La_0.7Sr_0.3Mn_0.9X_0.1O₃ perovskite oxides were synthesized using the modified Pechini method (Pechini, 1967). First, stoichiometric molar ratio of each perovskite material was dissolved in distilled water with metal nitrate. The reagents used in this study were La(NO₃)₃·6H₂O (Wako, 99.9% purity), Sr(NO₃)₂ (Wako, ≥98.0% purity), Cr(NO₃)₃·9H₂O (Strem Chemicals, 99.9% purity), Mn(NO₃)₃·6H₂O(Wako, ≥98.0% purity), Mg(NO₃)₂·6H₂O (Wako, ≥98.0% purity), and Cu(NO₃)₂·3H₂O (Wako, 99.9% purity), Fe(NO₃)₃·9H₂O (Kanto Chemical, ≥99% purity), Co(NO₃)₂·6H₂O (Kanto Chemical, ≥98% purity), Ni(NO₃)₃·6H₂O (Wako, 99.9% purity), and Ga(NO₃)₃·8H₂O (Kishida Chemical, ≥99% purity). Citric acid (Wako, ≥98.0% purity) and ethylene glycol (Wako, ≥99.5% purity) were added and dissolved in an aqueous solution. The molar ratio of the total amount of metallic cations, citric acid, and ethylene glycol was 1:5:5. These solutions were stirred and heated at 80°C for 1-h in an oil bath to evaporate water and then heated at 170°C for gel formation. The obtained gel was dried at 300°C for 5-h in an electric oven. The resulting precursor was pulverized using a mortar and pestle, and calcined at 1,200°C for 8-h in a muffle furnace. The following powder-like materials were synthesized: La_0.7Sr_0.3Mn_0.9X_0.1O₃ (X = Mg, Al, Cr, Fe, Co, Ni, Cu, and Ga).

Powder X-ray diffraction (XRD) was performed using a Bruker D2 PHASER with Cu Kα radiation (λ = 0.15418 nm, 30 kV–10 mA) at room temperature for crystallographic phase identification of the synthesized materials and materials obtained after thermochemical testing. Diffraction data were collected with an angular range of 2θ = 20°–80°, step interval of 0.02°, and recording time of 1 s. The crystalline phases were identified by comparison with standard reference databases [Inorganic Crystal Structure Database (ICSD) and Crystallography Open Database (COD)]. Rietveld refinement was performed using the whole pattern fitting (WPF) method of the FullProf package to evaluate the space group, crystal system, and lattice parameters of the synthesized materials.

Scanning electron microscopy (SEM, JCM-6000, JEOL) with an acceleration voltage of 15 kV was used to observe the powder morphology and size before and after the redox activity test of the thermochemical cycle.

X-ray photoelectron spectroscopy (XPS) analysis was performed using a Quantum 2000 (ULVAC-PHI. Inc.) with an aluminum anode (Al Kα, 25 W (15 kV, 1.67 mA), hv = 1,486.6 eV) to evaluate the valence states of the elements. For energy calibration of the XPS spectrum, the authors used the adventitious carbon C1s peak at 284.8 a of binding energy. The Gaussian function for peak fitting of the XPS spectrum, Shirley method (Shirley, 1972) for the baseline of the XPS spectrum, and the Savitzky–Golay method (Press and Teukolsky, 1990) for data smoothing of the XPS spectrum were used for analysis of the XPS spectrum.

The redox reactivity of La_0.7Sr_0.3Mn_0.9X_0.1O₃ perovskite oxides was evaluated through oxygen/carbon monoxide (CO) productivity, kinetics, and TR/CS temperature impacts for two-step thermochemical CO₂ splitting cycling. The redox activity performance was examined using a thermogravimetric reactor (NETZSCH STA2500 Regulus, weight resolution of 0.03 μg, temperature resolution of 0.3 K) equipped with a type S thermocouple (temperature resolution of ±0.0025 × |t| °C) for endothermic thermal reduction (TR) and subsequent exothermic carbon dioxide splitting (CS) steps of the sample powder.

In the thermogravimetric reactor used in our laboratory, O₂ gas evolved during the TR steps is quickly removed upward by gas flow from below the balance. However, the upward stream of gas flow is a disadvantage with regard to contacting CO₂ with the sample during the CS step. In order to estimate the impacts on the rates and amounts of O₂ release and CO production affected by gas diffusion limitations owing the sample mounting and gas flow orientation in the reactor, before starting the present study, the authors examined redox reactivity using some samples with different amounts (heights) and different sizes of platinum pan in order to determine the experimental procedure. As the results, the authors found that the present test conditions can realize reversible redox reaction under the gas flow orientation without governing diffusion limitation of CO₂ supply in the preliminary test. In the preliminary test, a blank test without any sample in a Pt pan was performed in order to check general drift of balance in the thermogravimetric reactor under the experimental procedure of redox reactivity test. The weight change of 0.0274% was observed in the test. Thus, an influence of general drift owing to equipment was relatively small compared with weight change derived from redox reaction.

The reactivity test was performed using the following procedure: Approximately 50 mg of sample powder was packed into a platinum pan (inner diameter of 5.2 mm, height of 2.6 mm) and mounted on the balance in a tubular ceramic chamber of the reactor. To eliminate adsorptive species on the surface of the sample powder, such as water, CO₂, and O₂, the sample powder was first heated at 300°C for 0.5-h at a rate of 50 K/min using an electric furnace while passing 99.999% high-purity nitrogen gas through the reactor at a flow rate of 100 cm³/min at standard state (unit of sccm). Subsequently, the sample powder was subjected to the TR step at 1,200°C–1,400°C for 1.5-h to thermally release oxygen from the sample power. The temperature of the sample powder was monitored using an S-type thermocouple in contact with the bottom of the platinum pan. The powder was cooled at 800°C–1,200°C at a rate of 20 K/min under a N₂ flow, as indicated above. The passing gas was switched into a gas mixture with a N₂ flow rate of 50 sccm and a CO₂ flow rate of 50 sccm to perform the CS step. The sample was maintained at 800°C–1,200°C for 0.5-h in the electric furnace of the reactor to complete the CS step. The TR and CS steps for the reactivity tests were repeated four times. The temperature impacts were evaluated by testing five times in sequence with the temperatures of the TR and CS steps. The sample powder weight changed in the reactor furnace by O₂ release during the TR step and O₂ uptake from CO₂ during the CS step. The weight change of the sample was recorded as a function of time to monitor the fractional extent of the reaction in the TR and CS steps.

The repeatability of the redox reaction for the sample power was evaluated using a thermogravimetric reactor using the following procedure. The TR step was performed at 1,400°C at a N₂ flow rate of 100 sccm for 10 min, and subsequently, the CS step was conducted at 1,200°C at a gas mixture flow rate of 100 sccm (CO₂ 50% and N₂ 50%) for 15 min. The TR and CS steps were repeated 50 times in sequence to examine the repeatability of O₂ release and CO production.

O₂ and CO productivity per sample weight [n_O2 and n_CO (mol/g)] were calculated using the molar weight of the oxygen atom [M_O (g/mol)], weight change of the sample [Δm (g)], and sample weight at time t = 0 [m_s (g)]. The weight change of the sample (Δm) during the TR and CS steps was calculated from the weight decrease and increase, respectively. The O₂ and CO production rates per sample weight [r_O2 and r_CO (mol/g)] were calculated from the time derivative of weight m (g) at time t = t (dm/dt), as represented in Eqs 7, 8. n O 2 = Δ m 2   ×   M O   ×     m s (5) n CO = Δ m M O   ×     m s (6) r O 2 = − dm /   dt 2   ×   M O   ×     m s (7)   r CO = dm   /   dt Mo   ×     m s (8)

Figure 1 illustrates the XRD patterns of La_0.7Sr_0.3Mn_0.9X_0.1O₃ [X = Mn (non-substituted LSM), Co-substituted LSM, and Ce-substituted LSM] at room temperature in the 2θ range. The XRD data for the other samples are shown in Supplementary Figure S1. The lattice parameters and space groups of the synthesized La_0.7Sr_0.3Mn_0.9X_0.1O₃ were determined by Rietveld refinement. All diffraction peaks assigned to the crystalline phase were indexed in a trigonal unit cell [space group R- 3C (167)], and the lattice parameters are listed in Table 2 (Petrov et al., 1999; Troyanchuk et al., 2015; Mostafa et al., 2008; El-Fadli et al., 2002; Yanchevskii et al., 2006; Reshmi et al., 2013; Paiva-Santos et al., 2002; Wu et al., 2014). As seen in the XRD pattern of the non-substituted LSM (Figure 1A), a series of peaks (blue vertical bars) correspond to the crystalline phase (red vertical bars below the XRD pattern), as cited in (Petrov et al., 1999). In addition, no secondary phases were observed in the pattern. As seen in Figure 1B, a series of peaks for Co-substituted LSM are identical to the crystalline phase with the same index as the Co-substituted reference (red vertical bars below the XRD pattern). The lattice parameter of the Co-substituted LSM is smaller than that of the non-substituted LSM (Table 2), because ionic radius of Co cation (Co³⁺: 0.61 Å, 6 coordination by O²⁻ ion) is smaller than Mn cation (Mn³⁺: 0.645 Å, at high spin and 6 coordination by O²⁻ ion) (Shannon and Prewitt, 1969). The XRD peaks for the Co-substituted LSM were shifted toward greater diffraction angles than those of the non-substituted LSM. The results indicate that the Co-substituted LSM was successfully synthesized as a solid solution of the perovskite structure without impurities. However, in the XRD pattern of the Ce-substituted LSM (Figure 1C), La_0.64Sr_0.36MnO₃ [a trigonal unit cell, space group R- 3C (167)] (Paiva-Santos et al., 2002), and La₂Ce₂O₇ [cubic unit cell, space group Fm-3m (225)] (Wu et al., 2014) were observed. The result indicates that Ce cation is not incorporated into a perovskite structure of non-substituted LSM owing to relatively large ionic radius of Ce cation (Ce³⁺: 1.034 Å, Ce⁴⁺: 0.87 Å in 6 co-ordination by O²⁻ ion) compared with Mn cation (Mn³⁺: 0.645 Å at high spin and 0.58 Å at low spin in 6 co-ordination by O²⁻ ion) (Shannon and Prewitt, 1969). Except for the Ce-substituted LSM, all sample powders were successfully synthesized as a single phase in a trigonal unit cell [space group R- 3C (167)]. The XRD patterns of as-synthesized sample and sample obtained after the redox activity test are shown in lower and upper side of Supplementary Figure S1. The original crystalline phase was maintained for the sample obtained after the redox activity test, except of Cu-substituted sample. The XRD peak at the 2θ = 28° observed for the sample obtained after the redox activity test were due to glass holder made by Si in the XRD measurement (COD data No. 9013105) (Dutta, 1962).

Figure 2 displays the TGA profiles of the weight change in the activity tests of two-step thermochemical CO₂ splitting using La_0.7Sr_0.3Mn_0.9X_0.1O₃ sample powder. The first run of all TGA profiles was excluded from analysis and discussion of the test results because the result in the first run might include non-reproducible effects of sample and TGA reactor preparations (moisture and air adsorbed on the surface of sample powder, and contamination in the TGA reactor). As shown in Figure 2A, the non-substituted LSM gradually decreased with time during the TR step and rapidly increased during the subsequent CS step. The variation in weight change was reproducibly repeated in the second to fourth runs. For the Cr-substituted LSM, we observed that the weight gain gradually increased during the CS step of the second to fourth runs when the run number increased. This result indicates that the Cr substitution into LSM perovskite contributes to the reaction rate of the oxidation reaction during the CS step. In the case of the Ce-substituted LSM, the behavior of the weight loss and gain during the second to fourth runs was repeatedly observed with similar levels of the non-substituted LSM. However, the Cu-substituted LSM showed that the weight loss during the TR step was greater than the weight gain during the CS step in first to second runs. The results indicate that Cu-substituted LSM partially decomposed into solid phases, providing irreversible redox properties in cyclic activity tests of two-step thermochemical CO₂ splitting. The XRD patterns of the synthesized Cu-substituted LSM and the material obtained after the activity test are shown in Supplementary Figure S2. As seen in Supplementary Figure S2A, the as-prepared sample is a single solid phase with a trigonal unit cell (space group R- 3C (167)), as listed in Table 2. After the activity test (Supplementary Figure S2B), the material was composed of a solid mixture of La_0.67Sr_0.33MnO₃ [a trigonal unit cell, space group R- 3C (167), COD data No. 1533312] (Zhang et al., 2003) and La_1.5Sr_0.5Mn_0.5Cu_0.5O₄ (a tetragonal unit cell, space group I4/mmm (139), COD data No. 4002411) (McCabe and Greaves, 2006). In previous studies, reversible or irreversible phase transition were observed when perovskites were heated at high temperatures (Vogt and Schmahl, 1993; Chen et al., 2014). Ideally, the molar ratio of oxygen atom in RP1 structure [A₂BO₄, oxygen/(metal cation) = 4/3] is less than that of the perovskite structure [ABO₃, oxygen/(metal cation) = 3/2]. The large weight loss of Cu-substituted LSM during the TR step may include the irreversible phase transition of the perovskite structure into the PR1 and thermal reduction of perovskite structure. Furthermore, there is a possibility that a vaporization of Cu-substituted sample occurs at high temperature. The incomplete redox reaction of Cu-substituted LSM may be that the latter solid phase (La_1.5Sr_0.5Mn_0.5Cu_0.5O₄) was not oxidized with CO₂ under the present test conditions. The TGA profiles of the Al-, Fe-, and Ga-substituted LSMs are shown in Figure 2B, and as a reference, the results of the non-substituted LSM are also plotted in the figure. The loss and gain in weight were reproducibly observed during the second to fourth runs of the two-step thermochemical CO₂ splitting activity tests. The extent of weight loss during the TR step for all samples was approximately the same as that of the non-substituted LSM, while that of the weight gain was greater than that of the non-substituted LSM and Cr- and Ce-substituted LSMs. Figure 2C illustrates the TGA profiles of the Mg-, Co-, and Ni-substituted LSMs. The weight change for all samples during the second to fourth runs was significantly greater than that for the non-substituted LSM. In addition, the loss and gain in weight were reproducibly observed. These results indicate that La_0.7Sr_0.3Mn_0.9X_0.1O₃ samples (except for Cu-substituted LSM) have high thermal stability and reactivity without degradation and sufficient repeatability for thermochemical CO₂ splitting under the present reaction conditions.

Figure 3A shows the O₂ and CO productivities of all samples during the first to fourth runs of the two-step thermochemical CO₂ splitting activity tests. Each value of productivity was calculated from the extent of the weight change in Figure 2. The non-substituted LSM demonstrated stable and reproducible O₂ and CO productivities in the second to fourth runs. The amounts of O₂ and CO produced were approximately 140 and 250 μmol/g-material, respectively. The authors categorized them into three groups in order of productivity: group A = Cu, Ce, and Cr (low productivity); group B =Fe, Al, and Ga (middle productivity); group C = Co, Ni, and Mg (high productivity levels). In group A, the O₂ and CO productivities were equivalent for Ce-substituted LSM or lower for Cr-substituted LSM than that of non-substituted LSM during the second to fourth runs. For group B, the O₂ productivity of the substituted LSM was similar to that of the non-substituted LSM, while the CO productivity was approximately the same or slightly higher than that of the non-substituted LSM during the second to fourth runs. In the case of group C, both productivities were higher than that of the non-substituted LSM during the second to fourth runs.

The average values of O₂ and CO production and CO/O₂ ratio in the second to fourth runs are shown in Figure 3B. In group A, O₂ production decreased in the order of Cu-, Ce-, and Cr-substituted LSMs, while CO production and CO/O₂ ratio increased. The average values of O₂ and CO production in the second to fourth runs for the Cr-substituted LSM were 106 and 248 μmol/g-material, respectively. The CO/O₂ (CO/O₂ = 2.33) value exceeded the stoichiometric ratio (CO/O₂ = 2). The results indicate that a large extent of thermal reduction obtained at the TR step of first run was oxidized through the CS step of the second to fourth runs. The CO/O₂ ratios of Cu-, and Ce substituted LSMs were lower than the stoichiometric ratio of 2. The results indicate that Cr substitution may improve CO production when the stoichiometric ratio increases as the number of runs increase. For group B, the average values of O₂ and CO production in the second to fourth runs were 145–150 (143 for non-substituted LSM) and 258–295 (248 for non-substituted LSM) μmol/g-material, respectively. The CO production was in order of Fe < Al < Ga, and CO/O₂ ratio was 1.79 (Fe-substitution), 1.97 (Al-substitution), and 1.96 (Ga-substitution). The results indicate that Al and Ga substitutions at the B-site of LSM enhanced CO production in the CS step. Shannon’s literature reported that Al and Ga ions normally have a common valence number of 3+ in 6 co-ordination by O²⁻ ions (Shannon and Prewitt, 1969). The following assumption is considered from the above results: when the valence number of substitution Al and Ga ions does not vary, thereby retaining the valency in the LSM perovskite structure, the substitution ion may promote thermal reduction of Mn ions in the LSM perovskite structure, while it plays a role in stabilizing the crystal structure with oxygen vacancies formed by the TR step. Then, in the subsequent CS step, the low-to-high valency transition of Mn ions in the LSM perovskite structure can be enhanced during the CS step.

In the case of group C, the O₂ production level of 188–197 μmol/g-material was the highest among all the samples tested in this study. The average CO production was 347 μmol/g-material for Co substitution, 351 μmol/g-material for Ni substitution, and 354 μmol/g-material for Mg substitution. Among all the samples, these three substituted LSMs exhibited similar O₂ and CO production levels. The CO/O₂ ratio was 1.85 (Co-substitution), 1.87 (Ni-substitution), and 1.79 (Mg-substitution). The impact of the test temperature on the O₂ and CO productivities is evaluated in the subsequent section. The results of group C indicate that Co, Ni, and Mg substitutions at the B-site of LSM enhanced both O₂ production during the TR step and CO production during the subsequent CS step. These ion species have a common valence number of 2+ in 6 coordinated by O²⁻ ions. In contrast, Co and Ni ions can also have valence numbers of 3+ and 4+ in 6 co-ordination by O²⁻ ions (Shannon and Prewitt, 1969). The following supposition is made for the Mg-substituted LSM: the lattice volume of the Mg-substituted LSM is smaller than that of the non-substituted LSM (Table 2). This result means that an ion species smaller than the Mn ion is substituted into the B-site of the LSM perovskite. However, the ionic radius (0.720 Å) of Mg²⁺ in 6 coordination by O²⁻ ion is the greatest among those of Mn ions with various valence numbers in six coordination. Thus, from the viewpoint that substitutional solid solution is formed in the Mg-substituted LSM perovskite oxide, the Mg ion has a small coordination number (<6), leading to a small ionic radius at retaining the Mg²⁺ state, or high valence numbers of Mn ions are induced owing to charge balance by Mg²⁺ substitution at the B-site. The variation of O²⁻ coordination or high valency Mn ions may enhance the mobility of O²⁻ ions in the crystal structure at high temperatures, leading to improved O₂ production during the TR step. As a result, the reduced phase containing oxygen vacancies in the Mg substitution of LSM perovskite contributes to the enhancement of CO production compared to the non-substituted LSM. Further investigations about kinetics of mobility of O^2- ion during the TR step are required to elucidate the detailed mechanism.

Figure 4A illustrates the O₂ production profiles of samples belonging to group A during the TR step in the first to fourth runs. When the temperature of the sample reached 1,400°C, the production rate of O₂ reached a maximum in all runs. The peak rates of O₂ production for non-substituted LSM were 6.5–6.7 µmol/(min·g-material) in the second to fourth runs. For the Cu-substituted LSM, the peak rate of O₂ production in the first run increased because of the irreversible decomposition of the solid phase. The peak rates of O₂ production in the second to fourth runs gradually decreased in comparison with the non-substituted LSM. The peak rates of the Cr- [5.1–6.5 µmol/(min·g-material)] and Ce-substituted LSMs [6.5–6.8 µmol/(min·g-material)] in the second to fourth runs were similar to the non-substituted LSM in each run. Thus, the substitution ions of group A did not affect the production rate of O₂ during the TR step of the two-step thermochemical CO₂ splitting cycle. Figure 4A illustrates the O₂ production profiles of samples belonging to group B during the TR step in the first-fourth runs. The peak rates of Fe-(7.2–7.4 µmol/(min·g-material)), Al-(7.3–8.7 µmol/(min·g-material)), Ga-substituted LSMs (8.1–8.2 µmol/(min·g-material)) were higher than those of non-substituted LSM in second to fourth runs. The decay trend of O₂ production after the peak rate for the samples in group B was similar to that of the non-substituted LSM. The results indicated that the substitution ions of group B enhanced the peak rate of O₂ production during the TR step. The O₂ production profiles of samples belonging to group C are shown in Figure 4C. The peak rates of Co-(9.3–10.2 µmol/(min·g-material)), Ni-(9.7–10.3 µmol/(min·g-material)), Mg-substituted LSMs (9.6–10.1 µmol/(min·g-material)) were highest among the samples tested in this study. The peak rates of group C were 1.4–1.6 times greater than those of non-substituted LSM. There was little difference in the decay trend of O₂ production in group C. The results indicated that the substitution ions in group C contributed to the enhancement of the production rate.

Figure 5A shows the CO production profiles of samples belonging to group A during the subsequent CS step in the first to fourth runs. The CO production for all samples tested in this study was immediately initiated at the target temperature when the gas stream was switched to a gas mixture of N₂ and CO₂. The peak rates of CO production for the non-substituted LSM were 69.8–70.8 µmol/(min·g-material) in the second to fourth runs. The time required for CO production was approximately 20 min. The order of peak rates of the CO production is Cu-(61.4–78.0 µmol/(min·g-material)), Cr-(71.7–75.9 µmol/(min·g-material)), and Ce-substituted LSMs (80.8–82.3 µmol/(min·g-material)) in second to fourth runs. The peak rates of the Cr-and Ce-substituted LSMs were higher than those of the non-substituted LSM in the second to fourth runs. The results of the Cr substitution agreed with the previous results that the Cr substitution in LSM enhanced the oxidation rate of the reduced material during the WS step of two-step thermochemical water-splitting cycling (Gokon et al., 2019). Compared with the non-substituted LSM, the production rates of Cu- and Ce-substituted LSM during the CS step rapidly decayed after reaching the peak rate. This leads to a decrease in the amount of CO production compared with the non-substituted LSM (Figure 3B). In the case of group B (Figure 5B), the order of peak rates of the CO production was Fe-[64.2–65.0 µmol/(min·g-material)] < Ga-[66.8–68.1 µmol/(min·g-material)] < Al-substituted LSMs [74.6–77.5 µmol/(min·g-material)] in second to fourth runs. The time required for CO production in the Al- and Ga-substituted LSMs was extended to approximately 25 min. These results correspond to the results that the Al- and Ga-substituted LSMs enhanced CO production during the CS step (Figure 3B). For the group C (Figure 5C), the peak rates of the CO production were Co-[65.6–66.5 µmol/(min·g-material)], Mg-[74.2–75.6 µmol/(min·g-material)], and Ni-substituted LSMs [72.9–76.1 µmol/(min·g-material)] in the second to fourth runs (Co < Mg ≅ Ni). The period of time for CO production in group C was extended to approximately 30 min. These results of increased peak rate and prolonged CO production lead to an increase in the amount of CO production in group C compared to the non-substituted LSM.

In this section, all samples of substituted and non-substituted LSMs were compared in terms of redox activity and reaction kinetics under the same test conditions. In particular, Mg-, Co-, and Ni-substituted LSMs in group C demonstrated higher CO productivity and superior reaction kinetics than the non-substituted LSM.

In this section, in order to examine the effect of temperature on the redox activity of substituted LSMs in group C, the samples were subjected to the TR step at a temperature of 1,200°C for 90 min. Subsequently, the CS step was performed at temperatures of 800°C–1,200°C for 30 min (Figure 6A). As shown in Figure 6B, the TR step was conducted at temperatures of 1,200°C–1,400°C for 90 min, and subsequently the CS step was performed at a temperature of 1,200°C for 30 min. The TGA profiles of Mg-, Co-, Ni-substituted LSMs were displayed in Supplementary Figure S3. As seen in Figure 6A, Average O₂ productivity at TR temperature of 1,400°C were approximately similar at 183.7 µmol/(min·g-material) (standard deviation (SD) = 0.26) for Mg-substituted LSM, and 183.4 µmol/(min·g-material) (SD = 5.19) for Co-substituted LSM, while 194.5 µmol/(min·g-material) (SD = 2.79) for Ni-substituted LSM was highest among the group C. Even if the CS temperature decreased from 1,200°C to 800°C, the average CO productivity at CS temperature of 800°C–1,200°C were 344.4 µmol/(min·g-material) (SD = 5.63) for Mg-substituted LSM, 347.1 µmol/(min·g-material) (SD = 15.84) for Co-substituted LSM, and 349.7 µmol/(min·g-material) (SD = 5.61) for Ni-substituted LSM. The temperatures in the exothermic CS step did not influence CO productivity in the present test conditions. However, the peak rate of CO production decreased as the CS temperature decreased. Thus, the results indicate that the CS temperature contributes to the oxidation kinetics of the reduced sample during the CS step. The average CO/O₂ ratios and test conditions are 1.79–2.10, 1.84–2.24, 1.67–2.00 for the Mg-, Co-, and Ni-substituted LSMs, respectively. On average, the stoichiometric CO/O₂ ratio was approximately attained for the samples in group C, meaning that the two-step reaction proceeds stoichiometrically under test conditions of different CS temperatures in Figure 6A. The impact of the TR temperature on the redox activity is shown in Figure 6B. As the TR temperature decreased from 1,400°C to 1,200°C, the productivities and peak rates of O₂ and CO decreased. The average CO/O₂ ratios were nearly equal to 2.0 for all samples in group C. These results indicate that the samples maintained stoichiometric two-step reactions under different TR temperatures. Thus, the temperatures in the endothermic TR step strongly affect the redox activity and reaction kinetics in terms of thermodynamics and kinetics for both steps of thermochemical cycling.

The effects of temperature on the redox activity of substituted LSMs in group C were studied in this section; whereby we elucidated that, the TR temperature of 1,400°C and subsequent CS temperatures of 800°C–1,200°C provided large O₂ and CO productivity and reaction rates without degradation of the sample, while the stoichiometric redox activity and reproducible reaction rate were maintained during the repeated test runs. In the following section, the thermal durability and stability of the samples in group C during thermochemical two-step cycling under severe test conditions are examined and evaluated.‬

To prevent the decrease in the reactivity of the redox material owing to sintering and coagulation at high temperatures, the presence of thermal durability and stability of redox materials for thermochemical two-step CO₂ splitting is necessary (Jiang et al., 2014b; Wang et al., 2020; Liu et al., 2021). In the previous section, the highest reactivity and CO production rate were observed in the redox cycles of Mg-, Co-, and Ni-substituted LSMs at TR and CS temperatures of 1,400°C and 1,200°C, respectively, while such high-temperature conditions may cause coagulation and sintering of the sample. Therefore, long-term continuous multiple cycling of the thermochemical two-step reaction was performed to evaluate the thermal durability of the samples. The TR step was performed at 1,400°C for 10 min, and the reduced sample was subjected to a subsequent CS step at 1,200°C for 15 min. The as-synthesized sample before testing and the sample obtained after cycling were observed by SEM to evaluate the thermal stability of the sample at high temperature.

Figure 7A illustrates the thermogravimetric measurements of the continuous TR-CS cycling of thermochemical CO₂ splitting using the Mg-substituted LSM for long-term thermal durability. The profiles of O₂ and CO production were reproducibly observed at the same level without degradation during continuous 50 runs. Similarly, the Co- and Ni-substituted LSMs exhibited stable and reproducible profiles of O₂ and CO production over long-term continuous multiple cycling (Figures 7B,C). The O₂ and CO productivity at each run are plotted against the run number in Figure 7D. The average values of O₂ and CO productivities were 103.7 (SD = 2.19), and 208.3 (SD = 3.76), respectively. The decreases in O₂ and CO productivity were 4.3% and 1.0% after the 50th run when the standard point for comparison was the second run because the result in the first run might include non-reproducible effects of sample and TGA reactor preparations (moisture and air adsorbed on the surface of sample powder, and contamination in the TGA reactor). The slight decreases in the productivities indicate the two following possibilities: 1) limited spillage of the power-shaped sample together with a releasing/accompanying gas flow and 2) low vaporization of the sample during the repeated runs. However, significant effusion and vaporization of the sample were absent during the continuous runs. The results indicated that the Mg-substituted LSM was a durable and reproducible redox material during long-term thermochemical cycling at high temperatures. As seen in Figures 7E,F, stable and reproducible O₂ and CO productivity over the long-term continuous multiple cycling were observed with the remaining stoichiometric ratio of CO/O₂ ≅ 2 without degradation of redox activity for the Co- and Ni-substituted LSMs, as well as the Mg-substituted LSM.

Figure 8 illustrates the secondary electron image (SEI) micrograph of the Mg-substituted LSM as follows: the initial sample (before testing) magnification of 1) 2,000 and 2) 5,000; after the long-term tests magnification of 3) 2,000 and 4) 5,000. A micrograph of the Co- and Ni-substituted LSMs is shown in Supplementary Figures S4, S5. As seen in the initial sample (Figures 8A,B), the size of particle was in the range of 0.1–0.5 µm, and non-porous and polygon-shaped particle were numerously observed. After the long-term test (Figures 8C,D), a large number of larger particle size 1–3 µm without variation in particle shape appeared in the SEI image. In addition, the particles appear to have coalesced as per the large particle size after the tests. The results indicated that the initial powder sample was partially sintered at high temperatures during long-term tests. However, as shown in Figure 7D, the redox performances of the production rate and productivities of O₂ and CO were maintained without degradation during the long-term test. Thus, the results indicate that over a long reaction time, the Mg-substituted LSM retains thermal stability and stable redox performance.

To evaluate the structural stability of the Mg-substituted LSM, we compared the XRD patterns of the samples before and after the long-term test, as shown in Supplementary Figure S6. As shown in Supplementary Figure S6A, all diffraction peaks assigned to the La_0.7Sr_0.3Mn_0.85Mg_0.15O₃ reference phase (red vertical bars below the XRD pattern) (Troyanchuk et al., 2015) were indexed in a trigonal unit cell [space group R- 3C (167)], for which the lattice parameters are listed in Table 2. The results indicate that the sample was synthesized as a single phase before the test. The diffraction pattern of the sample obtained after the test did not vary. However, a series of peaks resulting from the Si reference phase (Dutta, 1962) (green vertical bars below the XRD pattern) were observed together with the La_0.7Sr_0.3Mn_0.85Mg_0.15O₃ phase. This is owing to the sample holder made of glass for XRD measurements; no secondary phase occurred in the test. The XRD patterns of Co-, and Ni-substituted LSMs after the long-term test are displayed in Supplementary Figures S6B,C, respectively. No secondary phase was also observed in the patterns of both samples as well as the Mg-substituted LSM, although the XRD peaks derived from Si appeared. These results indicate that these substituted LSMs promoted the redox reaction without crystallographic dissociation during the long-term multi-cycling test.

The Mg-substituted LSM, which has the highest activity and long-term thermal stability among the LSM perovskites tested in the present study, was selected as a tentative redox material. The change in the valence state of all ionic species was examined for the synthesized sample, and samples obtained after the TR and subsequent CS steps of thermochemical two-step CO₂ splitting by XPS analysis.

Figure 9 shows the La 3d_5/2 and La 3d_3/2 x-ray photoelectron spectra with peak deconvolutions of the synthesized, reduced, and the oxidized samples. As seen in Figure 9A, the triplet peaks centered at 833.7, 835.7, and 837.9 eV for the synthesized sample can be assigned to the La 3d_5/2 main peak and La 3d_5/2 satellite peaks. The binding energies of La 3d_3/2 were located at 850.7 (main peak), 853.0 (satellite peak), and 855.1 eV (satellite peak), as shown in Figure 9B. The triplet peaks of La 3d_5/2 and La 3d_3/2 was observed for LaFe_xO_3-δ perovskites with different x values (0.7 ≤ x ≤ 1.3) in the previous work. The satellite peaks could be assigned to the antibonding and bonding components of the 3d state (Cao et al., 2016). These data of the present study indicate that lanthanum ions are present in the trivalent form for the synthesized sample. Negligible differences were observed in terms of the binding energies of the synthesized, reduced, and the oxidized samples, indicating that the valence state of La did not change during the thermochemical process.

The results of XPS analysis for Sr 3d and O 1s spectra with peak deconvolutions are shown in Figure 10. As shown in Figure 10A, the Sr 3d spectrum of the synthesized sample was fitted into two doublets (Sr 3d_5/2 and Sr 3d_3/2). The doublet peaks of Sr 3d_5/2 were centered at 132.4 (main peak) and 133.7 eV (satellite peak), while those of Sr 3d_3/2 were located at 134.5 (main peak) and 136.0 eV (satellite peak). The peaks of Sr 3d and O 1s were divided into two components resulting from the surface and lattice (Wang et al., 2017). The high binding energy component of Sr 3d_3/2 corresponds to the position of Sr 3d_3/2 in SrO segregated on the surface, while the low binding energy component of the Sr 3d_3/2 can be attributed to Sr in the manganite structure (Bertacco et al., 2002). The binding energy of Sr 3d_5/2 was approximately 2.1 eV higher than that of Sr 3d_3/2. The peak of Sr 3d_5/2 located in low-binding energy region (131.6–132.7 eV) is regarded as lattice component owing to the perovskite structure on the near-surface region, on the other hand, that located in high-binding energy region (133.4–133.6 eV) is considered as component of surface termination (strontium oxide/strontium hydroxide/strontium carbonate) (Wang et al., 2017). The peak area in the lattice component of Sr 3d_5/2 for the synthesized, reduced, and oxidized samples are 4.0, 4.9, and 5.1 times as much as that of the surface component. The Sr 3d data of the present study indicate that Sr resulting from the lattice component is more dominant than the surface termination component, and the binding energies of all components for the synthesized, reduced, and oxidized samples exhibit negligible variation. Thus, the valence state of Sr varies negligibly during the thermochemical process.

The Sr 3d data can be related to the results of O 1s data by separating the spectrum into two components: lattice and surface. The O 1s spectra with peak deconvolution are shown in Figure 10B. There are three types of oxygen: O_I (528.8 eV), O_II (531.3 eV), and O_III (534.0 eV), which represent the lattice oxygen, adsorbed O₂ ²⁻/O⁻, and oxide (or secondary phase). In the O 1s data, the lattice O_I and surface O_II do not vary in the binding energies of the synthesized, reduced, and oxidized samples, while the surface O_III does not change after the thermochemical process of the TR and CS steps. The area ratios of O_I/(O_II + O_III) in the O 1s data are 0.20, 0.22, and 0.25 for the synthesized, reduced, and oxidized samples, respectively. The results indicate that the oxygen resulting from the adsorbed and secondary phase area is predominant in the surface area relative to the oxygen corresponding to the lattice component. The O 1s located at 528.8 eV in the low-binding-energy region (528.7–529.7 eV) corresponds to lattice oxygen in the structure of perovskite oxide (McIntyre and Cook, 1975; Vasquez, 1991; Galenda et al., 2007; Wang et al., 2016). The peak of O 1s located at 531.3 eV in high-binding energy region (531.1–531.6 eV) is regarded as a component of surface termination resulting from an adsorbed oxygen and carbonate phase (McIntyre and Cook, 1975; Uwamino et al., 1984; Norman and Leach, 2011; Wang et al., 2016). The peaks of O 1s located at 533.8–534.0 eV are considered an oxygen component of adsorbed OH⁻ or absorbed bulk oxygen species and metal oxide (Sakairi et al., 2015). The O_III (534.0 eV) could be related to the presence of the Mg(OH)₂ species (Yao et al., 2000). Peak areas of lattice oxygen (s_I), adsorptive species (s_II), and surface oxygen (s_III) in O 1s spectra were listed in Table 3. A lattice peak ratio (R_lattice) was calculated using s_I and s_III data. R lattice = ( s l + s lll )

The value for R_lattice decreased after the TR step, and increased after the subsequent CS step. This change of R_lattice was in good agreement with the weight change obtained by thermochemical redox cycle with TGA.

The Mn 3s, Mn 2p, and Mg 2p spectra with peak deconvolutions are shown in Figure 11. As seen in Figure 11A, the Mn 3s spectrum for the synthesized sample fitted doublet peaks are centered at 82.9 and 88.0 eV. The peak centered at 92.2 eV may correspond to La 4d satellite signals (Galakhov et al., 2002). It was reported that the magnitude of the Mn 3s splitting decreases monotonically with an increase in the formal valence of the manganese ions (Galakhov et al., 2002). The Mn valence,   v M n   ( e V ) was estimated using the following linear equation: v M n = 9.67 − 1.27 Δ E M n 3 s   where Δ E M n 3 s is the difference in binding energy between the Mn 3s doublet (Beyreuther et al., 2006). The linear relationship between v M n and Δ E M n 3 s was derived for the valence range between +2 and +4 from XPS investigations of different bulk mixed-valent manganites and binary Mn oxides, respectively. The results of the Mn valence   v M n   for the synthesized, reduced, and oxidized samples are listed in Table 3; for which the values of Δ E M n 3 s were 5.12, 5.20, and 4.94 eV, respectively. The corresponding valence states of Mn are 3.20, 3.07, and 3.39, respectively. The Mn valence   v M n decreased after the TR step, releasing oxygen, and backwards increased after the subsequent CS step. The results based on Mn 3s spectra indicate that the valence states of Mn ions contribute to the redox reaction in the Mg-substituted LSM during the thermochemical two-step process.

According to previous research (Zhang et al., 2014; Ilango et al., 2016; Xia et al., 2021), the Mn valence in perovskites can also be estimated from the Mn 2p spectra. The Mn 2p spectra with peak deconvolution are shown in Figure 11B. For the synthesized sample, the doublet peaks centered at 641.6 and 653.2 eV corresponds to Mn 2p_3/2 and Mn 2p_5/2 of Mn³⁺, respectively, while the peaks at 644.1 and 656.1 eV represents Mn 2p_3/2 and Mn 2p_5/2 of Mn⁴⁺, respectively. According to the peak areas of the Mn³⁺ and Mn⁴⁺ curves, the corresponding Mn valences can be calculated; the results are shown in Table 3. The ratios of Mn³⁺/Mn⁴⁺ of the synthesized, reduced, and oxidized samples were 3.28, 3.33, and 3.37 for the as-synthesized samples, after oxygen release, and after CO₂ splitting, respectively. The variations in the Mn valence ratio calculated from the Mn 2p spectra could not describe the redox reaction in the thermochemical two-step process. This is owing to some valence states of Mn ions in the Mn 2p spectra in order to superimpose them in the limited binding energies of 640.0–645.0 eV. It is known that the doublet of the Mn 2p spectra cannot specifically distinguish and separate Mn²⁺/Mn³⁺/Mn⁴⁺, indicating that complicates the estimation of the valence states of Mn (Biesinger et al., 2011; Polfus et al., 2015). According to the results in Table 3, the change of R_lattice was in good agreement with the results of Mn 3s spectra. Thus, in order to evaluate the variations in Mn valence during the thermochemical two-step process, we considered a secure result based on Mn 3s spectra to estimate the variations in Mn valence.

Figure 11C shows the Mg 2p spectra with peak deconvolution. For the synthesized sample, the binding energies of 48.2 and 50.7 eV correspond to Mg(OH)₂ and Mg²⁺, respectively. Previous studies reported that the binding energies of Mg(OH)₂ and MgO were centered at 48.8 and 50.2 eV (Luo et al., 2015), and that of Mg²⁺ and metal Mg were centered at 51.0 and 49.7 eV (Yao et al., 2000). Thus, Mg ions in the Mg-substituted LSM adsorbed hydroxyl groups on the surface of perovskite and existed in the lattice of the crystal structure, indicating that Mg ions are in a bivalent state without metal (zero-valent) in the synthesized sample. Negligible differences (48.2–48.3 eV for Mg(OH)₂, 50.7 eV for Mg²⁺, respectively, were observed in the binding energies of the synthesized, reduced, and oxidized samples, indicating that the valence state of Mg did not vary during the thermochemical process. Therefore, Mg substitution in LSM perovskite enhances the amount of O₂ released and improves fuel production in order to induce variations in the Mn valence without changing the Mg valence during the thermochemical two-step process.

In the present study, in order to compare the redox reactivity of Mg, Ni, and Co-substituted LSMs the authors selected some literatures with regard to thermochemical two-step CO₂ splitting using LSM perovskites among the numerous previous reports. The main results and test conditions used in previous studies are listed in Table 4. It is necessary to be careful for the test equipment and test conditions (reaction time and temperature in both steps) when the results are compared. Thermogravimetric analysis (TGA) indirectly estimates the O₂ and CO productivities from the weight change of the sample during the thermochemical two-step CO₂ splitting cycle. Because the results may include physical gas adsorption/desorption on the surface of the sample, the TGA test may overestimate the productivities in comparison to some reactors (fixed bed, stagnation flow reactor et al.) directly measured O₂ and CO gases using gas chromatography and gas analysis. The test temperatures of the TR and CS steps and the reaction time strongly affected the O₂ and CO productivities. La_0.5Sr_0.5Mn_0.95Sc_0.05O₃, La_0.5Sr_0.5Mn_0.75Ga_0.25O₃, and La_0.6Sr_0.4MnO₃ were reported to have high CO productivities (506, 460, and 469 μmol/g-material, respectively) by TGA (Dey et al., 2016; Luciani et al., 2018). There are two possibilities for La_0.5Sr_0.5Mn_0.95Sc_0.05O₃ and La_0.5Sr_0.5Mn_0.75Ga_0.25O₃: 1) a reaction time period of 45 min during the CS step. This may lead to enhanced CO productivity owing to slow kinetics of the oxidation reaction for the reduced LSM perovskite, if the kinetics predominate as a rate-determining step in the entire process; 2) a small amount of B site substitution into LSM perovskites may enhance O₂ and CO productivities owing to an improvement of thermal stability under reductive and oxidative atmospheres during the thermochemical processes. However, the data for La_0.6Sr_0.4MnO₃ (Luciani et al., 2018) may be overestimated because O₂ and CO productivities were considerably high in comparison to the productivity levels of O₂ 100–215 and CO 194–295 μmol/g-material with similar compositions of LSM perovskites (Nair and Abanades, 2018; Takalkar et al., 2021).

To evaluate the small amount of substitution at the B-site of the LSM perovskite among the previous data, we compared the data between La_0.5Sr_0.5Mn_0.9Mg_0.1O₃ and La_0.5Sr_0.5MnO₃ (Demont and Abanades, 2015; Jouannaux et al., 2019). Although the thermochemical cycle was performed under similar test conditions, 10% Mg substitution decreased O₂ productivity and increased CO productivity. On the other hand, the data for La_0.7Sr_0.3MnO₃ in the present study were in the middle level of O₂ and CO productivities and were in good agreement with the previous data listed in Table 4. Furthermore, the substitution of 10% Mg, Ni, and Co into the LSM perovskite enhanced O₂ and CO productivities by approximately 1.3–1.4 times. The results indicate that the evaluation of the redox reactivity of the thermochemical cycle is required such that they can be systematically examined using the same test equipment and conditions.