Solar thermochemical hydrogen (STCH) production has been studied as a potential path to produce alternative fuels (Miller et al., 2014). This reaction is generally a two-step process that generates hydrogen (H₂) gas by splitting water using metal oxide materials and concentrated solar energy (Steinfeld, 2005; Smestad and Steinfeld, 2012). In the first step of a typical STCH cycle, the metal oxide is thermally reduced under inert atmosphere at high temperature (>1,200°C, achieved using concentrated solar-thermal flux), creating oxygen vacancies in the metal oxide and releasing oxygen gas (Scheffe and Steinfeld, 2014). In the second re-oxidation step at a lower temperature under steam, the oxygen-deficient metal oxide splits water to produce H₂ gas while regenerating the metal oxide for consecutive water splitting cycles. While a combined photovoltaic/electrolytic system can be an alternative method for generating H₂ from water (Ivy, 2004), the efficiency of STCH systems were predicted to potentially exceed the efficiency of the combined system (Siegel et al., 2013).

Different classes of materials have been studied demonstrating STCH capabilities, including fluorites (CeO₂) (Gauckler et al., 1997; Abanades et al., 2010; Chueh et al., 2010; Scheffe and Steinfeld, 2014; Rao and Dey, 2017), iron oxides (Fe₃O₄) (Nakamura, 1977; Kodama et al., 2004; Coker et al., 2011; Scheffe and Steinfeld, 2014), and spinel ferrites (MFe₂O₄, M = Cu, Ni, Zn, etc.) (Tamaura et al., 1995; Allendorf et al., 2008; Miller et al., 2008; Fresno et al., 2009; Scheffe and Steinfeld, 2014). Additionally, perovskite oxides (ABO₃) have been investigated heavily with many perovskite materials demonstrated to possess water splitting capabilities (McDaniel et al., 2013; Yang et al., 2014; Rao and Dey, 2017; Barcellos et al., 2018; Nair and Abanades, 2018; Qian et al., 2020a; Qian et al., 2020b). A large chemical space of perovskite materials is available due to the flexibility in chemical compositions and crystal structures the formula can stabilize (Vasala and Karppinen, 2015). For this reason, perovskite oxides are fertile ground for discovering new STCH materials with high efficiencies.

With the vast composition space of inorganic materials, preliminary computational materials screening has become an important tool for accelerating materials discovery. Computational methods were previously applied to materials screening for the solid-state hydrogen storage reaction and chemical looping process (Clary et al., 2020; Singstock et al., 2020). For STCH, ternary (ABO₃) and quaternary (AA’BO₃) perovskites were explored to computationally evaluate their oxygen vacancy formation energies, electronic properties, and thermodynamic stabilities, with several promising candidate materials predicted based on these results (Emery et al., 2016; Sai Gautam et al., 2020). In this work, we report Gd_0.5La_0.5Co_0.5Fe_0.5O₃ (GLCF) as a new STCH material that was discovered using a computational screening approach and experimentally demonstrated as a STCH producing material.

To evaluate compound stability relative to decomposition, the DFT computed energies of the La₂CoFeO₆ (L2CF), GdLaCoFeO₆ (GLCF), and Gd₂CoFeO₆ (G2CF) perovskites optimized in the monoclinic, triclinic, and monoclinic space groups, respectively, (Supplementary Figure S1) were compared to the DFT computed energies of their potential decomposition products tabulated in the MP database. The energy of a material relative to the convex hull, E_hull, quantifies a material’s stability relative to its decomposition products. Monoclinic L2CF, triclinic GLCF, and monoclinic G2CF exhibit E_hull of 0, 6.8, and 0 meV/atom, respectively, predicting that all three STCH candidates can be successfully synthesized as perovskites relative to their competing phases. The combination of the τ-predicted stabilities and E_hull values of L2CF, GLCF, and G2CF suggests that Gd-La-Co-Fe-O is a promising compositional space for experimentally synthesizable perovskite oxides. We thus selected L2CF, GLCF, and G2CF from our high-throughput screening and computed additional STCH-relevant properties for these compounds, including DOS and charge neutral oxygen vacancy formation enthalpies (ΔH_Ovac).

The distributions of ΔH_Ovac computed by DFT for the symmetrically unique sites of L2CF, GLCF, and G2CF are shown in Figure 1A. The ΔH_Ovac distribution of GLCF is bounded by the ΔH_Ovac distributions of L2CF (lower bound) and G2CF (upper bound). Herein, we use the DFT computed ΔH_Ovac of CeO₂ (3.95 eV/atom) (Abanades and Flamant, 2006; Chueh and Haile, 2010), the gold standard STCH redox mediator (Muhich et al., 2016), as the upper bound for the STCH active range and a liberal lower bound of 2 eV/atom to account for uncertainty in DFT energetics (Naghavi et al., 2020). Materials approaching the minimum ΔH_Ovac for STCH activity exhibit slow oxidation kinetics and/or degradation during redox cycling, whereas materials approaching the maximum ΔH_Ovac for STCH activity suffer from reduced H₂ capacity (Muhich et al., 2016). GLCF exhibits both O vacancies with low and high ΔH_Ovac, where vacancies with low ΔH_Ovac participate in H₂ production and those with high ΔH_Ovac are less likely to form under STCH operating conditions and therefore enable preservation of the perovskite lattice during redox cycling. This ΔH_Ovac distribution predicts enhanced cyclability of GLCF relative to L2CF and increased H₂ production capacity relative to G2CF.

Additionally, GLCF exhibits a large DOS effective mass, m _e , that arises from its large concentration of accessible electronic states near the conduction band minimum (CBM). Lany showed that large m _e corresponds with large electronic contributions to the entropy of reduction S_red in STCH processes (Lany, 2018) that benefits STCH performance at high temperatures (Meredig and Wolverton, 2009). Figure 1B shows the DFT computed DOS for L2CF, GLCF, and G2CF. GLCF exhibits a larger DOS m _e at 1000 K (m _e = 8.356) than both L2CF (m _e = 7.304) and G2CF (m _e = 3.209). The introduction of Gd into the nominally La-occupied A site of L2CF results in unoccupied states that lie closer to the CBM than those of La alone, which increases m _e . However, complete substitution of Gd for La increases the splitting of the unoccupied Gd-f states that results in lowering m _e relative to GLCF. Due to its favorable ΔH_Ovac distribution and larger m _e relative to L2CF and G2CF, GLCF was recommended for experimental synthesis and characterization of STCH performance.

The room temperature XRD pattern of the synthesized GLCF is shown in Figure 2. The resulting material is isostructural to LaFeO₃, GdFeO₃, and La_1-xGd_xFeO₃ (x = 0, 0.2, 0.5, 0.8, and 1), which are all orthorhombic perovskites (Proskurnina et al., 2007; Wiglusz et al., 2015; Orlov et al., 2016). LeBail profile fitting analysis was performed, confirming the GLCF crystalizes in the Pbnm space group (No. 62) with no additional secondary phases. The lattice parameters for GLCF (a = 5.3973(3) Å, b = 5.4605(3) Å, c = 7.6455(4) Å) are slightly smaller than LaFeO₃ (a = 5.5506(6) Å, b = 5.5608(5) Å, c = 7.8464(9) Å) (Proskurnina et al., 2007). This is likely due to the smaller atomic radii of Gd and Co compared to La and Fe, respectively (Shannon, 1976).

The redox behavior of GLCF was examined through thermogravimetric analysis (TGA), which monitors the mass change with respect to temperature and atmosphere (Ar or air). In the case of these materials, the mass change corresponds to gain/loss of oxygen. The mass change was converted to the extent of reduction (δ), assuming Gd_0.5La_0.5Co_0.5Fe_0.5O_3-δ during redox cycling. The as-prepared powder was subjected to redox cycles, comprising a reduction step at 1,250°C under Ar followed by a re-oxidation step at 1,100°C under air, corresponding to 0.16 atm O₂. The redox cycle was repeated twice with the two redox cycles showing similar behavior. The second redox cycle is shown in Figure 3. During thermal reduction, a shallow weight loss was observed at ∼750°C, followed by a steeper weight loss onset at ∼1,030°C. After the isotherm at 1,250°C, the extent of reduction (δ) was ∼0.13. During re-oxidation, the mass sharply increased when the gas was switched from Ar to air at 400°C, followed by a more gradual increase as the temperature was raised. The dip in the extent of reduction seen between 900 and 1,100°C is attributed to a phase change in the material, that is, formation of a phase at around 900°C that ejects oxygen to achieve stability. This change in δ is reproducible between cycles. The sample mass returns to the starting mass after the re-oxidation step.

The water splitting capability of GLCF was verified using a stagnation flow reactor (SFR). Figure 4 shows the water splitting results performed with a thermal reduction temperature (T_R) of 1,350°C for 330 s and re-oxidation temperatures (T_O) of 850 and 1,000°C for 1,200 s under 40 vol% H₂O. Water splitting was observed under both sets of conditions, with the amount of H₂ produced increasing with each cycle. For T_O = 850°C (Figure 4A), the amount of H₂ produced was 67, 90, and 101 μmol/g for each consecutive cycle. Moreover, significantly less O₂ was released in the second reduction cycle compared to the first cycle, likely due to the water splitting step being kinetically limited. This is also evident from the long tails of H₂ gas evolved during the re-oxidation steps. For T_O = 1,000°C (Figure 4B), the H₂ capacity increased (127 and 141 μmol/g), evident of an improvement in the rate of water splitting at higher temperature. Similar to the T_O = 850°C experiment, more H₂ was produced as cycle-number increased. In an ideal situation, once steady state has been achieved, the amount of H₂ should be twice of the amount of O₂ produced. From the SFR experiments with repeated cycles, GLCF is evolving and approaching this ideal steady state, though additional studies are needed to understand this behavior.

While LaFeO₃ was previously reported to have negligible solar thermochemical H₂O and CO₂ conversion behavior (Jiang et al., 2014; Chen et al., 2017), similar perovskites to GLCF, La_0.6Sr_0.4Co_0.2Fe_0.8O₃ and LaFe_0.75Co_0.25O₃, were shown to be active for solar thermochemical CO₂ conversion (STCH activity is unknown) (Nair and Abanades, 2018), indicating that the mixing of Co and Fe may have contributed to the solar thermochemical conversion activity. For La_0.6Sr_0.4Co_0.2Fe_0.8O₃ and LaFe_0.75Co_0.25O₃, however, CO production decreased substantially during subsequent cycles (Nair and Abanades, 2018). The substitution of Gd for La may have contributed to minimizing performance degradation as predicted by DFT calculations.

In terms of its water splitting ability, at T_R = 1,350°C and T_O = 850°C, GLCF produced more H₂ (101 μmol/g) than CeO₂ (50 μmol/g) (Barcellos et al., 2018). However, GLCF produced less H₂ compared to the previously studied perovskite materials BaCe_0.25Mn_0.75O₃ (BCM) and Sr_0.4La_0.6Mn_0.6Al_0.4O₃ (SLMA4664) (McDaniel et al., 2013; Barcellos et al., 2018). BCM and SLMA4664 produced 140 μmol/g (T_R = 1,350°C, T_O = 850°C) and 307 μmol/g (T_R = 1,350°C, T_O = 1,000°C) of H₂, respectively (McDaniel et al., 2013; Barcellos et al., 2018). However, direct comparisons of performance reported for different conditions (temperature, atmosphere, and time) for various materials that have different optimized conditions for STCH can lead to incorrect conclusions about the H₂ production capabilities of candidate materials. In the present case, the water splitting experimental conditions implemented for GLCF have not yet been optimized. Nevertheless, computational screening greatly accelerated the discovery of GLCF as a water splitting material, which would otherwise have been experimentally time consuming due to the vast chemical space of perovskite materials.

To understand the phase stability in GLCF during redox cycling, HT-XRD patterns were collected first from room temperature to 1,250°C to 50°C under He for thermal reduction, then with an identical temperature profile under air for re-oxidation. The GLCF sample was redox-cycled (T_R = 1,250°C under Ar, T_O = 1,100°C under air) before collecting the HT-XRD patterns. The HT-XRD pattens are shown in Figure 5. During the HT-XRD experiment, GLCF appears to maintain its perovskite phase as the major phase, but the major peaks broaden and exhibit some peak splitting occurring during reduction. In addition to the changes in the major GLCF phase, during cool down at ∼1,200°C under He, additional peaks appear corresponding to Gd₂O₃ phase. Due to the shift in peaks, it may have different chemistry (i.e. Gd_2-xLa_xO₃). The additional Gd₂O₃ phase remains in the pattern until heated under air at ∼700–800°C during re-oxidation. The disappearance of the Gd₂O₃ phase may correspond to the increase in the extent of reduction observed in the TG experiment at ∼900°C during reoxidation. Moreover, a spinel phase appears during cool down under air. Overall, during the redox cycle, the perovskite phase persists, agreeing with the stable STCH activity observed in the SFR results and TGA cycling tests.

Solar thermochemical hydrogen production (STCH) significantly contributes to the renewable energy portfolio. However, materials discovery with high efficiency is needed, and computational screening can greatly accelerate this process. This was demonstrated with Gd_0.5La_0.5Co_0.5Fe_0.5O₃ (GLCF), a perovskite oxide, that was computationally determined first, then experimentally demonstrated. When compared to LaCo_0.5Fe_0.5O₃, it was predicted that GLCF would have higher phase stability due to the incorporation of Gd on the La site. The synthesis of a single phase GLCF sample was achieved. With redox activity confirmed through thermogravimetric analysis, stable water splitting behavior over multiple cycles was also observed. When compared to previously reported La_0.6Sr_0.4Co_0.2Fe_0.8O₃ and LaFe_0.75Co_0.25O₃, which suffered from performance degradation with subsequent cycles, the Gd substitution of La plays a significant role to maintain water splitting performance. Based on high temperature x-ray diffraction experiments, the GLCF perovskite phase persists which potentially contributes to the stable water splitting performance. This work demonstrates that computational materials screening can greatly accelerate the discovery of new water splitting materials. Through computational screening, potential water splitting materials were narrowed down to a promising STCH material from the vast chemical space of perovskite materials, and a unique strategy of incorporating a rare earth element that improved the stability was demonstrated.