Atomic armor for thermal stability in nanoporous structures

Significance

Sintering has long limited the thermal stability and practical application of nanoporous materials. An atomic-level modification strategy that significantly suppresses sintering is introduced. This approach not only enhances the structural integrity of nanoporous structures but also unveils a fundamental mechanism for improving thermal stability in nanoporous structures. The present findings offer a transformative method for designing thermally robust nanoporous systems and pave the way for the rational selection and engineering of materials with tailored stability at the atomic scale. These insights open broad avenues in the development of advanced nanoporous materials for high-performance applications across catalysis, energy, and beyond. Also, it attracts interest in the glassy properties of nonglass formers at the atomic level.

Abstract

Nanoporous structures play a critical role in a wide range of applications, including catalysis, thermoelectrics, energy storage, gas adsorption, and thermal insulation. However, their thermal instability remains a persistent challenge. Inspired by the extraordinary resilience of tardigrades, an “atomic armor” strategy is introduced to enhance the stability of nanoporous structures. Applied to mesoporous silica at parts-per-million levels, the atomic armor provides thermal resistance exceeding that of existing stabilization techniques. Thermal treatment at 1,000 °C for 168 h results in a fivefold increase in specific surface area, 66% lower thermal conductivity, and a sixfold increase in pore volume compared to untreated samples. Surface viscosity is linked to sintering resistance, and glass transition temperature and fragility are introduced as design parameters. Machine-learned interatomic potentials and metabasin escape algorithm-assisted molecular dynamics simulations are employed to reveal that materials traditionally classified as nonglass formers can exhibit glass transition temperatures and display intrinsic fragility. Alumina is identified as having a record-high glass transition temperature. By modulating the surface viscosity of nanoparticles, this approach stabilizes nanoporous structures effectively. The proposed method offers a simple and universal posttreatment process for improving the thermal stability of nanoporous structures.

Nanoporous structures exhibit numerous desirable properties, such as high surface area, lightweight, selective adsorption, and low thermal conductivity, making them highly versatile across a broad range of applications, including catalysis (1–3), thermoelectric (4–6), energy storage (7–9), gas adsorption (10–12), and thermal insulation (13–15). These properties arise from the nano-scale pores and surfaces formed by interconnected nanoparticles. However, at elevated temperatures, the thermal stability of the nanoparticles is significantly compromised due to sintering and phase transitions. To address this, stabilization techniques such as doping (16–18) and mixing (19–21) with foreign elements have been extensively explored. Despite significant efforts, progress has plateaued, highlighting the need for new approaches and deeper mechanistic understanding.

Recently, nano armoring approaches have been applied to realize water oil separation membranes (22), functionalize single living cells (23), and enhance amelioration of antibiotic-associated diarrhea (24), which inspire us to explore the possibility of applying nano armor for thermal stability. In this context, the remarkable thermal resilience of tardigrades (Fig. 1A), microscopic organisms capable of surviving extreme temperatures up to 151 °C (25), provide valuable insights. Unlike most organisms, which can only tolerate temperatures up to ~60 °C, tardigrades employ intrinsically disordered proteins (TDPs) as protective molecular armor to shield cellular structures from heat-induced damage (26–28). These proteins adhere to cellular organelles, preventing thermal degradation by stabilizing bioactive structures (Fig. 1B). The increased thermal tolerance of tardigrades highlights two key factors: amorphous states and unique surface properties. This observation raises the possibility that similar principles could be applied to improve the thermal stability of nanoporous materials.

Fig. 1.

Bioinspired pathway of amorphous nonglass atomic armor. (A) Comparative thermal stability between tardigrades and their cells. (B) Identification of TDPs as the key molecular stabilizers of cell organelles. (C) Atomic armor concept inspired by Tardigrades. (D) Structure of mesoporous silica highlighting their limited thermal stability. (E) Divergence between bulk and surface viscosity in silica. (F) T_g of known glass formers (35), with values observed to have the highest T_g in this work.

To explore whether the thermal stability insights from tardigrades can provide a guideline for nanoporous material enhancement, we applied the concept of “atomic armor” (Fig. 1C) to mesoporous silica (Fig. 1D), a widely used material across multiple fields (29–32), which serves as an ideal prototype due to their small particle size and the highest known glass transition temperature among glass-forming materials. By addressing these extremes, we aim to advance the fundamental understanding of nanoporous material stability and develop universal stabilization strategies.

Unlike crystalline materials, whose thermal stability is determined by their melting points, the stability of amorphous materials depends on their viscosity, relative to the glass transition temperature (T_g), which marks the transition between soft rubbery (liquid-like) and hard brittle (solid-like) states. Notably, the viscosity of amorphous surfaces differs significantly from that of the bulk. Using a surface viscosity model (33), we demonstrated that the silica surface becomes liquid-like at approximately 800 °C, aligning well with experimental findings (34) (Fig. 1E). This indicates that the thermal stability of amorphous nanoporous materials and nanoporous materials with amorphous surfaces is fundamentally linked to surface viscosity. Consequently, we hypothesize that an atomic armor approach stabilizes nanoporous materials by modulating surface viscosity.

This prompts the question: Can we design atomic-level armor for silica nanoparticles to tune surface viscosity effectively? A straightforward approach might involve identifying a material with a higher T_g than silica. However, it is impractical, as silica already possesses the highest T_g among glass-forming materials (35) (Fig. 1F). Nonglass formers, on the other hand, crystallize at high temperatures before their T_g can even be evaluated. A critical insight addressing this challenge is that nonglass formers could remain amorphous at atomic scales (<10 nm) under high-temperature conditions. This observation opens the possibility of using nonglass formers as potential candidates for creating such atomic-level armor.

In this study, we demonstrate that applying atomic armor at the parts-per-million (ppm) level significantly enhances the thermal stability of silica nanoparticles. Furthermore, the present findings reveal the existence of materials with a higher T_g than silica and the inherent fragility of nonglass formers. By leveraging the combined effects of high T_g and fragility, we propose a unique approach to control surface viscosity, enabling effective stabilization of nanoparticles.

Thermal Stability Enhancement by Atomic Armor.

To identify potential candidates for enhancing the thermal stability of mesoporous silica, we employed a hybrid approach combining molecular dynamics (MD) simulations and big data analysis. This enables the efficient screening of potential materials from a vast dataset (SI Appendix, Fig. S1, Table S1, and Supporting Information Text). Among the candidates, Al₂O₃, ZrO₂, and HfO₂ emerged as the most promising. Due to the similar properties of ZrO₂ and HfO₂, further investigation was concentrated on Al₂O₃ and ZrO₂ (SI Appendix, Fig. S1). Given the nanoscale dimensions of silica nanoparticles, atomic layer deposition was chosen to apply the atomic armor, owing to its precision in depositing materials at the atomic scale. We then demonstrate the improvements in thermal conductivity, specific surface area, pore volume, and morphology, key characteristics for applications in catalysis, adsorption, and thermal insulation.

The application of atomic armor had a negligible impact on the thermal conductivity of mesoporous silica, with changes of less than 0.002 W m⁻¹ K⁻¹, remaining below the conductivity of free air (Fig. 2A). Initially, the specific surface area of armored mesoporous silica exceeded that of pure mesoporous silica. However, after approximately 30 deposition cycles, the specific surface area decreased to below that of unarmored mesoporous silica. This suggests a progressive construction process for the atomic armor: initially forming particle-wise, subsequently merging into a continuous film, and thickening with additional deposition cycles (Fig. 2B).

Fig. 2.

Characterization of mesoporous silica with and without armor. (A) Thermal conductivity of atomic armored mesoporous silica with different deposition layers and elements. The dashed line indicates the thermal conductivity of air. (B) Specific surface area of atomic armored mesoporous silica with different deposition layers and elements. (C) Specific surface area of atomic armored mesoporous silica after heat treatment under different temperatures, compared with mesoporous structures with other stabilization methods in the last 5 y (SI Appendix, Table S2). The filled symbols indicate a heat treatment time of 2 h or more, while the hollow symbols represent a heat treatment time of less than 2 h. (D and E) Thermal conductivity of atomic armored mesoporous silica with different deposition layers and elements after long-term heat treatment under 900 °C (D) and 1,000 °C (E). (F) Thermal conductivity of mesoporous silica with and without armor after heat treatment under 1,100 °C (Top) and 1,200 °C (Bottom).

High-temperature treatments were conducted to assess the effectiveness of the atomic armor. After 2 h of heat treatment, the armored mesoporous silica consistently exhibited lower thermal conductivity compared to mesoporous silica (SI Appendix, Fig. S2A). Compared to mesoporous silica with alternative thermal stabilization methods, the atomic-armored mesoporous silica demonstrated the highest specific surface area at temperatures up to 1,100 °C and maintained superior performance even at 1,200 °C (Fig. 2C and SI Appendix, Table S2). These results confirm the effectiveness of the atomic armor approach in addressing the key factor of thermal instability, providing a robust method to inhibit the sintering process and significantly enhancing the thermal stability of silica nanoparticles.

While most studies assess thermal stability using heat treatments of 2 h or less, long-term evaluation is essential for understanding the performance of high-temperature materials, particularly in commercial applications. To address this, we extended the heat treatment duration (t_heat) to 168 h to evaluate near-converging, long-term performance.

Extended t_heat at 900 and 1,000 °C corroborated the findings derived from the specific surface area measurements across different deposition cycles. After approximately 30 deposition cycles, the surface of silica nanoparticles was fully modified. Before reaching this threshold, the antisintering performance showed divergence with increasing t_heat. Beyond this point, performance began to converge (Fig. 2 D and E). To further validate the formation of the atomic armor, the morphology change of 100 nm silica nanoparticles with different atomic layer deposition (ALD) cycles was observed under transmission electron microscopy (TEM) (SI Appendix, Fig. S3). Before 30 cycles, some particle fusion is observed, which shows that partial coverage leads to incomplete sintering resistance. However, after 30 cycles, the nanoparticles retain their discrete morphology after high-temperature treatment, clearly indicating the formation of a protective and continuous armor. These results help illustrate the evolution and effectiveness of the atomic armor layer. Note that the needed cycles to form film-wise atomic armor are consistent with those of mesoporous silica.

Interestingly, mesoporous silica with 5-cycle ZrO₂ atomic armor exhibited higher thermal conductivity than mesoporous silica at 1,000 °C (Fig. 2E). This suggests that effective antisintering requires more than simply mixing elements. Inappropriately applied foreign elements can act as glass modifiers, introducing nonbridged oxygen species that reduce viscosity and, consequently, accelerate sintering. This highlights the critical importance of precise application in achieving the desired thermal stability.

After 168 h of t_heat at 900 and 1,000 °C, the Al₂O₃ atomic armored mesoporous silica exhibited lower thermal conductivity by 59% and 66%, respectively, compared to mesoporous silica. In contrast, ZrO₂ atomic-armored mesoporous silica showed only 38% and 16% lower thermal conductivity under the same conditions. The superior inhibition performance of Al₂O₃, despite ZrO₂’ higher melting temperature, is particularly intriguing. A higher melting temperature usually suggests better thermal stability. However, this discrepancy highlights that T_g is not directly correlated with melting temperature. This finding highlights the need for further investigation into the relationship among T_g, melting temperature, and their respective influences on thermal stability.

Heat treatments at 1,100 and 1,200 °C were exclusively conducted on mesoporous silica and Al₂O₃ atomic armored mesoporous silica with 120 deposition cycles. The armored mesoporous silica demonstrated thermal conductivity reductions of 56% and 45%, respectively, compared to mesoporous silica after prolonged heat treatment (Fig. 2F). This remarkable enhancement in thermal stability, even beyond the T_g of silica, indicates the observation of a material with a higher T_g. The pronounced difference in thermal conductivity between armored and unarmored mesoporous silica after extended t_heat (SI Appendix, Fig. S2B) highlights the substantial impact of atomic armor on thermal stability. The reduced thermal conductivity of the atomic armored mesoporous silica is attributed to its enhanced antisintering properties, which preserve smaller nanoparticle sizes and a greater number of mesopores. This conclusion is further validated and visualized through Brunauer–Emmett–Teller (BET) analysis (Fig. 3A) and scanning electron microscopy (SEM) imaging (Fig. 3B).

Fig. 3.

Structural change of mesoporous silica with and without armor. (A) Specific surface area of mesoporous silica with and without armor after heat treatment under different temperatures. (B) SEM images of mesoporous silica with and without atomic armor after heat treatment under different temperatures. Note that the heat treatment durations are 168, 168, 8, and 2 h for 900, 1,000, 1,100, and 1,200 °C, respectively.

After heat treatment, mesoporous silica with atomic armor exhibits a significantly higher specific surface area compared to unarmored mesoporous silica (Fig. 3A). This difference becomes more pronounced at higher temperatures and after extended t_heat. Specifically, the specific surface area of the armored mesoporous silica is 2.0, 5.1, 8.3, and 57.2 times higher than unarmored mesoporous silica at 900, 1,000, 1,100, and 1,200 °C, respectively. This exceptionally high enhancement at 1,200 °C results from the rapid densification of silica above its T_g, highlighting the significant improvement in thermal stability conferred by the atomic armor. The high specific surface area values are attributed to well-maintained nanoparticles and small connections (Fig. 3B). High specific surface area values are typically associated with well-maintained mesopores. Following heat treatment, the pore volume of atomic armored mesoporous silica is 1.8, 5.9, 6.8, and 10.7 times higher than that of mesoporous silica at 900, 1,000, 1,100, and 1,200 °C, respectively (SI Appendix, Fig. S4). Enhanced thermal stability is further evidenced by the pore distribution of mesoporous silica after heat treatment from 1,000 to 1,200 °C. Armored mesoporous silica retain a substantial number of mesopores, whereas mesoporous silica loses most of their mesopores, becoming nearly dense (Fig. 3B).

Atomic Armor Characterization and Antisintering Mechanism.

These substantial performance improvements achieved through the application of atomic armor, not only outperform those of other stabilization methods but also provide significantly higher long-term thermal stability compared to mesoporous silica. While the enhanced thermal stability provided by the atomic armor has been extensively validated, and preliminary insights into the underlying mechanisms have been proposed, the complete mechanism remains to be fully elucidated. The following section explores the antisintering mechanism facilitated by the atomic armor, providing a deeper understanding of its role in stabilizing the nanoporous structure.

Before discussing the antisintering mechanism, it is essential to verify key preconditions of the atomic armor theory. For the atomic armor to effectively tune surface viscosity, it must be amorphous and uniformly deposited on silica nanoparticles. X-ray diffraction (XRD) results confirm the amorphous nature of the atomic armor, as no characteristic diffraction peaks are observed prior to heat treatment (Fig. 4A). Energy-dispersive X-ray spectroscopy (EDS) imaging further demonstrates uniform deposition of both Al₂O₃ (Fig. 4B) and ZrO₂ (Fig. 4C) on the mesoporous silica. To better visualize the uniformity and extent of surface coverage, the atomic armor approach was also applied to 100 nm silica nanoparticles. These particles can be clearly observed under lower magnification levels, allowing for high-resolution EDS mapping. The alumina armor appears uniformly distributed across the surface of the nanoparticles, providing clearer evidence of consistent surface coverage (SI Appendix, Fig. S5).

Fig. 4.

Structure and enhancement mechanism of atomic armor. (A) XRD of mesoporous silica with and without armor before and after heat treatment. (B and C) EDS images of Al₂O₃ (B) and ZrO₂ (C) atomic armored mesoporous silica before heat treatment. (D and E) EDS (D) and HRTEM and FFT (E) images of atomic armored mesoporous silica after heat treatment. (F) Specific surface area of mesoporous silica with and without atomic armor after heat treatment. (G) Viscosity–temperature relationship of SiO₂, Al₂O₃, and ZrO₂ for bulk and surface.

Fourier transform infrared (FTIR) spectroscopy and solid-state NMR (SSNMR) analyses on both mesoporous silica and 100 nm silica nanoparticles were conducted to gain a deeper understanding of the atomic interface formed by deposition. The FTIR spectra (SI Appendix, Fig. S6) of both materials showed characteristic bands for silica and surface hydroxyls (36, 37). However, no pronounced spectral changes were observed postdeposition. This is likely due to overlapping absorption bands between silica and alumina, the amorphous nature of the deposited alumina, and its low concentration. Notably, literature reports have shown that even in the systems with a significantly higher aluminum content (Si: Al = 3:1), only minor peak shifts were observed in the 300 to 1,400 cm⁻¹ region (38). According to inductively coupled plasma (ICP) analysis, the foreign element weight content of the atomic armored mesoporous silica is at ppm levels (SI Appendix, Table S3), showing that the Si: Al ratio is approximately 400:1, which further limits the detectability of interfacial changes by FTIR.

In contrast, SSNMR provided more understanding of the surface modification. The ²⁹Si spectra (SI Appendix, Fig. S7) revealed an increase in Q⁴ species (SI Appendix, Table S4) and a shift of the main peak toward higher chemical shifts after alumina deposition, indicating the transformation of surface Q³ species into Q⁴_(1Al) species. This confirms successful interfacial interaction between alumina and silica. Furthermore, ²⁷Al SSNMR (SI Appendix, Fig. S8) revealed that the silica–alumina interface is primarily composed of ^[5]Al_(1Si) and ^[6]Al_(1Si) environments, with a minor presence of ^[6]Al_(2Si). The spectral features closely resemble those reported for amorphous alumina (39), further supporting the amorphous character of the deposited layer. Also, based on the percentage of interfacial species (those connected to Si) in the ²⁷Al SSNMR result (SI Appendix, Table S5), the thickness of the atomic armor is estimated to be around 1 to 2 nm thick.

Additionally, after high-temperature heat treatment, the atomic armor retains its amorphous structure. This is confirmed by the continued absence of characteristic diffraction peaks in XRD (Fig. 4A), the lack of lattice fringes in high-resolution TEM (HRTEM) images, and the absence of diffraction spots in fast Fourier transform (FFT) images (Fig. 4 D and E and SI Appendix, Fig. S9). These findings validate the fundamental requirements for the atomic armor to function effectively.

The sintering of amorphous nanoparticles arises from the surface flow of atoms, a process closely tied to surface viscosity. Therefore, the key to inhibiting sintering lies in effectively tuning the surface viscosity of the silica nanoparticles. Viscosity is temperature-dependent, and this relationship varies across materials. Two primary factors influence this behavior: T_g defined as the point where viscosity reaches 10¹² Pa s and fragility which determines how sharply viscosity increases near T_g. Higher fragility corresponds to a steeper viscosity–temperature relationship near T_g.

Al₂O₃ and ZrO₂ atomic armors enhance the thermal stability of mesoporous silica to 1,200 °C (Fig. 2F) and 1,000 °C (Fig. 4F). The observation that Al₂O₃ enhances stability beyond the T_g of SiO₂ (1,128 °C) suggests that Al₂O₃ has a higher T_g. In contrast, the thermal stability provided by ZrO₂ remains below the T_g of SiO₂, potentially indicating higher fragility. To verify these hypotheses, further analysis is required.

As nonglass formers, Al₂O₃ and ZrO₂ remain amorphous at high temperatures only when applied as ultrathin layers (<10 nm), making direct experimental viscosity measurements challenging. To overcome this, we employed MD simulations to estimate viscosity. To ensure the accuracy of the simulation results, we trained interatomic potentials for Al₂O₃ and ZrO₂ using over one million frames of ab initio simulation data (SI Appendix, Fig. S10), which were validated against density functional theory (DFT) and experimental data (SI Appendix, Figs. S11 and S12). However, obtaining high-viscosity data directly from classic MD simulation is difficult due to the limited sampling of configuration space (SI Appendix, Fig. S13). To overcome this limitation, we applied a metabasin escape algorithm, which enables exploration of larger configuration spaces by sampling more potential energy minima (40, 41). Activation barriers (SI Appendix, Fig. S14) were then calculated from these minima and used to estimate viscosity. The resulting data were fitted to the viscosity–temperature model developed by Mauro et al. (42) to derive the viscosity–temperature relationship and determine the T_g. Results revealed that the T_g of Al₂O₃ and ZrO₂ are 1,221 and 1,055 °C (Fig. 4G, dashed line), respectively, confirming that the T_g of Al₂O₃ exceeds that of SiO₂ (1,128 °C), providing a plausible explanation for why the Al₂O₃ atomic armor enhances the thermal stability of the mesoporous silica. In contrast, the lower T_g of ZrO₂ suggests that the increase in thermal stability may be potentially due to its fragility.

While bulk viscosity offers a general understanding of how these materials inhibit sintering, it does not directly illuminate the underlying mechanism. Since mesoporous silica consist of nanoparticles with atomic armor confined to their surfaces, a comprehensive analysis of sintering inhibition must focus on surface properties. To address this, we applied the surface viscosity model (34), deriving surface viscosity values from the previously calculated bulk viscosity data. This approach provides a more precise understanding of the role that surface dynamics play in inhibiting sintering.

We compared the surface viscosities of Al₂O₃, ZrO₂, and SiO₂ (Fig. 4G, solid line) and correlated the results with experimental observations (Fig. 4F). At 1,000 °C, the surface viscosity of Al₂O₃ is higher than that of ZrO₂, which in turn exceeds that of SiO₂. This is consistent with experimental observations showing that Al₂O₃ significantly enhances thermal stability, while ZrO₂ provides moderate improvement. At 1,100 °C, ZrO₂ accelerates sintering, as its surface viscosity drops below that of SiO₂, whereas Al₂O₃ continues to inhibit sintering due to its superior surface viscosity. These observations align with the experimental findings, reinforcing the validity of the proposed antisintering mechanism.

The fragility of Al₂O₃ and ZrO₂ (Fig. 4G) plays a crucial role in enhancing thermal stability, as viscosity near the T_g can vary by orders of magnitude, exceeding that of strong glasses like SiO₂. This property enables ZrO₂ to enhance thermal stability at 1,000 °C, a temperature close to its T_g. Additionally, fragile glasses can achieve ultrastable states when applied using deposition techniques (43, 44), further contributing to the sintering inhibition. The superior inhibition effect of Al₂O₃ is attributed to its higher T_g relative to ZrO₂. Thus, the enhanced thermal stability arises from the combined effects of high T_g and the fragility of nonglass formers. These findings quantitatively explain the experimental observations, demonstrating that the effectiveness of atomic armor depends on selecting materials with high surface viscosity at the target temperature. This is primarily governed by two critical factors: T_g and fragility.

To better elucidate the independent effects of T_g and fragility on surface viscosity, a comparative plot that categorizes materials into high, medium, and low T_g and fragility groups was introduced (SI Appendix, Fig. S15). To aid interpretation, the plot was further decomposed into subsets in which each parameter (T_g and fragility) varies separately (SI Appendix, Fig. S16) with respect to the medium baseline (SI Appendix, Fig. S16E). From these comparisons, it can be observed that an increase in either T_g or fragility relative to the base material generally leads to increased viscosity and enhanced thermal stability (SI Appendix, Fig. S16 A, B, and D). Conversely, a decrease in either parameter tends to reduce viscosity and thermal stability (SI Appendix, Fig. S16 F, H, and I). Interestingly, in the cases where one parameter increases while the other decreases, the resulting viscosity can either increase or decrease depending on the magnitude of change in T_g and fragility (SI Appendix, Fig. S16 C and G). These trends highlight how T_g and fragility tune the viscosity–temperature relationship and underscore the need for an understanding of both properties when designing high-performance atomic armors.

The application of Al₂O₃ and ZrO₂ on mesoporous silica are good examples of SI Appendix, Fig. S16 A and C, respectively. To further validate the generalizability of the proposed mechanism, SiO₂ and Al₂O₃ were deposited onto mesoporous borosilicate, which features lower T_g and higher fragility than SiO₂. Upon heat treatment at 800 °C, the SiO₂ atomic armored mesoporous borosilicate exhibits a minor decrease in specific surface area (SI Appendix, Fig. S17). However, at 900 °C and beyond, the SiO₂ armor began to enhance thermal stability, with the inhibition effect becoming more pronounced at higher temperatures (SI Appendix, Fig. S17). This thermal stabilization trend closely matches the viscosity trend shown in SI Appendix, Fig. S16G, providing further support to the present mechanism. Additionally, Al₂O₃ atomic armor consistently demonstrated strong stabilizing effects, attributed to its high T_g and fragility (SI Appendix, Fig. S17). These results extend the applicability of the atomic armor concept to other systems and verify the prediction capability of T_g and fragility in determining stabilization behavior.

Discussion

This research not only advances the understanding of stability of nanoporous materials but also provides a pathway for improving their thermal performance to remarkable levels. By addressing the thermal stability extremes of mesoporous silica, we offer insights into the T_g and fragility of nonglass formers and the pivotal role of surface viscosity in stabilizing nanoporous materials. We anticipate that these findings will inspire further research into surface viscosity optimization in nanoporous systems, leveraging variations in T_g and fragility across different materials, including both glass and nonglass formers. Furthermore, the inhibition mechanisms explored here may provide insights into the protective strategies of tardigrade cells involving TDPs, which suggests that the atomic armor approach could potentially be extended beyond oxides to metals, polymers, and other molecular systems. Ultimately, this work could lead to the development of a simple and universal posttreatment method for enhancing nanoporous material stability.

Materials and Methods

Fabrication of Mesoporous Structures.

Mesoporous silica was synthesized via a multistep process. Ethyl silicate, anhydrous ethanol, and deionized water were mixed in a beaker with a molar ratio of 1:10:5. The pH of the mixture was adjusted to 2 to 3 using hydrochloric acid, followed by magnetic stirring for 2 h at room temperature to achieve complete hydrolysis. Then, ammonia was introduced to raise the pH to 9-10, initiating the formation of the gel. The gel was aged in a solution of ethyl silicate and ethanol in a 1:1 ratio at 55 °C for 24 h, followed by immersion in anhydrous ethanol for an additional 24 h. To modify the gel, it was submerged in a solution of trimethylchlorosilane, ethanol, and n-hexane in a ratio of 2:1:8 for 48 h, followed by a 24 h immersion in n-hexane for solvent exchange and washing. The mesoporous silica was ultimately obtained through sequential drying under atmospheric pressure at temperatures of 60, 80, and 120 °C, maintaining for 2, 2, and 12 h, respectively. Note that the same fabrication process was used for mesoporous borosilicate, with ethyl silicate, trimethyl borate, anhydrous ethanol, and deionized water at a molar ratio of 3:1:40:20.

Atomic Layer Deposition.

The atomic armor was fabricated using plasma-enhanced ALD (PEALD). The sample was first evenly distributed onto an 8-inch silicon wafer, which was then loaded into a vacuum ALD reaction chamber (ALD 004, SENTECH, Germany). One cycle of PEALD begins with a pulse of a gaseous precursor, which chemisorbs onto the substrate surface in a self-limiting reaction. This is followed by a purge step using nitrogen to remove excess precursor and by-products. Next, a plasma pulse is introduced using oxygen that react with the adsorbed precursor to form the desired film. Finally, another purge step clears the chamber of remaining plasma species and reaction by-products, preparing the surface for the next cycle. Trimethylaluminum, tetrakis(dimethylamino)zirconium, and tris(dimethylamino)silane were employed as the precursors for aluminum, zirconium, and silicon, respectively, while deionized water served as the oxygen source. High-purity nitrogen was used as the carrier and purge gas throughout the process. The deposition was carried out at a chamber temperature of 250 °C to ensure effective precursor volatilization and surface reactions. For each Al₂O₃ deposition cycle, 0.06 s of trimethylaluminum was injected into the chamber, followed by a 9.94 s nitrogen purge to remove excess precursor and reaction byproducts. This was followed by a 5 s plasma pulse to activate surface reactions between the adsorbed precursor and oxidizing species, and then a final 1 s nitrogen purge. For each ZrO₂ deposition cycle, 0.6 s of tetrakis(dimethylamino)zirconium was injected, followed by an 11.4 s nitrogen purge, a 5 s plasma pulse, and a 3 s final nitrogen purge. Note that except for the change of precursor, the remaining parameters used are the same for the deposition of SiO₂. The number of deposition cycles was adjusted to control the deposition quantity of the resulting oxide films. The plasma step in each cycle enhanced the surface reactivity and contributed to the formation of uniform layers.

Characterization.

Various techniques were employed to comprehensively assess the properties of the samples. The mesoporous silica used for characterization was heat-treated in a muffle furnace. Initially, the samples were ground into small fragments and subsequently positioned at the base of a crucible. Subsequently, the crucibles were inserted into the muffle furnace. The temperature was increased to the target temperature with a heating rate of 5 °C min⁻¹ and maintained for a predetermined duration. After which, the crucibles were removed from the furnace and cooled under ambient conditions. For HRTEM, the samples were pulverized in an agate mortar and dispersed in anhydrous ethanol. Subsequently, a drop of this suspension was placed onto an ultrathin copper mesh and heated to 60 °C using a heating lamp. The sample was observed using a Lorentz transmission electron microscope (Talos-F200X, ThermoFisher Scientific, USA). For SEM, the samples were uniformly distributed on a conducting resin, followed by the deposition of a thin layer of carbon using a high vacuum coating machine (Leica EM ACE600, Leica, Germany). The sample was observed using a field emission scanning electron microscope (Gemini SEM 500, Zeiss, Germany). ICP analysis involved first dissolving the samples in acid, followed by testing using the plasma emission spectrometer (Agilent 5110, Agilent, USA). The specific surface area of the samples was determined from the adsorption isotherm using BET theory. This measurement was conducted using a surface and porosity analyzer (ASAP 2460, Micromeritics, USA) in a nitrogen environment. The pore size distribution of the samples was derived from the desorption branches of the isotherms using the Barret-Joyner-Halenda model. XRD patterns were obtained by a D8 Advance diffractometer (Bruker, Germany) with Cu–Kα radiation (40 kV, 40 mA). The thermal conductivity was measured by the transient plane source method using the hot disk (TPS 2500, Hot Disk, Sweden). Kapton sensors (Hot Disk Kapton 5501) were used with a heating power of 8 to 60 mW for a duration of 40 to 160 s. The chemical structures were analyzed using Fourier transform infrared (FTIR) spectroscopy (Nicolet iS50, Thermo Fisher, USA) within the wavenumber range of 4,000 to 400 cm⁻¹. The SSNMR experiments were conducted using an AVANCE NEO 500 MHz spectrometer (Bruker, Germany) for ²⁹Si and on an AVANCE NEO 600 MHz spectrometer (Bruker, Germany) for ²⁷Al. The ²⁹Si NMR spectra were recorded at a magic angle spinning (MAS) rate of 5 kHz using a 7 mm zirconia rotor, while the ²⁷Al NMR spectra were recorded with a MAS rate of 12 kHz using a 4 mm zirconia rotor.

Machine Learned Interatomic Potential Training.

Molecular dynamic simulations were performed using the graphics processing units MD (45) package with the neuroevolution potential (NEP) (46) model. Datasets for training the NEP model (SI Appendix, Fig. S10) were generated by AIMD simulations using the DFT with the Vienna ab initio simulation package (47). The projected augmented wave (48) method and the Perdew–Burke–Ernzerhof (49) generalized gradient approximation were used for the electronic exchange-correlation. The energy cutoff was set to 600 eV and the Γ point was used to sample the Brillouin zone.

The Nose–Hoover thermostat was employed to control the temperature for the NVT ensemble, while the Langevin thermostat was used for controlling the temperature for the NpT ensemble. The timestep was set to 1 fs. The initial structure, obtained from the materials project (50), was replicated to a cubic cell containing 80 and 96 atoms for Al₂O₃ and ZrO₂. This structure was melted at 5,000 K for 10 ps under the NpT ensemble. Subsequently, the structures were relaxed under the NpT ensemble at various temperatures (from 300 to 5,000 K) for 10 ps to determine the correct cell size for the corresponding temperatures. The cell size of the last 5 ps of the NpT simulations was averaged and used for the NVT ensemble simulations. Random trajectories from the melting process were selected, adjusted to the determined cell sizes, and simulated under the corresponding temperatures for 10 ps.

This process was repeated five times to comprehensively sample the space configuration. Only frames from the NVT calculations were used as datasets, resulting in a total of 650,000 trajectories. For computational efficiency, one frame was extracted for every 500 frames, creating a training set of 1,350 frames for the NEP model. Then this model was used to run a melt quench process with cubic cells containing 270 and 324 atoms, to further refine the dataset, and crystalline structures were also taken from the materials project (50), perturbed and added to the dataset, then retrained (SI Appendix, Fig. S11). The NEP model was then validated by comparing the radial distribution and structural factor with experimental data for Al₂O₃ and ab initio results for ZrO₂ (SI Appendix, Fig. S12), due to the lack of experimental data for amorphous ZrO₂.

Viscosity–Temperature Correlation.

The system was melted at 4,000 to 5,000 K under the NpT ensemble using the Berendsen thermostat, then quenched to 300 K at a cooling rate of 1 × 10¹² K s⁻¹ under the NpT ensemble using the Berendsen thermostat, and further relaxed under the NpT and NVT ensemble at 300 K for 100 ps using the Berendsen thermostat. The viscosity is calculated by the Green–Kubo relation:

$${\eta }_{\alpha \beta }=\frac{V}{k_{\mathrm{B}}T}{\int }_0^t〈\left(P_{\alpha \beta }\left(0\right)-〈P_{\alpha \beta }〉\right)\left(P_{\alpha \beta }\left(t^{\prime }\right)-〈P_{\alpha \beta }〉\right)〉d{t}^{\prime },$$ [1]

where V is the volume of the system, k_B is the Boltzmann constant, T is the temperature, and the integrand is called the pressure autocorrelation function. Note that before calculation, the cell was relaxed at the target temperature under the NpT ensemble using the Berendsen thermostat for 5 ns.

Due to the time scale limitation of MD simulation, we applied the self-learning metabasin escape algorithm (40), a speed up approach of the autonomous basin climbing algorithm (41), to lift the system out of potential wells and sample a larger configuration space. This method is like metadynamics (51) but without the need of defining controlled variables, and the effectiveness of this method has been well benchmarked by Kushima et al. (41, 52), who initially developed the autonomous basin climbing algorithm. The self-learning metabasin escape algorithm was used to sample the minima and saddles, and the single path approximation was used to calculate the viscosity (SI Appendix, Fig. S13).

The viscosity–temperature correlation was fitted by the Mauro model (42):

$$\begin{array}{c}log\eta (T) = log{\eta }_{\infty }+\left(12-log{\eta }_{\infty }\right)\frac{T_{\mathrm{g}}}{T}\text{exp}\left[\left(\frac{m}{12-log{\eta }_{\infty }}-1\right)\left(\frac{T_{\mathrm{g}}}{T}-1\right)\right],\hfill \end{array}$$ [2]

where η_∞ is the extrapolated infinite temperature viscosity, T_g is the glass transition temperature, and m is the fragility.

Surface Viscosity Calculation.

The surface viscosity can be calculated by

$$log{\eta }_{\mathrm{s}}=log\eta +\epsilon {\left(\frac{T_g}{T}\right)}^{\alpha }(s^{\alpha }-1),$$ [3]

where η is the bulk viscosity, ε is a dimensionless parameter given by $\epsilon \equiv log\left(\frac{{\eta }_{\mathrm{g}}}{{\eta }_{\infty }}\right), {\eta }_{\mathrm{g}}={10}^{12} \text{Pa} \mathrm{s}$, is the viscosity at the glass transition temperature, α is a dimensionless parameter given by $\alpha \equiv \frac{2C}{\mathit{ZR}}$, R is the ideal gas constant, z is the coordination number of the main building units of the system, c is the upper limit of the specific heat, and s is a dimensionless parameter showing the surface effect. Further information of the derivation and parameter choosing is given in ref. 34.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Datasets have been deposited in Figshare (https://doi.org/10.6084/m9.figshare.29923850) (53). All other data are included in the manuscript and/or SI Appendix.