Grain boundary segregation of ceramic aerogels for reliable thermal insulation at extreme conditions

Abstract

Thermal insulation under extreme conditions requires materials to have high thermal stability and insulating properties to guarantee the critical system functionality and safety in aerospace, military, and high-temperature industries. Ceramic aerogels are the most attractive candidate to date, but due to the grain growth effect at high temperatures, they still suffer from insufficient structural stability and degraded thermal insulation that can lead to catastrophic failures. Here, we report an unusual grain boundary segregation in Y₆Zr₁O₁₁ fibrous aerogels with zirconium segregated at the yttria-zirconia grain boundary to inhibit boundary migration for fine-grain enhancement. The aerogel, fabricated by an airflow-assisted centrifugal spinning method, exhibits a grain size of only 172 nm and high thermal stability with volume shrinkage <3% at 1,500 °C in air, as well as near-infrared reflectivity up to 98.12% and low thermal conductivity of only 89 mW m⁻¹ K⁻¹ at 1,000 °C under a density of 50 mg cm⁻³. This work not only establishes a set of grain boundary segregation considerations in material design of ceramic aerogels, but also promotes the large-scale and low-cost practical applications of such aerogels for reliable thermal insulation at extreme conditions.

Grain-boundary segregation enables ceramic aerogels with high thermal stability and reliable thermal insulation at temperatures up to 1500 °C, offering a strategy for designing lightweight thermal protection materials for extreme environments.

Introduction

Ceramic aerogels, well known for their excellent properties such as light weight, high thermal stability and ultralow thermal conductivity (κ), have become the most attractive material family for thermal insulation and protection under extreme conditions in aerospace, military and high-temperature industries^1–5. To overcome the intrinsic brittleness of ceramic aerogels, deliberate structural engineering in building units has been developed^6–12, and the fibrous aerogels, constructed from randomly entangled one-dimensional (1D) ceramic fibers, have exhibited enhanced mechanical properties in compressibility, bendability, stretchability and twistability^13–15. However, due to the significant grain growth behaviors at high temperatures, this type of ceramic aerogels still suffers from severe fiber fracture with large volume shrinkage or even structural collapse, which can lead to the degraded thermal insulation and serious thermal runaway^16,17. As a result, thermal stability has become the key roadblock to use fibrous ceramic aerogels for practical applications under extreme conditions.

To inhibit grain growth behaviors in fibrous ceramic aerogels, several strategies have been developed such as entropy engineering and crystal/phase design. For the entropy engineering, a medium-entropy (Ti_0.42Zr_0.42Y_0.08Si_0.08)O_2.08 was reported to maintain a small grain size at 1400 °C due to the intensive crystalline energy caused by lattice distortion and sluggish diffusion¹⁶. A single-phase high-entropy (La_1/4Sm_1/4Gd_1/4Y_1/4)₂Zr₂O₇ fibrous aerogel was designed to maintain a high structural stability during long-term calcination at 1400 °C owing to the further enlarged lattice distortion and sluggish diffusion by high-entropy effect¹⁸. A high-entropy (Zr_0.2Hf_0.2Ti_0.2Gd_0.2Y_0.2)O_2-δ fiber membrane was reported with good flexibility and strength after heat treatment at 1200 °C for 50 h and 1400 °C for 180 s¹⁹. For the crystal/phase design, a hypocrystalline zircon nanofibrous aerogel was reported with excellent thermomechanical properties of 50% compressive strain under blowtorch flames (~ 1300 °C) due to the pinning effect of crystals on amorphous state¹³. A nanograin-glassy dual-phase mullite fiber spongy with ceramic grains embedded in amorphous shells was developed to possess temperature-invariant superelasticity under temperatures below 1500 °C²⁰. A multi-phase Al₂O₃/ZrO₂/Y₂O₃ composite was designed to induce a confinement effect, endowing the resulting nanofibrous membranes with high flexibility and compressibility at 1400 °C²¹. However, these above-mentioned strategies mainly focus on the structural stabilization within crystal grains, while the grain boundary (GB), another key factor to enhance the thermal stability of ceramic aerogels, has received few attentions due to its more complex structural configurations and more difficulty regulation methods^22,23, leading to a slow grain growth at high temperatures (>1200 °C) and a rapid growth at high temperatures (≥1500 °C).

Ceramic aerogels usually demonstrate a thermal super-insulating performance at room temperature (thermal conductivity ≤ 26 mW m⁻¹ K⁻¹) owing to their excellent suppression effect on thermal conduction and convection²⁴. However, according to the Stefan-Boltzmann law^25,26, thermal radiation rises with a third power of temperature and becomes the dominant factor of heat transfer at temperatures over 500 °C. To this end, the design of ceramic aerogels for practical applications under extreme conditions must consider how to inhibit thermal radiation at high temperatures, such as infrared absorption or reflection^13,27. Carbonaceous aerogels, including graphene aerogels and carbon nanotube aerogels, exhibited an excellent infrared absorption ability, but their poor oxidation resistance largely restrict their applications within vacuum or temperature below 600 °C in air^28,29. Ceramic aerogels, especially the oxide ceramic ones, possessed a much higher working temperature in air and usually adopted the infrared reflection pattern. For example, Y₂O₃ and TiO₂ were well-known for their high infrared reflectance, exhibiting infrared reflectivity up to 83% and 89.9%, respectively^30,31. The rare-earth zirconates, such as La₂Zr₂O₇ and Y₂Zr₂O₇, demonstrated even higher reflectivity up to 93.2% and 96.34%, respectively^32,33. Such high infrared reflectance endowed these ceramics with promising thermal insulating performance at high temperatures. However, TiO₂ usually showed a terrible softening effect at high temperatures with large volume shrinkage, while the Y₂O₃ and rare-earth zirconates often suffered from a serious grain-growth problem^33,34. Therefore, it is still very challenging to realize a synergetic enhancement between thermal stability and thermal insulation in ceramic aerogels for reliable thermal insulation at extreme conditions.

Herein, we report an unusual grain boundary segregation in the yttria-zirconia fibrous aerogels fabricated by a large-scale and low-cost airflow-assisted centrifugal spinning method to simultaneously enhance the thermal stability and thermal insulation at high temperatures. Owing to the significant segregation of zirconium at the yttria-zirconia grain boundary, the Y₆Zr₁O₁₁ aerogels demonstrate a fine-grain enhancement with high thermal stability (grain size of 172 nm and volume shrinkage <3%) at 1500 °C in air and NIR reflectivity up to 98.12% with low thermal conductivity of only 89 mW m⁻¹ K⁻¹ at 1000 °C under a density of 50 mg cm⁻³, presenting a reliable material for thermal insulation at extreme conditions.

Results

Design and fabrication

Even though a GB segregation phenomenon of yttrium segregated at yttria-stabilized zirconia (YSZ) boundary was reported^35–40, its inhibition effect on boundary migration and grain growth was still unclear and insufficient, leading to a low thermal stability of YSZ fibrous aerogels under high temperatures below 1200 °C. Meanwhile, considering the high infrared reflectivity of Y₂O₃ and rare-earth zirconates, we selected the yttria-zirconia material system to investigate the potential GB segregation enhancement on thermal stability and high infrared reflectivity on thermal insulation at extreme conditions.

To systematically investigate the yttria-zirconia aerogels (Y-Zr ratio from 10:1 to 1:1), we developed a large-scale and low-cost airflow-assisted centrifugal spinning method (Fig. 1). Centrifugal spinning is a typical fiber technique in which polymer or inorganic precursor solutions are spun and solidified under centrifugal force^41,42. This method is safe, easy and highly productive, enabling a large-scale and low-cost fabrication of fibers. However, conventional centrifugal spinning usually yields doughnut loops or densified membranes (Supplementary Fig. 1a), which can’t be directly employed to make the three-dimensional (3D) fibrous aerogels. To this end, we introduce perpendicular air flows to regulate the centrifugal effect for aerogel assembly (Fig. 1a). To be specific, four symmetrically inward-curved grooves were set on the thin-walled cylindrical spinning disk (Supplementary Fig. 2). During a high-speed rotation (~ 10,000 rpm), the precursor solution was continuously supplied to the rotating disk and ejected through the orifices under centrifugal force, and the grooves generated strong air jet flows with speed of 8 m s⁻¹ (Supplementary Fig. 3). Subsequently, driven by the combined effects of centrifugal stretching and near-field air jet flow, the as-spun precursor was fully dried and stretched to form continuous fibers. These fibers were then randomly entangled and assembled layer-by-layer under the far-field air turbulent field, forming a fluffy 3D lamellar fibrous aerogel (Supplementary Fig. 1b). Finally, the aerogel was annealed at 1000 °C to remove organics and solvents to realize the ceramic crystallization process. Benefiting from this near-field air jet and far-field air turbulent fields, the as-spun fibers could be well assembled into a low-density (30–50 mg cm⁻³) aerogel architecture (Fig. 1b and Supplementary Fig. 4). In particular, ceramic fibers were randomly interwoven within each layer, and adjacent layers were further connected through surface entangled fibers (Supplementary Fig. 5), forming a uniform laminar fibrous network. In addition, the as-prepared aerogels could be easily cut into various shapes (e.g., square, circular, or hexagonal) to facilitate practical applications (Supplementary Fig. 6).

Fig. 1. Design and fabrication process of ceramic aerogels.

a Schematic illustration of the airflow-assisted centrifugal spinning method and fabrication process of aerogels. Stage I: Precursor spinning, where precursor solutions containing PAY (polyacetylacetone-yttrium), PAZ (polyacetylacetone-zirconium), and PVP (polyvinylpyrrolidone) are loaded onto the spinning disk. Stage II: Airflow-assisted aerogel formation, with the central gray cylindrical component representing the high-speed spinning disk rotating at ~10,000 rpm. Stage III: High-temperature calcination of the as-spun aerogel to obtain the final fibrous ceramic aerogel monolith. b Optical image of an aerogel sample placed on a flower, and the SEM images of its microscopic structure.

Grain boundary segregation behavior

We investigated the grain boundary segregation behavior in yttria-zirconia aerogels. Interestingly, with the increase in Y-Zr molar ratio from 1:0 to 6:1, an unusually upward trend of Zr segregation at boundary was found (Supplementary Fig. 7a, b), while further increasing the ratio from 5:1 to 1:1, the yttria-zirconia ceramics began to generate phase transition (Supplementary Figs. 7c, 8). As a result, the Y₆Zr₁O₁₁ aerogel (denoted according to the nominal chemical composition defined by the Y-Zr molar ratio) demonstrated the optimal Zr segregation at the yttria-zirconia GB. These observations indicated that Zr segregation at grain boundaries occurred within a specific Y-Zr ratio range (10:1-6:1) and maximized at 6:1. Further increase in Zr content would mainly lead to the phase transition of crystal grains from cubic Y₂O₃ to Y₂Zr₂O₇ (defect-fluorite structure).

The Y₆Zr₁O₁₁ sample, annealed at 1000 °C, retained a C-type cubic Y₂O₃ structure with an average grain size of 14.62 nm (Fig. 2a and Supplementary Figs. 9, 10). After annealing at 1500 °C, HAADF-STEM and EDS results showed that the elemental signals of Y and Zr exhibited complete atomic-scale coherence and were highly colocalized with oxygen, indicating a homogeneous elemental distribution with no phase segregation or local enrichment inside the grains (Fig. 2b). This demonstrated that Zr⁴⁺ ions were substitutionally incorporated into the Y₂O₃ lattice in the bulk regions. In contrast, pronounced Zr enrichment was clearly detected along GBs, forming a Zr-rich GB (with a thickness of ~5 nm) with a slight decrease in Y concentration. Line-profile (K-edge) analysis showed that the Zr signal at GBs was ~20% higher than that in the grain interior (Y-Zr molar ratio of ~6:1), supporting the presence of a grain boundary segregation phenomenon (Fig. 2c, Supplementary Fig. 12).

Fig. 2. GB segregation behavior of Y₆Zr₁O₁₁ aerogel.

a Bright-field TEM image of the polycrystalline morphology. (Insets) Corresponding selected-area electron diffraction pattern. b Atomic-resolution HAADF-STEM images and corresponding EDS maps within the Y₆Zr₁O₁₁ grains. c High-resolution HAADF-STEM images and compositional distributions at the grain boundaries, all scale bars in inset, 5 nm. Gray shaded region in charts denoted grain interior domains. d Illustration of the MCMD processes in Y₆Zr₁O₁₁. The snapshot showed the initial atomic configuration. Below, three diagrams illustrated the profiles of the Zr fraction (f_Zr) along the x direction (perpendicular to GBs), at initial state, intermediate stage, and final state during simulation, respectively. Blue shaded areas corresponded to GB core regions. e Results of MCMD simulations for systems with different Y-Zr molar ratios. Y₆Zr₁O₁₁ exhibited the most pronounced Zr segregation as well as lowest segregation energy. Symbols represented mean values and standard deviations estimated over the final 1 ns of the MCMD trajectory for each composition. Orange shaded bands marked the composition of Y₆Zr₁O₁₁.

This significant GB segregation behavior of Zr atoms in Y₆Zr₁O₁₁ was attributed to the synergistic interplay between thermodynamic driving forces and kinetic stabilization mechanisms^43,44. Grain boundaries, characterized by atomic disorder and high interfacial energy, served as energetically favorable sinks for solute atoms³⁵. Due to the ionic radius and charge mismatch between Zr⁴⁺ (0.92 Å) and Y³⁺ (1.16 Å), Zr exhibited a pronounced energetic preference for grain boundary sites over bulk lattice sites. Our density functional theory (DFT) calculations therefore focused on the site-dependent segregation energy, E_seg, by comparing the energies of Zr atoms located at different distances from the grain boundary (Supplementary Fig. 13). The results indicated that Zr segregation was energetically favorable and contributed to grain boundary stabilization. Given the intrinsic size and timescale limitations of DFT, we further employed hybrid Monte Carlo-Molecular Dynamics (MCMD) simulations based on the universal neuroevolution potential NEP89⁴⁵ to capture the segregation process in more realistic GB configurations. Figure 2d illustrated the evolution of the Zr fraction (f_Zr) of Y₆Zr₁O₁₁ during MCMD simulation, where Zr displayed a distinct tendency of segregation behavior at GBs. Among all investigated compositions, Y₆Zr₁O₁₁ displayed the strongest segregation tendency, as evidenced by the largest difference in Zr fraction between the grain boundary and bulk regions, which was in good agreement with the experimental observations (Fig. 2e and Supplementary Figs. 14, 15). In addition, the energy difference between the segregated and homogeneous configurations, ΔE, was most favorable for Y₆Zr₁O₁₁, indicating that Zr segregation yielded the greatest energetic stabilization in this composition and thus provided the strongest thermodynamic driving force for grain boundary stabilization.

The Y₆Zr₁O₁₁ thus defined an optimal molar ratio (Y:Zr = 6:1) to ensure the largest segregation behavior of Zr while preventing phase transition problem. Rather than the conventional surface-energy-driven particle coalescence effect^46–48, the GB segregation behavior in our aerogel firstly ensured the favorable GB migration pattern for grain growth. Furthermore, the optimal Y-Zr ratio of 6:1 generated the largest segregation behavior with enhanced Cahn’s impurity-drag and Zener’s pinning effects^38,49,50 to inhibit boundary migrations (Supplementary Fig. 16). This Y-Zr ratio was also the turning point for phase transition, which could largely aggravate the grain growth and offset the segregation stabilization effect. Then, the combined structural, compositional, and thermodynamic optimization of Y₆Zr₁O₁₁ might realize an effective suppression of grain boundary migration, resulting in a potentially high thermal stability.

Material characterization

To investigate the influence of this unusual GB segregation behavior on thermal stability of ceramic aerogels, we systematically compared the grain sizes of ceramic fibers with different Y-Zr molar ratio after calcination at 1500 °C for 2 h in air (Fig. 3a, c and Supplementary Fig. 17, 18). The Y₆Zr₁O₁₁ aerogel showed the smallest grain size of only ~ 172 nm, while any change of the Y-Zr molar ratio would lead to an increase in grain size, such as ~ 419 nm of Y₂O₃ (no segregation) and ~ 288 nm of Y₂Zr₂O₇ (phase transition). Compared to the little GB segregation behaviors in other Y-Zr molar ratios (Supplementary Fig. 7a), the significant Zr segregation in Y₆Zr₁O₁₁ aerogel could result in a more stable boundary structure and more effective pinning effect, which finally generated the largest fine-grain enhancement at high-temperature conditions. Although grain growth became pronounced above 1500 °C (Supplementary Figs. 19, 20), our Y₆Zr₁O₁₁ fibers still exhibited a smaller average grain size compared with the counterparts, confirming the effectiveness of GB segregation to suppress grain growth at high temperatures.

Fig. 3. Material characterization of ceramic aerogels.

a, c Comparative analysis of grain size in ceramic fibers with different Y-Zr molar ratios after annealing at 1500 °C for 2 h, scale bar in inset, 400 nm. b XRD results of Y₂O₃ and Y₆Zr₁O₁₁ sintered at different temperatures. The dotted line indicated the (222) peak position of pure Y₂O₃ for comparison of peak shifts in Y₆Zr₁O₁₁ solid solutions. d Rietveld refinement results showing temperature-dependent lattice constants of Y₆Zr₁O₁₁ and Zr occupancy of Y₆Zr₁O₁₁. e Raman results of Y₂O₃ and Y₆Zr₁O₁₁. f TG-DSC results of Y₆Zr₁O₁₁. DTG represented derivative thermogravimetry and “Endo” arrow indicated endothermic events.

Then, the chemical composition and material stability of Y₆Zr₁O₁₁ aerogel were further investigated using X-ray photoelectron spectroscopy (XPS), X-ray diffraction (XRD), Raman spectroscopy and thermogravimetric-differential scanning calorimetry (TG-DSC). The XPS spectra confirmed the expected elemental composition without any detectable impurity (Supplementary Fig. 21). The XRD patterns revealed a highly crystalline cubic Y₂O₃ phase without any phase transition (Fig. 3b). Owing to the incorporation of Zr, the diffraction peaks slightly shifted toward higher 2θ angles compared with pure Y₂O₃, which indicated that Zr⁴⁺ ions were substitutionally dissolved into the Y₂O₃ lattice to form the Y₆Zr₁O₁₁ grains. Notably, with increased annealing temperature, the diffraction peaks exhibited a gradual shift with a consistent decrease in the refined lattice parameter. This peak shift could be attributed to the progressive crystallization of the single-phase solid solution. Moreover, the diffusion kinetics of Zr⁴⁺ were further enlarged at elevated temperatures from 1000 °C to 1500 °C, triggering the reverse solution of Zr back into the Y₆Zr₁O₁₁ grains (lattice parameter of grains decreased from 10.57 Å to 10.55 Å) from the Zr-rich boundaries (Fig. 3d and Supplementary Fig. 22, Supplementary Table 1), which provided evidence for the inhabitation effect of GB segregation on grain growth of ceramic aerogels. Raman spectroscopy showed a single-phase solid solution in Y₆Zr₁O₁₁ aerogel across all annealing temperatures (Fig. 3e). The broad peaks of characteristic F_g + A_g mode (~378 cm⁻¹) was due to the solution of Zr⁴⁺ into the cubic Y₂O₃ crystals^51,52. TG-DSC analysis showed no endothermic/exothermic peaks and mass changes with elevated temperatures up to 1500 °C (Fig. 3f). These results jointly confirmed the high material stability of our Y₆Zr₁O₁₁ aerogel at extreme conditions.

Mechanical and thermal properties

As the individual ceramic fiber played a critical role in determining the bulk properties of aerogels, we firstly investigated the mechanical properties of single Y₆Zr₁O₁₁ fiber at room temperature. Owing to the GB segregation induced fine-grain enhancement, the fiber exhibited a dense and compact cross-section structure with a smooth and defect-free surface morphology after annealing at 1000 °C (Supplementary Figs. 11, 17). Under uniaxial tension, the fiber displayed a large fracture strength and strain up to 1.53 GPa and 8%, respectively (insets in Supplementary Fig. 23). The fiber also showed a good bendability under circular bending test (Supplementary Fig. 24), offering a robust building unit to construct the aerogel monolith. Then we investigated the compressive, tensile and bending properties of Y₆Zr₁O₁₁ aerogel using a universal mechanical testing system. For the compression test, the aerogel could be compressed from 2.0 cm to 0.4 cm in height and fully recovered to its original configuration after pressure release, presenting a large compressive strain of 80% (Fig. 4a, Supplementary Fig. 25 and Movie S1). Owing to the well assembled laminar fibrous structure (Fig. 1b), no lateral expansion or contraction was observed during compression test, indicating a near-zero Poisson’s ratio behavior of our aerogel. This compatible deformation pattern prevented the stress concentration and strain mismatch of aerogel network, realizing a high fatigue resistance. As a result, even after 500 compression cycles at 50% strain, the aerogel could retain a full recoverability with small stress degradation of only 2.6%, while the energy loss coefficient remained nearly constant (0.417-0.432) (Supplementary Fig. 26, 27). For the tension test, the Y₆Zr₁O₁₁ aerogel could withstand a nearly linear tensile strain up to 23.5% followed by a nonlinear tensile strain up to 60% (insets in Fig. 4b and Supplementary Movie S2). This high tensile ductility could be attributed to the high fracture resistance of ceramic fiber and the well assembled laminar network. Bending test of our fibrous aerogel were conducted using a typical two-point buckling method. The results showed that the aerogel exhibited high flexibility with a bending strain up to 90% without obvious structure fracture (Supplementary Fig. 28 and Supplementary Movie S3). We also evaluated the torsibility of aerogel by a 360° cyclic torsion setup. After ten cycles, the torsional stress reached a steady state with the residual stress ratio stabilized at ~60% (Fig. 4c and Supplementary Fig. 29, Supplementary Movie S4).

Fig. 4. Mechanical and thermal properties of Y₆Zr₁O₁₁ aerogel.

a Compressive stress-strain curves of Y₆Zr₁O₁₁ aerogel with strain up to 80%. b Tensile stress-strain curve. (Insets) Optical snapshots of the tensile test. c 360° cyclic torsional test. d Tensile test of single fiber calcinated at 1,500 °C for 2 h. e Long-term thermal stability test at 1500 °C for 24 h. (Insets) Optical images of the aerogel. f Stress-strain curves before and after 100 cycles of sharp thermal shocks. (Insets) Infrared images in one thermal shock cycle. g Optical images (left) and infrared images (right) of the aerogel under 80% compressive strain in bilateral flame of butane blowtorch. h Optical images of the aerogel under uniaxial tension in bilateral flame of butane blowtorch, scale bar in inset, 5 mm.

We next investigated the thermal stability of Y₆Zr₁O₁₁ aerogel (Fig. 4d-h). After calcination at 1500 °C for 2 h, due to the GB segregation triggered fine-grain effect, the Y₆Zr₁O₁₁ fibers showed an average grain size of only ~ 172 nm and retained a large fracture strength and strain up to ~ 1 GPa and 4.5%, which were higher than those of intermediate compositions (Y₈Zr₁O₁₄ and Y₄Zr₁O₈), and about twice as high as those of Y₂O₃ (no segregation) and Y₂Zr₂O₇ (phase transition) (Fig. 4d). Then we evaluated the thermal stability of Y₆Zr₁O₁₁ aerogel after long-term high-temperature exposure at 1500 °C in air (Fig. 4e). Owing to the fine-grain effect, the aerogel exhibited a volume shrinkage of only 2.37% after 24 h, while the intermediate samples showed shrinkage of 10–15%, and the Y₂O₃ aerogel with no segregation and Y₂Zr₂O₇ with phase transition exhibited much larger volume shrinkage exceeding 50% (Fig. 4e and Supplementary Fig. 30). In addition, a home-make pneumatic system was used to test the thermal stability under rapid thermal shocks between 1000 °C and 25 °C (Supplementary Fig. 31, 32). After 100 cycles of thermal shock, the aerogel retained its original morphology with negligible strength degradation (<1%), indicating a high tolerance to abrupt thermal fluctuations (insets in Fig. 4f and Supplementary Movie S5). Furthermore, under bilateral butane flame ablation (~1300 °C), the aerogel sustained compressive strains up to 80% with full recovery (Fig. 4g, Supplementary Fig. 33 and Movie S6). Meanwhile, the aerogel exhibited a near-linear tensile strain exceeding 20% under the same ablation condition (Fig. 4h, Supplementary Fig. 34 and Movie S7). Liquid-nitrogen immersion tests (−196 °C) also showed stable compressive recoverability of the aerogel (Supplementary Fig. 35, Movie S8), confirming its reliable thermomechanical performance across a broad temperature range.

Thermal superinsulation at high temperatures

Based on the excellent mechanical flexibility and thermal stability, we then evaluated the thermal insulating properties of Y₆Zr₁O₁₁ aerogel. Heat transfer from room to high temperatures in air depends on solid/air conduction, air convection and thermal radiation. Prior studies have demonstrated that when the bulk density of a porous fibrous assembly exceeded 24 mg cm⁻³, the confined pores were sufficiently small to suppress air convection⁵³. Therefore, our aerogels, with a density of 50 mg cm⁻³, could be identified as a convection-free monolith, in which solid/air conduction and thermal radiation were considered in the overall heat transfer.

At high temperatures >500 °C, thermal radiation rose rapidly with the third power of temperature (${\kappa }_r=T_r^3{\rho }^{-1}{e}^{-1}$)²⁶, becoming the dominant factor in heat transfer. To this end, radiation absorption or reflection should be critically considered in design of thermal insulating materials at extreme conditions. According to the Wien’s displacement law⁵⁴, thermal radiation at high temperatures primarily lied within the near-infrared (NIR) region (780–2500 nm). Then we measured the UV-Vis-NIR diffuse reflectance spectrum of Y₆Zr₁O₁₁ aerogel, which exhibited a reflectivity up to 98.12% in the NIR region (Fig. 5a). The Y₂O₃ and Y₈Zr₁O₁₄ aerogels showed slightly higher reflectivity due to the higher contents of yttrium, but they exhibited lower thermal stability at extreme conditions as reflected by high-temperature volume shrinkages of about 65% and 10%, respectively. These results confirmed the intrinsic infrared-reflective advantage of yttria, which could substantially improve the thermal insulating capability of aerogel at high temperatures. In addition, the fiber diameter of Y₆Zr₁O₁₁ aerogel were around 2 μm, which were comparable to the NIR wavelength, leading to a resonant Mie scattering⁵⁵ with enhanced NIR reflection.

Fig. 5. Thermal insulation performances of Y₆Zr₁O₁₁ aerogel.

a UV-Vis-NIR reflection spectra of ceramic aerogels. b Schematic illustration of radiation reflection and solid conduction in fibrous aerogels with lamellar and random structure. c κ of Y₆Zr₁O₁₁ aerogel from 100 °C to 1000 °C. (Insets) SEM images of fiber morphology inside the aerogel. Error bars indicated standard deviations of κ. d Comparative analysis of the thermal conductivity and working temperature with previously reported aerogels under ρ ≤ 50 mg cm⁻³ at 1000 °C. The corresponding data were provided in Supplementary Table 2. e Schematic demonstration of a practical application in spacecraft and infrared images of the backside of Y₆Zr₁O₁₁ aerogel as a TIL. f Temperature difference (front vs. back side) versus heating time for Y₆Zr₁O₁₁ fibrous aerogel in the thermal insulation test. g Comparative analysis of backside temperature of different ceramic aerogels.

Compared to the randomly distributed fibrous structure of conventional ceramic aerogels^56,57, the well assembled lamellar structure in our Y₆Zr₁O₁₁ aerogel could induce an enhanced layer-by-layer radiation reflection effect (Fig. 5b and Supplementary Fig. 36). This layer-by-layer reflection effect could effectively reduce the penetration depth of thermal radiation, while the randomly distributed structure needed more thickness and higher density to obtain the same result (Supplementary Figs. 37, 38). In addition, this lamellar structure could also mitigate the solid conduction of fibrous aerogel. The solid conduction pathways along fibers within each layer were mainly perpendicular to the heat flux to endow a low out-of-plane heat conduction, while the significant GB segregation behaviors within fibers and physical connecting points among fibers could further decrease the in-plane heat propagation. As a result, our Y₆Zr₁O₁₁ aerogel demonstrated a low κ of 28 mW m⁻¹ K⁻¹ at 25 °C and only 89 mW m^-1 K^-1 at 1000 °C with density of 50 mg cm^-3 (Fig. 5c). Owing to the GB segregation triggered high thermal stability and radiation reflection induced low κ, our Y₆Zr₁O₁₁ aerogel demonstrated a more reliable thermal insulating performance at extreme conditions (1500 °C) (Fig. 5d). Conventional ceramic aerogels such as silica and alumina showed a lower κ at 25 °C but limited working temperature below 1000 °C^58–62, while zirconate and mullite showed a higher κ at 1000 °C with working temperature below 1400 °C^63–66.

To evaluate the practical thermal insulation performance at extreme conditions, the Y₆Zr₁O₁₁ aerogel was integrated as a thermal insulation layer (TIL) into the thermal protection system (TPS) model of spacecraft. For the next-generation reusable spacecrafts, the TPS acted as a vital shield against severe aerodynamic heating during atmospheric entry and reentry (Fig. 5e). However, the high-speed airflow impingement and high-temperature exposure may induce a serious volume shrinkage or structural collapse of TIL, leading to a catastrophic thermal runaway. To this challenge, a Y₆Zr₁O₁₁ aerogel TIL, with a typical thickness of 3 cm and shape of hexagon, was subjected to a butane flame ablation (~1300 °C) to test its insulating performance (Fig. 5e and Supplementary Fig. 39a). After 30 mins of flame ablation, the backside temperature remained below 100 °C (maximum 97.2 °C) with TIL density of only 50 mg cm^-3, presenting a large temperature attenuation exceeding 1150 °C between the front and back sides (Fig. 5f). In contrast, the intermediate compositions (Y₈Zr₁O₁₄ and Y₄Zr₁O₈) showed surface shrinkage during ablation, and the Y₂O₃ (no segregation), Y₂Zr₂O₇ (phase transition), and conventional 3YSZ aerogels exhibited severe shrinkage and surface damage (Supplementary Fig. 40). As a result, we observed dramatic temperature rises up to 146.9 °C, 136.7 °C, 121.3 °C, 117.9 °C, 109.8 °C at the backsides of Y₂O₃, 3YSZ, Y₂Zr₂O₇, Y₄Zr₁O₈, and Y₈Zr₁O₁₄ TIL, respectively, which were due to the degraded thermal insulation performance or thermal runaway induced by structural damages (Fig. 5g and Supplementary Fig. 41). Moreover, to test its thermal insulation performance under mechanical loadings, a 3 cm thickness of Y₆Zr₁O₁₁ aerogel was exposed to ~1300 °C flame from the bottom under vibration conditions (100 cycles of 50% compressive strain within 500 s). The top side temperature remained around ~78 °C (Supplementary Figs. 39b, 42), demonstrating an excellent thermal insulation performance under high-temperature and mechanical-loading conditions. Therefore, benefited from the GB segregation triggered high thermal stability and radiation reflection induced low κ, our Y₆Zr₁O₁₁ aerogel presented a reliable thermal insulation performance at extreme conditions ready for practical applications.

Discussion

Together, we have shown that a GB segregation design and synthesis of Y₆Zr₁O₁₁ aerogels can realize a high thermal stability up to 1500 °C by the Zr segregation triggered fine-grain effect and a low thermal conductivity of only 89 mW m⁻¹ K⁻¹ at 1000 °C by high thermal radiation reflectivity. The resulting aerogel features a unique combination of good mechanical flexibility, high thermal stability and reliable thermal insulation performance beyond the reach of typical aerogels, thus defining an attractive material for thermal insulation under extreme conditions, such as aerospace, military, and high-temperature industries. Moreover, this GB segregation design also provides opportunities to develop high-performance structural and functional ceramics for a wide spectrum of practical applications.

Methods

Materials

Zirconia chloride (ZrOCl₂·8H₂O, 99.9%), yttrium chloride (YCl₃·6H₂O, 99.9%), acetylacetone (C₅H₈O₂, AR), tetrahydrofuran (C₄H₈O, AR), triethylamine (C₆H₁₅N, AR), polyvinylpyrrolidone (PVP, Mv≈1,300,000), and methanol were purchased from Macklin Co., Ltd, China. All these chemicals were used as received without further purification.

PAZ solution was produced from a mixture with a molar composition of zirconia chloride: acetylacetone: triethylamine: tetrahydrofuran: methanol=1: (about1−1.35): (about 20-23): (about 60-65): 135. The mixture was stirred for 10 h to get a transparent solution. Then, the solution was rotary evaporated at 50 °C for 1 h to obtain PAZ powder.

PAY solution was produced from a mixture with a molar composition of yttrium chloride: acetylacetone: tetrahydrofuran: triethylamine: methanol=1: (about1.55−1.75): (about 40-43): (about 60-65): 135. Then, the solution was rotary evaporated at 75 °C for 3 h to obtain PAY powder.

Preparation of aerogels

The precursor solution was prepared by dissolving PAZ and PAY in methanol with a molar ratio of Y:Zr = 1:0 (Y₂O₃), 10:1, 9:1, 8:1, 7:1, 6:1, 5:1, 4:1, 3:1, 2:1, 1:1 (Y₂Zr₂O₇). Polyvinylpyrrolidone (PVP) was then added to the precursor solution at a precursor sol/PVP weight ratio of 40:1 and a temperature of 60 °C, followed by stirring for 2 h to obtain a golden transparent solution with a viscosity of 10 Pa s.

Subsequently, the self-developed airflow-assisted centrifugal spinning method was used for the preparation of aerogels. By adopting a structural aerodynamic grooving design on the wall of the cylindrical spinning disk, the device can spontaneously generate a strong outward-expanding airflow field exceeding 8 m s^-1 during high-speed rotation. Four spinneret orifices (0.3 mm in diameter) were uniformly fabricated on the spinning disk wall. The spinning solution was pumped into the high-speed rotating spinning disk via a liquid delivery tube, then ejected from the spinneret orifices by centrifugal force, dried into fibers in air, and deposited on the collection wall (with the spinning disk as the center and a radius of 60 cm), thus forming a lamellar fibrous aerogel structure.

Finally, the collected aerogels were neatly stacked and subjected to a sintering process in a muffle furnace under flowing air. The temperature was ramped up to 1000 °C at a heating rate of 2 °C per minute and holding for 2 h to fully ablate and remove organic components, complete crystallization, and form oxide ceramics. Unless otherwise specified, all specimens in mechanical and thermal tests were prepared using samples with a density of 50 mg cm⁻³.

Density functional theory calculations

To investigate the atomistic origin of Zr segregation and its influence on the GB energetics, we performed ab initio simulations with density functional theory (DFT)⁶⁷ methods by using Quickstep⁶⁸ module of the CP2K⁶⁹ program. A symmetric Σ3{111} GB model was constructed for the cubic Y₆Zr₁O₁₁ composition as shown in Supplementary Fig. 13a. The simulation cell had dimensions of approximately 37 × 15 × 13 ų and contained 494 atoms. This model served as the initial configuration for subsequent segregation-energy analysis. The initial GB model was first fully relaxed using DFT. Based on the optimized configuration, four Zr atoms located at a distance of approximately 6 Å from the GB plane were selected. Each of these atoms was successively exchanged with a neighboring Y atom positioned closer to the boundary, generating a series of configurations in which the substituted Zr atom occupied sites at intermediate distances (~3 Å) and directly at the GB. All exchanged structures were subsequently re-optimized to ensure local energetic stability. The relative segregation energy E_seg was then evaluated as the energy difference between the exchanged configurations and the initial configuration defined as,

$$E_{seg}=E_{exchanged}-E_{init}$$ 1

The resulting total energies of all configurations relative to the initial structure, together with the fitted dependence of E_seg on the distance from the GB, are summarized in Supplementary Fig. 13b.

The exchange-correlation interactions were treated with generalized gradient approximation (GGA) using Perdew–Burke–Ernzerhof (PBE)⁷⁰ functional. Double-ζ polarization quality Gaussian basis sets introduced by VandeVondele (DZVP-MOLOPOT-SR-GTH)⁷¹ were used. The charge density cutoff of the finest grid level was set to 800 Ry and 5 multigrid levels were employed. The convergence criterion of self-consistent field energy was set to 1.0 × 10^-6eV. Structural optimization was performed by simultaneous relaxation of atomic positions and lattice parameters using the direct cell optimization scheme. Periodic boundary conditions were applied in all directions. Geometry relaxations were carried out with the BFGS optimizer under an external pressure of 1 atm. During the optimization, the cell angles were fixed, whereas the lattice constants and internal atomic coordinates were fully relaxed. Convergence thresholds of 0.023 eV/Å and 0.0016 Å were adopted for the maximum forces and atomic displacements, respectively.

Hybrid Monte Carlo-molecular dynamics simulations

To investigate large-scale cation segregation behavior beyond the size limitation of DFT calculations, hybrid MCMD simulations were performed with the GPUMD⁷² package. The universal neuroevolution potential NEP89⁴⁵ was employed to accurately describe interatomic interactions. The temperature and pressure were regulated using a Nosé-Hoover thermostat⁷³ with a relaxation time of 0.1 ps and a Parrinello-Rahman barostat⁷⁴ with a relaxation time of 10 ps to maintain the target isothermal-isobaric (NPT) conditions. A time step of 1 fs was used to integrate the equations of motion.

To start with, a series of symmetric Σ5{310} GB models with cation compositions ranging from Y₁₀Zr₁ to Y₁Zr₁ were constructed. The simulation cells had dimensions of 100×31×32 ų and contained approximate 8000 atoms. During the construction of the initial configurations, the spatial distribution of Zr was controlled to be homogeneous along the direction normal to the GB in order to avoid artificial segregation bias.

Each system was first equilibrated by conventional MD for 200 ps to relax the GB structure and remove residual stresses. After equilibration, hybrid MCMD simulations were conducted for 10 ns to enable long-range cation redistribution and capture segregation kinetics. In the hybrid scheme, MD evolution was periodically interrupted by canonical MC swap attempts between randomly selected Zr and Y cations. Specifically, 10 trial swap moves were attempted every 100 MD steps at temperature of 1273 K. Accepted swaps were determined according to the Metropolis⁷⁵ criterion.

To quantify segregation, we first defined one-dimensional atomic density profiles ${\rho }_{Zr}\left(x\right)$ along the grain-boundary normal by superimposing Gaussian functions centered at the x-coordinates of individual Zr and cation atoms. The local Zr fraction $f_{Zr}\left(x\right)$ was defined as the ratio of Zr density to total cation density at each position,

$$f_{Zr}\left(x\right)={\rho }_{Zr}\left(x\right)/{\rho }_{cations}\left(x\right)$$ 2

During the MCMD simulation, Zr atoms gradually accumulated near the GB and eventually reached a quasi-stationary distribution (Fig. 2d). To robustly identify the GB region and minimize the influence of boundary migration, the GB Zr fraction $f_{Zr}^{GB}$ was defined as the average value within the top 15% region of the $f_{Zr}\left(x\right)$ profile. Considering the GB fraction in the present MCMD simulation cell is considerably larger than that in real polycrystalline materials, a sufficiently bulk-dominated region cannot be rigorously isolated from the model. Therefore, the bulk reference value $f_{Zr}^{Bulk}$ was approximated by the spatial average of $f_{Zr}\left(x\right)$ over the entire system. As shown in Supplementary Figs. 14, 15, both the segregation metric $f_{Zr}^{GB}-f_{Zr}^{Bulk}$ and the associated energy change ΔE converge after 10 ns of hybrid MCMD simulation, indicating that the segregation state is statistically well sampled and thermodynamically stabilized.

Computational fluid dynamics simulation

In the numerical simulation, the Multiple Reference Frame (MRF) formulation was employed as the steady-state modeling framework to characterize the flow generated by a rotating disk inside a sealed test chamber. The computational domain consisted of a closed rectangular outer chamber (1000 × 1000 × 1000 mm³) and a cylindrical rotating sub-domain enclosing the disk (Supplementary Fig. 2), which was operated at 10,000 rpm (Fig. 1a and Supplementary Fig. 3). The two zones were coupled through a general grid interface (GGI) using a frozen-rotor treatment to retain circumferential non-uniformity in a steady manner. The finite-volume method was used in the solver, and pressure-velocity coupling employed the SIMPLE algorithm. Gradients were computed using a least-squares cell-based approach. Second-order upwind schemes were used to discretise the convective terms in both the momentum and turbulence equations, while pressure was interpolated with a second-order scheme. The SST k-ω turbulence model was adopted to improve predictions in near-wall and separated-flow regions.

Characterization and performance tests

The morphology of the aerogels and fibers was characterized using scanning electron microscopy (SEM, ZEISS Sigma 560) equipped with energy-dispersive X-ray spectroscopy (EDS) mapping, and transmission electron microscopy (TEM, FEI Talos F200X G2). Structural analysis was performed by X-ray diffraction (XRD, Panalytical X’Pert Pro diffractometer), focused ion beam-scanning electron microscopy (FIB-SEM, FEI Helios 5 CX), and aberration-corrected scanning transmission electron microscopy (AC-STEM, FEI Spectra 300, 300 kV).Chemical compositions and phase states were analyzed using X-ray photoelectron spectroscopy (XPS, FEI K-Alpha), thermogravimetry-differential scanning calorimetry (TG-DSC, Netzsch STA449F3), and Raman spectroscopy (Horiba LabRAM HR Evolution).

Ceramic fibers used for characterization and analysis were randomly selected from the aerogel samples to ensure statistical reliability, and more than 300 grains were measured for each composition. Grain size quantification was performed using Nano Measure 1.2 software, which automatically generated the statistical results, including average size and distribution width.

The tensile strength of individual ceramic fibers was measured with an in-situ micro/nano tensile testing system (Suzhou ShengTe Technology Co., Ltd.). Torsional performance was evaluated using a multimodal mechanical testing platform for flexible materials and devices (Suzhou ShengTe Technology Co., Ltd.). The compressive and tensile properties of the fibrous aerogels were tested on a universal mechanical testing machine (Instron 3365) equipped with a 100 N load cell at a loading rate of 10 mm min^-1.

Thermal shock performance was examined in a self-designed pneumatic thermal shock testing system, in which the samples were cyclically transferred between high- and low-temperature zones (1000 °C and 25 °C). Additional high-temperature exposure tests were carried out in a tube furnace (OTF-1,200X, HF-Kejing). UV-Visible-near-infrared (UV-Vis-NIR) diffuse reflectance spectra were recorded using a Shimadzu UV-3600i Plus spectrophotometer with a polytetrafluoroethylene (PTFE) backing.

Thermal conductivity (κ) was measured using the plate heat-flux meter method (DRPL-III, DRPL-V; Xiangtan Xiangyi Instrument Co., Ltd.). The practical thermal insulation performance was evaluated under direct butane-torch ablation (~ 1300 °C), with surface temperatures recorded via thermocouples and corresponding infrared images captured by a FLIR A615 thermal imager.

Author contributions

X.X., H.L. and J.G. designed the research and advised on manuscript preparation. Z.Y. and Y.Z. conducted experiments with assistance from H.Z., S.D., D.W., C.S., H.Y., J.Z., D.L., J.C., C.L., T.L., H.Y.Z., S.Z.; Y.D. conducted the DFT calculations and MCMD simulations. J.X. and W.C. conducted the computational fluid dynamics simulation. X.X. and Z.Y. analyzed the data and wrote the manuscript with input from all co-authors. All authors reviewed and commented on the manuscript.

Peer review

Peer review information

Nature Communications thanks Jessica Krogstad, Hongjie Wang, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This research was supported by the National Natural Science Foundation of China (Grant No. 52322803), the Scientific Research Innovation Capability Support Project for Young Faculty (Grant No. ZYGXQNJSKYCXNLZCXM-M8), the Key Program of National Natural Science Foundation of China (Grant No. 52192661) and the Fundamental Research Funds for the Central Universities (HIT. DZJJ. 2023139).

Data availability

All other data supporting the findings of this study are included within this article and its Supplementary files. 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 https://doi.org/10.1038/s41467-026-77371-w.