Li-modified BaCoO_3−δ for thermochemical energy storage: enhanced reaction performance and modification mechanism

Abstract

Perovskite materials are promising candidates for thermochemical energy storage, yet conventional substitutional doping has not effectively increased their reactivity at lower temperatures (600–900 °C), limiting practical applications. This study synthesized Li-modified BaCoO_3−δ to enhance gas–solid reaction activity by introducing structural defects. XRD and ICP analyses confirmed the incorporation of Li into the BaCoO_3−δ lattice. TG and DSC experiments demonstrated that Li doping significantly improved the redox activity of the material within the 600–900 °C range, increasing the thermochemical storage density by approximately 75% from 199.1 kJ kg⁻¹ to 348.4 kJ kg⁻¹. Van't Hoff analysis indicates that Li doping increases the entropy and enthalpy of the thermochemical reactions. Cycling experiments showed stable performance enhancement, retaining over 95% (and even up to 99%) of activity after 450 cycles, still significantly outperforming fresh BaCoO_3−δ. DFT calculations, XPS, and EPR analysis revealed that Li doping stabilizes surface oxygen vacancy structures, increasing surface defect oxygen content and enabling stronger redox reactions at lower temperatures. This study elucidates how Li doping enhances the thermochemical heat storage performance of BaCoO_3−δ, providing valuable insights for designing perovskite materials.

---

Introduction

Thermal energy storage (TES) is essential for balancing the supply and demand of thermal energy, thereby improving the efficiency of its production and consumption.¹ TES is widely used in solar thermal power generation and heating systems,^2–4 thermal management of equipment,^5,6 and industrial and building energy efficiency.⁷ TES technologies generally fall into sensible heat storage,^8,9 latent heat storage,^10–12 and thermochemical heat storage (TCES).^12,13 TCES, which relies on reversible chemical reactions to store and release heat, offers high energy density within a narrow temperature range, making it ideal for long-term storage.¹⁴ Key TCES systems include amino energy storage,^15,16 metal hydride energy storage,^17–19 carbonate energy storage,^20,21 metal oxide energy storage,^22–26 and hydroxide energy storage.^27–29 Among these, metal oxide systems are notable for their suitability in air environments due to their simplicity, flexibility, and controllability.^{22–26,30,31} However, they are constrained by a limited operational temperature range and challenges in finding suitable materials for specific temperature requirements.^32,33

Recently, mixed metal oxides with a perovskite structure (ABO₃) have gained significant attention. These materials are valued for their wide thermal storage temperature range,³⁴ excellent cycling stability,³⁵ and favorable doping properties.³⁶ By doping elements into the A or B sites of perovskites, their reaction properties can be modified, expanding their temperature range and application potential.^35–39 Perovskites such as the doped CaMnO₃ family, La-family, Sr-family, and Ba-family have been widely studied for high-temperature thermochemical energy storage (TCES) applications. By introducing dopants, such as Sr at the A-site or Al, Ti, Fe, Co, Mn, and Cr at the B-site, researchers have improved the stability of their crystal structures and enhanced their reaction performance.^36,40–46 Zhang et al.³⁴ conducted a study on 12 different types of Ba/Sr-based perovskite materials. They discovered that BaCoO_3−δ exhibited the highest reactivity and reversibility, with an energy storage capacity of 292 kJ kg⁻¹ under a redox cycle between 600 °C/0.2 atm O₂ and 1050 °C/10⁻⁶ atm O₂. They further investigated modifications with Sr doping at the A-site and Fe doping at the B-site. It was found that Sr doping at the A-site did not alter the redox mode of BaCoO_3−δ but affected its O₂ exchange capacity, slightly weakening it while improving reversibility. In contrast, Fe doping at the B-site significantly changed the redox mode, lowering the reaction temperature but greatly limiting the oxygen exchange capacity. Yuan et al.^31,47 studied BaCoO_3−δ materials with Sr doping at the A-site and Mn doping at the B-site, demonstrating that the modified BaCoO_3−δ materials can undergo stable redox cycling in air. One of the composites, Sr_0.5Ba_0.5CoO_3−δ, had a reduction reaction enthalpy of 202 kJ kg⁻¹. Sr doping at the A-site improved the micro-porous structure, promoting oxygen diffusion and resulting in higher reversibility. Mn doping at the B-site inhibited the formation of oxygen vacancies, thereby increasing the redox reaction temperature, with BaCo_1−xMn_xO_3−δ (x = 0–0.4) achieving an adjustable reaction onset temperature range of 426–702 °C. BaCoO_3−δ is a promising air-cycled perovskite material. However, its redox reaction activity is significantly reduced below 900 °C, adversely affecting its thermal storage efficiency. Conventional A-site and B-site substitutional doping mainly improve the material's structural stability and microscopic bonding energy. While this can modify the onset reaction temperature and overall reaction extent, it does not enhance the reaction activity at lower temperatures.

A key factor influencing the material's reactivity is the presence of oxygen vacancies on its surface.⁴⁵ Introducing foreign heteroatoms is a common technique to create vacancies in solid metal oxide materials.⁴⁸ These vacancies facilitate lattice oxygen migration during gas–solid reactions, thereby increasing the number of intrinsic active sites.⁴⁹ The well-documented self-diffusive properties of alkali metal elements make them suitable for doping into the crystal structure, thus generating more active sites that enhance redox reactions.⁵⁰ In this study, we introduced Li ions into perovskites to improve their redox activity at lower temperatures. To understand the microscopic mechanisms by which these alkali metal elements influence redox behavior, we employed Density Functional Theory (DFT) calculations. These calculations provide reliable theoretical guidance for new approaches to material modification.

Experimental

Sample preparation

All the samples were synthesized by the modified Pechini process,⁵¹ and their abbreviations are listed in Table 1. Barium nitrate, cobalt nitrate, and lithium nitrate were dissolved in deionized water according to stoichiometric ratios. To compensate for lithium evaporation, an additional 30% wt lithium nitrate was added. Subsequently, a specific amount of citric acid, EDTA, and ethylene glycol was added. The molar ratios of citric acid, EDTA, and ethylene glycol to total metal ions were 1.2 : 0.6 : 0.8 : 1. The pH of the solution was adjusted to 8 by using ammonia. Then, the mixed solution was stirred in a thermostatic water bath at 80 °C for 5 hours to obtain the mixed resin. The resin was placed in a blast drying oven at 200 °C for 3 hours. The dried precursor was transferred to a muffle furnace and pre-calcined at 300 °C for 3 hours to remove organic components, and then calcined at 1000 °C for 5 hours. After the sample had cooled, it needed to be reheated to 1000 °C and then cooled to room temperature to complete a redox cycle. Subsequently, the sample was ground into a powder, marking the completion of the material preparation process. To analyze the material's crystalline phase composition and surface chemistry changes during warming, reduced samples must be characterized and analyzed at different temperatures. The samples were heated in an air atmosphere, from room temperature to 800 °C, 850 °C, 900 °C, 1000 °C, and 1050 °C, respectively. The heating rate was 20 °C min⁻¹, and each sample was maintained at the targeted temperature for 10 minutes. After that, the samples were passed through nitrogen and cooled to room temperature at a rate of −20 °C min⁻¹. This process resulted in the production of reduced samples at each temperature.

Table 1 Samples prepared and abbreviations

Sample name	Abbreviation
BaCoO3−δ	BC
Li0.03Ba0.97CoO3−δ	LBC3
Li0.0625Ba0.9375CoO3−δ	LBC6.25
Li0.1Ba0.9CoO3−δ	LBC10
Li0.15Ba0.85CoO3−δ	LBC15
Li0.2Ba0.8CoO3−δ	LBC20
Li0.25Ba0.75CoO3−δ	LBC25

---

Material characterization

Lattice and surface characterization. X-ray diffraction data were obtained using the X-pert Powder instrument by PANalytical B.V. The samples were scanned using Cu Kα radiation (λ = 0.1540598 nm) produced with an extraction voltage of 40 kV at 40 mA current, in a range of 10° < 2θ < 90° with a step size of 0.01° and a scan rate of 10° min⁻¹. The samples' morphology changes after redox cycles were observed using a Hitachi Regulus 8100 (SEM). Additionally, an Energy Dispersive Spectrometer (EDS) was utilized to identify the distribution of various elements in the synthesized samples. The distribution of A-site and B-site elements in the samples before and after cycling was analyzed using a Hitachi HT-7700 120 kV transmission electron microscope (TEM). Due to the low ionization energy of Li, EDS could not provide Li elemental distribution data. The distribution of Li elements can be inferred from the changes in the elemental distribution of Ba and Co. Finally, X-ray Photoelectron Spectroscopy (XPS) was conducted on a Thermo Fisher K-Alpha spectrometer, with C1s (284.8 eV) used for energy calibration. Low-temperature electron paramagnetic resonance (EPR) measurements were conducted using a Bruker ESRA-300 spectrometer to detect the presence of oxygen vacancies in the perovskite samples.

Thermodynamics measurements. Perovskite is capable of releasing oxygen during reduction and absorbing oxygen during oxidation, which causes a change in the sample's mass. To test its reaction properties, thermal gravimetric (TG) experiments were conducted using TGA55 equipment produced by TA. The sample used for testing weighed approximately 10 mg and was chosen to quickly achieve the equilibrium of the redox reaction and obtain precise experimental data. The change in mass during testing, denoted as Δm, evaluates thermal storage material reactivity. The thermogravimetric experiment encompassed a single redox cycle. The process entailed reducing the sample in the air through gradual heating from 50 to 1050 °C at a rate of 20 °C min⁻¹, followed by a 10 minute hold at 1050 °C. The sample was then cooled down to 400 °C in the air at a rate of −20 °C min⁻¹, undergoing a re-oxidation process. The gas flow was constant at 100 N mL min⁻¹ at atmospheric pressure throughout the process. A blank run was performed using the same measurement conditions to correct for buoyancy. The samples' reaction extent, reoxidation rate, initial reaction temperature, etc., were analyzed in the TG curves. The DSC used the same temperature and gas conditions as the TG, and the data from the second redox period were used to calculate the reaction enthalpy of the sample. Besides, this study employed thermodynamic equilibrium tests under varying oxygen partial pressure environments to analyze the heat storage reactions of the samples. The results from two thermogravimetric (TG) experiments conducted at different oxygen partial pressures (pO₂ = 0.8, 0.5, 0.1, 0.00001) provided insights into the non-chemical stoichiometry of the material's oxygen content (δ). This value can be calculated using the eqn (1):where m₀ is the initial mass of the sample, Δm is the mass change during the reaction, M is the relative molecular mass of the sample, and M_O is the relative atomic mass of the oxygen atom. The value of δ at the isothermal plateau can be considered constant, allowing for the extraction of a dataset that includes δ, pO₂, and temperature (T) once equilibrium is reached at each temperature. Using the Van't Hoff method, the molar enthalpy change (ΔH^red_O) and molar entropy change (ΔS^red_O) associated with the redox reactions were determined. According to the Arrhenius equation, the relationships between the enthalpy change, entropy change, oxygen partial pressure (pO₂), and temperature (T) are represented as follows:

The values of ΔH^red_O and ΔS^red_O can be derived from the slope and intercept of the plotted curves of versus. Furthermore, by integrating the molar reaction enthalpy corresponding to each value of δ, the total chemical heat storage density can be obtained, as illustrated in eqn (3) and (4).

Cycle experiment. To investigate the cyclic stability of doped BaCoO₃, three samples with the best reaction performance were selected: LBC6.25, LBC10, and LBC15. These samples were cycled 450 times between 600 and 1050 °C under an air atmosphere in a tubular furnace. Every 150 cycles were chosen as a test point. TG tested the redox performance, and SEM observed the microstructures of the samples at every test point.

Computational methods

The Vienna Ab initio Simulation Package (VASP 5.4.4) was used for density-functional theory calculations in crystallography. The calculations utilized the Projector-Augmented Wave (PAW) approach and the Perdew-Wang 91 (GGA-type) functional, and the DFT + U method was proposed to improve the description of systems with strongly correlated d or f electrons. Effective U_Co = 3.32 eV taken in Materials Project database⁵² were used for the 3d orbital of Co. Initial hexagonal phase BaCoO₃ lattice data (a = b = 5.74 Å, c = 4.78 Å; α = β = 90°, γ = 120°) were obtained from the Materials Project database. A bulk model consisting of 80 atoms arranged in a 2 × 2 × 2 supercluster was created based on the results of the convergence test. To create Li_xBa_1−xCoO₃ models, some of the Ba atoms were replaced with Li atoms. Lattice optimization was performed first. A 9 × 9 × 9 k-point grid was created using the Monkhorst–Pack method. The kinetic energy cut-off for the plane wave basis set was 510 eV with an energy convergence value of 1 × 10⁻⁵ eV. A conjugate gradient algorithm was used to relax the ions to their instantaneous ground state until the force applied to each ion was less than 0.05 eV Å⁻¹. The computational procedure took into account the spin polarization. After optimizing the lattice, slab models with different end faces for various crystallographic indices were obtained by surface-cutting the crystal models. The surface energies were calculated by relaxation optimization of these slab models, and then the most stable surfaces were screened for BaCoO₃versus Li_xBa_1−xCoO₃. The Slab model optimization process was performed using a Monkhorst–Pack 3 × 3 × 1 k-point mesh, and the convergence parameters were kept consistent with the optimization process. Calculations of the surface and subsurface oxygen vacancy formation energies are performed on the screened stable surfaces.

Results and discussion

Microstructural structure and composition

To understand the presence pattern of Li in the microstructure of the samples, the composition and crystal structure of the samples were first investigated. Fig. 1a demonstrates the XRD patterns of BaCoO_3−δ samples doped with different proportions of Li. The synthesized Li_xBa_1−xCoO_3−δ samples (x = 0, 0.03, 0.0625, 0.1, 0.15, 0.2, 0.25) all exhibit the hexagonal crystal phases BaCoO_3−δ (ICCD PDF #97-001-5257 P6₃/mmc). However, BaO (ICCD PDF #04-005-4400) was detected in the Li-doped samples, likely due to excess Li displacing Ba during the synthesis process. The peak position details are provided in Table S1.† Since BaO does not participate in the reduction reactions and its content is minimal, its influence can be disregarded. When the Li doping ratio exceeds 0.15, a new crystalline phase, identified as the tripartite crystal system of LiCoO₂ (ICCD PDF #04-008-6330, R3̄m) appears. This indicates that Li may reach its solubility limit in the BaCoO_3−δ lattice and begin to precipitate in the form of the LiCoO₂ phase. Fig. 1b presents a magnified view of the main XRD diffraction peaks of the BaCoO_3−δ phase. The incorporation of Li causes a significant shift in these peaks towards lower angles, particularly in the (2 −1 −1 0) and (1 0 −1 7) planes, indicating that Li likely enters the BaCoO_3−δ lattice and induces unit cell expansion. To assess the stability of Li in the Li_xBa_1−xCoO_3−δ samples during high-temperature preparation, the metal element contents were determined using the ICP method, as shown in Fig. 1c. The levels of Ba, Co, and Li are consistent with the preset values during material preparation, with no significant Li evaporation⁵³ during high-temperature processing. This indicates that Li is present in a stable structure within the samples, successfully incorporating into the BaCoO_3−δ lattice, especially at doping ratios below 0.15. Fig. 1d illustrates the microstructure and elemental distribution of BaCoO_3−δ and two Li-doped BaCoO_3−δ samples. The LBC10 sample exhibits more uniform grain sizes, smoother surfaces, and looser packing compared to the BC sample, indicating improved material homogeneity and potentially enhanced reactivity and stability. In contrast, the LBC15 sample displays some fragmented small particles on its surface. While the distribution of Li cannot be determined through EDS spectroscopy, the LBC15 sample shows Ba-deficient and Co-rich regions on its surface. Combined with XRD results, these findings suggest that the smaller particles attached to the BaCoO_3−δ grains in the SEM images are precipitated LiCoO₂. The impact of LiCoO₂ on the material's reactive properties will be clarified in the following sections.

Fig. 1 Structure and composition of Li-doped BaCoO_3−δ samples: (a) XRD patterns of BaCoO_3−δ doped with varying proportions of Li; (b) magnified view of the main XRD diffraction peaks of the BaCoO_3−δ phase; (c) ICP results showing the metal element content of Li_xBa_1−xCoO_3−δ samples; (d) microstructural images of BC, LBC10, and LBC15, along with the metal element distribution mapping for the lithium-doped samples (SEM-EDS).

Thermal storage performance

Reaction extent, reaction enthalpy, reaction kinetics, and cyclic stability are important indicators for evaluating the performance of thermochemical heat storage materials for practical applications.⁵⁴ TG and DSC experiments were used to investigate the reaction properties of the synthesized Li_xBa_1−xCoO_3−δ samples, with mass change (Δm) used to measure the degree of reaction.⁵⁵Fig. 2a shows the TG curves of the reduction and oxidation reactions of the Li_xBa_1−xCoO_3−δ sample. The Li-doped samples exhibit significantly enhanced reduction and oxidation reaction extents compared to the undoped BC sample. Specifically, the Li_xBa_1−xCoO_3−δ materials show higher activity and better kinetics for reduction at 700–900 °C and oxidation at 600–800 °C. Notably, the Li-doped materials display significantly enhanced redox activity near 860 °C during reduction and 770 °C during oxidation. The raw TG data and the thermogravimetric hysteresis curve are presented in Fig. S1,† demonstrating that the reaction reversibility for the Li_xBa_1−xCoO_3−δ samples remains above 99.9%, with enhanced redox activity in the 600–900 °C range. Two-cycle TG tests, shown in Fig. S2,† confirm that the Li-doped samples maintain excellent reversibility in the second cycle, and the increased reactivity in the specific temperature range persists stably. To investigate the impact of LiCoO₂ as an impurity on the redox reactions, TG tests were conducted on pure LiCoO₂ samples, with the results shown in Fig. S3.† Pure LiCoO₂ exhibits no substantial mass change between 600 and 900 °C, indicating that the enhanced reactivity is primarily due to the Li_xBa_1−xCoO_3−δ solid solution.

Fig. 2 Thermal reaction characteristics: (a) TG curves of Li_xBa_1−xCoO_3−δ samples (showing reduction reaction during heating and oxidation reaction during cooling); (b) TG and DSC curves of BC sample (pre-Li doping); (c) TG and DSC curves of LBC6.25 sample (post-Li doping); (d) comparison of oxidation/reduction reaction enthalpies for Li_xBa_1−xCoO_3−δ samples.

To elucidate the enhancement in reaction enthalpy, a comprehensive analysis of the thermal storage performance has been conducted. The DSC results for BC and LBC6.25 are depicted in Fig. 2b and c. The main endothermic peak for BC occurs around 950 °C during heating, while the main exothermic peak appears around 890 °C during cooling. These peaks correspond to the most rapid reaction stages in the TG curve. In contrast, LBC6.25 exhibits a new endothermic peak at around 860 °C during reduction and a new exothermic peak at around 770 °C during oxidation, indicating enhanced reactivity in this localized temperature zone, as shown in the TG curve. This increased reactivity significantly raises the reaction enthalpy of the doped material. Fig. S4† demonstrates the thermal storage performance of LBC10, LBC15, LBC20 and LBC25, whose DSC curves show similar characteristics to LBC6.25. Doping with Li increases the likelihood of oxygen loss and uptake, lowering the temperature required for the onset of vigorous reduction–oxidation reactions. Consequently, the redox reaction kinetics and reaction enthalpies are improved in the 600–900 °C temperature range, enhancing the material's practical applications. The enthalpies of reduction and oxidation reactions for Li_xBa_1−xCoO_3−δ samples, as measured by DSC, are presented in Fig. 2d. The difference between reduction and oxidation enthalpies for each material is relatively small, indicating that the processes of heat storage and release are highly reversible. Compared to BC (199.1 kJ kg⁻¹), all LBC samples exhibit enhanced heat storage density, with LBC6.25 and LBC10 showing notable improvements, reaching 348.4 kJ kg⁻¹ and 305.1 kJ kg⁻¹, respectively an increase of 75%. However, as the level of Li doping increases, the heat storage density of LBC samples declines. This suggests that higher Li content may alter the bond energy within the lattice, reducing the energy required for oxygen release and slightly lowering the heat storage capacity. Additionally, the formation of LiCoO₂ as an inert phase also contributes to this reduction in heat storage density.

From a thermodynamic perspective, the chemical equilibrium of the heat storage reaction under various oxygen partial pressures was investigated. Using the Van't Hoff method, the changes in enthalpy (ΔH) and entropy (ΔS) during the reaction process were quantified, as shown in Fig. 3. The data reveal that LBC undergoes a higher degree of reduction compared to BC across all oxygen partial pressure conditions. Fig. 3c illustrates the variation in partial molar enthalpy and partial molar entropy with respect to δ during the heat storage reaction. Notably, both ΔH^red_O and ΔS^red_O for LBC are slightly higher than for BC, particularly in the range of 0.06 < δ < 0.2. This range corresponds to the stage where LBC's reactivity is significantly enhanced. These results suggest that Li doping increases the entropy change during the heat storage reaction, which enhances the reduction activity of oxygen ions. As a result, more oxygen is released, and a possible phase transition occurs, leading to a higher enthalpy change during the reaction. Finally, the enthalpy change for the reduction reactions of both BC and LBC was determined through the integration of the partial molar enthalpy, as shown in Fig. 3d. The Li-doped sample exhibits a higher heat storage enthalpy change, consistent with the results obtained from the DSC measurements. This agreement further supports the conclusion that Li doping improves the overall thermochemical performance of the material.

Fig. 3 Enthalpy and entropy change tests for heat storage reactions: equilibrium oxygen non-stoichiometry versus temperature for (a) BC and (b) LBC10 under different oxygen partial pressures; (c) partial molar enthalpy and partial molar entropy of samples as a function of δ; (d) chemical heat storage reaction enthalpies of BC and LBC10.

Table 2 compares the thermochemical performance of the most promising perovskites. The selection of reaction atmosphere and enthalpy testing methods significantly influences the determination of heat storage density. The air cycle exhibited inferior redox activity compared to the oxygen-concentration-controlled cycle, and the reaction enthalpies measured via DSC tended to be lower than those calculated using the Van't Hoff and Point-defect model. Notably, the chemical reaction enthalpy (ΔH_ch) of the Li_0.0625Ba_0.9375CoO_3−δ material demonstrated exceptional performance among air-cycled perovskites, establishing itself as a highly promising thermochemical heat storage material.

Table 2 Key parameters of perovskite oxides for the thermochemical energy storage

Compositions	T (°C)	pO2	ΔHch (kJ kg^−1)	Approach	Reference
CaCo0.05Mn0.95O3−δ	500–1000	0.21–0.00001	571	Van't Hoff calculation	45
CaCr0.05Mn0.95O3−δ	500–1000	0.17–0.0001	392	Point-defect model	43
CaCr0.1Mn0.9O3−δ	400–1000	0.21-argon	174	DSC	46
SrCoO3−δ	400–1000	0.21–0.0001	48	DSC	56
SrFe0.2Co0.8O3−δ	600–950	0.2–0.000001	45	DSC	34
Sr0.7Ba0.3CoO3−δ	500–1100	0.21 (air)	186	DSC	57
La0.3Sr0.7Co0.9Mn0.1O3−δ	200–1250	0.9–0.001	245	Van't Hoff calculation	36
BaCoO3−δ	600–950	0.2–0.000001	292	DSC	34
BaMn0.15Co0.85O3−δ	608–1050	0.21 (air)	125	DSC	31
BaCoO3−δ	600–1050	0.21 (air)	199	DSC	This work
Li0.0625Ba0.9375CoO3−δ	600–1050	0.21 (air)	348	DSC	This work

---

Cycling stability

To eliminate experimental uncertainties and test the lifespan of the high-performance composite materials, further studies on the stability and crystalline phase evolution of the LBC materials during long-term cycling were conducted. Fig. 4a and b illustrate the reduction and oxidation activity of LBC6.25, LBC10, and LBC15 after 150, 300, and 450 cycles. These results are derived from the analysis of TG data presented in Fig. S5.† After 450 cycles, the redox reaction extents of all three materials decreased but remained significantly higher than that of freshly prepared BC, tending to stabilize after 150 cycles. Specifically, Fig. 4b shows that the oxidation rates of the three materials demonstrate varying degrees of decay. The oxidation performance of LBC6.25 and LBC10 declined significantly in the first 150 cycles but then stabilized over the last 300 cycles. In contrast, the LBC15 material decayed at a comparatively slower rate and maintained a consistently high oxidation reaction extent even after 450 cycles. Combined with the XRD results, this may be attributed to the activation role of the LiCoO₂ phase during high-temperature cycling. Fig. S6† shows the DSC test results of LBC6.25, LBC10, and LBC15 after 450 cycles. The oxidation reaction enthalpies for all three materials exceeded 240 kJ kg⁻¹, with the highest heat storage density reaching 296.1 kJ kg⁻¹, nearly 50% higher than that of fresh BC. Due to their better cycling performance, LBC10 and LBC15 were selected for further study.

Fig. 4 Evolution of the reactivity, crystal structure, and morphology of samples during cycling: (a) reduction activity of LBC6.25, LBC10, and LBC15 before and after 150, 300, and 450 cycles; (b) oxidation activity of LBC6.25, LBC10, and LBC15 before and after 150, 300, and 450 cycles; (c) XRD patterns of LBC10 before and after 150, 300, and 450 cycles; (d) XRD patterns of LBC15 before and after 150, 300, and 450 cycles; (e) TEM morphology and EDS elemental distribution images of raw LBC15; (f) TEM morphology and EDS elemental distribution images of LBC15 after several cycles; (g) SEM images of LBC10 and LBC15 before and after 150 and 450 cycles.

Fig. 4c and d show the XRD results of LBC10 and LBC15 before and after 150, 300, and 450 cycles. The crystal compositions of both materials did not significantly change after cycling, and their structures remained relatively stable. Notably, the LiCoO₂ phase in LBC15 gradually vanished during the cycling process, with no LiCoO₂ detected after 150 cycles. To further investigate, higher precision TEM-EDS morphology tests were conducted on LBC15 samples before and after cycling, as shown in Fig. 4e and f. These results confirm that the LiCoO₂ initially attached to BaCoO_3−δ grains disappeared during cycling. This disappearance could be due to Li evaporation causing decomposition or the formation of a solid solution with BaCoO_3−δ. To clarify the fate of LiCoO₂, further studies were conducted. After 450 cycles, ICP testing on LBC15 compared to the raw sample (shown in Table S2†) indicated that the Li content did not significantly change, with its mass fraction among metallic elements decreasing by only 0.017%. This suggests that the disappearance of LiCoO₂ is not due to Li evaporation but is likely a result of ion diffusion. This phenomenon is most likely due to the Kirkendall effect occurring at the junction of LiCoO₂ and the BaCoO_3−δ grain surface under high temperatures.⁵⁶ Li and Ba diffuse with each other to form a homogeneous Li_xBa_1−xCoO_3−δ solid solution. The reaction formula is as follows:

xLiCoO₂ + (1 − x)BaCoO_3−δ1 → Li_xBa_1−xCoO_3−δ2 (5)

Fig. 4g shows the micro-morphological evolution of LBC10 and LBC15 samples. Before cycling, the microstructures of LBC10 and LBC15 are loose with more developed pore structures. After 150 cycles, both samples exhibit denser structures. XRD results suggest that sintering may have occurred, with grains fusing to form larger ones and smaller grains attaching, reducing the overall pore structure.³³ LBC15 shows more severe sintering than LBC10, likely due to ion diffusion between LiCoO₂ and BaCoO_3−δ. After 450 cycles, both samples have significantly denser structures and reduced porosity. This sintering decreases the gas–solid reaction contact area, potentially contributing to the decline in cycling performance.

Based on the solid solution behavior between LiCoO₂ and BaCoO_3−δ, high-temperature treatment presents a promising method for synthesizing purer LBC15, LBC20, and LBC25 samples. Additional thermal treatments were applied to these samples, resulting in the formation of pure LBC15 and a reduction in LiCoO₂ content in LBC20 and LBC25, as demonstrated by the XRD results in Fig. S7.† Fig. S7c and d† show that no changes were observed in the XRD diffraction patterns after the second thermal treatment and cycling tests, indicating that the solubility of LiCoO₂ in the BaCoO_3−δ crystal structure has an inherent limit. Fig. S8† compares the thermal storage capacities of the three materials before and after treatment. The results show a noticeable decrease in thermal storage density for all samples, likely due to sintering effects associated with prolonged high-temperature exposure. This decrease suggests that excessive Li content can lead to phase separation within the crystal structure, exacerbating ionic migration at the interfaces and intensifying the sintering process. Such alterations introduce greater uncertainty into the material's crystal structure, ultimately impairing its thermal storage performance. Therefore, it is recommended to limit Li doping to below 15% to optimize the thermal characteristics of the materials.

In summary, Li doping significantly enhances the thermal storage performance of the materials, with reliable stability in cycling experiments. The redox reaction activities of the three Li-doped materials showed varying degrees of attenuation during cycling. Two primary reasons contribute to the attenuation of material properties: the primary reason is the sintering of crystals, which reduces the reactive surface area, and a secondary reason is the loss of active sites due to slight Li evaporation during high-temperature cycling. After 450 cycles, the thermal storage performance of the three Li-doped materials remains significantly better than that of freshly prepared BC. The Li_xBa_1−xCoO_3−δ (x ≤ 0.15) material demonstrates comprehensive and stable improvement in thermal storage performance in the 600–1000 °C temperature range, showing promise as a high-performance thermal storage material.

Microstructural evolution

Thermophysical and cycling test experiments have verified that Li doping stably enhances the reactivity of BaCoO_3−δ materials. To investigate how Li facilitates the reduction and oxidation reactions, it is essential to study the crystalline phase and morphology transformation of the samples during these processes. LBC10 was selected as the test sample due to its superior thermal storage performance and crystalline phase purity. Additionally, its Li content is sufficient for detection, providing better comparative results. A comparison of the crystal structures of BC and LBC10 before and after reduction and oxidation is presented in Fig. S9.† After reduction, both samples produced a homogeneous BaCoO_2.22 (ICCD PDF #97-023-8817 Pm3̄m) phase. The peak position of LBC10 significantly shifted to a lower angle compared to BC, indicating that Li was incorporated into the BaCoO_2.22 solid solution, leading to lattice expansion. After oxidation, the main crystal phase peaks of LBC10 also shifted to a lower angle, confirming the stable presence of Li within the BaCoO_3−δ lattice. XRD results indicate that both BC and LBC10 maintained excellent lattice reversibility during the redox process.

Fig. 5a demonstrates the morphological changes of BC and LBC10 during the redox cycle. Initially, both raw BC and LBC10 exhibit loose grain accumulations. LBC10 shows smaller grain sizes, a developed pore structure, and a significantly improved specific surface area compared to BC. After the reduction reaction, the grain morphology of both samples became homogeneous and smooth, forming a homogeneous BaCoO_2.22 solid solution. LBC10 grains were larger and exhibited signs of sintering, possibly due to the more intense oxygen release reaction. After oxidation, BC and LBC10 returned to smaller grain sizes and smoother surfaces, without the angular heterogeneous phases seen on the raw samples. It was also observed that BC exhibited more severe sintering compared to LBC10, suggesting that the oxygen uptake of BC was less sufficient, affecting the reversibility of the reaction. In conclusion, Li doping not only enhances reduction sufficiently, leading to some sintering, but it also dramatically improves the oxidation reaction activity, which can reverse the sintering. Ultimately, LBC10 exhibits superior reversible redox ability during cyclic reactions.

Fig. 5 Phase evolution during one redox cycle: (a) SEM morphology images of BC and LBC10 after reduction and oxidation; (b) XRD pattern evolution of BC during reduction process (800–1050 °C); (c) XRD pattern evolution of LBC10 during reduction process (800–1050 °C).

To achieve a more precise understanding of the lattice evolution during the heat storage reaction, the crystal phase changes of BC and LBC10 during reduction at 800–1050 °C were characterized by XRD, as illustrated in Fig. 5b and c (refer to Fig. S10† for the original data). During the heating process, both BC and LBC10 underwent lattice expansion and formed a homogeneous solid solution with a hexagonal BaCoO₃-like structure at 800 °C. As the temperature increases, the main diffraction peaks of the (1 0 −1 1) and (2 −1 −1 0) crystal planes of BaCoO₃ become closer to each other. Combined with the TG data, it can be inferred that the lattice expanded while a reduction reaction occurred to release oxygen, resulting in significant changes in the lattice constant, suggesting an impending transition to the cubic phase. It's worth noting that LBC10 exhibits two phase transitions at approximately 840 °C and 950 °C, whereas BC shows only a single phase transition around 950 °C. Additionally, the shift in the diffraction peak angle of LBC10 (2 −1 −1 0) is significantly larger than that of BC around 840 °C. This suggests that Li doping induces new oxygen release reactions on the (2 −1 −1 0) plane within this temperature range, potentially triggering phase transitions that significantly affect the material's heat storage capacity. When the temperature exceeds 950 °C, both BC and LBC10 release a large amount of lattice oxygen and undergo a drastic crystal phase evolution, forming a homogeneous cubic phase BaCoO_2.22. The crystal lattice remained relatively stable at 950–1050 °C, with almost no further oxygen release, consistent with the TG data. By studying the crystalline phase and morphology evolution of the material during redox cycling, it was found that Li significantly affected the crystal surface structure during the reaction process. It created a new oxygen release pathway on the hexagonal crystal surface while ensuring the reversibility of the reaction, thereby significantly enhancing the redox activity of the hexagonal phase.

Surface chemistry

The surface chemistry directly determines the nature of the gas–solid reaction.⁵⁷ To explore the modulation mechanism of the reaction properties of the material by Li doping, the changes in the surface properties of the material need to be investigated. LBC10 was selected for a series of XPS characterization. Fig. S11a† shows the Ba 3d/Co 2p spectra of BC and LBC10, revealing a slight increase in the higher binding energy components. However, due to the overlap of Co 2p and Ba 3d, the chemical states of Ba and Co cannot be directly confirmed.⁵⁸ Therefore, Co 3p and Ba 4d spectra were analyzed. As shown in Fig. S11b,† the Co 3p spectrum is divided into Co⁴⁺ (∼62.4 eV) and Co³⁺ (∼60.8 eV),⁵⁸ with Li doping having no significant effect on the Co oxidation state. The differences between the raw data of BC and LBC10 are within the acceptable error range. Consequently, the Ba 4d spectra was used to determine the Li doping effect. Fig. 6a compares the Ba 4d spectra of both BC and LBC10 materials. Due to the presence of Ba ions in different coordination environments on the surface of BaCoO_3−δ crystals, the Ba 4d spectra are divided into two components corresponding to BaII (∼88.7 eV) and BaI (∼87.3 eV).^59,60 It can be observed that the binding energies of the two types of Ba remain largely unchanged before and after doping. However, Li doping increases the content of BaII with higher binding energy. This indicates an increase in the number of Ba²⁺ ions with higher coordination numbers on the crystal surface, leading to an enrichment of surface oxygen. This phenomenon is likely due to Li carrying fewer charges than Ba, necessitating more BaII at the A-site to coordinate with oxygen to maintain electroneutrality. Consequently, this may contribute to the formation of oxygen defects on the surface of the Li-doped material.

Fig. 6 XPS spectra analysis: (a) comparison of XPS Ba4d spectra for raw BC and LBC10; (b) comparison of XPS O1s spectra for raw BC and LBC10; (c) comparison of XPS O1s spectra for reduced BC and LBC10 at 850 °C; (d) comparison of XPS O1s spectra for reduced BC and LBC10 at 1050 °C.

The O1s spectra of BC and LBC10 are compared in Fig. 6b. The O1s spectra were fitted into four components corresponding to lattice oxygen (lattice O, ∼528.3 eV), defective oxygen (O defect, ∼529.2 eV), surface adsorbed OH⁻ and O²⁻ (surface OH⁻/O²⁻, ∼531.0 eV), and surface adsorbed H₂O species (surface H₂O, ∼532.2 eV).^61,62 Surface-adsorbed species account for the vast majority (55–66%) of the O1s spectra of BC and LBC10 samples. Surface defect oxygen and lattice oxygen, although not in high amounts, are the main components affecting the gas–solid reaction activity. Surface defect oxygen is a weakly charged species, and perovskite materials containing this oxygen species are prone to form oxygen vacancies on the surface.⁴⁷ It can be found that the proportion of O defects in LBC10 (24.8%) is significantly more than that of BC (18.2%), implying that LBC10 is more prone to form oxygen vacancies on the surface. It is the possible reason why the hexagonal phase of LBC10 possesses higher reactivity. To better understand the changes in the surface chemical state of O during the reaction process, the reduced samples at 850 °C and 1050 °C, were analyzed by XPS, as shown in Fig. 6c and d, respectively. From Fig. 6c, it can be learned that the lattice oxygen and defect oxygen contents of the reduced hexagonal phases BC and LBC10 at 850 °C are significantly reduced. The lattice oxygen content of BC and LBC10 are very close to each other at 4.1%, while the defect oxygen content of LBC10 is obviously higher than that of BC. With the increase in temperature, the defective oxygen on the crystal surface of the material was continuously released, and the lattice began to collapse inward at 950 °C to form the BaCoO_2.22 solid solution in cubic phase. Fig. 6d illustrates the O1s spectra of reduced BC and LBC10 at 1050 °C. The sample composition at this temperature is dominated by fully reduced cubic-phase BaCoO_2.22, and the electronegativity of O is significantly enhanced within this system. Therefore, the defective oxygen species are no longer present, and there is a significant increase in the binding energy of the lattice oxygen (∼529.5 eV).⁶³ Considering that the lattice oxygen contents of BC and LBC10 are also closer, the oxygen release behavior within 850–1050 °C should be dominated by defective oxygen. This implies that the oxygen release ability of LBC10 is stronger than that of BC before the phase transition occurs, which is consistent with the earlier experimental findings.

Fig. 7 presents the low-temperature EPR results for BC and LBC10, showing oxygen vacancies detected at g = 2.003. The oxygen vacancy signal in LBC10 is significantly stronger than in BC, further supporting the conclusion that lithium doping promotes the formation of oxygen vacancies. In summary, Li doping stabilizes the defective oxygen structure on the surface at room temperature, leading to a higher content of reactive oxygen that can participate in gas–solid reactions. Consequently, in the heat storage reaction, LBC10 not only demonstrates a stronger potential for oxygen release but also effectively absorbs more defective oxygen and maintains stability during oxidation.

Fig. 7 Low-temperature EPR spectra of BC and LBC10.

Enhancing mechanisms of thermal storage performance

The experimental results show that Li doping enhances the reaction kinetics and reaction enthalpy in the mid-temperature range (600–900 °C) of BaCoO_3−δ samples. It appears that the increase can be attributed to a change in the defect structure of the crystal surface. The accuracy of the mechanistic explanation of the above phenomena by experimental means is not sufficient. To provide a more reliable and accurate analysis of the mechanism, the effect of Li doping on the surface properties of the material was analyzed using DFT calculations. Based on XRD analysis showing hexagonal BaCoO₃-like phases for both BC and LBC at 700–900 °C, 2 × 2 × 2 hexagonal phase superlattice models for BaCoO₃ (BC) and Li_xBa_1−xCoO₃ (LBC) were established. The BC and LBC structures were optimized to obtain the most stable lattice structure. The Li in the LBC model satisfies the minimum energy distribution when x = 0.125. A series of SLAB models were created using the optimized crystal cells to characterize the surface reaction during the gas–solid reaction of the materials. Fig. 8 presents the surface energy calculations for various crystalline surfaces of BC and LBC materials. The results indicate that the (1 0 −1 0) and (2 −1 −1 0) surfaces of the BC material are the most stable, with surface energies of 0.55 J m⁻² and 0.57 J m⁻², respectively. For the LBC material, surface energy calculations were conducted for the (1 0 −1 0) and (2 −1 −1 0) surfaces with different end facets. The Li-facet generally exhibited lower surface energy compared to the Ba-facet, with values of 0.39 J m⁻² for the (1 0 −1 0)-Li surface and 0.37 J m⁻² for the (2 −1 −1 0)-Li surface. This suggests that Li doping lowers the surface energy, enhancing the stability of the material's surfaces. Considering the XRD results showing that the preferred crystallographic plane of BC and LBC is (2 −1 −1 0), it indicates that the surface structure of LBC is significantly more stable than that of BC. The presence of Li on the surface may be key to improving the material's redox activity.

Fig. 8 Surface energy calculation of different SLAB models of BC and LBC.

The thermal storage reaction process of metal oxide materials involves the transfer of oxygen atoms. Previous studies have indicated that the energy required to form oxygen vacancies is linked to the reaction temperature.⁴⁷ Therefore, the calculation of the surface oxygen vacancy formation energy was carried out for the most stable (1 0 −1 0) and (2 −1 −1 0) surfaces of BC and LBC. Fig. 9a shows the vacancy formation energies of oxygen at different locations on the BC surface. A1 and B1 represent two types of oxygen on the BC (1 0 −1 0) surface, while C1 and D1 represent types on the BC (2 −1 −1 0) surface. The oxygen vacancy formation energy for D1-O is high at 1.90 eV. In contrast, the energies for A1-O (0.73 eV), B1-O (0.57 eV), and C1-O (0.93 eV) on the BC (1 0 −1 0) surface are relatively low. This suggests that oxygen ions are most likely to escape from these sites, particularly from the B1 site on the BC (1 0 −1 0) surface. For LBC, the oxygen vacancy formation energies on the (1 0 −1 0) and (2 −1 −1 0) surfaces were calculated, as shown in Fig. 9b. The addition of Li increased the diversity of oxygen types on the LBC surface. Among them, a1 and a2 correspond to A1 sites in BC, b1 and b2 to B1 sites, and c1, c2, and c3 to C1 sites. The oxygen vacancy formation energies on the LBC (1 0 −1 0) surface range from 0.74 to 1.48 eV. The energies for a1-O (0.74 eV) and b1-O (0.89 eV), closest to Li, are lower than those for a2-O (1.07 eV) and b2-O (1.28 eV), farther from Li. The a1-O vacancy formation energy is similar to that of A1-O in BC. Notably, the oxygen vacancy formation energies on the LBC (2 −1 −1 0) surface are lower than those on the (1 0 −1 0) surface. The c3-O, closest to Li, has the lowest vacancy formation energy at 0.45 eV, significantly lower than the most active B1-O in BC (1 0 −1 0). This suggests that the LBC (2 −1 −1 0) surface has more reductively active oxygen than the BC (1 0 −1 0) surface. Doping with Li not only stabilizes the BaCoO₃ crystal surface but also reduces the vacancy formation energy of oxygen atoms adjacent to Li. This indicates that the Li_xBa_1−xCoO_3−δ solid solution can undergo oxygen-releasing reactions at lower temperatures, explaining the increased reduction activity near 770 °C.

Fig. 9 Formation energy of oxygen vacancy at different sites E(O_vac): (a) BC (1 0 −1 0) and BC (2 −1 −1 0) slabs (A1, B1, C1, D1 represent different sites); (b) LBC (1 0 −1 0) and LBC (2 −1 −1 0) slabs.

Additionally, the well-established microscopic oxygen release mechanism suggests that after releasing surface oxygen, internal lattice oxygen migrates to the surface to sustain the reaction.⁶⁴ Therefore, subsurface oxygen vacancy formation energy calculations were performed for the most stable surface of BC and the more stable LBC (2 −1 −1 0) and LBC (1 0 −1 0) surfaces, as shown in Fig. 10. Compared with BC (1 0 −1 0), LBC (2 −1 −1 0) has lower surface oxygen vacancy formation energy and higher subsurface oxygen vacancy formation energy, while LBC (1 0 −1 0) has higher surface and subsurface oxygen vacancy formation energies. This suggests that while Li doping enhances surface oxygen reactivity, it also makes it harder to release subsurface oxygen ions. Thus, the enhancement of reduction reactivity in the 700–900 °C range is mainly due to the surface activity of LBC (2 −1 −1 0), and the effect is limited by the blockage of inner oxygen ion migration, resulting in a limited overall improvement.

Fig. 10 The step diagram of formation energies of surface and subsurface oxygen vacancies for three SLAB models.

Conclusions

In this study, Li_xBa_1−xCoO_3−δ was synthesized using an optimized sol–gel method to enhance its redox activity. Performance results demonstrate that lithium doping significantly improves the redox activity of BaCoO_3−δ within the 600–900 °C range. This improvement is attributed to the generation of new oxygen release reactions, which also increase both the entropy and enthalpy changes of the energy storage process. As a result, the material's thermal storage density increases by more than 50%. Specifically, the Li_0.0625Ba_0.9375CoO_3−δ solid solution achieved a thermal storage density of 348.4 kJ kg⁻¹. Cyclic testing demonstrated superior stability, with the Li_0.15Ba_0.85CoO_3−δ material showing almost no degradation in reactivity after 450 cycles. Although sintering at high temperatures may pose a potential risk for long-term cycle performance degradation, the Li_xBa_1−xCoO_3−δ solid solution remains a highly promising thermochemical storage material.

Material characterization revealed that the enhanced reaction activity between 600 and 900 °C is due to the hexagonal Li_xBa_1−xCoO_3−δ solid solution, which exhibits an increased presence of defect oxygen species on the surface. DFT + U calculations showed that Li prefers to appear on the surface of LBC, enhancing the stability of the hexagonal BaCoO_3−δ (2 −1 −1 0) crystal plane and reducing the vacancy formation energy of surrounding oxygen ions. Thus, doping BaCoO_3−δ with Li introduces point defects that alter the lattice energy state and charge balance, promoting the formation of oxygen vacancies on the surface of hexagonal BaCoO_3−δ. This facilitates vigorous redox reactions at lower temperatures.

Data availability

The data supporting this article have been included as part of the ESI.†

Author contributions

Zeyu Ning: conceptualization, methodology, investigation, sample preparations, experiments, software, DFT calculations and analysis, writing – original draft. Changdong Gu: writing – review & editing. Yibin He: validation, writing – review & editing. Haoran Xu: validation, writing – review & editing. Peiwang Zhu: writing – review & editing. Jinsong Zhou: supervision, writing – review & editing. Gang Xiao: resources, project administration, funding acquisition.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

The authors gratefully acknowledge the support from the National Natural Science Foundation of China [No. 52176207], the National Natural Science Foundation of China [No. 52325605], and the Fundamental Research Funds for the Central Universities [2022ZFJH004].

Notes and references

1. G. Sadeghi, Energy Storage Mater., 2022, 46, 192–222.
2. J. Woods, A. Mahvi, A. Goyal, E. Kozubal, A. Odukomaiya and R. Jackson, Nat. Energy, 2021, 6, 295–302.
3. P. Sharan, C. Turchi and P. Kurup, Sol. Energy, 2019, 185, 494–507.
4. C. Amy, H. R. Seyf, M. A. Steiner, D. J. Friedman and A. Henry, Energy Environ. Sci., 2019, 12, 1426.
5. Z. Y. Ling, Z. G. Zhang, G. Q. Shi, X. M. Fang, L. Wang, X. N. Gao, Y. T. Fang, T. Xu, S. F. Wang and X. H. Liu, Renewable Sustainable Energy Rev., 2014, 31, 427–438.
6. T. Y. Yang, J. G. Kang, P. B. Weisensee, B. Kwon, P. V. Braun, N. Miljkovic and W. P. King, Appl. Phys. Lett., 2020, 116, 071901.
7. A. Mavrigiannaki and E. Ampatzi, Renewable Sustainable Energy Rev., 2016, 60, 852–866.
8. G. Li, Renewable Sustainable Energy Rev., 2016, 53, 897–923.
9. L. Seyitini, B. Belgasim and C. C. Enweremadu, J. Energy Storage, 2023, 62, 106919.
10. D. Raut, A. K. Tiwari and V. R. Kalamkar, J. Braz. Soc. Mech. Sci. Eng., 2022, 44, 446.
11. M. Opolot, C. R. Zhao, M. Liu, S. Mancin, F. Bruno and K. Hooman, Renewable Sustainable Energy Rev., 2022, 160, 112293.
12. Y. Zhao, C. Y. Zhao, C. N. Markides, H. Wang and W. Li, Appl. Energy, 2020, 280, 115950.
13. J. S. Prasad, P. Muthukumar, F. Desai, D. N. Basu and M. M. Rahman, Appl. Energy, 2019, 254, 113733.
14. Z. H. Pan and C. Y. Zhao, Sci. Bull., 2017, 62, 256–265.
15. R. Dunn, K. Lovegrove and G. Burgess, Proc. IEEE, 2012, 100, 391–400.
16. M. Tawalbeh, S. Z. M. Murtaza, A. Al-Othman, A. H. Alami, K. Singh and A. G. Olabi, Renewable Energy, 2022, 194, 955–977.
17. S. K. Dewangan, M. Mohan, V. Kumar, A. Sharma and B. Ahn, Int. J. Energy Res., 2022, 46, 16150–16177.
18. K. Malleswararao, P. Dutta and M. S. Srinivasa, Appl. Therm. Eng., 2022, 215, 118816.
19. K. T. Moller, T. R. Jensen, E. Akiba and H. W. Li, Prog. Nat. Sci.: Mater. Int., 2017, 27, 34–40.
20. S. Guo, X. Tian, Y. Xu, S. Ju, C. Y. Zhao and J. Yan, Chem. Eng. J., 2023, 468, 143691.
21. D. Choi, S. Noh and Y. Park, Chem. Eng. J., 2023, 470, 144036.
22. H. B. Dizaji and H. Hosseini, Renewable Sustainable Energy Rev., 2018, 98, 9–26.
23. S. K. Wu, C. Zhou, E. Doroodchi, R. Nellore and B. Moghtaderi, Energy Convers. Manage., 2018, 168, 421–453.
24. L. Andre, S. Abanades and L. Cassayre, Appl. Sci., 2018, 8, 2618.
25. A. J. Carrillo, D. Sastre, D. P. Serrano, P. Pizarro and J. M. Coronado, Phys. Chem. Chem. Phys., 2016, 18, 8039–8048.
26. U. Pelay, L. A. Lu, Y. L. Fan, D. Stitou and M. Rood, Renewable Sustainable Energy Rev., 2017, 79, 82–100.
27. X. Y. Yang, S. J. Li, J. G. Zhao, H. Y. Huang and L. S. Deng, J. Energy Chem., 2022, 73, 301–310.
28. Z. H. Pan and C. Y. Zhao, Energy, 2015, 82, 611–618.
29. M. Schmidt, A. Gutierrez and M. Linder, Appl. Energy, 2017, 188, 672–681.
30. A. W. Sun, W. Shuai, N. Zheng, Y. Han, G. Xiao, M. Ni and H. R. Xu, Energy Convers. Manage., 2022, 271, 116310.
31. P. Yuan, C. D. Gu, H. R. Xu, Z. Y. Ning, K. F. Cen and G. Xiao, Iscience, 2021, 24, 103039.
32. C. Pagkoura, G. Karagiannakis, A. Zygogianni, S. Lorentzou and A. G. Konstandopoulos, Energy Procedia, 2015, 69, 978–987.
33. E. Alonso, C. Pérez-Rábago, J. Licurgo, E. Fuentealba and C. A. Estrada, Sol. Energy, 2015, 115, 297–305.
34. Z. K. Zhang, L. Andre and S. Abanades, Sol. Energy, 2016, 134, 494–502.
35. J. Vieten, B. Bulfin, P. Huck, M. Horton, D. Guban, L. Y. Zhu, Y. J. Lu, K. A. Persson, M. Roeb and C. Sattler, Energy Environ. Sci., 2019, 12, 1369–1384.
36. S. M. Babiniec, E. N. Coker, A. Ambrosini and J. E. Miller, AIP Conf. Proc., 2016, 1734, 050006.
37. C. W. Sun, J. A. Alonso and J. J. Bian, Adv. Energy Mater., 2021, 11, 2000459.
38. E. J. Popczun, D. Tafen, S. Natesakhawat, C. M. Marin, T. D. Nguyen-Phan, Y. Y. Zhou, D. Alfonso and J. W. Lekse, J. Mater. Chem. A, 2020, 8, 2602–2612.
39. M. A. Pena and J. L. G. Fierro, Chem. Rev., 2001, 101, 1981–2017.
40. R. X. Cai, H. Bektas, X. J. Wang, K. McClintock, L. Teague, K. R. Yang and F. X. Li, Adv. Energy Mater., 2023, 13, 2203833.
41. S. M. Babiniec, E. N. Coker, J. E. Miller and A. Ambrosini, Int. J. Energy Res., 2016, 40, 280–284.
42. B. Bulfin, J. Vieten, D. E. Starr, A. Azarpira, C. Zachaus, M. Havecker, K. Skorupska, M. Schmucker, M. Roeb and C. Sattler, J. Mater. Chem. A, 2017, 5, 7912–7919.
43. L. Imponenti, K. J. Albrecht, R. J. Braun and G. S. Jackson, ECS Trans., 2016, 72, 11–22.
44. E. Mastronardo, X. Qian, J. M. Coronado and S. M. Haile, J. Mater. Chem. A, 2020, 8, 8503–8517.
45. F. Jin, C. Xu, H. Y. Yu, X. Xia, F. Ye, X. Li, X. Z. Du and Y. P. Yang, ACS Appl. Mater. Interfaces, 2021, 13, 3856–3866.
46. B. Lucio, M. Romero and J. Gonzalez-Aguilar, Solid State Ionics, 2019, 338, 47–57.
47. P. Yuan, H. R. Xu, Z. Y. Ning and G. Xiao, J. Energy Storage, 2023, 61, 106695.
48. J. A. Peña, E. Lorente, E. Romero and J. Herguido, Catal. Today, 2006, 116, 439–444.
49. H. M. Wang, G. C. Liu, A. Veksha, X. M. Dou, A. Giannis, T. T. Lim and G. Lisak, J. Cleaner Prod., 2021, 282, 124467.
50. Z. Y. Zheng, Y. L. Li, Q. Guo, L. Zhang and T. Qi, Int. J. Hydrogen Energy, 2023, 48, 18177–18186.
51. Q. W. Shen, Y. Zheng, S. A. Li, H. R. Ding, Y. Q. Xu, C. G. Zheng and M. Thern, J. Alloys Compd., 2016, 658, 125–131.
52. A. Jain, S. P. Ong, G. Hautier, W. Chen, W. D. Richards, S. Dacek, S. Cholia, D. Gunter, D. Skinner, G. Ceder and K. A. Persson, APL Mater., 2013, 1, 011002.
53. A. Paolella, W. Zhu, G. Bertoni, S. Savoie, Z. Feng, H. Demers, V. Gariepy, G. Girard, E. Rivard, N. Delaporte, A. Guerfi, H. Lorrmann, C. George and K. Zaghib, ACS Appl. Energy Mater., 2020, 3, 3415–3424.
54. A. J. Carrillo, D. P. Serrano, P. Pizarro and J. M. Coronado, J. Phys. Chem. C, 2016, 120, 27800–27812.
55. M. A. Pena and J. L. G. Fierro, Chem. Rev., 2001, 101, 1981–2017.
56. N. Pegios, V. Bliznuk, S. A. Theofanidis, V. V. Galvita, G. B. Marin, R. Palkovits and K. Simeonov, Appl. Surf. Sci., 2018, 452, 239–247.
57. G. Ertl, Angew. Chem., Int. Ed., 2008, 47, 3524–3535.
58. D. H. Zhou, F. H. Le, W. Jia and X. H. Chen, Inorg. Chem., 2023, 62, 8001–8009.
59. N. S. Mcintyre and M. G. Cook, Anal. Chem., 1975, 47, 2208–2213.
60. R. P. Vasquez, M. Rupp, A. Gupta and C. C. Tsuei, Phys. Rev. B: Condens. Matter Mater. Phys., 1995, 51, 15657–15660.
61. C. H. Jia, X. P. Xiang, J. Zhang, Z. Y. He, Z. H. Gong, H. J. Chen, N. Zhang, X. W. Wang, S. J. Zhao and Y. Chen, Adv. Funct. Mater., 2023, 33, 2301981.
62. D. Kim, I. Jeong, S. Ahn, S. Oh, H. Im, H. Bae, S. J. Song, C. W. Lee, W. C. Jung and K. T. Lee, Adv. Energy Mater., 2024, 14, 2304059.
63. D. Xiang, C. D. Gu, H. R. Xu and G. Xiao, Small, 2021, 17, 2101524.
64. G. S. Jackson, L. Imponenti, K. J. Albrecht, D. C. Miller and R. J. Braun, J. Sol. Energy Eng., 2019, 141, 021016.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4ta05176a

---