2 million (middle), and 9 million (right) steps. e In situ FTIR spectra, f SEM images, and g zeta potential of the cross-linked mixture of ANFs and DBH assembled in formic acid for different times (as indicated). h 2D-WAXS patterns and i the calculated degree of orientation of LS, HS, and CS.

The cross-linking-induced CS was further studied by coarse-grained molecular dynamics simulation (Details seen in Supplementary Information, Fig. 2d and Supplementary Fig. 6). ANFs were described using a Bead‒Spring model, each comprising 200 chains of 42 beads, with a reactive bead density of 26.2%. Both the solvent and DBH cross-linker were simplified as single beads, given their relatively shorter chain lengths compared to ANFs. As shown in Fig. 2d (left), the deprotonated ANFs are initially distributed randomly due to electrostatic repulsion. Upon the introduction of DBH beads, they first attach to the reactive beads on ANFs, subsequently pulling proximal chains together. This induces localized aggregation, while leading to the formation of voids in regions with fewer reactions (Fig. 2d, middle). With extended simulation time, the spatial heterogeneity becomes more pronounced, characterized by dense clustered regions interspersed with sparse large voids, ultimately giving rise to a cellular structure (Fig. 2d, right). These results provide theoretical support for the pivotal role of DBH in directing ANF assembly in aqueous environments to form such a special structure.

We next explored the self-assembly kinetics of cross-linked ANFs in acidic conditions. About 1 mL of the sol mixture of ANFs and DBH, obtained after 2 min of reaction, was introduced with 1 mL of 40 vol% formic acid on an infrared spectrometer, with the spectra collected on site. FTIR spectra show that the ‒CH₂‒ band at 2924 cm⁻¹ remains nearly unchanged, whereas the C=O stretching vibrations and N–H bending vibrations at 1645 and 1509 cm⁻¹ shift to lower and higher wavenumbers, respectively (Fig. 2e), indicating the formation of hydrogen bonds³⁶. The sol mixture of ANFs and DBH was also immersed in an excess of 40 vol% formic acid. At different time intervals, the resulting gels were taken out, washed with water, and dried for SEM measurement. It displays that the outmost ANFs start to align and stack parallel to the surface direction within 2 s, which transforms to a well-ordered, layer-by-layer architecture in 30 s (Fig. 2f). Prolonging the immersing time to 10 min does not further alter the morphology (Supplementary Fig. 7). Meanwhile, the surface charges of the ANFs first see a decrease in the negative value from −17 to −9 within 10 s, and turn into positive after 30 s of exposure to the acid (Fig. 2g).

The microstructures of the three assembled architectures including HS, CS, and layered structure (LS) were studied by wide- and small-angle X-ray scattering (SAXS/WAXS) analysis. The 2D-WAXS patterns of all samples exhibit a circular shape, indicating the presence of crystallites (Fig. 2h). Among them, LS displays significantly higher scattering intensity in both the 2D-WAXS pattern and 1D profile compared to HS and CS (Fig. 2h and Supplementary Fig. 8a), confirming its highest crystallinity. The intensity of the circular patterns of HS and CS does not show a discernible difference along their respective azimuth circle (Supplementary Fig. 8b), reflecting the highly isotropic orientation of the crystalline ANFs. In contrast, LS exhibits two distinct arcs at the top and bottom of the circle (Fig. 2h), demonstrating alignment of ANFs along that direction. Quantitative analysis of the 2D patterns by extracting azimuthal profile (Supplementary Fig. 8c), fitting intensity against angle, and applying Hermans orientation function allows us to calculate the degree of orientation (Fig. 2i)³⁷. The results show that LS has the largest orientation degree of 55%, which resonates well with its highly aligned and layered architecture observed by SEM (Fig. 2f). This is then followed by 30% of HS, and 11% of CS (Fig. 2i). The low value for CS is primarily attributed to the random distribution of ANFs in it. Comparison of the SAXS data further shows that CS exhibits weaker scattering intensity than HS, probably due to the looser packing of ANFs within its network (Supplementary Fig. 8d).

Altogether, the above results demonstrate that ANFs can undergo diversified self-assembly processes, giving rise to structures that range from isotropic to highly ordered, and from porous to densely stacked. The resulting morphologies can be precisely tuned by varying reaction time, selecting the assembly solvent, and introducing cross-linkers as needed. Such kinetic and thermodynamic control provides a versatile platform for exploiting ANFs as building blocks to construct structurally sophisticated and functional materials.

Fabrication and characterization of cellular aerogel fibers and fabrics

Having elucidated the self-assembly characteristics of ANFs, we next sought to translate these behaviors into the fabrication of technologically relevant materials. Of particular interest is the cellular structure, which is expected to impart high softness to the resulting materials, similar to that in penguin feathers (Fig. 1a). Considering the high softness intrinsically required for textile applications, we here develop CAFs that are mechanically robust and yet soft, while also offering desirable thermal insulation properties.

Prior to fiber spinning, the rheological properties of the deprotonated ANF sol were measured (Fig. 3a). Initially, a 10 mg mL⁻¹ sol exhibits a steady fluid-like feature, with the loss modulus (G”) exceeding the storage modulus (G’). Upon gentle injection of DBH at 225 s, the mixture retains the fluid state for ca. 125 s, during which G’ sees a rapid increase and then intersects with G” (Fig. 3a), marking the sol‒gel transition. This crossover point roughly coincides with the onset of a sharp rise in viscosity (η*). This is then followed by the attainment of a plateau for both G’ and η* in 130 s of cross-linking. The finding teaches us to spin the ANF sol within this time window, namely, no later than 130 s of its mixture with DBH. Further addition of 40 vol% formic acid to the ANF gel intensifies its solidification, inferred from the simultaneous increases in G’, G”, and η* (Fig. 3a), confirming the significance of acid in enhancing the rigidity of the ANF gel.

Fig. 3. Fabrication, structure, and textile performance of CAFs.

a Time sweeps of the ANFs sol upon the sequential additions of DBH at 225 s and 40 vol% formic acids at 500 s. SEM images of the (b) cross-section and core zoom-in (c, d) at different magnifications of a CAF. e SEM images of knotted and twisted CAFs. Photographs showing (f) the dyed aerogel fibers, g a fabric produced on a loom, h a freestanding, and i a kneaded fabric.

The insights gained from the chemical and rheological studies guided us to slightly adapt the industrially relevant wet-spinning process for the fabrication of ANF-based aerogel fibers (Fig. 1b). The process began with the exfoliation of Kevlar fibers into ANF sol using KOH in a DMSO–H₂O mixture (25/1, v/v) (Supplementary Fig. 9). The resulting sol was mixed with DBH (10 mol% relative to the aramid repeat unit) for 2 min, and then extruded through an 800-µm-diameter spinneret. The extruded fibers were subsequently passed through two sequential coagulation baths, consisting of formic acid and deionized water. These two baths, as explored above, induce distinct self-assembly pathways of ANFs (Fig. 2b, f), and by controlling the corresponding coagulation times, it should allow to create a core‒shell architecture, in which ANFs adopt differentiated structural organizations. We show that this is indeed the case, and following the solvent exchange and freeze drying, core‒shell structured aerogel fibers are formed with a diameter of 504.4‒741.2 µm (Fig. 3b‒d). Analysis of the cross-section of the fiber reveals that the shell, formed in the acid bath, displays a densely packed architecture, where ANFs are stacked in a highly ordered and layer-by-layer pattern with an interlayer spacing of ca. 181.6 nm (Supplementary Fig. 10a, b). By contrast, the core is in a cellular structure comprising ca. 2.1 µm large pores with their walls constituted by nanoscale porous networks of ANFs (Fig. 3c, d and Supplementary Fig. 10c, d). These morphological characteristics of CAFs closely resemble those of penguin feathers (Supplementary Fig. 1).

The CAFs thus produced exhibit good flexibility, enabling knotting, twisting, and bending free of damage (Fig. 3e and Supplementary Fig. 11a). Moreover, a single CAF with a diameter of ~700 µm can support a 500 g weight without fracture (Supplementary Fig. 11b). The gel fibers are readily colorable using both cationic and anionic dyes, such as methylene blue, methyl orange, and acid green, imparting diverse colors to the final CAFs (Fig. 3f). The chemical stability of the aerogel fibers was evaluated by immersion in various organic solvents including DMSO, THF, DMF, acetone, DCM, and DMAc for 3 months. In all cases, the fibers retain their structural dimensions, demonstrating strong resistance against these solvents (Supplementary Fig. 12). Even in the ANF-solubilizing KOH/DMSO–H₂O mixture, the fibers do not fully dissolve thanks to their covalently cross-linked network (Supplementary Fig. 12). These results support that our aerogel fibers combine high mechanical robustness, chemical stability, flexibility, and colorability, which may serve as promising raw materials for weaving into durable, customizable, and esthetically versatile fabrics.

The weavability of the aerogel fibers was next evaluated using an automatic rapier loom. The fibers demonstrate good processing performance, enabling the successful production of a large fabric with a dimension of 1.5 × 0.45 m (l × w) and a surface density 37.0 g m⁻² (Fig. 3g) without any detectable abrasion or damage on either surface (Supplementary Fig. 13a‒c). Further analyses show that the fabrics exhibit several compelling features. First, they are hydrophobic (Supplementary Fig. 13d, e), which confers good stain resistance, preventing contamination from common household liquids such as dyed water, milk, cola, and tea (Supplementary Fig. 14). The fabrics can withstand high temperature, showing a high 5% decomposition temperature of 546 °C and a char ratio of 45% at 700 °C, as measured by thermogravimetric analysis (Supplementary Fig. 15a). This characteristic further allows a fabric to stay above a flame created by an alcohol lamp for 3 s without burning and significant scorching (Supplementary Fig. 15b, c). Moreover, the fabrics can be easily molded and shaped by hand (Fig. 3h, i and Supplementary Movie 1), suggesting the good flexibility and softness inherited from the constituent CAFs.

We next assessed the durability of the CAFs fabrics under harsh conditions, including exposure to high relative humidity, repeated laundering, and UV irradiation. Upon exposure to 100%RH for 24 h, no discernible damage was found in both the fabric surface and the internal structure, as evidenced by optical microscopy and SEM analyses (Supplementary Fig. 16a, c–e), and the tensile behavior is comparable to that of the untreated sample (Supplementary Fig. 16b). These results support the strong moisture-resistance of CAFs. Aerogels are notoriously vulnerable to repeated washing^4,38. To avoid this issue in the CAFs, we first treated a CAFs fabric with poly(dimethylsiloxane) (PDMS) for water repellency—a commonly adopted treatment in improving the washability of textiles—³⁹, before washing it at a speed of 1000 r min⁻¹ for 35 min, with the washing process repeated three times (Supplementary Fig. 17a). To our delight, such the treatment confers good washability on the CAFs fabric, allowing it to preserve its morphology and mechanical properties successfully (Supplementary Fig. 17b–k). The CAFs fabric was also subjected to UV aging analysis. It was irradiated at a wavelength of 340 nm with an intensity of 850 μW cm⁻². Following the treatment for 12 h, the fabric turns from a light yellow into a brownish yellow (Supplementary Fig. 18a, b). Concurrently, the mechanical properties see significant decreases, including 77 and 50% losses in stress and strain at break, respectively (Supplementary Fig. 18c), even though the internal structural morphology remains intact (Supplementary Fig. 18d–f). The performance loss is primarily attributed to the photo-oxidation of amide bonds into carboxylic acids and aldehydes, causing extensive polymer chain scissions⁴⁰. The issue could be mitigated by coating tinuvin, a UV absorber, onto the fabric surface, as reflected by the tiny variations in fabric color and tensile strength after the 12h-UV irradiation (Supplementary Fig. 18g–i).

Tunable pore sizes and mechanical properties of cellular aerogel fibers

Using the adapted wet-spinning approach, a series of cellular aerogel fibers (CAF-xD) were fabricated by varying either the amount or molecular length of the cross-linker. Specifically, DBH in amounts of 6, 10, and 14 mol% relative to the aramid repeat unit, or cross-linkers with different carbon spacings including 4 (1,4-dibromobutane, DBB), 6 (1,6-dibromohexane, DBH), 8 (1,8-dibromooctane, DBO), and 10 (1,10-dibromodecane, DBD) carbons were employed. In CAF-xD, the parameter x denotes the DBH amount, and “D” represents the cross-linker type.

For comparison, homogeneous porous aerogel fibers (HPAFs) were prepared following the same spinning protocol but without the addition of any cross-linker to the ANF sol. The resulting HPAFs also show a core‒shell structure, with the shell similar to that of CAFs (Supplementary Fig. 19a, b). The inner core consists of a 3D homogeneous porous ANF network (Supplementary Fig. 19c), with pore sizes comparable to the nanoscale pores observed in the walls of CAF-10DBH, as confirmed by Brunauer–Emmett–Teller (BET) measurements (Supplementary Fig. 20). Introduction of a DBH cross-linker transforms the inner core into a cellular structure comprising macropores of 0.7–6.1 µm and nanopores of 10.3–464.8 nm (Supplementary Fig. 21). Increasing the amount of DBH from 6 mol% to 10 mol% leads to an increase in the average sizes of macropores from 1.8 to 2.3 µm (Supplementary Fig. 21d, e), but a decrease in nanopore sizes from 137.2 to 101.3 nm (Supplementary Fig. 21g, h). As the DBH amount reaches 14 mol%, the macropore sizes further increase to 2.7 µm, while the nanopores shrink to a level of 76.0 nm (Supplementary Figs. 21f, i). The length of the cross-linker also influences pore architecture: increasing the chain length from 4 to 10 carbons while keeping the amount as 14 mol% causes an initial increase followed by a decrease in the size of both macropores and nanopores (Supplementary Fig. 22). Among them, CAF-14BDH exhibits the largest macropore size of 2.7 µm, while CAF-14DBO was the largest in nanopore size with an average diameter of 104.1 nm. (Supplementary Fig. 22e, i). These findings show that the pore sizes of CAFs can be systematically tuned by controlling the cross-link density and distance.

The mechanical properties of HPAFs and CAFs were first determined by axial tensile testing. The stress‒strain curves show that HPAFs exhibit stress and strain at break of 57.1 MPa and 27.4%, respectively (Fig. 4a). Upon the introduction of DBH, both the stress and strain at break increase to 60.5 MPa, 31.8% of CAF-6DBH, which see further increases to 74.6 MPa, 38.1% as the amount increases to 14 mol% (Fig. 4a and Supplementary Fig. 23a). This is an impressive finding, since increasing cross-link density typically increases the strength of the materials but at the expense of the extensibility^41–44. We attribute this enhanced elongation to the presence of larger macropores in the highly cross-linked materials (Supplementary Fig. 21d–f). Elimination of the acid coagulation treatment during wet spinning produces aerogel fibers lacking the shell (Supplementary Fig. 23b), which exhibit a considerably lower stress at break of 22.2 MPa (Supplementary Fig. 23c), highlighting the reinforcing contribution of the shell to CAFs. Within the CAF series, optical microscopy analysis reveals that they have a shell thickness of 56–65 μm, with the shell-to-diameter ratio lying in a narrow range of 7.4–8.6% (Supplementary Fig. 24 and Supplementary Table 1), ruling out the probability of the variation in shell thickness to influence the mechanical properties of CAFs. In contrast to the increased strength and extensibility, Young’s modulus displays an inverse trend: the most rigid fibers are HPAFs, with a modulus of 0.78 GPa, which gradually decreases to 0.57 GPa for CAF-6DBH, 0.50 GPa for CAF-10DBH, and 0.32 GPa for CAF-14DBH (Supplementary Fig. 23a, d). It suggests the increased softness present in the aerogel fibers featuring the larger pore sizes.

Fig. 4. Mechanical properties and structural flexibility mechanisms of CAFs.

a Axial tensile and b bending stress–strain curves of HPAFs and CAF-14DBH, with the maximum bending strain of 70%. c Cyclic bending curve of CAF-14DBH at a maximum strain of 70% for 100 consecutive cycles. d Compression stress–strain curves of HPAFs and CAF-14DBH, measured at increased maximum compression strain from 2.5 to 50% (as indicated), with five consecutive cycles performed at each strain. e Cyclic compression curves of CAF-14DBH at a 25% compression strain for 100 consecutive cycles. f Comparisons of tensile and compression stresses between CAF-14DBH and other reported aerogel materials. Simulated stress distribution images (g, i) and the stress distribution along the selected line (h, j) for the cellular (g, h), and homogeneous (i, j) structures before and after being compressed to a 30% strain.

The softness of HPAFs and CAFs was further assessed through axial bending and radial compression tests. Indeed, as the DBH amount increases, both the maximum bending and compression stresses decrease, from 162.0 and 81.6 kPa of CAF-6DBH to 33.8 and 39.8 kPa of CAF-14DBH, respectively (Supplementary Fig. 25a). Notably, the bending stress of 35.4 kPa seen in CAF-14DBH is significantly lower than that (350.6 kPa) of HPAFs (Fig. 4b), reflecting the importance of the cellular structure in softening the aerogel fibers. The robustness of the bending behavior was examined over 100 cycles by axially compressing CAF-14DBH to a maximum strain of 70% (Supplementary Fig. 25b). The maximum bending stress exhibits slight fluctuations during the initial ten cycles, then stabilizes at ~38 kPa (Fig. 4c). For radial compression, the applied strain was progressively increased from 2.5 to 50%, with five cycles performed at each strain. The maximum compression stresses of both HPAFs and CAF-14DBH show gradual increases from 15.8 and 2.3 kPa at 2.5% strain to 268.8 and 148.9 kPa at 50% strain, respectively. Across the whole strains tested, the corresponding maximum compression stresses of CAF-14DBH are consistently lower than those of HPAFs. This discrepancy is more pronounced in the low compression regime, as HPAFs show a logarithmic increase, while CAF-14DBH exhibits a nearly linear response (Fig. 4d). This indicates that CAF-14DBH exhibits lower resistance to deformation under external compression, particularly at the low amplitude. Furthermore, repeated compression–release tests at 25% strain over 100 cycles demonstrate that the maximum compression stress of CAF-14DBH remains low and stabilizes at 42.0 kPa (Fig. 4e and Supplementary Fig. 25c). After the cyclic compression testing, the fiber preserves its microstructure, as evidenced by SEM analysis (Supplementary Fig. 25e, f). These experiments jointly confirm that the soft feature is stable and the materials are resilient under repeated compressions.

Keeping the amount of cross-linkers constant at 14 mol% while increasing their chain length also has an impact on the softness, but only moderately. The four CAF-14D variants exhibit both low bending and compression stresses of 33.8‒49.2 kPa at 70% strain and 39.8‒58.0 kPa at 25% strain, respectively (Supplementary Fig. 25d). Among them, CAF-14DBH are most susceptible to both bending and compression (Supplementary Fig. 25d), probably due to its largest macropore size (Supplementary Fig. 22e). The above results demonstrate that the joint presence of a dense shell and a cellular core enables the decoupling of high softness from high tensile strength, allowing both properties—typically conflicting—to be enhanced simultaneously. This compelling finding further encourages us to benchmark our CAFs against previously reported (fibrous) aerogels in terms of axial tensile strength and transverse compression stress (Fig. 4f). Clearly, the tensile strength of CAF-14DBH exceeds those of most other aerogels regardless of the material type (e.g., polyimides, ceramics, cellulose, aramid) or structural form (fibers, films, and monoliths)^{11,17,45–57}, except one case reported by Liu, Fan, and coworkers¹¹. Their polyimide aerogel fibers show impressive stress at break of up to 93.1 MPa, but this trades off the materials’ softness, displaying a compression stress of ca. 2 MPa at 25% strain, which is over 45-fold that (42 kPa) of our CAF-14DBH¹¹. The 42 kPa compression stress, to our knowledge, is among the lowest reported for aerogel materials (Fig. 4f) ^{11,17,45–57}.

We next employed Finite Element Simulations to elucidate the underlying mechanism behind the high softness of CAFs. The cellular structure, based on SEM observations (Fig. 2c), was modeled to comprise two distinct pore scales. Under compression, the larger pores deform preferentially at strains below 30% (Fig. 4g and Supplementary Fig. 26a), with the stress distribution along a selected line showing pronounced heterogeneity (Fig. 4h). At 40% strain, the large pores undergo significant distortion, while smaller pores begin to deform, progressively shrinking and ultimately flattening at 70% strain (Supplementary Fig. 26a). These simulated deformation behaviors of the cellular structure were experimentally supported by SEM measurements that show an initial shrinkage in macropore at the compression strain of 30%, followed by its collapse at 70% strain and the nanopore deformation (Supplementary Fig. 26b). By contrast, the homogeneous porous structure of HPAFs, simulated to possess only small pores, exhibits a more uniform load-bearing behavior under both low and high compression strains (Fig. 4i and Supplementary Fig. 27). Stress along the selected line is distributed relatively evenly, with peak values of approximately 3.5 × 10⁻³ kPa (Fig. 4j)⁵⁸. This uniform stress distribution renders HPAFs less susceptible to deformation, thus correlating with their higher rigidity.

Thermal insulation performance of CAFs

The cellular core of CAFs is porous, which should, in principle, provide good thermal insulation to the materials. To verify this, we first used an infrared thermal camera to observe the surface temperature of CAFs placed on a 200 °C hot stage. They exhibit an obvious temperature gradient between the shell and core, with the inner core maintaining a lower temperature (Fig. 5a), indicating the good thermal insulation. Quantitative analysis of the temperature difference |ΔT| between the hot stage and the fiber core reveals that the |ΔT| is the lowest at 13.8 °C in CAF-14DBH, while it is the highest for CAFs-10DBH (19.8 °C) (Fig. 5a, b). This trend is also relevant upon decreasing the measurement temperature to 150 and 100 °C (Fig. 5b). Moreover, CAFs exhibit a thermal conductivity (λ) of 28.4‒32.1 mW m⁻¹ K⁻¹ (Fig. 5c). Among them, CAF-10DBH has the lowest value, and such a superior performance is probably attributed to its balanced macro- and nano-pore sizes, both in the medium level compared to CAF-6DBH and CAF-14DBH (Supplementary Fig. 21). Moreover, as the relative humidity increases, the λ of CAF-10DBH only slightly increases from 28.4 mW m⁻¹ K⁻¹ at 30%RH to 29.7 mW m⁻¹ K⁻¹ at 90%RH (Supplementary Fig. 28). The good thermal insulation of CAFs is inherited to their woven fabrics, as confirmed by a large temperature difference of 72.8 °C observed in the CAF-10%DBH-woven fabric with a thickness of 0.9 mm on the 200 °C hot stage (Fig. 5d). Under the same conditions, commercially available materials of a same thickness including cotton, down, and knitted fabrics, only exhibit temperature differences of 52.5‒66.5 °C (Fig. 5d). Such good insulation is persistent over time; As shown in Fig. 5e, upon exposure to 200 °C, in situ measurements reveal that a CAF-10DBH fabric sustains a lower temperature steadily over a 10 min period compared to other three materials (127.2 vs 133.5‒147.5 °C). The measurements were also conducted at a wide range of temperatures from 34 to 200 °C. It shows that the CAF-10%DBH fabric consistently exhibits the largest ΔT across all conditions measured (Fig. 5f).

Fig. 5. Thermal insulation performance of CAFs and fabrics.

a Infrared images of CAFs on a 200 °C hot stage. b Temperature difference |ΔT| between the core of CAFs and the hot stage at different temperatures. c Thermal conductivity (λ) of CAFs. d Photographs (top) and infrared thermal images (bottom), and e the temperature variation as a function of time of a CAF-10DBH fabric, cotton, down, and a knitted fabric placed on a 200 °C hot stage, and f also on other temperatures. g Photographs of a volunteer wearing a commercial knitted sweater, a down jacket, and the tailored garment out of CAF-10%DBH fabrics and h the corresponding infrared images measured at −20 °C. All error bars denote mean ± standard deviation, number of replicates n  =  3.

We last explored the potential of applying CAFs in the real-life scenario by tailoring the 0.9-mm-thick CAF-10%DBH fabric into a sweater (Fig. 5g). When placed in a ‒20 °C environmental chamber, infrared imaging shows that the tailored garment appears blue, with measured temperatures of 0.3 °C at the chest and ‒0.5 °C at the abdomen, indicating minimal heat loss. In comparison, a ~15-mm-thick down jacket exhibits slightly higher temperature of 0.6 °C in the chest region and lower temperature of ‒2.1 °C in the abdomen, while a ~2.5-mm-thick knitted sweater appears yellow and displays substantially higher surface temperatures (Fig. 5h). Furthermore, we put these garments on a heat source maintained at 40 °C (simulating body temperature), and then measured the heat flux density on their surfaces. It shows that the CAFs sweater preserves heat well and exhibits a low heat flux density of 69.3 W m⁻², comparable to that of a down jacket (66.4 W m⁻²), both outperforming the knitted sweater (Supplementary Fig. 29). It reaffirms the high thermal insulation performance present in CAFs.

Discussion

In summary, we are inspired by penguin feathers and harness the divergent self-assembly behaviors of aramid nanofibers to fabricate core‒shell structured aerogel fibers. The outer shell adopts a well-ordered, layered architecture, formed in acid via reprotonation-induced hydrogen bonding, and imparts high mechanical strength to the fibers. The core part exhibits a cellular structure, with its formation driven by the thermodynamic interplay between covalent cross-linking and electrostatic repulsion. This cellular architecture combines macropores with nanoporous walls, providing exceptional softness. The integration of high strength and high softness within a single aerogel material—a combination rarely reported—renders these fibers particularly suitable for textile applications. Further weaving the aerogel fibers into fabrics shows that they exhibit good thermal insulation, and a 0.9-thick sweater delivers superior thermal resistance compared to much thicker commercial counterparts, including a 2.5 mm knitted sweater and a 15 mm down jacket. These results establish the cellular aerogel textile as a promising candidate for next-generation soft yet thermally protective clothing. Importantly, we envision that the cellular structure present in our aerogels and/or the molecular self-assembly approach governing its formation could be extrapolated to other material systems, including but not limited to elastomers, foams, metals, and concretes, thus overcoming the trade-off between high strength/rigidity and high softness/flexibility.

Methods

Materials

Kevlar fibers were purchased from DuPont Co., Ltd. (USA). DMSO (99.7%) was purchased from Shanghai Aladdin Biochemical Technology Co., Ltd. (China). Potassium hydroxide (95%, KOH), 1,4-dibromobutane (98%, DBB), 1,6-dibromohexane (97%, DBH), 1,8-dibromooctane (98%, DBO), 1,10-dibromodecane (97%, DBD), tert-butyl alcohol (99.5%, TBA), formic acid (88%), tetrahydrofuran (99%, THF), N,N-dimethylformamide (99.8%, DMF), acetone (AR), dichloromethane (99%, DCM) and N,N-dimethylacetamide (99.8%, DMAC) were purchased from Kaimat (Tianjin) Chemical Technology Co., Ltd. (China). Poly(dimethylsiloxane) (99%, PDMS) were purchased from Dow Corning Corporation (USA).

Deprotonation kinetics of commercial Kevlar fibers

Investigation on the deprotonation kinetics of Kevlar fibers adopted a protocol as follows. Under room temperature (ca. 22 °C), first, 2.5 g of Kevlar fibers was added to 104 mL KOH/DMSO/H₂O (3.75/100/4, w/v/v) mixture solvent in a bottle, and the mixture was heated to 80 °C while being stirred at 800 rpm. During the deprotonation process, 1 mL aliquot of the mixture was taken out at 5, 10, 30, or 35 min, and poured into 20 mL of deionized water in a beaker for 30 min. Following solvent exchange with 20 mL TBA for 12 h and subsequent freeze drying at −50 °C with a ramp rate of 2 °C min⁻¹ under a vacuum of 10 Pa for 24 h, the fibers at different deprotonation stages were obtained and subjected to SEM analysis.

Self-assembly kinetics of ANF sol cross-linked by DBH

Under room temperature (ca. 22 °C), 5 mL ANF sol with a concentration of 25 mg mL⁻¹ was mixed with 13.2 mg of DBH (10 mol% relative to the aramid repeat unit) in a 15 mL beaker. The mixture was magnetically stirred at 800 rpm. After 30 s, 1 min or 2 min of reaction time, 1 mL aliquot was taken out and poured into 20 mL of deionized water for 30 min. Following the same drying treatment, including solvent exchange and freeze drying as described above, a series of aerogels were obtained and subjected to SEM analysis.

Self-assembly kinetics of the cross-linked mixture of ANF sol and DBH

Under room temperature (ca. 22 °C), 5 mL ANF sol (25 mg mL⁻¹) was mixed with 13.2 mg of DBH in a 15 mL beaker. The mixture was stirred at 800 rpm and reacted for 2 min. After that, it was extruded using a 10 mL syringe into 10 mL of 40 vol% formic acid solution in a beaker. At 2 s, 30 s, or 10 min, a small amount of the sol was taken out, transferred to 20 mL of deionized water and kept for 30 min. Following the same drying treatment as described above, a series of aerogels were obtained and subjected to SEM analysis.

Preparation of CAFs

Under room temperature (ca. 22 °C), first, different amounts of DBH, including 6, 10, and 14 mol% relative to the aramid repeat unit, were added to 50 mL of 25 mg mL⁻¹ ANF sol and stirred for 2 min. The resulting mixture was then extruded through a 0.8-mm-diameter spinning needle at a rate of 80 mL h⁻¹. The fibrous sol was first passed through a 40 vol% formic acid coagulation bath for ca. 2 s with a draft ratio of 1.2, followed by a deionized water bath for 1 min. The process affords gel fibers that were collected on a rolling cylinder. After undergoing post-curing in water overnight, solvent exchange with TBA for 12 h, and freeze drying at ‒50 °C for 24 h, the gel fibers were transformed to CAF-xDBH. The parameter x is the amount of DBH employed. Other cross-linkers, including DBB, DBO, and DBD in an amount of 14 mol% were also used to react with the ANF sol following the same wet-spinning processes, which afford the corresponding CAF-14DBB/DBO/DBD. For fiber dyeing, the gel fibers were immersed in 10 mg L^–1 of aqueous dyeing solution overnight prior to the drying treatment.

Preparation of CAFs fabrics

The CAFs thus obtained were woven into fabrics on an automatic rapier loom (ASL3100-24D, Tianjin Longda Electromechanical Technology Development Co., Ltd., China). The fabric pattern was set as twill weave, and the resulting fabrics show a width of 45 cm, a warp and weft density of 50 and 180 threads per 10 cm, respectively.

Characterization

The morphologies of the CAFs were observed using a field-emission SEM (Gemini 9SEM 500, Zeiss, Germany) with an accelerating voltage of 10 kV. The specific surface areas of the aerogel fibers were determined via Brunauer–Emmett–Teller (BSD-660S A6S, BSD Instrument, China) methods. The pore size distribution was determined by the analysis of SEM images via Nano Measure software. In situ infrared absorption spectroscopy was obtained via Fourier transform infrared spectroscopy (Nicolet iS5, Thermo Fisher, USA) with a spectral range of 4000–400 cm⁻¹ and a resolution of 4 cm⁻¹. The zeta potential was tested by Nano Particle Size and Zeta Potential Analyzer Malvern Zeta sizer (Nanolink ZS90, Zhuhai Truth Optical Instrument Co., Ltd., China). The dynamic rheometer measurements were determined by a rotational rheometer (Haake Mars40, Thermo Scientific, Germany) at a constant strain of 0.5% and a frequency of 1 Hz. X-ray diffraction analysis was conducted on an X-ray diffractometer (D8 Discover, Bruker, Germany). Mechanical tests were performed on the electronic strength tester (YM-06A, Shanghai Chugong Industrial Co., Ltd., China). Thermal conductivity at different humidity levels was evaluated by a thermal conductivity meter (TPS 2500S, Hot Disk, Switzerland) at room temperature. Infrared thermal images were taken by a thermal imaging camera (343, Fotric, China). Water contact angle measurements were performed on a contact angle meter (JC2000D1, Shanghai Zhongchen Digital Technology Apparatus Co., Ltd., China), with an injection volume of 100 μL. Thermal analysis was performed on a thermogravimetric analyzer (TG 209 F3 Tarsus, Netzsch, Germany) under a N₂ atmosphere with the measurement temperature increasing at a rate of 10 °C min⁻¹. The surface morphology of the fabric was observed using an optical microscope (SK2700, Light Instrument World Machine Forest Rain, China). Thermal insulation performance of the fabric samples was recorded by a multi-channel heat flux detector (TNL-3RL, JinZhou TINEL Environment Energy Instrument Co., Ltd., China). The UV aging analysis was tested using a UV lamp (T5 8 W UVA-340, YIXIAN, China), operating at 340 nm ± 2% and 850 μW cm⁻².

Small-angle X-ray scattering (SAXS) and wide-angle X-ray scattering (WAXS) were tested using a small-angle/wide-angle X-ray scattering instrument (Xeuss 2.0, Xenocs, France). The X-ray tube power was 30 W, and the detector model was Pilatus 3R 300 K. The X-ray wavelength was set to 1.54 Å, with the distance between the detector and the sample of 1188 mm (SAXS) and 96.5 mm (WAXS). The orientation factor (f) was utilized to quantitatively assess the degree of aerogel fiber. A value of 0 signifies a completely random orientation, while a value of 1 indicates that the nanofiber crystals were perfectly aligned along the fiber axis. The value of f can be calculated quantitatively by adapting the following equation³⁷:

$$f=\frac{180^{\circ }-FWHM}{180^{\circ }}$$ 1

Here, FWHM refers to the full width at half-maximum of the azimuthal angle profiles. Among them, the azimuthal angle data were obtained by analysing the 2D-WAXS image using Fit 2D software, and the FWHM data were obtained by analysing the azimuthal angle image.

Human body thermal insulation performance test

The human thermal insulation test was conducted in an artificial climate chamber at ‒20 °C and a humidity of 40%. Volunteers wore knitted sweaters, down jackets, and aerogel fabrics and stayed in the environment for 10 min. Subsequently, infrared images were captured using a thermal imaging camera at a distance of 1 m from the volunteers.

Author contributions

Y.H. contributed methodology, investigation, writing–original draft preparation; G.Z. contributed investigation, software, methodology; G.Y. contributed formal analysis, technical guidance; Z.Z. and C.L. contributed investigation, testing, data curation; S.Y. contributed methodology supervision; Y.D., H.L., and L.X. contributed investigation, testing; Y.M. and X.Z. contributed funding acquisition, conceptualization, methodology supervision, writing–reviewing and editing; B.C. contributed data analysis and paper reviews.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data are available from the corresponding author upon request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-71723-2.