Synergistic shape-interface-channel ordering enhances solar-driven thermochemical energy storage

Abstract

Solar-driven thermochemical energy storage based on salt hydrates offers high energy density and long-duration storage, providing a promising route to mitigate the temporal mismatch between renewable energy supply and thermal demand. However, sluggish reaction kinetics, poor cycle stability, and low solar utilization efficiency remain long-standing challenges. Here, we demonstrate a synergistic shape-interface-channel ordering strategy that substantially enhances thermochemical energy storage performance through rational engineering of CaCl₂-based pellets featuring red blood cell-mimic shape, nanoengineered hydrophilic interfaces, and hierarchical leaf veins-inspired channel networks. The optimized sample exhibits an ultrafast sorption rate coefficient of 0.03973 min⁻¹, a high sorption capacity of 885 mg g⁻¹, and a high energy storage density of 988.38 kJ kg⁻¹, with only 5.65% capacity decay over 1000 cycles (~3722 h). With further assistance of high solar absorptance of 97.72%, the desorption efficiency achieves as high as 93% within 60 min under one-Sun irradiation. This work establishes a general design paradigm for thermochemical energy storage pellets based on multiscale structural ordering, offering a pathway towards scalable solar-driven thermochemical energy storage.

A synergistic shape-interface-channel ordering strategy is proposed to overcome sluggish heat and mass transfer in salt-based materials, enabling fast, high-energy-density, and durable solar-driven thermochemical energy storage.

Introduction

The rapidly increasing consumption of fossil fuels has led to critical challenges of energy shortages and climate change. A shift toward renewable energy has emerged as a global consensus to mitigate these issues¹. In particular, solar energy has attracted extensive attention due to its abundance, wide distribution, and environmental friendliness^2,3. However, its inherent intermittency and instability lead to a mismatch between energy supply and demand⁴. In this context, energy storage technologies play a pivotal role in decoupling energy supply and demand across temporal and spatial scales, providing effective solutions for the fluctuations associated with solar energy utilization^5,6. Among various energy storage technologies, thermal energy storage (TES) is particularly significant, as thermal energy accounts for approximately 50% of the total end-use energy consumption^7–9. In particular, thermochemical energy storage (TCES) based on the sorption-desorption of salt hydrates, which uses water vapor as the working medium, has gained increasing attention due to its unique advantages of nontoxicity, eco-friendliness, low cost, high energy density, and minimal thermal losses¹⁰. In the sorption process, the sorbent captures water vapor, releasing sorption heat. Reversely, solar energy can drive the water desorption process, storing thermal energy as the bond energy of sorbent-adsorbate working pairs^11,12. Considering the abundant solar irradiation and ubiquitous moisture, solar-driven TCES based on salt hydrates is regarded as a highly promising pathway for achieving stable and sustainable solar thermal energy harvesting^13–15.

Nevertheless, solar-driven TCES based on salt hydrates still suffers from poor stability, slow reaction kinetics, and low solar utilization efficiency. The deliquescence and agglomeration of salts severely impede internal heat and mass transfer, leading to slow kinetics and poor cycling stability^16,17. To overcome these issues, impregnating salt into a matrix has emerged as one of the most common and effective approaches¹⁸. The pores of the matrix not only provide space to hold the salt, limit leakage of the salt solution, but also prevent salt agglomeration into clusters. Moreover, the high specific surface area can enhance heat and mass transfer, helping to accelerate reaction kinetics. Various matrices such as silica gel¹⁹, zeolite²⁰, mineral materials²¹, porous carbon²², nanoporous materials²³, and metal organic frameworks (MOFs)²⁴, have been employed to fabricate TCES composites. However, most TCES composites possess disordered pore networks, leading to tortuous mass transport pathways and limited kinetic. Quantitatively, disordered TCES composites typically exhibit a low sorption rate coefficient of 0.006-0.025 min⁻¹, and capacity degradation of 1.65–28.21% within tens of cycles. Such performance falls short of industrial requirements, which demand rapid charge-discharge kinetics and stability over hundreds to thousands of cycles to ensure decade-long operation. These limitations originate from coupled transport constraints across multiple scales: spherical shapes impose long diffusion lengths; disordered channels introduce tortuosity; and poorly engineered interfaces restrict vapor accessibility and active site utilization. Although efforts have been made to introduce aligned channels to accelerate vapor transport, interfacial properties and macroscopic shape remain largely overlooked^25–27. In parallel, direct solar capture by composites can effectively convert solar energy into thermal energy, eliminating the need for intermediate heat transfer processes, thereby minimizing thermal losses for solar-driven TCES²⁸, but the intrinsically low absorptance of salt hydrates leads to significant optical losses²⁹. Achieving fast kinetics, long-term stability and efficient solar utilization simultaneously, therefore, remains a central challenge^16,30,31.

Nature has evolved over billions of years to create highly efficient mass and energy transfer structures that offer transformative inspiration for designing high-performance solar-driven TCES pellets (Fig. 1)³². A compelling example is the red blood cell (RBC), responsible for transporting oxygen and carbon dioxide throughout the body, and it exhibits an optimized biconcave disc shape that maximizes the surface-to-volume ratio and minimizes mass transfer distance, thereby ensuring highly efficient gas exchange^33,34. Similarly, leaf veins form hierarchical channel networks with a central main vein branching into lateral veins and further subdividing into fine veins, enabling rapid transport of nutrients and water with minimal energy losses while maintaining structural integrity^35–37. Inspired by these biological paradigms, we engineer CaCl₂-based pellets with synergistically ordered shape, interface and channel structures to overcome the intrinsic limitations of conventional TCES materials. Using a scalable solvent-induced phase separation approach, which is a mature method widely used for porous material production, we fabricate millimeter-sized (~ 1.7 mm) robust monoconcave disc-shaped pellets that mimic RBC morphology to enhance surface area to volume ratio, while solvent-nonsolvent exchange drives the formation of leaf veins-mimic oriented internal channels to promote rapid water vapor diffusion and structural stability. Subsequent chemical activation introduces micropores and oxygen-containing functional groups, increasing specific surface area, improving hydrophilicity, and leveraging the intrinsic photothermal properties of carbon. This multiscale structural design enables efficient and directional heat and mass transfer, leading to markedly improved sorption-desorption kinetics. The optimized RL800@25%CaCl₂ achieves a water uptake of 885 mg g⁻¹ and reaches 90% of its equilibrium capacity within 60 min under 20 °C and 60% RH. It delivers a high sorption rate coefficient of 0.03973 min⁻¹ with only 5.65% capacity decay after 1000 cycles (~ 3722 h), substantially outperforming benchmarks reported in the literature. This work provides a design paradigm of solar-driven TCES pellets based on synergistic shape-interface-channel ordering, offering a pathway towards high-performance and scalable thermochemical energy storage technologies.

Fig. 1. Comparison of disordered and ordered solar-driven thermochemical energy storage (TCES) pellets.

Conceptual comparison between conventional disordered pellets (left) and bioinspired ordered pellets (right), the latter is featured with a red blood cell-mimic shape, highly hydrophilic surface, and leaf veins-mimic channels.

Results

Characterization of ordered solar-driven TCES pellets

Figure 2a schematically illustrates the preparation of carbon pellets with RBC-mimic shape and leaf veins-mimic channels, encompassing solvent-induced phase separation, oxidative stabilization, carbonization, and activation steps, with photos provided in Supplementary Fig. 1. N, N-Dimethylformamide (DMF) and water-ethanol were used as the solvent and nonsolvent (coagulation bath) for Polyacrylonitrile (PAN), respectively. To ensure adequate mechanical strength for stacked pellets in practical applications, the water-to-ethanol ratio is fixed at 7:3, which yields the highest pellet strength³⁸. Driven by an injection pump, the PAN solution drips into the coagulation bath. As the droplets descend under gravity, a low-pressure vortex forms at the trailing edge. Combined with the droplet’s relatively low viscosity, this low-pressure region causes the upper surface of the droplet to collapse inward, resulting in a concave depression³⁹. Upon solidification, the droplets evolve into a monoconcave disc shape that mimics the morphology of RBC. Another key feature of this approach is the spontaneous formation of leaf veins-mimic oriented macroporous channels within the polymer-lean phase. This occurs due to the strong mutual solubility between DMF and water-ethanol mixture, which promotes rapid and large-scale phase separation³⁸. The resulting high driving force directs the DMF and water-ethanol to migrate along the shortest paths, yielding leaf veins-mimic oriented channel structures. Meanwhile, a relatively dense shell forms on the droplet surface, ensuring both shape retention and mechanical integrity. The resulting robust pellets with RBC-mimic shape and leaf veins-mimic channels obtained after oxidation and carbonization are referred as RL, while samples further activated at 700 °C, 800 °C, and 900 °C are denoted as RL700, RL800, and RL900, respectively.

Fig. 2. Preparation process, morphology, and internal architecture of pellets.

a Schematic illustration of the fabrication process, including solvent-induced phase separation, oxidation, carbonization, and activation. The illustrated structures are schematically drawn from experimentally observed morphologies. b–e Macroscopic and cross-sectional SEM images of RL (b, c) and RL800 (d, e). f SEM image of multiple RL800 pellets. g–i Three-dimensional micro-CT reconstructions of RL800, showing the overall structure (g), longitudinal section (h) and cross-section (i), highlighting the RBC-mimic shape and the leaf-vein-inspired hierarchical “main vein–branch” channel architecture. RL800 is colored red in (g–i). The SEM imaging experiments shown in b-e were independently repeated three times with similar results. RL and RL800 denote the carbonized pellets and the pellets activated at 800 °C, respectively.

Figure 2b–i and Supplementary Figs. 2 and 3 show the overall morphology and internal structure of the RL and activated samples, as revealed by SEM and micro-CT scans. All activated samples maintain the monoconcave disc-shaped structure with surface wrinkles, which enhance the surface area. This shape resembles the biconcave disc of RBC, evolved to maximize gas transport efficiency. The increased surface area to volume ratio and reduced diffusion path length enable highly efficient mass transport. In cross-section view, both RL and activated samples exhibit a central main channel bisecting the pellets, with symmetrically aligned branches extending outward. Near the center, these branches are perpendicular to the main channel and gradually transition to angles ranging from 0° to 90° near the edge. This “main vein–branch” architecture mimics natural leaf venation, providing highly efficient transport pathways and mechanical support. Activation temperature further modulates structural integrity, whereas RL700 and RL800 preserve well-defined architectures, while RL900 exhibits surface cracking, indicative of over-activation and structural degradation.

To quantify the effect of shape on sorption performance, numerical simulations were conducted for spherical, RBC-mimic and ideal RBC shapes (Supplementary Note 1 and Supplementary Figs. 4–6). The RBC-mimic shape exhibits a 13.6% higher surface area to volume ratio (0.005 μm⁻¹) compared to the spherical shape (0.0044 μm⁻¹). Correspondingly, the RBC-mimic shape reduces the sorption time to 90% conversion ratio (α = 0.9) by 10.7% relative to the sphere. The RBC-mimic and RBC shape show nearly identical surface area to volume ratio and sorption kinetics, indicating that the fabricated morphology effectively captures the intrinsic advantages of the biological prototype.

The role of channel architecture was further evaluated by comparing ordered leaf veins-mimic channels and random channels at equivalent pore volumes (Supplementary Note 1 and Supplementary Figs. 7 and 8). Channels ordering reduces the sorption time to reach α = 0.9 by 8.4% for the RBC-mimic shape and by 10.7% for the spherical shape. At fixed channel structure, RBC-mimic pellets consistently outperform spherical ones, achieving 11.1% and 13.4% faster sorption for ordered and random channels, respectively. Notably, the combination of RBC-mimic shape and ordered channels yields the fastest kinetics, reducing response time by 20.7% compared with spherical pellets with random channels. Together, these results demonstrate that coupling RBC-mimic geometry with leaf-vein-like channel networks enables hierarchical optimization of mass transport. This synergistic structural design significantly enhances water vapor transport in TCES systems.

The carbonized RL sample shows a low specific surface area of 0.84 m²g⁻¹, necessitating activation for further enhancement. Given the intrinsic hydrophobicity of carbon materials, KOH activation was employed to simultaneously increase surface area and tailor surface chemistry. Supplementary Fig. 9a, b present the N₂ adsorption/desorption isotherm curves and corresponding pore size distributions for RL and activated samples. At low relative pressures, the curves rise sharply and closely trace the Y-axis, indicating a type I isotherm and a predominance of micropores⁴⁰. In RL700 and RL800, the pore structure is dominated by micropores, whereas RL900 shifts toward mesopores. With increasing activation temperature from 700 °C to 900 °C, the specific surface area increases from 743.42 m² g⁻¹ to 1551.48 m²g⁻¹ due to KOH activation (Supplementary Equations (1)-(5)), which effectively enhance the surface area⁴¹ (Supplementary Table 3). This increase in porosity is accompanied by a trade-off in mechanical strength. The crush strength decreases from 5.86 N (RL700) to 4.69 N (RL800) and further to 2.56 N (RL900), consistent with the surface cracking observed for RL900. Therefore, RL800 is selected as an optimal balance, providing a high specific surface area (901.07 m²g⁻¹) while retaining acceptable crush strength.

As nitrogen adsorption probes pores below 50 nm, it does not capture the micrometer-scale channels observed by SEM. Mercury intrusion porosimetry was therefore used for further analysis^21,38. Supplementary Fig. 10a shows similar Hg intrusion curves for RL and RL800, but RL800 exhibits higher intrusion volume, indicating increased porosity. As shown in Supplementary Fig. 10b, both samples contain pores in the 100–4800 nm and 4800–60000 nm ranges, indicating that activation treatment modifies microscopic pores without compromising macropore architecture. Overall, KOH activation increases the specific surface area of RL800 by 1077 times compared to RL and optimizes the pore structure. The resulting three-dimensional and directionally interconnected structure offers abundant salt-loading sites, enhances vapor diffusion, and accommodates salt solution via capillarity⁴².

Another major benefit of KOH activation lies in its ability to tailor surface chemistry by introducing oxygen-containing functional groups and significantly improving surface hydrophilicity. FTIR spectra reveals a peak near 3450 cm⁻¹, indicating -OH groups (Fig. 3a). Contact angle measurements confirm this shift, with RL exhibits a contact angle greater than 90°, indicating strong hydrophobicity. In a stark contrast, RL800 demonstrates a contact angle of approximately 46.9°, confirming its conversion to a hydrophilic state (Fig. 3b). This enhanced wettability is critical for salt loading and vapor diffusion⁴³. In summary, KOH activation plays a dominant role in generating pores and introducing oxygen-containing functional groups. Up to now, the RBC-mimic external shape and leaf veins-mimic internal channels of RL800 deliver high specific surface area, short diffusion distance, hierarchical porosity across the nano- and micrometer scales, favorable mechanical strength, and strong surface hydrophilicity, making it highly suitable for serving as supports of salt hydrates.

Fig. 3. Pore structure, surface chemistry and wettability of pellets.

a FTIR spectra and b Contact angle measurements of RL and RL800. c XRD patterns and d TGA curves of CaCl₂, RL, RL800, and RL800@CaCl₂ composites. Pore size distribution profiles of RL800@CaCl₂ composites derived from e N₂ adsorption-desorption analysis and f Mercury intrusion porosimetry. g Picture of RL800@CaCl₂ composites with different salt loadings. RL800@CaCl₂ denotes the corresponding CaCl₂-loaded composites.

In this study, RL800@x%CaCl₂ refers to the composite obtained by impregnating RL800 with an aqueous CaCl₂ solution of concentration x wt.%. XRD analysis confirms that RL possesses a graphite-mimic nanocrystalline structure, and peaks located at approximately 26°, 42° and 44° are assigned to the (002), (100) and (101) planes of C (PDF#41-1487), respectively. Upon KOH activation, the (002) peak becomes weaker and broader, indicating the disruption of the graphite lattice and introduction of more pores⁴⁴. To assess the composition in detail, XRD patterns of RL800@CaCl₂ composites are compared with that of analytically pure CaCl₂ (PDF#24-0223) (Fig. 3c). The main diffraction peaks of the composites match those of CaCl₂, indicating a physical mixture without new phase formation, and these peaks become more pronounced with increasing CaCl₂ content.

Thermogravimetric analysis under air further quantifies the carbon and salt contents (Fig. 3d and Supplementary Fig. 11). As carbon combusts while CaCl₂ remains thermally stable, the weight loss is used to estimate the carbon fraction. All samples, except pure CaCl₂, exhibit two major weight loss stages: the first is due to adsorbed and crystallization water, and the second is attributed to carbon combustion. Notably, RL800 loses 98.71 wt.% of its mass, while pure CaCl₂ shows a minimal loss of 0.78 wt.%, confirming that the weight loss primarily stems from the carbon component. Furthermore, composites demonstrate good thermal stability up to 400 °C. The salt loading (mass fraction, wt.%) increases with the concentration of the impregnation solution, reaching 22.09 wt.%, 39.01 wt.%, 47.42 wt.%, 52.23 wt.%, and 55.27 wt.% for RL800@10%CaCl₂, RL800@20%CaCl₂, RL800@25%CaCl₂, RL800@30%CaCl₂, and RL800@40%CaCl₂, respectively. However, the growth trend plateaus with higher concentrations, for example, only 3.04 wt.% increase is observed between the RL800@30%CaCl₂ and RL800@40%CaCl₂, due to the reduced infiltration capacity of high-concentration salt solutions caused by increased surface tension.

N₂ adsorption/desorption isotherms (Supplementary Fig. 12) reveal that all RL800@CaCl₂ composites retain the type I profile typical of microporous materials, with pore size distributions primarily below 2 nm (Fig. 3e). As the salt loading increases, the specific surface area decreases from 608.99 m²g⁻¹ to 125.59 m²g⁻¹, indicating pore occupation by CaCl₂ (Supplementary Table 4). Mercury intrusion porosimetry (Supplementary Fig. 13 and Fig. 3f) supports this trend, showing reduced intrusion volumes with higher salt contents, while confirming the preservation of the hierarchical pore structure across the 100–4800 nm and 4800–60000 nm range.

As shown in Fig. 3g, the composites appear as black pellets, indicative of strong solar absorptance. However, faint white aggregates appear on RL800@30%CaCl₂ and RL800@40%CaCl₂, suggesting the presence of excess salt. SEM and EDS mapping (Supplementary Fig. 14) confirm overall uniform salt dispersion, although slight clustering is observed at higher loadings. Although increased salt enhances energy storage density, excessive loading may cause agglomeration and leakage, impairing sorption kinetics and cycling stability (Supplementary Figs. 15 and 16). In contrast, Supplementary Fig. 17 demonstrates that pure CaCl₂ undergoes a transition to a transparent liquid upon sorption and subsequently forms white agglomerated solids after desorption, which leads to leakage, corrosion, mass transfer resistance, and poor cycling stability, thereby hindering its practical application in solar-driven TCES. These results highlight the necessity of optimizing salt loading to balance storage capacity with structural stability and mass transport.

Water sorption properties of RL800@CaCl₂ composites

For salt hydrate-based solar-driven TCES, water sorption behavior directly determines thermal energy output⁴⁵. The sorption process of composites involves three steps: initial physisorption via van der Waals forces, followed by salt chemisorption forming coordinate bonds, and finally, at RH above the salt’s deliquescence threshold, absorption through deliquescence into a saturated or dilute salt solution⁴⁶. In this process, both temperature and RH critically influence the sorption kinetics and capacity.

To investigate the effect of salt loading, we measure the sorption performance of RL800@CaCl₂ composites. For instance, at 20 °C and 60% RH, the equilibrium sorption capacity increases with CaCl₂ content, rising from 414 mg g⁻¹ for RL800@10%CaCl₂ to 1167 mg g⁻¹ for RL800@40%CaCl₂ (Fig. 4a), due to the increased active salt content. As shown in Fig. 4b, the time required to reach α = 0.9 increases markedly with salt loading, extending from 22 min for RL800@10%CaCl₂ to 82 min for RL800@40%CaCl₂, while RL800 and pure CaCl₂ reach α = 0.9 at 70 min and 127 min, respectively. Kinetic analysis is conducted to further clarify the sorption mechanism, and the F1 reaction model (f(α) = 1-α) is identified as the most probable mechanism based on isothermal fitting (Fig. 4g and Supplementary Note 2). Similarly, the sorption rate coefficient k decreases with increasing salt loading (Supplementary Fig. 18). For example, k drops sharply from 0.106 min⁻¹ for RL800@10%CaCl₂ to 0.04096 min⁻¹ for RL800@20%CaCl₂ and further decreases to 0.02733 min⁻¹ for RL800@40%CaCl₂. This deterioration is mainly caused by reduced accessible surface area and porosity, together with blocked diffusion pathways and deliquescence, which impose substantial mass transfer resistance⁴⁷ (Supplementary Tables 4 and 5). For reference, RL800 exhibits an equilibrium capacity of 434 mg g⁻¹ with a k of 0.03137 min⁻¹, while pure CaCl₂ reaches 2120 mg g⁻¹ but has a much slower k of 0.01825 min⁻¹ under identical conditions (Fig. 4b and Supplementary Table 5). Thus, the composites integrate the high capacity of CaCl₂ with the faster kinetics of RL800. Considering the trade-off between capacity, kinetics and structural stability at high loadings, RL800@25%CaCl₂ is selected as the optimal composition for further investigation (A quantitative selection rationale is provided in Supplementary Note 3 and Supplementary Table 6).

Fig. 4. Water sorption characteristics of ordered pellets.

a, b Water sorption profiles (a) and corresponding kinetics (b) of RL800, RL800@CaCl₂ composites and pure CaCl₂ at 20 °C and 60% RH. c, d Water sorption profiles (c) and kinetics (d) of RL800@25%CaCl₂ at 20 °C under RH of 20%, 40%, 60% and 80%. e Sorption kinetics of RL800@25%CaCl₂ at 60% RH and temperatures of 10, 20 and 30 °C. f Equilibrium water uptake of RL800, RL800@CaCl₂ composites, and pure CaCl₂ as a function of RH and temperature. g Sorption mechanism function. The smooth traces represent theoretical kinetic mechanism functions (R2, R3, F1, D2, D3 and D4) calculated from the corresponding equations listed in Supplementary Table 2. Comparison of h water sorption kinetics and i sorption rate coefficient between RL800@25%CaCl₂ and state‑of‑the‑art salt‑based composites under similar conditions (conversion ratio calculated as m/m_eq; data in Supplementary Table 8). j Long‑term cycling performance of RL800@25%CaCl₂ over 1000 cycles at 20 °C/60% RH (first 500 cycles) and 30 °C/60%RH (last 500 cycles). k Comparison of capacity retention ratio after hydration-dehydration cycles between RL800@25%CaCl₂ and state‑of‑the‑art salt‑based composites (Supplementary Table 9).

To clarify the role of interfacial properties, a control sample RL@25%CaCl₂ was prepared using unactivated, hydrophobic RL. Owing to poor wettability, salt solution infiltration is largely confined to the outer surface, as evidenced by visible surface deposits (Supplementary Fig. 19) and EDS mapping (Supplementary Fig. 20). Consequently, significant leakage occurs during sorption (Supplementary Fig. 21), leading to a substantially reduced uptake (257 mg g⁻¹) compared to RL800@25%CaCl₂ (885 mg g⁻¹) under identical conditions (Supplementary Fig. 22). These results underscore the critical role of interface engineering in enabling uniform salt incorporation and retention.

Figure 4c presents the water sorption profiles of RL800@25%CaCl₂ at 20 °C across RH levels of 20%, 40%, 60%, and 80%. As expected, higher RH enhances the sorption capacity due to increased vapor pressure. For instance, at 20 °C and RH of 20%, 40%, 60%, and 80%, RL800@25%CaCl₂ achieves sorption capacities of 422, 683, 885, and 1446 mg g⁻¹, respectively. At low RH (20%), sorption proceeds slowly due to limited driving force, whereas at high RH (80%), equilibrium is also not reached within 360 min because of continued deliquescence and slow ion diffusion⁴⁸ (Supplementary Fig. 16). As shown in Fig. 4d, the time to reach α = 0.9 first decreases from 136 min (20% RH) to 60 min (60% RH) and then increases to 120 min (80% RH). Fitted k values exhibit a similar non-monotonic trend (Supplementary Fig. 23), increasing from 0.02703 min⁻¹ (20% RH) to 0.03973 min⁻¹ (60% RH) and then decreasing to 0.02833 min⁻¹ (80% RH). This behavior reflects the dual effect of RH, moderate RH promotes mass transport, while excessive RH induces deliquescence, which hinders vapor diffusion (Supplementary Table 7).

Temperature also influences the sorption kinetics, as shown in Fig. 4e. With increasing temperature from 10 °C to 30 °C, the time to reach α ≈ 0.9 shortens significantly from 74 min to 37 min, indicating accelerated sorption. Meanwhile, the equilibrium capacity remains relatively stable⁴⁹. The fitted k values also rise from 0.02919 to 0.06502 min⁻¹ across this range, indicating that higher temperature enhances molecular motion and vapor diffusion (Supplementary Fig. 24 and Supplementary Table 7). Furthermore, the sorption behaviors of other samples under different temperatures and RH are provided in Supplementary Fig. 25 for comparison. Figure 4f summarizes the equilibrium sorption capacity of RL800 and RL800@CaCl₂ composites under various RH and temperature conditions, confirming that capacity primarily depends on RH, with minimal variation across the tested temperature range. Compared with state-of-the-art salt-based composites, RL800@25%CaCl₂ exhibits ultrafast sorption kinetics (k = 0.03973 min⁻¹), highlighting the benefits of its “shape-interface-channel” design (Fig. 4h, i and Supplementary Table 8)^{6,12,21,25,26,29,50–53}.

Cycling stability is essential for practical deployment. Long-term durability was evaluated over 1000 sorption-desorption cycles (~ 3722 h), including 500 cycles at 20 °C and 60% RH followed by 500 cycles at 30 °C and 60% RH (Detailed cycling protocols are provided in Supplementary Figs. 26–28). As shown in Fig. 4j and Supplementary Fig. 29, both RL800 and RL800@25%CaCl₂ exhibit minimal capacity decay after 1000 cycles, with capacity retentions of 96.13% (3.87% decay) for RL800 and 94.35% (5.65% decay) for RL800@25%CaCl₂. Kinetic profiles from representative cycles show nearly identical behavior between cycles 1 and 500, with only slight slowing observed at extended cycling (Supplementary Figs. 30 and 31). Post-cycling characterization reveals that the overall pellet architecture and salt distribution are largely preserved despite minor surface cracking (Supplementary Figs. 32 and 33). In stark contrast, conventional salt-based composites typically suffer severe degradation within tens of cycles (Fig. 4k and Supplementary Table 9)^{21,22,24,29,51,53–64}. The superior stability observed here arises from uniform salt dispersion and the robust, hierarchical structure. Collectively, these results demonstrate that RL800@25%CaCl₂ achieves an optimal balance between capacity, kinetics and long-term stability, making it a promising candidate for efficient and durable thermochemical energy storage.

Thermochemical energy storage performance

The thermochemical energy storage density of RL800 and RL800@CaCl₂ composites is evaluated using DSC analyses (Fig. 5a). RL800 shows a single endothermic dehydration peak at 83 °C and an energy storage density of 599.79 kJ kg⁻¹, corresponding to the removal of physically adsorbed water. The desorption process of salt hydrate is carried out in several stages, corresponding to different endothermic peaks. Pure CaCl₂ displays two endothermic peaks at 117 °C and 174 °C, yielding an energy storage density of 1507.9 kJ kg⁻¹, which reflects its multi-step desorption mechanism. For RL800@CaCl₂ composites, RL800@10%CaCl₂ shows a single endothermic peak, while composites with higher salt contents present multiple endothermic peaks. This stems from the matrix adsorbing water at low CaCl₂ loadings, enabling robust sorption and forming abundant solutions. During desorption, the dehydration peak of the dilute solution overlaid that of the crystalline hydrate, yielding a broader peak. As salt content increases, the peak temperature shifts to higher values, attributed to the salt’s strong water affinity and diffusion resistance that increase with salt loading⁶⁵. The energy storage density rises accordingly, reaching 988.38 kJ kg⁻¹ for RL800@25%CaCl₂ (Fig. 5b). Notably, RL800@30%CaCl₂ and RL800@40%CaCl₂ experience solution leakage upon water sorption (Supplementary Fig. 15), precluding further thermal analyses.

Fig. 5. Thermochemical energy storage performances of RL800@CaCl₂ composites.

a Differential scanning calorimetry (DSC) thermogram for RL800 and RL800@CaCl₂ composites with varying salt loadings. b Dehydration enthalpy. Data are presented as mean ± standard deviation. The sample size (n = 3) represents three independently prepared and tested samples for each group. c UV-Vis-NIR absorptance spectra of RL800, RL800@CaCl₂ composites, and pure CaCl₂. d Average absorptance. Data are presented as mean ± SD, n = 3 independent replicates. e Desorption efficiency and surface temperature for fully hydrated RL800@25%CaCl₂ and CaCl₂ under simulated one-Sun illumination (1000 W m⁻²). f Comparison of solar-driven desorption efficiency between RL800@25%CaCl₂ and state‑of‑the‑art salt‑based composites (at 60 min under 1000 W m⁻² irradiation) (Supplementary Table 10). g Infrared thermal images of RL800@25%CaCl₂ under direct solar irradiation.

As shown in Fig. 5c, d, pure CaCl₂ possesses very low solar absorptance, whereas RL800@25%CaCl₂ exhibits a high average solar absorptance of 97.72%. This guarantees efficient conversion of solar irradiation into thermal energy, providing the heat required for the composites’ desorption and enabling rapid water release⁶⁶. Under simulated solar illumination, the fully hydrated RL800@25%CaCl₂ embedded in thermal insulation undergoes fast desorption (Supplementary Figs. 34 and 35). The surface temperature rises rapidly, reaching 82°C within 30 min and achieving an 84% desorption efficiency. The temperature then climbs gradually to 90 °C, with the desorption efficiency reaching 93% after 60 min (Fig. 5e). In contrast, pure CaCl₂ exhibits much slower heating and desorption under identical conditions, reaching only 50 °C (30 min) and 54 °C (60 min), with corresponding desorption efficiencies of 32% and 54%, respectively. This pronounced difference highlights the critical role of the synergistic shape-interface-channel ordering architecture in promoting solar-driven desorption. Infrared thermography (Fig. 5g) further confirms the rapid temperature rise of RL800@25%CaCl₂, whereas pure CaCl₂ shows significantly lower temperatures and slower desorption under the same conditions (Supplementary Fig. 36). Compared with state‑of‑the‑art solar-driven TCES materials (Fig. 5f and Supplementary Table 10)^{6,29,50,53,67,68}, RL800@25%CaCl₂ delivers competitive desorption kinetics and efficiency, making it a promising candidate for efficient solar-driven TCES.

Limitations and outlook

Despite the promising performance demonstrated here, several challenges remain for practical implementation. The current fabrication route involves DMF-based phase separation, KOH activation and high-temperature treatments, which entail non-negligible solvent, chemical and energy inputs; thus, further optimization toward greener, lower-cost and energy-efficient processing will be necessary. In addition, although RL800@25%CaCl₂ exhibits fast kinetics and excellent durability, its long-term stability under realistic operating conditions, such as fluctuating humidity, temperature variations and prolonged outdoor exposure, requires further validation, particularly regarding potential structural degradation and salt redistribution. Moreover, this study primarily focuses on material-level performance, while the overall efficiency of solar-driven TCES systems depends critically on reactor design, heat and mass transfer coupling, and thermal management, which remain to be systematically evaluated. Future work should therefore integrate the present ordered materials into practical reactor configurations, establish a full energy balance under realistic conditions, and develop robust operational strategies to ensure stable performance under dynamic environments. Addressing these challenges will be essential to translate the present material design into scalable and efficient solar thermochemical energy storage technologies.

Discussion

In this work, we establish a synergistic shape-interface-channel ordering strategy for solar-driven thermochemical energy storage. By integrating a monoconcave RBC-mimic shape, leaf-veins-inspired aligned channels and hydrophilic surface chemistry, the resulting RL800 matrix enables simultaneous optimization of mass transport, structural stability and solar absorption. This multiscale ordering leads to a high specific surface area (901 m²g⁻¹), robust mechanical strength (4.69 N) and strong wettability, providing an effective scaffold for salt incorporation. Upon combining RL800 with CaCl₂, the resulting RL800@25%CaCl₂ delivers an ultrafast water sorption rate coefficient of 0.03973 min⁻¹ and a high equilibrium capacity of 885 mg g⁻¹ at 20 °C and 60% RH, corresponding to an energy storage density of 988.38 kJ kg⁻¹. Importantly, it maintains superior cycling stability with only 5.65% capacity decay over 1000 successive sorption-desorption cycles (~ 3722 h). With further assistance of high solar absorptance of 97.72%, desorption efficiency as high as 93% within 60 min is demonstrated under one-Sun irradiation. Compared with state-of-the-art salt-based composites, which typically exhibit slower kinetics (0.006–0.025 min⁻¹), severe capacity degradation within tens of cycles, and limited solar-driven regeneration efficiency, the proposed RL800@25%CaCl₂ demonstrates a prominent enhancement in kinetics, stability and solar utilization. This improvement arises from the synergistic coupling of external shape, internal channel architecture and interfacial properties, which collectively reduce diffusion resistance, enhance vapor accessibility and suppress salt agglomeration. The proposed design strategy provides a generalizable framework for engineering TCES materials via synergistic shape-interface-channel ordering and advances the practical implementation of solar-driven TCES technologies.

Methods

Materials and chemicals

Calcium chloride anhydrous (AR, 96%, CaCl₂) was purchased from Shanghai Aladdin Industrial Co., Ltd. Polyacrylonitrile (PAN, average Mw = 85,000) and Potassium hydroxide (95%, KOH) were purchased from Shanghai Macklin Biochemical Technology Co., Ltd. N, N-Dimethylformamide (DMF) and absolute ethanol were purchased from Sinopharm Chemical Reagent Co., Ltd. Deionized water was used throughout the experiments.

Preparation of RL and activated samples

RL pellets were synthesized via a solvent-induced phase separation technique, followed by carbonization and activation. In brief, PAN powder was dissolved in DMF to obtain an 8 wt.% PAN solution, which was then heated to 80 °C and stirred until a uniform solution was formed. The solution was transferred to a 20 ml syringe and injected at a rate of 5 mm min⁻¹ through a 21 G needle located 15 cm above the surface of water-ethanol coagulation bath (mass ratio of water to ethanol is 7:3). After 24 h, the resulting pellets were removed into deionized water for 24 h, with the water replaced several times to remove any residual solvent. The wet pellets were dried in an air-drying oven at 80 °C for 12 h. The dried pellets were then subjected to oxidation stabilization at 280 °C for 2 h with a heating rate of 0.5 °C min⁻¹ in a muffle furnace under an air atmosphere. The oxidized pellets were subsequently carbonized at 900 °C for 2 h with a heating rate of 5 °C min^-1 in a tubular furnace under an Ar atmosphere and named as RL. To obtain activated samples, typically, 1 g RL pellets were immersed in a KOH solution prepared by dissolving 2 g of KOH in 17 g deionized water (mass ratio of KOH to RL is 2:1) and stirred for 3 h. After impregnation, the entire slurry was transferred into a corundum crucible and dried at 120°C for 2 h. The dried mixture was activated in a tubular furnace under Ar flow at a heating rate of 5 °C min^-1 for 1 h at 700 °C, 800 °C, and 900 °C, and named as RL700, RL800, and RL900, respectively. After activation, the samples were thoroughly washed with deionized water until a neutral pH was reached.

Preparation of RL800@CaCl₂ composites

RL800@CaCl₂ composites were prepared by the wet impregnation method. RL800 was first dried at 100°C for 1 h to remove adsorbed water. The dried RL800 was then immersed in five aqueous CaCl₂ solutions with different concentrations (10 wt.%, 20 wt.%, 25 wt.%, 30 wt.%, and 40 wt.%) and stood for 24 h. Typically, RL800 pellets were immersed in CaCl₂ aqueous solution with a solid-to-solution mass ratio of 1:40 for 24 h. The resulting samples were named as RL800@10%CaCl₂, RL800@20%CaCl₂, RL800@25%CaCl₂, RL800@30%CaCl₂, and RL800@40%CaCl₂, respectively. The impregnated materials were filtered using a Büchner funnel and dried in an oven at 200 °C for 2 h.

Characterizations of samples

The microscopic and macroscopic morphologies of samples were observed by scanning electron microscopy (SEM, ZEISS, Geminin 300), while the sample elemental compositions were analyzed using an energy dispersive spectrometer (EDS). The three-dimensional reconstruction topography of the sample was obtained by micro-CT (X-ray micro-CT, Skyscan 1272, Bruker). X-ray diffraction (XRD) patterns were obtained on the diffractometer (Bruker D8 ADVANCE). Thermogravimetric analysis (TGA) was carried out using a thermogravimetric analyzer (TA Instrument, SDT650) by arranging the sample as a single layer covering the bottom of the alumina crucible and heated from room temperature to 800 °C with a heating rate of 10 °C min^-1 in the airflow of 100 mL min⁻¹. The porous structure was analyzed using an N₂ adsorption/desorption analyzer (ASAP 2020) and mercury intrusion porosimetry (AutoPore V 9600). The Brunauer-Emmett-Teller (BET) model is adopted to calculate the specific surface area of original samples, and the Barrett-Joyner-Halenda (BJH) model is used to obtain pore size distribution. Fourier-transform infrared spectroscopy (FTIR) was measured by Nicolet iS50 (Thermo Fisher Scientific) in the range of 400 cm⁻¹ and 4000 cm⁻¹. The contact angles were measured by a surface contact angle meter (DSA25, Kruss). The energy storage density was measured by a differential scanning calorimeter (DSC, Themys One, SETARAM) using samples fully hydrated at 20 °C and 60% RH for 6 h. The samples were arranged as a single layer covering the bottom of a vented aluminium crucible and heated from 20 °C to 250 °C with a heating rate of 5 °C min⁻¹ in a N₂ flow rate of 40 mL min⁻¹. The water sorption behavior and the cyclic stability tests were evaluated using a gravimetric method gas/vapor sorption analyzer (BeiShiDe, 3H-2000PW). Cycles 1-500 were conducted at 20 °C and 60% RH for 180 min, followed by heating for 40 min to ~ 200 °C for dehydration and cooling for 40 min back to the sorption temperature (20 °C). For checking cycle stability at different conditions, cycles 501-1000 were performed at 30 °C and 60% RH for 100 min, followed by identical dehydration (heating for 40 min to ~ 200 °C) and cooling steps (40 min to 30 °C). (Supplementary Figs. 26–28). The spectral reflectance ρ_w,s in the range of 300-2500 nm was measured using an Ultraviolet–Visible–Near infrared (UV-Vis-NIR) spectrometer (PerkinElmer, Lambda 1050 + ). During the test, the sample was poured naturally into a custom cylindrical quartz sample holder (the inner cavity of the holder had a diameter of 19 mm and a depth of 7 mm) equipped with a window made of quartz glass. Since the quartz glass would affect the result, the method in literature was used to correct this, calculated by the following formula⁶⁹:

$$R\left(\lambda \right)=\frac{{\rho }_{\mathrm{w},\mathrm{s}}-{\rho }_{\mathrm{w}}}{{\tau }_{\mathrm{w}}^2+{\rho }_{\mathrm{w}}\left({\rho }_{\mathrm{w},\mathrm{s}}-{\rho }_{\mathrm{w}}\right)}$$ 1

where R(λ) represents the true reflectance of the sample, ρ_w,s represents the reflectance measured by covering the sample surface with a quartz glass window, ρ_w represents the reflectance of the quartz glass window, ${\tau }_{\mathrm{w}}$ represents the transmittance of a quartz glass window. The spectral absorptance A(λ) is calculated by:

$$A\left(\lambda \right)=1-R\left(\lambda \right)$$ 2

The average solar absorptance A_ave is employed to evaluate solar capture properties of samples, as shown by the following formula:

$$A_{\mathrm{ave}}=\frac{{\int }_{300\mathrm{nm}}^{2500\mathrm{nm}}A\left(\lambda \right)I{\left(\lambda \right)}_{\mathrm{AM1}\mathrm{.5D}}d\lambda }{{\int }_{300\mathrm{nm}}^{2500\mathrm{nm}}I{\left(\lambda \right)}_{\mathrm{AM1}\mathrm{.5D}}d\lambda }$$ 3

where I(λ)_AM1.5D represents the spectral solar irradiance under AM1.5D conditions.

Hydrated samples were also measured after fully hydrated at 20 °C and 60% RH for 6 h. The crush strength is employed to evaluate the mechanical performance, which is recorded by a manual compression dynamometer (JINKELIDA, GWJ-3) before single pellet crushes. Unless otherwise specified, all samples were fully regenerated at 200 °C for 2 h before testing.

Solar-driven thermochemical energy storage performance measurement

Desorption performance was evaluated on a photothermal platform. The samples were first fully hydrated at 20 °C and 60% RH for 6 h, and then placed in a circular container and naturally packed as a single layer and left loose without external compaction. Under direct irradiation of simulated sunlight (MERRY CHANGE, MC-PF300C) at 1000 W m⁻², solar-driven desorption was conducted under ambient atmospheric conditions, without additional purge gas flow. Surface temperature and mass loss were monitored using an infrared camera (TELOPS, FAST M80hd IRC-HD-MC-SX) and an electronic balance (SARTORIUS, BSA323S), respectively. The infrared camera was positioned directly above the sample with an unobstructed optical path. The emissivity was set to 0.96 for RL800@25%CaCl₂ and 0.98 for hydrated CaCl₂. The desorption efficiency η (%) was calculated as

$$\eta =\frac{m_{eq}-m}{m_{eq}-m_0}$$ 4

where m_eq (g g⁻¹) is the equilibrium sorption amount, m (g g⁻¹) is the dynamic water uptake, and m₀ (g g⁻¹) is the initial water uptake.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

E.C.: Conceptualization, methodology, validation, investigation, data curation, formal analysis, visualization, writing-original draft. Q.Z.: Validation, investigation, formal analysis, visualization. Q.L.: Software, formal analysis, visualization. Y.W.: Supervision. X.L.: Writing-Review & Editing, Supervision, Funding acquisition. H.Y.: Formal analysis, visualization, revision. J.W.: Visualization, revision. C.S.: Formal analysis, visualization. Z.J.: Visualization. S.L.: Formal analysis. T.R.: Investigation. Y.L.: Supervision.

Peer review

Peer review information

Nature Communications thanks Lingling 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 work is financially supported by the National Foundation Science of China (No. 52488201 and No. U24A20143) and as well as Scientific Research Innovation Capability Support Project for Young Faculty (No. ZYGXQNJSKYCXNLZCXM-E6). X.L. also wants to thank the support from Jiangsu Province (No. BK20232022, BE2022024, BT2024009, and BE2023094) and the National Key Laboratory of Spacecraft Thermal Control (No. NKLST-JJ-202401022).

Data availability

The data supporting the findings of this study are available within the main text and its Supplementary Information files. Source data are provided in 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/s41467-026-76588-z.