1 Introduction

Insulating materials that can withstand harsh conditions are urgently needed to meet growing needs in energy automobiles, high-rise buildings, aircraft, aerospace, etc. [1,2,3]. The frontier of next-generation thermal insulators is a group of ultralight porous ceramics, so-called ceramic aerogels (CAs). One of the well-known CAs is a monolithic silica aerogel synthesized using sol–gel techniques. Its thermal conductivity (TC) less than 13.5 mW m−1 K−1, i.e., lower than that of stationary air (ca. 25 mW m−1 K−1) [4]. This excellent thermal insulation behaviour is ascribed to the ultralow solid content and the tortuous solid network made of interconnected amorphous ceramic particles [5, 6]. However, the problems of monolithic CAs are also prominent. They are brittle and fragile and can only accommodate small compressive strain (< 10 %) before cracking or breaking into pieces [4, 7]. These materials also suffer from particle sintering, structure collapse, and size shrinkage under thermal shock or high-temperature exposure (e.g., > 750 °C for SiO2 aerogel) [8,9,10]. Moreover, CAs continuously absorb and accumulate moisture from the environment, eventually leading to catastrophic structural failure [11]. Therefore, despite their attractive properties, conventional CAs fail to meet the requirements for reliable thermal insulation, especially in buildings, portable devices, and vehicles.

Extensive efforts have been made to develop high-performance insulating materials through fundamental structural design. Most achievements were reported in constructing CA using flexible low-dimensional nanostructured ceramics, including hexagonal boron nitride nanosheets [2, 12], silicon carbide nanowires [13], alumina nanolattices [14], and various ceramic nanofibers [15,16,17,18]. The one-dimensional (1D) ceramic fibers stand out from the others due to the broad materials selection and the ease of mass production. Ultra-lightweight flexible materials made of fibers are now well-recognized as ceramic fibrous aerogels (CFAs).

Electrospinning is the most well-developed method to produce micro and nanofibers with reproducible characteristics. When joining with sol–gel synthesis, the so-called ‘sol–gel electrospinning’ technique enables various ceramic fibers with controllable properties [19, 20]. However, the electrospun fibers tend to assemble only into densely packed thin mats with typical thicknesses up to several hundreds of micrometers and densities far beyond the definition of ultralight materials, i.e., < 10 mg cm−3 [21, 22]. Therefore, most reports on CFA fabrication are based on a multistep procedure developed by Ding’s group [16, 23, 24], which includes mechanical chopping of the ceramic fibers followed by freeze-drying the fiber-containing slurries into desired shapes. The produced CFAs exhibit significantly better elasticity and flexibility than the conventional CAs. However, because short fiber segments physically do not get sufficiently entangled during the freeze-drying process, binders must be used to bond the short fibers, which severely limits the bendability and stretchability of the resulting aerogel [25]. Also, the binder usually cannot withstand high temperatures and intense thermal shocks, causing the fiber network and hence the aerogel to disintegrate at temperatures above 900 °C [26].

Fortunately, recent studies show that only by modifying the spinning solution can ultralight three-dimensional (3D) fiber sponges be generated without needing any post-treatment or specialist equipment beyond the standard electrospinning setup [27, 28]. Calcination of such electrospun fibers directly creates CFAs consisting of ultralong fibers entangled into a continuous 3D network. This so-called ‘3D electrospinning’ technique enables timesaving and cost-efficient CFA fabrication and could be readily combined with any commercial spinning equipment for up-scaling production. However, the development of 3D electrospinning technique is still in its infancy. Due to the complex dynamic sol–gel chemistry and solution-electric field interactions, identifying ‘3D spinnable solutions’ that can consistently and repeatably yield 3D fiber macro-assembles remains very challenging.

In this study, we report the fabrication of alumina/zirconia (Al2O3/ZrO2) fibrous aerogel (AZFA) using 3D sol–gel electrospinning technique. Correlations between solution properties and fiber macro and microstructures are established. The as-spun 3D fiber assembles have continuous and uniform fiber networks, which are calcined into flexible AZFA with ultra-low density (> 3.4 mg cm−3) and low TC (> 21.5 mW m−1 K1). The whole fabrication process takes less than 10 h and requires no more than classical electrospinning setup and a furnace, which significantly eases the CFA fabrication process. The produced AZFA, with a hypocrystalline microstructure, can remain highly flexible at 1300 °C, and no fiber melting or welding occurs even at 1500 °C. We highlight the correlation between fiber structure and various heat transfer mechanisms, establishing structure-TC relationships that are especially beneficial for tailoring and customizing heat conduction in 3D fiber networks. Furthermore, the AZFA is fluorinated to endow hydrophobicity and non-wetting self-cleaning properties. The created AZFA-F provides reliable insulation under extreme conditions, including high temperatures, high humidity, and underwater environments. They are ideal replacements for traditional insulating materials in, e.g., automobiles, aircraft, buildings, oil pipelines, etc.

2 Fabrication of AZFA by 3D sol–gel electrospinning

Al2O3/ZrO2 precursor fibers were fabricated by 3D sol–gel electrospinning using a standard electrospinning setup (Fig. 1a). The spinning solutions contained three Al sources, i.e., aluminum isopropoxide, aluminum nitrate, and aluminum chloride, and one Zr source, i.e., zirconium acetate, which are mixed in a polyvinylpyrrolidone (PVP)-containing aqueous medium. Modulating the ratio of ceramic precursor alters the solution viscosity and electrical conductivity simultaneously, resulting in electrospun fiber assemblies with distinct macro- and microstructures (Fig. 1b, Details in Supplementary Text, Table S1, and Figure S1, Supplementary information). The fiber assemblies were categorized into three types, (b1) cloud-like fiber piles consisting of microfibers and solution droplets visible to the naked eye, (b2) thin mats of densely packed sub-micron fibers, and (b3) centimeter-high self-supporting 3D assemblies. (b3) was used for further studies due to the ultrahigh porosity, uniform fiber morphology, and interweaved continuous fiber network. The significant differences between the fiber macrostructures are attributed to (i) the different electrodynamic behavior of solution jet, which is affected by the solution electrical conductivity, and (ii) the mechanical properties of single fibers, requiring the individual fibers to be stiff enough to support a free-standing 3D structure [27]. Interestingly, a similar trend has been observed in the 3D electrospinning of polymer fibers, where silver nanoparticles were incorporated to tailor the solution conductivity [29], which indicates that the solution modification is potentially a universal method to tune the macrostructure of electrospun fibers.

Fig. 1
figure 1

Fabrication of alumina/zirconia fibrous aerogel (AZFA) using 3D sol–gel electrospinning. (a) Schematics of AZFA fabrication steps: preparing the solution containing sol–gel reactants, 3D electrospinning, and air calcination. (b) A series of solutions were obtained by tuning the ratios of the sol–gel reactants shown in (a). The viscosity and electrical conductivity of each solution are plotted as correlated parameters, which define three regions that lead to three types of fiber macrostructures: (b1) a cloud-shaped mixed fiber/droplet macrostructure, (b2) a flat 2D mat, and (b3) a 3D assembly with highly entangled continuous fiber network. The red star marks the solution chosen for AZFA fabrication. (c) AZFA obtained after calcining the precursor fiber in (b3) can stand freely on the pistils of cherry blossoms exemplifying its ultra-light nature. (d) Scanning electron microscope (SEM) image shows the ribbon-like Al2O3/ZrO2 fibers. (e) Comparison of the processing step and time consumption of various CAs and CFAs. The time consumption of each is calculated assuming spinning 1 mL solution, and the heating and cooling rates of the furnaces are 10 °C s−1 unless specified otherwise. Every protocol starts with either solution (sol) or scaffold preparation, and the fabrication of CFAs includes furnace treatment under either vacuum or heat. Ref: [2, 16, 26, 28, 35,36,37,38, 40, 43, 44]

Full size image

Alumina/zirconia fibrous aerogels (AZFAs) with ultralow density down to 3.4 mg cm−3 (overall porosity exceeding 99.95 vol%) were obtained after calcination of the precursor fibers at 800 °C to remove the organic components. It can stand freely on, for instance, delicate flower pistils (Fig. 1c). Compared to the conventional CAs, our AZFAs show superior compressibility and bendability because it consists of ultralong ceramic fibers that are physically entangled to form a 3D continuous fibrous network (Movie S1, Figure S2, S3 Supplementary information). The individual fibers have a ribbon-like morphology with an average cross-section width and length of 2.72 μm and 1.18 μm, respectively (Fig. 1d, Figure S4, Supplementary information). Similar ribbon-shaped fibers have been reported in a wide range of ceramic systems, including ZrO2/SiO2 [30], SnO2 [31], Fe–Al–oxides [32], etc. The proposed mechanism is that such fiber morphology is most likely formed at the very beginning of the electrospinning process when the surface of the solution jet dries to form a skin layer that rapidly collapses and changes the fiber cross-section from circular to flat [33, 34].

X-ray photoelectron spectroscopy (XPS) spectra suggest that the elemental composition of AZFA surface is 56 at% O, 15.2 at% Al, 12.8 at% Zr, and 15.3 at% C (Figure S5 Supplementary information). The deconvolution of peaks confirms the presence of Zr-O Al-O, and C-O bonds. The energy-dispersive X-ray spectroscopy mapping (EDS) suggests the homogeneous distribution of Al and Zr elements in fiber structure (Figure S6 Supplementary information).

Compared with producing other ultralight ceramic insulators, the fabrication of AZFA is significantly more time-saving and cost-effective. This 3D sol–gel electrospinning (3D ES) approach only consists of three steps: (i) solution preparation, (ii) electrospinning, and (iii) calcination, taking less than 10 h in total (Fig. 1e, details in Table S2 Supplementary information). In contrast, the processes that include chemical vapor deposition (CVD), hydrothermal reaction, or freeze drying (FD) usually require several days of continuous operation [2, 16, 26, 35,36,37]. Solution blow spinning (SBS) method is as simple as 3D electrospinning, but sol–gel SBS has only been realized in a few groups and applied to highly limited systems because it has strict requirements on the solution properties [15, 38,39,40,41,42]. Also, while SBS and other electrospinning-based methods employ customized setups [43,44,45], 3D electrospinning is compatible with any commercialized electrospinning machine, and its production rate (currently 6 mL h−1 feeding rate through a single nozzle) could be easily multiplied by jointing with the multi-nozzle electrospinning setup. Therefore, due to a combination of fewer steps, short fabrication time, and conventional setup, AZFA is technically and economically feasible for up-scaling fabrication.

3 Thermal stability of AZFA

Figure 2a–d presents the microstructure of AZFA fibers. The ceramic fiber comprises nanocrystals embedded in an amorphous matrix, i.e., a hypocrystalline structure, and each nanocrystal contains one or several nanocrystallites. N2 isotherm of AZFA presents characters of an H5-type hysteresis loop [46], suggesting the presence of a combination of open and partially blocked mesopores (Fig. 2i, Figure S7 Supplementary information). Mesopores with sizes between 5 and 30 nm are formed because the alumina and zirconia reactants in solution undergo hydrolysis and condensation into ceramic condensates (Figure S8 Supplementary information). The ceramic condensates segregate from the polymer component, forming ceramic-rich and polymer-rich phases [47, 48]. The polymer component in the polymer-rich phase decomposes during calcination, leaving mesopores.

Fig. 2
figure 2

Microstructure characterization of AZFA before and after heat treatment at 1000–1300 °C. (a) SEM and (b–d) transmission electron microscopy (TEM) images of AZFA show the hypocrystalline structure of the mesoporous fibers. (e–h) SEM images of AZFA fibers after one-hour heat treatment at varied temperatures. (Scale bar 1 μm) (i) N2 adsorption–desorption isotherms and (j) XRD patterns of the heat-treated AZFAs (intensity normalized to the most intensive diffraction peak). (k) The recovered height of fiber samples after cyclic compression to strains from 25 to 95 %. Inset: photos of AZFA-1300 before and after uniaxial compressing to 90 % strain, showing the near-zero Poisson’s ratio. (l) SEM image displays the continuous fiber network of AZFA-1300 fiber under bending stress. Inset: a single AZFA fiber with a small curvature of about 2 μm (scale bar: 1 μm). (m–o) TEM images show that a typical grain in AZFA-1300 contains multiple crystals fused at interfaces, and each crystal consists of one or several crystallites

Full size image

To test the thermal stability of our material, we conducted one-hour heat treatment at 1000 to 1500 °C on AZFA (denoted as AZFA-temperature). The solid-state diffusion of atoms under high-temperature environments causes an amorphous-to-crystalline transition and change of fiber morphology. AZFA-1000 has a porous core composed of nanograin aggregates (Fig. 2e, S9 Supplementary information), which displays an H3-type hysteresis corresponding to slit-shaped pores with larger pore sizes than AZFA. AZFA-1100 shows a less distinct H3-type loop because the coalescence of fine mesopores generates meso- and macropores uniformly distributed throughout the fiber body (Fig. 2f). AZFA-1200 has even larger pores and a prominent grain structure (Fig. 2g). AZFA-1300 possesses coarser and partially coalesced grains, corresponding to a reversible Type II isotherm typical for nonporous or macroporous structures (Fig. 2h).

The crystalline structures of AZFA samples are indicated by XRD patterns (Fig. 2j). The AZFA fibers have tetragonal ZrO2 (t-ZrO2, ICDD 01–072-7115) nanocrystallite with a size of 8.9 nm, in agreement with the hypocrystalline microstructure observed by TEM. Monoclinic ZrO2 (m-ZrO2, Baddeleyite, 00–037-1484) emerges in AZFA-1000 and becomes dominant in AZFA-1100. Most of the t-ZrO2 crystalline phase transforms to m-ZrO2 in AZFA-1200. AZFA-1300 shows traces of trigonal Al2O3 crystalline peaks (α-Al2O3, Corundum, 00–046-1212), suggesting that Al2O3 starts to crystallize at 1300 °C. Fourier transform infrared (FTIR) attenuated total reflection (ATR) spectra of AZFA samples also show a broad absorption band in the range of 1000 to 600 cm−1, consistent with the XRD results that Al2O3 mainly exists in the form of amorphous or nanocrystallites (Fig. S10 Supplementary information). Compared to the previous sol–gel synthesized Al2O3 fibers in which α-Al2O3 arises at 1000–1100 °C, this retardant Al2O3 crystallization allows AZFA to remain flexible under higher temperatures [49, 50].

Table S3 (Supplementary information) summarizes the crystallite sizes of each crystalline phase, the Brunauer–Emmett–Teller (BET) surface areas, grain sizes measured from SEM, and the crystallinity calculated by fitting the XRD spectra. The increase in fiber crystallinity and grain size suggests that the amorphous-to-crystalline transition and grain growth of ZrO2 phase occur at 1100–1200 °C. AZFA-1200 and AZFA-1300 have similar crystallinity of approximately 62 %, which matches the XPS analysis of AZFA, i.e., 15.2 at% Al and 12.8 at% Zr correspond to 59 vol% well-crystalized ZrO2 and 41 vol% hypocrystalline Al2O3. The BET specific surface areas of AZFAs decrease against the heat treatment temperature because the crystal growth, coalescence, and coarsening are accompanied by the reduction of high-energy surface and interface.

Despite the microstructure changes, AZFA shows no obvious size shrinkage at 1300 °C, suggesting the preservation of the macroporous morphology, i.e., similar fiber-to-fiber distance (Fig. S11 Supplementary information). Cyclic compression tests were conducted, where the cubic-shaped fiber materials were uniaxially pressed with gradually increasing strain from 20 to 95 % (Movie S1, S2 Supplementary information). AZFA recovers more than 90 % of the original height from 60 % strain, while AZFA-1300 recovers to about 85 % (Fig. 2k). After decompressing from a large strain of 95 %, AZFA recovers to more than 50 % of the height, while AZFA-1300 displays a comparable recovery of about 44 %. No identifiable fracture, e.g., brokerage, crack, or crush, was observed among all test samples. Compared to CFAs produced by SBS that suffer from structural damage at 50 % strain [15], AZFA by 3D electrospinning illustrates greatly enhanced compressibility due to beadless fiber morphology, uniform fiber network, and sufficient inter-fiber entanglement. Note that the fiber materials compressed herein have densities of only approximately 4 mg cm−3, much lighter than the samples for mechanical tests in other reports. For example, the tested CFAs made of similar materials, e.g., zircon, Al2O3-stabilized ZrO2, and ZrO2-Al2O3, have a typical density above 20 mg cm−3 [26, 43, 51].

AZFA also remains highly flexible after 1300 °C heat treatment, being able to bend at 180˚ without cracking (Fig. 2l). The inset displays one fiber with a small curvature of about 2 μm, which survives through the 95 % strain compression without showing any fine crack or brittle fracture. Such excellent flexibility is attributed to the polycrystalline microstructure and the very high aspect ratio of the ultralong fibers forming a continuous fiber network. These properties are inaccessible to aerogels produced by the freeze-drying method [23]. Figure 2m–o reveals that AZFA-1300 contains grains with an average size of 162 nm, and each grain is a fusion of multiple crystals with sizes typically < 50 nm. The amorphous region, contributing 38 % of the XRD signal, acts as ‘glue’ to connect the grains and crystals and endow soft mechanical response by spreading the stress at the grain interfaces [52]. Each crystal contains one or several crystallites with an average size of 38.7 nm (Table S3 Supplementary information). Such small-sized crystallites are connected by small-angle grain boundaries, twin boundaries, and amorphous regions, which scatter the strain and promote stress dissipation to allow large deformation [52]. Moreover, it has been suggested by finite element modeling that nanobelt-shaped fibers can lower the stress levels better than cylindrical fibers [30], which jointly contributes to its superior flexibility.

Above 1300 °C, AZFA bulks present linear shrinkage of approximately 10.8 % at 1400 °C and 12.5 % at 1500 °C, corresponding to the increases of density from 3.4 to 4.8 mg cm−3 and 5.1 mg cm−3 respectively. Yet, there is no noticeable difference in the fiber size, length, and arrangement, and no welding or sintering between fibers can be observed, suggesting that AZFA remains thermally stable at 1500 °C (Fig. S12 Supplementary information). Such high-temperature tolerance significantly outperforms CFAs based on silica fibers that show adhesion and fusion among fibers after heat treatment at 1200 °C [16]. However, due to the increased crystallinity and larger grain size, AZFA-1400 and AZFA-1500 show a loss of mechanical resilience. AZFA-1500 presents approximately 75 % recovery from 60 % strain, 33 % recovery from 95 % strain, and the pristine cubic shape cannot be restored after unloading (Movie S3 Supplementary information). Also, large variations in the recovery were recorded among the three test samples, indicating that the recovery of AZFA-1500 is less reliable.

4 Structure-thermal conductivity (TC) relationships and thermal insulation performance

In the past decades, significant progress has been made in understanding the heat transfer mechanisms in conventional silica aerogels and developing structure–TC relationships [53, 54]. Yet, such knowledge of fiber-based flexible aerogels remains limited. Herein, our AZFAs, constructed by highly continuous ceramic fibers without any post-induced binding materials or post-synthesized hierarchical structure, provide a simple platform to investigate how the heat transfer process is related to both micro- and macrostructure of fibrous aerogel.

TCs at ambient conditions were measured using a transient plane source (TPS) method. A Kapton-insulated sensor was sandwiched between two AZFA pieces during the measurement. The obtained TCs are the sum of various heat-conducting pathways, including gas convection, thermal radiation, solid conduction, gas conduction, and the coupling between gaseous and solid conduction [55, 56]. Among these, heat transfer by thermal radiation is insignificant at room temperature [3]. Air convection is neglectable because the macropore sizes in AZFAs are much smaller than 4 mm, i.e., the Rayleigh number calculated based on such pores size is in the order of 10−10, far below the Rayleigh number of 1708 that is the threshold for natural convection to be considered considerable [57, 58]. Thus, the main factors contributing to TC are solid conduction through the fiber matrix (λs) and gas conduction via the interconnected macropores (λg), labeled in pink and blue arrows in Fig. 3a.

Fig. 3
figure 3

TC of AZFA in relation to the micro and macrostructure of ceramic fibers. (a) Schematic of the two primary heat transfer pathways in CFAs, gas conduction (λg) through the macropores between fibers and solid conduction (λs) along the fiber network. (b) Ambient TC of AZFAs before and after heat treatment at 1000 to 1500 °C and after fluorination treatment (AZFA-F). The calculated λg0 value at the aerogel density of 6 mg cm−3 is plotted. Inset: the fiber sample/sensor configuration in TPS measurement with pressure applied from top. (c) Schematics of the phonon propagation along the hypocrystalline fiber and crystalized fiber highlight various phonon scattering events that give rise to thermal resistance. (d) TC against the surface area of each heat-treated AZFA sample. The calculated λg0 values at AZFA of 10 or 15 mg cm−3 density are plotted, suggesting the suppression of gas conduction in mesoporous fiber. (g, h) AZFA is posed on a 200 °C hot plate while being pressed along the z direction to reach the targeted density. (e) The surface temperature of AZFA is recorded using an infrared camera and plotted as profiles along the z direction. Inset: the temperature dropping rates calculated from the initial linear region of the temperature profile. (f) The tortuosity factors (τ) along z axis (τz) or parallel to xy plane (τx/y) at each aerogel density

Full size image

The as-produced AZFA with a density of 3.4 mg cm−3 possesses ultralow TC of 21.6 mW m−1 K−1, lower than that of stationary air. To compare the TC of various heat-treated AZFAs, we measured the density of each sample and then pressed the sample/sensor assembly accordingly to regulate the densities to 6 mg cm−3 (Fig. S13 Supplementary information). This strain would not damage or change the microstructure of individual fibers. As shown in Fig. 3b, the overall TC increases with the heat treatment temperature. Because the AZFA samples have similar fiber arrangements and near isotropic macropore configurations, the increase in TC is mainly ascribed to larger λs, which is governed by (i) solid conductivity, an intrinsic property of solid material related to the chemical composition and micro/nanostructure, and (ii) tortuosity, a description of the intricacy of solid phase heat transfer path [53, 54]. In this case, the fiber morphology and crystalline microstructures affect the solid conductivity by deciding the phonon scattering events that give rise to the thermal resistance. The main phonon scattering events in mesoporous hypocrystalline AZFA and AZFA-1000 fibers are the pore scattering at the solid/gas interfaces, and the interfacial scattering between amorphous and crystalline domains (Fig. 3c) [59, 60]. Previous molecular dynamics simulations suggest that the phonon scattering at grain boundaries of small-sized grains (< 100 nm) can effectively suppress thermal transport [2]. For example, the TC of crystal zircon is 1.35 mW m−1 K−1, but decreased to 0.55 mW m−1 K−1 in hypocrystalline zircon [43]. In comparison, the interfacial thermal resistance in AZFA treated at more elevated temperatures is mainly contributed by the grain boundary scattering. The density of grain boundary is much lower than that of the nanocrystallite/amorphous interfaces, therefore, less effective in scattering the phonons.

To further investigate the factors influencing the solid conductivity, we measured the TCs of AZFA samples at either 10 or 15 mg cm−3 and plotted the values against the surface area (Fig. 3d). We calculated the theoretical air conductivity in the macroporous medium (λg0) based on [61]

$${\uplambda\;}_{g0}=\frac{{\uplambda\;}_{g,\;air}\Pi\;}{1+2\beta {K}_{n}}=\frac{{\uplambda\;}_{g,\;air}\Pi }{1+2\beta {L}_{g}/D}$$
(1)

where λg, air is the TC of free air at 20 °C [62], and \(\Pi\) is the porosity of AZFA at the density of 6, 10 or 15 mg cm−3 respectively, β ≅ 2 for aerogel, Kn is the Knudsen number, defined by the mean free path (MFP) of gas molecules in free space (Lg ≅ 69 nm [63]) divided by the pore size (D). When only considering the macropores in fiber networks that are at least two orders of magnitude larger than Lg, Eq. (1) is simplified into

$${\uplambda\;}_{g0}\cong\;{\uplambda\;}_{g,\;air}\Pi$$
(2)

The calculated λg0 values at each porosity are indicated Fig. 3b, d. Interestingly, the mesoporous AZFA and AZFA-1000 have TC lower than λg0. However, by fluorinating AZFA to diminish the mesopores at the fiber surface, the resulting AZFA-F has decreased surface area from 14.43 to 4.4 m2 g−1 and increased TCs to above λg0. Since fluorination is a low-temperature surface treatment process that would not modify the crystalline structure, AZFA-F and AZFA have the same λs, and the lower TC of AZFA is attributed to lower λg. Both AZFA and AZFA-1000 are mesoporous, with pore size (D) of 5–30 nm and > 10 nm respectively, corresponding to Kn < < 1. This indicates that although the total volume of the mesopores is neglectable compared to the > 99.9 vol% of macropores between fibers, these mesopores can trap the gas molecules and significantly suppress the air conduction, causing λg < λg0 [53, 64]. Our results agree with the previous observations that the heat transfer via the solid backbone can interact with gaseous conduction, and such a coupling effect deviates the measured TC from the theoretical values. Therefore, inducing mesopores into fibers is a promising strategy to decrease TC effectively and make CFAs more insulating than stationary air.

Pressing AZFA along z axis not only increases the overall density but also modifies the fiber packing configuration, especially the fiber orientation and fiber-to-fiber distance. To evaluate the influence of these macrostructural changes on the heat transfer process, we placed an AZFA piece on a 200 °C hot plate and pressed it to tune the density from 6 to 30 mg cm−3. The surface temperature was recorded as a function of the distance from the heating source (Fig. 3e, g, h). Interestingly, the denser aerogel has surface temperature drops faster along the z axis. The temperature decreasing rate on the surface of AZFA at 30 mg cm−3 is about three times higher than that at 6 mg cm−3. This indicates that although denser AZFA has larger overall TC when measured as a bulk isotropic material, the directional TC along z axis decreases. We attribute this to the more tortuous phonon propagation solid pathway since uniaxial pressing orients fibers perpendicularly to the heat flux direction [65]. To validate this assumption, we used a Matlab app, called TauFactor [66], to calculate the tortuosity factors (τ) along different directions based on 3D reconstructed tomographic data (Fig. 3f, S15 Supplementary information). τ of solid phase quantifies the reduction in diffusive transport along the intricate fiber network. At density of 6 mg cm−3, the calculated τ along z axis (τz) is approximately equal to τ parallel to xy plane (τx/y), suggesting isotropic heat conduction in the solid phase. τz increases significantly upon compressing along the z axis, reaching 5 times of τx/y at the density of 30 mg cm−3. Using these values, the effective conductivity due to solid phase conduction can be estimated by [67]

$${\uplambda\;}_{s}=\frac{{\uplambda\;}_{\mathrm{s}0}}{\uptau\;}(1-\Pi\;)$$
(3)

where λs0 is the TC of solid bulk material and (1-Π) is the volume fraction of the solid phase. λs along z direction decreases with density, whereas λg remains less affected since the macropore size is much larger than the MFP of air molecule. This suggests that pressing the aerogel could change the tortuosity of the fiber network and induce directional heat transfer, which is beneficial in developing functional thermal insulating materials with anisotropic TCs.

To demonstrate the thermal insulation performance of AZFA, we placed two water-filled vials on a hot plate at 200 °C; the vial on the left-hand side was heated directly, while the vial on the right side was wrapped by two pieces of AZFAs at the bottom and back of the vial (Fig. 4a). Without insulation, the water started to boil after 23.6 min, whereas the water in the insulated vial showed a slower heating rate and did not reach the boiling point even after 95 min (Fig. 4b). The flame resistance of AZFA was demonstrated by burning tests using a butane blowtorch (Fig. 4c, d, Movie S4 Supplementary information). Due to the refractory nature of Al2O3 and ZrO2, AZFA can withstand a blue flame (ca. 1350 °C) for over 2 min without visible crack or shrinkage and remain elastic in the fire. Movie S5 (Supplementary information) also presents the rapid heat dissipation of AZFA, which takes less than 20 s to cool down from 600 °C to room temperature.

Fig. 4
figure 4

Thermal insulation performance of AZFA. (a) Photographs and infrared images of two water-filled glass vials on a 200 °C hot plate. (Same temperature scale as in (e)). (b) The profiles of water temperature as a function of heating time. (c) AZFA is exposed to butane flame for over 2 min without visible damage or size shrinkage. (d) Photos of AZFA being pressed and recovering to the original height in flame. (e) AZFA is used to insulate a water-filled vial from flame. The water temperature only rises to 27.8 °C after 1 min of fire exposure. (f) Comparison of the ambient TC and maximum operating temperature (Tmax) of AZFA with other insulting materials, including polymer aerogel [68,69,70], carbon aerogel [71,72,73], cellulose aerogel [74, 75], fiberglass [76], silica aerogel [7, 77], composite aerogel [78,79,80], ceramic fiber mat [81,82,83], CA by a CVD method [2, 13, 18, 37], CFA [24, 26, 39], and commercial product Fiberfrax® [84]

Full size image

Furthermore, when using AZFA to protect a water-filled vial from flame, the AZFA surface that was exposed to blue flame (700 °C) only heated to a maximum of 170 °C, and the water temperature only increased from 23.8 to 27.8 °C after flaming for 1 min (Fig. 4e and Movie S6 Supplementary information). Compared to other thermal insulators (Fig. 4f), our material stands out by its ultralow TC, retaining the mechanical elasticity at 1300 °C, and remaining thermally stable at 1500 °C. Thus, our AZFA shows a distinct advantage for thermal protection under high-temperature environments such as furnaces, fire suits, and space vehicles.

5 Hydrophobic CAF for insulation in humid and underwater conditions

AZFA can absorb water droplets immediately (Movie S7 Supplementary information), i.e., water contact angle (WCA) of 0°, because of the capillary effects of the interconnected macropores and the hydrophilic Al–OH and Zr-OH groups on the fiber surface [85]. Such hydrophilicity is commonly seen in most sol–gel synthesized CAs, regarded advantageous for absorbing aqueous solution. However, in real-life insulation, the moisture tends to condense and accumulate within the aerogel material, which not only raises TC but also allows the detrimental growth of fungus and molds during long-term use [86]. To solve this, we modified the fiber surface via a two-step vapor treatment [87]. Glutaraldehyde (GA) was first evaporated to react with the hydroxyl groups on the alumina and zirconia surface, and 1H,1H,2H,2H-perfluorodecyltriethoxysilane was vaporized to bond the GA (inset in Fig. 5f). A conformal fluorocarbon coating layer was formed on the material surface, resulting in fluorinated AZFA (AZFA-F).

Fig. 5
figure 5

The properties and insulation application of hydrophobic AZFA-F. (a) Representative images of the water droplets in water contact angle measurement of AZFA-F. (b) Photograph of water droplets and acid or basic aqueous solution droplets standing on a flattened AZFA-F. HCl and NaOH were used to tune the pH. Methylene orange and methylene blue were used to dye the droplets. (c) The contact angle of various water-based liquids and oils on AZFA-F. (d) The contaminants on the surface of our AZFA-F could be easily washed away by pouring water onto it. (e) AZFA-F presents non-wetting surfaces when immersed in water. (f) XPS survey scan with the schematic of functional groups on the fiber surface (inset). (g) Peak deconvolution of C 1 s spectra suggests the -CF2 and -CF3 groups bonded to the fiber. (h) Infrared image of AZFA and AZFA-F posed on an alumina plate heated to 300 °C. Water droplets, either staying atop AZFA-F or being absorbed into AZFA, remain at room temperature after 5 min heating. (i) In the underwater thermal insulation test, a heating element (at 120°) was wrapped by AZFA-F and immersed in water with a pH value of 8 (tuned by NaOH). A control experiment was conducted without AZFA-F. (j) Infrared image of the insulated heating element and the surrounding water. Three white crosses label the positions where the water temperature was recorded. (k) The temperature profile indicates that AZFA-F shield could effectively reduce the water heating rate

Full size image

The as-produced AZFA-F with a density of 3.6 mg cm−3 has an ultralow TC of 25.2 mW m−1 K−1, which increases with density (Fig. 13b Supplementary information). In sharp contrast to the superhydrophilic AZFA, AZFA-F shows a high WCA of 140° despite the capillary effects of the macropores (Fig. 5a). AZFA-F repels both acidic and basic water droplets, shown in different colors in Fig. 5b. Removing the droplets after 10 min left no residue dye, suggesting no penetration of corrosive agents. AZFA-F also repels various water-based liquids such as saturated saline water, milk, and corrosive aqueous solutions. The contact angle of each solution was measured three times, and the averaged results are summarized in Fig. 5c. Meanwhile, AZFA-F is superoleophilic, showing instant uptake of hexane, toluene, chloroform, and silicone oil. Such features make it promising for insulation of, e.g., oil pipelines, where insulators can effectively mitigate the oil temperature variations along the pipeline and absorb leaking oil in case of any pipeline damage [88]. AZFA-F also possesses self-cleaning property that helps maintain a clean surface to avoid the build-up of contaminants and microspecies, minimizing moisture-induced corrosion and degradation (Fig. 5d). When forcibly immersed in water, the surfaces of AZFA-F remain non-wetting (Fig. 5e). The hydrophobicity of the fiber material is evidenced by the ‘mirror-reflection’ phenomenon, referred to as Cassie–Baxter non-wetting behavior [89]. The entrapped air between the material surface and the surrounding water causes total reflection at the air–water boundary.

The water droplet contacts the surface of AZFA-F following the Cassie–Baxter regime, where the apparent WAC is jointly decided by the properties of the solid surface and the air trapped between the water and fibers [90]. Therefore, hydrophobicity originates from the surface energy of the solid fibers and the surface roughness of the fiber assembly. The chemical bonding state of AZFA-F was examined by XPS (Fig. 5f), showing 43.9 at% F and 25 at% C, versus 15 at% C in AZFA. The pronounced CF2 and CF3 peaks are deconvoluted from the C 1 s spectrum, confirming the grafting of fluorocarbon chains (Fig. 5g). Replacing the surface hydroxyl groups with perfluorodecyl chains can effectively reduce surface energy. When analyzing the surface roughness of the soft, porous materials, two additional factors should be considered: (i) the change of fiber arrangement due to the gravity of the water droplet, and (ii) the sagging of water droplet into the space between adjacent fibers [87, 91]. The contact angle hysteresis and rolling-off angle of AZFA-F were approximately 29.2° and 61°, respectively, measured by tilting the sample (Fig. S16 Supplementary information). These relatively high angles are because the highly rough fibrous surface pins the water droplets and prevents them from rolling off, known as the “petal effect” [92]. To further evaluate these effects, the highly porous AZFA and AZFA-F were pressed into two dense pellets (Fig. S17, S18 Supplementary information). The water contacts the pellet surface with a minimized air gap, i.e., in a Wenzel regime, so the difference between the wettability of the two pellets is attributed to the reduction of surface energy after fluorination. The WCA of AZFA-F pellet decreased to 114° owing to its hugely reduced roughness compared to the original highly ‘furry’ surface. This value approaches the WCA of 119–120° for an ideally flat, dense surface covered with -CF3 groups [93]. The AZFA pellet absorbed water droplets much slower than AZFA because the decreased porosity and pore size mitigate the capillary effect (Movie S8 Supplementary information). Our results agree with observations in the polyurea aerogels that the WCA is reduced proportionally upon compression to reduce the porosity [94].

Figure 5h presents the infrared images of AZFA and AZFA-F placed on a 300 °C heating source. Their similar surface temperature distributions suggest a similarly superior insulating performance. Water droplets stay at the top of AZFA-F while being instantly absorbed by AZFA. As proof of concept, a 120 °C heating element was wrapped by a 5 mm-thick flexible AZFA-F mat and immersed in water with a pH of 8, i.e., mimicking seawater (Fig. 5i). We monitored the water temperature at three fixed measurement points and recorded the temperature as a function of time (Fig. 5j, k). Within the 15-min test duration, the overall heating rate of AZFA-F insulated water was 0.79 °C min−1, while that of the directly heated water reached 1.75 °C min−1. Thus, the fibrous shield could effectively block the heat transfer from the heating element to water.

The advantages of AZFA-F are manifested by comparing it with various hydrophobic aerogel insulators (Table 4 Supplementary information). Polymeric aerogel insulators, such as polyimide, polyurethane, polybenzoxazine, polysilsesquioxane, and their composites usually exhibit high WCA and robust mechanical properties [95,96,97,98,99]. However, these materials are relatively dense, having heat tolerance limited to about 300–500 °C and releasing toxic gases when ignited. Cellulose-based bio-aerogels are known for environmental friendliness and biocompatibility. Yet, extracting cellulose from natural resources such as bamboo, wood, and bagasse consumes harsh and toxic chemicals, making the preparation complicated and rather environmentally unfriendly [100,101,102,103]. Hybrid silica aerogels or saline-coated clay aerogels provide higher thermal stability and low TC, but they are relatively dense, rigid, and lack flexibility [104, 105]. In comparison, AZFA-F in this work exhibits the lowest density, among the lowest TC, excellent hydrophobicity at elevated temperature, and highest thermal stability. Therefore, AZFA-F is more attractive as insulators for buildings in highly humid environments, underwater pipelines, reactors in high-temperature and corrosive environments, and other conditions beyond the reach of conventional porous insulation materials.

6 Conclusion

Ultralight and flexible alumina/zirconia fibrous aerogels (AZFAs) were created using one of the most time-saving, cost-effective, and scalable techniques — 3D sol–gel electrospinning. AZFA exhibits low thermal conductivity (TC) down to 21.6 mW m−1 K−1, robust flexibility and compressibility, excellent fire resistance, high thermal resistance over 1300 °C, and remains structurally integrated at 1500 °C. This ultralow TC is attributed to reduced solid conduction, tortuous thermal pathway, and the mesopores on fiber that suppress air conduction. The structure-heat transport relationship of fibrous aerogels was elucidated in terms of the macrostructure of fiber assembly and microstructure of single fiber, which should enlighten the development of more fibrous aerogels with lower TC and tunable isotropic or anisotropic properties. Furthermore, we transformed the superhydrophilic AZFA to hydrophobic AZFA-F using a fluorination treatment. The resultant AZFA-F, with a low TC of 25.2 mW m−1 K−1, presents outstanding water repellency, self-cleaning properties, and excellent chemical stability. Such material enables thermal insulation not only in high-temperature conditions such as in furnaces, aircraft, and vehicles but also in wet humid environments and underwater conditions commonly in natural reservoirs.

7 Experimental sections

7.1 Materials

Polyvinylpyrrolidone (PVP, M.W. 1,300,000) and aluminum chloride (AlCl3, anhydrous, granular) were purchased from Alfa Aesar. Acetic acid was purchased from Honeywell. Zirconium (IV) acetate hydroxide (ZrAc), aluminum nitrate nonahydrate (AN), aluminum isopropoxide (AiP), ethanol (> = 99.8 % GC) was purchased from Sigma-Aldrich. For fluorinated coating, glutaraldehyde (GA, 50 wt% in water) and 1H,1H,2H,2H-perfluorodecyltriethoxysilane (PDTS) were purchased from Sigma-Aldrich. All reagents were used without further purification.

7.2 Solution preparation

The solution that generated 3D fiber assembly was prepared by the following steps. 0.2 g AlCl3, 0.35 g AN, and 0.265 g AiP were dissolved in a mixture of 1 mL H2O, 1.2 mL EtOH, and 0.5 mL AcOH by stirring at 80 °C for 0.5 h. 2.3 g ZrAc was added into the solution, further stirring at 80 °C for 1.5 h to obtain a dilute Al solution. Meanwhile, 1 g PVP powder was dissolved in 10 ml EtOH by stirring at 55 °C for 2 h. Then, PVP solution was slowly added to Al solution with a weight ratio of 1:2.68. After stirring at 70 °C for 2 h, a transparent and homogeneous solution was obtained. Tunning the relative ratios of AlCl3, AN, AiP, and ZrAc could change the solution properties, yielding precursor fibers with varied micro and macrostructures.

7.3 Fabrication of ceramic fibrous aerogels

Electrospinning was performed by a home-built setup. The solution was fed through a gauge 17 needle at 6 mL h−1. The nozzle-to-collector distance was fixed at 30 cm. A high voltage (Genvolt High Voltage Power Supply) of 22 kV was applied. The temperature and humidity were 21 ± 2 °C and 45 ± 15 %, respectively, without extra control. Within such ranges, 3D fiber macro-assemblies could be repeatedly produced. The fibers were collected on a ground Cu plate collector. As-electrospun precursor fiber assemblies were calcined at 800 °C in air for 2.5 h at 10 °C min−1 heating rate in an Elite horizontal tube furnace.

7.4 Fluorinated coating

The procedure for fluorinated treatment was described in the previous report [87]. In short, the calcined fiber assemblies were placed in a vacuum oven, co-heated under vacuum with 1 mL GA aqueous solution at 75 °C for 10 h, and then with 250 μL PDTS at 125 °C for 10 h.

7.5 Characterizations and modeling

The electrical conductivities of solutions were measured using a benchtop conductivity meter (Oakton CON 550) at room temperature. Each solution was measured three times. The viscosity was measured by Brookfield DV-II Viscometer using a cone spindle CP-52. A fixed amount of 0.5 mL solution was loaded into the sample cup. A ramping of rotation speed was used from 10 to 70 rpm, and the averaged viscosity values were reported.

X-ray diffraction (XRD) was carried out on a Bruker D8 ADVANCE Eco X-ray diffractometer using Cu Kα radiation (wavelength λ = 0.154 nm). Crystallinity was determined as the ratio of the crystalline peak area to the total area under the diffraction curve. The amorphous background was obtained by air calcinating the precursor fiber at 650 °C for 3 h. Under such conditions, PVP was completely decomposed while the ceramic components remained purely amorphous. Lorentz functions were used to fit the crystalline peaks after subtracting the amorphous background. The full width at half maximums (FWHMs) of the peaks could be extracted from the fitting. The crystallite sizes (D) were obtained using Scherrer equation D = Kλ/(FWHM × cosθ), where the shape factor K is 0.9.

SEM images were acquired using a Zeiss Merlin. Energy-dispersive spectroscopy (EDS) elemental mapping was collected by a Zeiss Evo at an accelerating voltage of 4 kV. The samples were coated with 5 nm thickness Pt. Fourier transform infrared (FTIR) attenuated total reflection (ATR) spectra were recorded in the range of 600 to 4000 cm−1 using a Varian Excalibur FTS 3500 FT-IR Spectrometer. Nitrogen adsorption–desorption isotherms were recorded on a Micromeritics ASAP 2420 M system. The surface areas and pore size distributions were determined using Brunauer–Emmett–Teller (BET) and Barrett-Joyner-Halenda (BJH) methods. X-ray photoelectron spectroscopy (XPS) measurements were performed on a Thermo Scientific K-Alpha XPS. TEM images were obtained by JEOL JEM-2100F at an acceleration voltage of 200 kV.

The densities of the AZFAs reported here were the ‘apparent densities’, calculated using the weight of the solid contents divided by the bulk volume of the test sample. The samples were cut into cubic shapes, and the length of each edge was measured by a digital caliper. The mass was measured by an analytical balance with a precision of 0.01 mg. At least three cubic-shaped material pieces were measured for each AZFA sample. The porosity was calculated as the volume percentage of air inside AZFA, using 4.89 g cm−3 as the density of solid net fiber, i.e., 45.7 wt% Al2O3 and 54.3 wt% ZrO based on the XPS result.

The thermal conductivities (TCs) of AZFAs at ambient conditions (19.5 ± 1 °C) were measured using a Hot Disk TPS 3500. A two-sided Kapton-insulated sensor (TPS Hot Disk 7577 with 2.0 mm in radius) was placed between a pair of identical samples. The ‘Bulk isotropic’ module was used under the assumption that the sensor is located in an infinite isotropic material, so the sample thickness should be more than the radius of the hot disk sensor, and the sample size should be larger than the heat penetration depth. To meet these requirements, AZFA samples were prepared with centimeter scales in all dimensions, with at least 2.5 cm in height before pressing. The optimal measurement time of 5 s and heating powder of 5 mW were used to give the best accuracy and stability. The density of each sample was measured prior to the experiment, and the top sample surface was pressed to regulate the density and minimize the thermal contact resistance between the sensor and the samples. Six tests were conducted on each sample. The error bar of each data point was obtained as the standard deviation of the TC values.

The elasticity and flexibility of the aerogel bulks were tested using a pressing machine. Samples were cut into cubic shapes and compressed by a metal plate along the z direction. At least three specimens per AZFA sample were characterized for generating the strain-recovery curves and error bars.

The fire resistance of AZFA was demonstrated using a butane blowtorch (PT-500 RS Component). The dynamic temperature distributions upon heating or flaming were recorded by an infrared camera (FLIR A65, FOV 45). The contact angle measurements were performed in an ambient environment using an Ossila Contact Angle Goniometer. Contact angle hysteresis was calculated as the difference between the advancing and receding angles when the water droplet starts to roll off from the tilted sample surface. At least three measurements were conducted on each sample to minimize the influence of the water droplet size. A pellet pressing machine (Atlas Manual Hydraulic Press 15 T, Specac) was used to shape the highly porous AZFA and AZFA-F into dense pellets.

The tortuosity of the fiber network was evaluated using an open-source Matlab app, TauFactor [66]. The input microstructural data was taken from our previous work, which was the 3D-reconstructed tomography of an electrospun 3D fiber macro-assembly with density of 3.6 mg cm−3 [27]. Although our previous fiber assembly was made of different ceramic materials, the structure of the fiber network shows a great resemblance to the AZFA in this work. The X-ray images, with voxel size of 965 nm, were segmented into two phases (solid and gas) and constructed into a vertical tiff. stack. When processed by TauFactor software, the voxel along x and y direction remained 965 nm (direction 2 and 3 in Fig. S14 Supplementary information), whereas the voxel along z axis was decreased to 579, 347.4, 231.6, and 115.8 nm to match the uniaxially compressed aerogels with densities of 6, 10, 15, and 30 mg cm−3 respectively.