Solar thermochemical splitting of water to generate hydrogen

Abstract

Solar photochemical means of splitting water (artificial photosynthesis) to generate hydrogen is emerging as a viable process. The solar thermochemical route also promises to be an attractive means of achieving this objective. In this paper we present different types of thermochemical cycles that one can use for the purpose. These include the low-temperature multistep process as well as the high-temperature two-step process. It is noteworthy that the multistep process based on the Mn(II)/Mn(III) oxide system can be carried out at 700 °C or 750 °C. The two-step process has been achieved at 1,300 °C/900 °C by using yttrium-based rare earth manganites. It seems possible to render this high-temperature process as an isothermal process. Thermodynamics and kinetics of H₂O splitting are largely controlled by the inherent redox properties of the materials. Interestingly, under the conditions of H₂O splitting in the high-temperature process CO₂ can also be decomposed to CO, providing a feasible method for generating the industrially important syngas (CO+H₂). Although carbonate formation can be addressed as a hurdle during CO₂ splitting, the problem can be avoided by a suitable choice of experimental conditions. The choice of the solar reactor holds the key for the commercialization of thermochemical fuel production.

The impact of global climate change as well as the likely shortage of fossil fuels demand harvesting of energy using renewable sources. Although solar energy captured by the Earth in an hour is expected to satisfy the energy demand of the world for a year, the high energy density and nonpolluting end product renders hydrogen a viable alternative to fossil fuels (1). In this context, conversion of solar power to H₂ and synfuels with the utilization of renewable H₂O and CO₂ seems to be a sound option. Artificial photosynthesis and photovoltaic-powered electrolysis of water are promising approaches, although their implementation is somewhat restricted because of the low solar-to-fuel conversion efficiency (η_{solar-to-fuel}) of <5% and <15%, respectively (2, 3). The other strategy would be a solar-thermochemical process that provides a high theoretical efficiency and enables large-scale production of H₂ by using the entire solar spectrum (4). Research in thermochemical splitting of H₂O made a beginning in the early 1980s (5, 6) and several thermochemical cycles have been examined. Thermochemical methods come under two main categories, the low-temperature multistep processes and the high-temperature two-step processes. The two-step process involving the thermal decomposition of metal oxides followed by reoxidation by reacting with H₂O to yield H₂ is an attractive and viable process that can be rendered to become an isothermal process. Thermochemical splitting of H₂O at low temperatures (<1,000 °C) is accomplished by a minimum of three steps as dictated by thermodynamic energy constraints (7, 8). In this paper we present the highlights of recent investigations of H₂O splitting by the low-temperature multistep process as well as the high temperature two-step process.

Low-Temperature Multistep Cycles

Low-temperature cycles are advantageous due to low radiative losses and availability of more heat resources, including nuclear waste heat sources. There has been a good deal of research in the past decades in this area and the performance of a few cycles such as the S–I and S–Br cycles as well as Fe–Cl, Hg–Br, and Cu–Cl cycles is noteworthy (5, 9). Although they produce H₂ steadily, these cycles suffer from environmental issues associated with the separation of acid mixtures, decomposition of acids, heavy-metal processing, production of toxic or corrosive intermediates, and so on. Even the most-studied sulfur–iodine cycle suffers from the disadvantage of corrosive intermediates (9).

Manganese Oxide-Based Cycles.

Transition metal oxide-based multistep cycles have attracted attention. One of these is the four-step cycle based on manganese oxides wherein Mn₂O₃ reduces to MnO with the evolution of O₂ above 1,500 °C, and NaOH then oxidizes Mn(II) back to Mn(III) as NaMnO₂(s), with the evolution of H₂ above 600 °C. The best efficiency for this cycle was 74% and 16–22% considering 100% and 0% heat recovery, respectively, as calculated by Sturzenegger and Nuesch (10). Volatility of NaOH above 800 °C and the high Mn₂O₃-to-MnO conversion temperature hinders the practical implementation of this cycle. Instead of NaOH, the use of Na₂CO₃ for the production of H₂ from redox active MnFe₂O₄ has been suggested by Tamaura et al. (11), as shown in Eqs. S1 and S2.

Although this cycle closes below 1,000 °C, stoichiometric quantity of O₂ is not evolved due to the incomplete extraction of Na⁺ even in the presence of CO₂(g). Modification involving the introduction of Fe₂O₃ as a sacrificial agent has been proposed, but the use of Fe₂O₃ does not close the cycle either (12).

Mn₃O₄-Na₂CO₃-MnO–Based Cycle.

In this context, the thermochemical cycle based on Mn₃O₄/MnO oxides proposed by Davis and coworkers is noteworthy (8). The four reactions in this cycle are shown in Fig. 1 (Eqs. S3–S6). Generation of H₂ occurs at 850 °C in this cycle (Fig. 1) and the cycle is devoid of corrosive products (8). Thermal oxidation of Mn₃O₄ to Mn₂O₃ is thermodynamically unfavorable, but the use of Na₂CO₃ drastically decreases the ΔG and forms MnO(s) and α-NaMnO₂(s) at 850 °C with the evolution of CO₂(g), as shown in step 1 (Eq. S3) (8, 9). The CO₂ evolution temperature can be decreased drastically (∼600 °C) with the use of nanoparticles of Mn₃O₄ and Na₂CO₃ obtained with ball milling (Fig. S1) (13). When nanoparticles of both Mn₃O₄ and Na₂CO₃ are used (Fig. S1), the weight loss occurs sharply below 600 °C, releasing ∼100% of CO₂. During step 2 (Eq. S4), introduction of H₂O(g) at 850 °C oxidizes MnO to α-NaMnO₂ with the evolution of H₂(g) (8). The main hurdle in this cycle is the slow H₂ evolution. The rate of H₂ evolution is significantly enhanced by the use of nanoparticles due to their high surface area (Fig. 2A). Thus, ball-milled samples show a high rate of H₂ production soon after the entry of H₂O, with an overall increase of 1.5–2 times in comparison with the bulk samples (13). Interestingly, the use of nanoparticles effectively brings down the H₂ evolution temperature to 750 °C or 700 °C as shown for the 60-min ball-milled (MnNa60) sample in Fig. 2B, although the H₂ evolution rate is somewhat slower in comparison with that at 850 °C (Fig. 2). The yield of H₂ at 750 °C is ∼40% (over 80 min) with an extended tail due to further H₂ evolution over a period (13).

Fig. 1.

Schematic presentation of Mn(II)/Mn(III) based low-temperature multistep thermochemical cycle. Reproduced with permission from ref. 13, copyright 2016, Royal Society of Chemistry.

Fig. S1.

Thermogravimetric weight loss plotted as a function of CO₂ evolution temperature (Eq. S3) due to the reaction of Mn₃O₄ with Na₂CO₃. Curves (1) and (2) represent a mixture of bulk Na₂CO₃ with commercial (C) and sintered (HT) Mn₃O₄, respectively. Curves (3) and (4) and (5) and (6) represent a mixture of ball-milled Mn₃O₄ with bulk (designated as Mn) and ball-milled Na₂CO₃ (designated as MnNa), respectively. The 30 and 60 represent the ball milling time in minutes. Reproduced with permission from ref. 13, copyright 2016, Royal Society of Chemistry.

Fig. 2.

H₂ evolution plots of reaction of Mn₃O₄ with Na₂CO₃ at (A) 850 °C and (B) 750 °C and 700 °C. Bulk and ball-milled Mn₃O₄ + Na₂CO₃ are designated as Mn-C and MnNa, respectively. The 30 and 60 represent ball milling times in minutes. Reproduced with permission from ref. 13, copyright 2016, Royal Society of Chemistry.

As shown in Fig. 1, extraction of Na⁺ from layered NaMnO₂(s) is performed by hydrolysis under a CO₂(g) atmosphere at 80 °C for 3 h (step 3, Eq. S5), giving rise to protonic birnessite as the product (8). It has been noted recently that hydrolysis at 50 °C for 1 h extracts Na⁺ completely (13). Birnessite reduces to Mn₃O₄ with stoichiometric O₂ evolution (step 4, Eq. S6). Investigations of this cycle with other metal oxide (Fe₃O₄ and Co₃O₄)–carbonate pairs (Li₂CO₃ and K₂CO₃) have been conducted by Xu et al. (14) and the H₂ evolution rate was found to vary in the order Fe₃O₄ > Mn₃O₄ > Co₃O₄ and Li₂CO₃ > Na₂CO₃ > K₂CO₃. In this method, the complete extraction of Li⁺ ions seems to be impossible and the Fe(III)–Fe(II) reduction temperature is high (∼1,150 °C). The manganese oxide-based system, Mn₃O₄/ Na₂CO₃/MnO, seems to be the best combination (14).

Two-Step Thermochemical Processes

The two-step metal oxide process carried out with the aid of solar concentrators eliminates the necessity of separating of H₂ and O₂. The metal oxide (MO_oxd) reduces to the metal or to a lower valent metal oxide (MO_red) (Eq. S7) with the release of O₂(g) during the endothermic step (T_red), and in the next step it gets reoxidized (T_oxd) on reaction with H₂O (Eq. S8), releasing a stoichiometric amount of H₂(g), as shown in Fig. 3.

Fig. 3.

Schematic representation of the two-step solar thermochemical splitting of H₂O using nonstoichiometric metal oxide redox pairs and concentrated solar energy. During the endothermal reaction, the metal oxide (MO_n) reduces to oxygen-vacant MO_n-δ with the evolution O₂(g) along with inert gas. In the exothermal step, MO_n-δ takes back oxide ions from H₂O and goes back to initial MO_n for cycling with the evolution of H₂(g).

T_red > T_oxd is the thermodynamic driving force in the two-step process for it to become feasible as described by Eq. 1 in terms of the free energy change of formation of H₂O $\left({\mathrm{\Delta }\text{G}}_{\text{f},{\text{T}}_{\text{oxd}}}^{{\text{H}}_2\text{O}}\right)$ and the entropy of O₂ $\left({\text{S}}_{{\text{T}}_{\text{red}}}^{{\text{O}}_2}\right)$. Increasing T_red increases ${\text{S}}_{{\text{T}}_{\text{red}}}^{{\text{O}}_2}$, so the decrease in $\mathrm{\Delta }\text{T}$ window can be permitted at higher T_red (7). Conversely, at lower T_red, $\mathrm{\Delta }\text{T}$ has to be higher (7):

$$\mathrm{\Delta }\text{T}={\text{T}}_{\text{red}}-{\text{T}}_{\text{oxd}}=-2{\mathrm{\Delta }\text{G}}_{\text{f},{\text{T}}_{\text{oxd}}}^{{\text{H}}_2\text{O}}/\left({\text{S}}_{{\text{T}}_{\text{red}}}^{{\text{O}}_2}\right).$$ [1]

Thermal reduction of the oxide and H₂O splitting have been conducted at the same temperature (T_red = T_oxd = T_iso) in which the large pressure swing in the gas composition between reduction and oxidation processes acts as the driving force (15).

Splitting of CO₂ is analogous to that of H₂O and releases a stoichiometric quantity of CO, making this process useful for the production of syngas (H₂+CO+CO₂) that can be converted into a liquid fuel as the end product. In view of the global climate change due to CO₂ emission, splitting of CO₂ provides an additional means of cutting its concentration. The important distinction between CO₂ and H₂O splitting is as follows. Above 1,100 K, CO₂ splitting is thermodynamically more favored and gets kinetic advantages of reduction. The water–gas shift reaction ($\text{WGS}:{\text{H}}_2\text{O}+\text{CO}\to {\text{CO}}_2+{\text{H}}_2$) and the reverse water–gas shift reaction ($\text{RWGS}:{\text{H}}_2+{\text{CO}}_2\to \text{CO}+{\text{H}}_2\text{O}$) appear as alternative approaches to liquid fuel production via syngas production. H₂ production through the low-temperature WGS reaction is thermodynamically favorable due to this cross-over in stability. Maravelias and coworkers (16, 17) assessed the three main thermochemical fuel production systems shown in Fig. S2. In these systems, syngas is produce via (i) solar CO₂ splitting followed by the WGS reaction, (ii) solar H₂O splitting followed by the RWGS reaction, and (iii) splitting of CO₂ and H₂O without the WGS/RWGS reaction. The loss of efficiency of systems i and ii is due to the energy-intensive CO₂/CO separation and to the unfavorable thermodynamics of the RWGS reaction, respectively. The efficiency of system iii is greater than that of the other two because the requirement of CO₂/CO separation is reduced significantly and the H₂:CO ratio is set by the distribution of the solar heating units (16, 17).

Fig. S2.

Various process configurations to produce methanol and Fisher–Tropsch (FT) fuels as a part of the Sunshine to Petrol framework. Adapted with permission from ref. 17, copyright 2012, Royal Society of Chemistry.

Two-step cycles can make use of stoichiometric or nonstoichiometric oxides. Stoichiometric oxides can be volatile and nonvolatile oxides during the redox process. Typical volatile cycles are ZnO/Zn, SnO/SnO₂, In₂O₃/In, and CdO/Cd redox pairs that exhibit reversible solid–gas phase transitions during cycling. Volatile cycles are thermodynamically favorable due to the high entropic gain related to the formation of the gaseous product during the reduction (18). Recombination of product gases is a practical challenge in these cycles and needs to be addressed. In nonvolatile cycles, the redox pairs remain in the condensed state and bypass the problems of product recombination. The Fe₃O₄/FeO cycle, first reported by Nakamura (19), operates at 2,500 K. The melting point of Fe₃O₄ (1,870 K) and FeO (1,650 K) is lower than the reduction temperature of Fe₃O₄, which results in severe coarsening of particles and further deactivates the cycles due to the alternating fusion and solidification of iron oxides (20). To lower the reduction temperature further, divalent metal ions (M = Zn, Ni, Co, Mg, and Mn) were incorporated in the Fe₃O₄ matrix to form mixed ferrites (M_xFe_3-xO₄) (9, 21). Incorporation of Co and Ni improves the reduction capability of ferrites substantially. Notably, NiO (melting point of 2,271 K) incorporated in the Fe₃O₄ matrix boosts the performance (21, 22). Slow Fe²⁺ diffusion retards the redox activity of ferrites. To overcome this limitation, either nanostructuring of iron oxides or introduction of additional supports (YSZ, ZrO₂, CaSZ, etc.) has been proposed (23). Nanostructuring has been carried out by atomic layer deposited CoFe₂O₄ and NiFe₂O₄ on high surface area Al₂O₃ to improve the reaction kinetics (24). However, because coarsening of ferrite particles is a matter of concern, nanostructuring is not a practical solution. Addition of supports improves the reaction kinetics and solves the sintering of ferrite particles, although it is also not the ideal solution because of heat loss due to the high mass percentages of supports.

Nonstoichiometric oxide cycles are nonvolatile. In step 1 (Fig. 3), thermal reduction generates nonstoichiometric compositions; in step 2, the oxygen vacancies are eliminated by reaction with H₂O or CO₂. The oxygen exchange capacity is dictated by the nonstoichiometry (δ), which further controls the H₂ production as well as η_solar-fuel. Knowledge of the partial molar enthalpy $\mathrm{\Delta }{\text{H}}_{\text{redox}}\left(\mathrm{\delta }\right)$ and the partial molar entropy ${\mathrm{\Delta }\text{S}}_{\text{redox}}\left(\mathrm{\delta }\right)$ of oxygen vacancy formation as a function of δ allows us to deduce oxygen nonstoichiometry under different ${\text{p}}_{{\text{O}}_2}$ and T (Eq. S9). The equilibrium H₂ yields (${\text{n}}_{{\text{H}}_2}$) and the molar ratio of the oxidant (${\text{n}}_{{\text{H}}_2\text{O}}$) needed for that purpose are obtained from a knowledge of $\mathrm{\Delta }{\text{G}}_{\text{oxd}}$ of reduced oxides and ${\mathrm{\Delta }\text{G}}_{\text{f},{\text{T}}_{\text{oxd}}}^{{\text{H}}_2\text{O}}$ (Eqs. S10 and S11). The solar-to-fuel conversion efficiency is defined by Eq. 2, where HHV stands for higher heating value of the fuel produced:

$${\mathrm{\eta }}_{\text{solar}-\text{to}-\text{fuel}}=\frac{{\text{HHV}}_{{\text{H}}_2}{\text{n}}_{{\text{H}}_2}}{{\text{Q}}_{\text{solar}}+{\text{E}}_{\text{penalties}}}.$$ [2]

Furthermore, the solar input energy (${\text{Q}}_{\text{solar}}$), which is mainly for heating water (first term) and the redox material from T_oxd to T_red (second term) as well as creating oxygen nonstoichiometry (third term), depends on solar energy absorption efficiency of receiver reactor (${\mathrm{\eta }}_{\text{abs}}$), as shown in Eq. S12. Tamaura et al. (25) used Ni-Mn-ferrite–based nonstoichiometric two-step cycles (1,073–1,373 K), but the quantity of H₂ produced was small.

CeO₂ and Doped CeO₂.

Chueh et al. (26) and Furler et al. (27) have investigated the two-step cycle based on the reduction of ceria (CeO₂ $\to$ CeO_2-δ). In such a nonstoichiometric cycle oxygen diffusion is higher than the Fe²⁺ diffusion in ferrite cycles, rendering the use of porous monolithic CeO₂ in state-of-the-art solar cavity receivers (26). The Gibbs free energy of CeO_2-δ $\to$ CeO₂ oxidation with H₂O is negative at all accessible temperatures but the drawback is the poor reducibility of CeO₂. CeO₂ has been modified with several metal ions such as divalent Ca, Sr, and Mg (28), trivalent La, Sc, Gd, Y, Cr, and Sm (29, 30), and tetravalent Zr, Ti, Hf, and Sn ions (29). Substitution with Zr⁴⁺ is found to increase the reduction capability significantly (29, 31). Fig. 4A shows how the partial molar Gibbs free energy of oxygen vacancy formation reduces gradually with increasing Zr content in CeO₂. However, the Gibbs free energy of oxidation of Zr-substituted CeO₂ is more positive than that of pure CeO₂, becoming more positive with increasing Zr content (Fig. 4B) (32). The effect becomes more pronounced at high oxidation temperatures. An increase in water concentration can promote the complete oxidation of Zr_xCe_1-xO₂ at all temperatures, but there is an energy penalty (32). A drawback of CeO₂-based oxides is the poor reducibility even at 1,500 °C as well as sublimation of the oxides at high temperatures. Syngas production with CeO₂ by the high temperature two-step cycles has been realized experimentally using the solar cavity receiver setup wherein the variation of the input H₂O:CO₂ molar ratio changes the H₂:CO in the range of 0.25–2.34 (27).

Fig. 4.

Gibbs energy change as a function of temperature for the (A) reduction and (B) oxidation of Ce_1-xZr_xO₂ for x = 0 (CeO₂), 0.05 (CZO_5), and 0.2 (CZO_20). Reproduced with permission from ref. 32, copyright 2015, Royal Society of Chemistry.

Manganite Perovskites.

In perovskite oxides of the type Ln_1-xA_xMnO₃ (LnAM_x), substitution of the trivalent Ln³⁺ by the divalent A²⁺ creates Mn³⁺/ Mn⁴⁺ redox active pairs, which assist H₂O splitting in accordance with Eqs. 3 and 4:

$${\text{Ln}}_{1-\text{x}}{\text{A}}_{\text{x}}{\text{Mn}}_{1-\text{x}}^{3+}{\text{Mn}}_{\text{x}}^{4+}{\text{O}}_3\to {\text{Ln}}_{1-\text{x}}{\text{A}}_{\text{x}}{\text{Mn}}_{1-\text{x}+\mathrm{\delta }}^{3+}{\text{Mn}}_{\text{x}-\mathrm{\delta }}^{4+}{\text{O}}_{3-\mathrm{\delta }/2}+\frac{\mathrm{\delta }}{4}{\text{O}}_2$$ [3]

$${\text{Ln}}_{1-\text{x}}{\text{A}}_{\text{x}}{\text{Mn}}_{1-\text{x}+\mathrm{\delta }}^{3+}{\text{Mn}}_{\text{x}-\mathrm{\delta }}^{4+}{\text{O}}_{3-\mathrm{\delta }/2}+\frac{\mathrm{\delta }}{2}{\text{H}}_2\text{O}\to {\text{Ln}}_{1-\text{x}}{\text{A}}_{\text{x}}{\text{Mn}}_{1-\text{x}}^{3+}{\text{Mn}}_{\text{x}}^{4+}{\text{O}}_3+\frac{\mathrm{\delta }}{2}{\text{H}}_2.$$ [4]

High-temperature oxygen nonstoichiometry analysis of La_1-xSr_xMnO₃ perovskites by Scheffe et al. (33) obtained by extrapolating the low-temperature experimental nonstoichiometry data estimates the oxygen exchange capacity of LaSM_x perovskites to be higher than that of CeO₂. The use of La_1-xSr_xMO₃ (M = Mn, Fe) perovskites for H₂O splitting started with Nalbandian et al. (34), who constructed a membrane reactor consisting of two compartments separated by dense membranes of mixed ionic–electronic perovskites. Rare earth manganite-based H₂O splitting has been investigated thoroughly in recent months.

A-Site Substitution in Manganites.

La_0.65Sr_0.35MnO₃ gets reduced to a greater extent than CeO₂ under similar ${\text{p}}_{{\text{O}}_2}$ and temperature, as reported by Scheffe et al. (33). Thermodynamic analysis by Yang et al. (35) predicts superior oxygen exchange capacity of La_1-xSr_xMnO₃ (x = 0–0.5) than CeO₂ that further increases with increasing Sr²⁺ substitution, as supported subsequently by thermochemical H₂O splitting resulting in an H₂/O₂ production ratio close to 2. Increasing the Sr²⁺ content decreases the partial molar enthalpy and the entropy of reduction. In other words, the oxidation thermodynamics becomes increasingly unfavorable. In particular, the more positive Gibbs free energy of oxidation of La_1-xSr_xMnO_3-δ relative to CeO_2-δ requires the use of a large excess of H₂O for oxidation (33). Nevertheless, considering the complete heat recovery from H₂O, the η_{solar-to-fuel} of LaSM_x is predicted to be higher than CeO₂ (33). In this context, it is noteworthy that La_1-xCa_xMnO₃ (x = 0.35, 0.5, 0.65) perovskites show an increase in oxygen nonstoichiometry with increasing Ca content, accompanied by a gradual decrease in the reduction temperature (36). A typical thermogravimetric analysis (TGA) plot reveals the oxygen nonstoichiometry generated after reduction (T_red = 1,400 °C, 99.9995% Ar) is 2.97, 2.91, and 2.88, respectively, for x of 0.35, 0.5, and 0.65. Although complete reoxidation seems to be difficult with increasing Ca content, La_0.5Ca_0.5MnO₃ (LaCM50) is found to reoxidize stoichiometrically (T_oxd = 1,100 °C, ${\text{p}}_{{\text{CO}}_2}=0.4\hspace{0.17em}\text{atm}$) with almost 1.6 and 5 times higher O₂ and CO production than La_0.5Sr_0.5MnO₃ (LaSM50) and CeO₂, respectively (Fig. 5) (36).

Fig. 5.

Representative TGA curves of thermochemical CO₂ splitting of La_0.5Ca_0.5MnO₃ (LaCM 50) in comparison with La_0.5Sr_0.5MnO₃ (LaSM 50) and CeO₂. T_red and T_oxd are 1,400 °C [99.9995% Ar, flow rate 40 standard cm³/min (sccm)] and 1,100 °C (pCO₂ = 0.4 atm, flow rate 40 sccm), respectively. Green arrow indicates the point of CO₂ injection. Adapted with permission from ref. 36, copyright 2015, Royal Society of Chemistry. Black arrow indicates change in temperature with time.

As shown in Fig. 6A, in the experiments performed with a laboratory-fabricated furnace, O₂ evolution starts at 1,000 °C with LaCM50 and is complete within 30 min of reaching the plateau near 1,400 °C (99.9995% Ar). O₂ produced by LaCM50 (272 μmol/g) is 1.4 times higher than LaSM50 (193 μmol/g). In particular, LaCM50 produces nearly 42% of O₂ predicted theoretically. The amount of O₂ produced in TGA measurements is slightly higher than that obtained using the laboratory-fabricated furnace due to differences in the reactor dimensions, sample weight, and gas flow rate. Thus, differences in the heat transfer rate and mass diffusion occur, making only a qualitative comparison possible. Interestingly, thermochemical H₂O splitting carried out at 1,000 °C shows the amount of H₂ produced in a span of 100 min by LaCM50 (407 µmol/g) to be greater than that of LaSM50 (308 µmol/g) (Fig. 6B). However, H₂ production is not complete even after 100 min and exhibits an extended tail over a long duration (36). The slow kinetics of H₂ evolution can arise due to the decrease in chemical diffusivity, changes in the surface reaction constant, and steam concentration as well as the intrinsic thermodynamic driving forces (35). The superior activity of LaCM50 can be related to its crystal structure. Due to the smaller ionic radii of Ca²⁺ in comparison with Sr²⁺, the tolerance factor (τ) of LaCM50 (τ = 0.978) is lower relative to LaSM50 (τ = 0.996) which emulates higher structural distortion in LaCM50 and helps to produce a greater amount O₂ as well fuel (36). Thermodynamic analysis by Steinfeld and coworkers (37) based on the extraction of partial enthalpy and entropy for oxygen vacancy formation suggests a higher oxygen exchange capacity of LaCM_x than of LaSM_x and identifies LaCM40 as a promising material. Investigations on La_1-xBa_xMnO₃ perovskites shows the production of O₂ as well as fuel production to be similar or somewhat less than that of LaSM_x and LaCM_x perovskites (38).

Fig. 6.

Rate of (A) oxygen (T_red = 1400 °C) and (B) hydrogen (T_oxd = 1,000 °C) production of La_0.5A_0.5MnO₃ (A = Sr, Ca). A laboratory-fabricated furnace was used for this particular measurement. Adapted with permission from ref. 36, copyright 2015, Royal Society of Chemistry. Black arrows indicate change in temperature with time.

B-Site Substitution in Manganites.

Partially substituted La_1-xSr_xFeO₃ with transition metals such as Cr, Co, Ni, and Cu has been investigated for syngas production using CH₄ and H₂O as reactants (34, 39). McDaniel et al. (40) first reported production of nine times more H₂ by Al-doped LaSM_x than CeO₂ using two-step H₂O splitting. A recent investigation by Demont and Abanades (38) finds almost no increase in the performance of LaSM_x due to Al doping, whereas a significant increment is reported by Steinfeld and coworkers (37, 41). Systematic doping of Al from 25 to 50% to LaSM50 perovskites causes increasing oxygen nonstoichiometry and subsequent fuel production, although the CO:O₂ ratio decreases from 1.5 to 1.2 (42). The performance of the Al-doped sample is stable during multiple cycling. Ga³⁺- (up to 35%) and Sc³⁺- (up to 10%) doped LaSM50 has also been examined for the two-step process, as shown in Fig. S3 (T_red = 1,400 °C reached at a heating rate of 20 °C/min, 99.9995% Ar; T_oxd = 1,100 °C, ${\text{p}}_{{\text{CO}}_2}=0.4\hspace{0.17em}\text{atm}$). Increasing the amount of the +3 substituents (Al, Ga, and Sc) increases the oxygen nonstoichiometry and shows better performance than the parent LaSM50. The effect of the trivalent ions varies as Sc > Ga > Al (42). Overdoping decreases the fuel production, although CO production with 25% Ga substitution and 5% Sc substitution is, respectively, 1.5 times and 1.7 times that of the parent LaSM50. The Sc³⁺-doped perovskite shows the best performance due to the enhanced oxygen mobility related to the bigger size of Sc³⁺ and higher structural distortion is further applied for H₂O splitting (Fig. 7) (42, 43). The CO:O₂ ratio decreases upon substitution of 25% Ga (∼1.57) as well as 5% Sc (∼1.3) in comparison with the undoped perovskites (∼1.62). The unfavorable oxidation thermodynamics and slow oxidation kinetics could cause this discrepancy. Sintering of these oxide materials observed during high temperature cycling could be a reason for the slow oxidation rate. The production of O₂ starts at 900 °C and completes within 40 min after reaching 1,400 °C with the total production of 390 µmol/g (percent reduction = 67%). H₂ is detected immediately after the entrance of the H₂O vapor (T_oxd = 1,100 °C), with the amount produced in a span of 100 min being ∼250 µmol/g and with an extended tail due to the slow kinetics (Fig. 7B) (42).

Fig. S3.

Comparative study of (A) O₂ (T_red = 1,400 °C reached at a heating rate of 20 °C/min, 99.9995% Ar, flow rate 40 sccm) and (B) CO (T_oxd = 1,100 °C, ${\text{p}}_{{\text{CO}}_2}=0.4\hspace{0.17em}\text{atm}$, flow rate 40 sccm) evolution of La_0.5Sr_0.5M_xMn_1-xO₃ (M = Al, Ga, Sc; 100× = % of substituent). Reproduced with permission from ref. 42, copyright 2016, Royal Society of Chemistry.

Fig. 7.

(A) Oxygen (T_red = 1400 °C) and (B) hydrogen (T_oxd = 1,100 °C) evolution profile of La_0.5Sr_0.5Sc_0.05Mn_0.95O₃. A laboratory-fabricated furnace was used for this particular measurement. Reproduced with permission from ref. 42, copyright 2016, Royal Society of Chemistry.

Deml et al. (44) analyzed the performance of La_1-xSr_xMn_1-yAl_yO₃ perovskites correlating the oxygen vacancy formation energy using density functional theory calculations. Steinfeld and coworkers have shown that Al doping leads to superior oxygen nonstoichiometry and decreases the partial molar enthalpy of vacancy formation compared with the undoped manganites (37, 41). In particular, production of oxygen by Al-doped LaSM_x and LaCM_x is at least two times higher than that of the undoped manganites at lower oxidation temperature ranges (42). Al³⁺ and Sc³⁺ dopants are resistant to carbonate formation, whereas incorporation of Mg²⁺ shows greater resistance to sintering (38, 42, 45). Few other transition metal- (Fe, Co, and Cr) based perovskites also show good fuel production activity (46, 47).

Changing the Rare Earth in the A-Site.

Properties of rare earth manganites depend strongly on the rare earth ion in the A site (48, 49). Two series of perovskites, Ln_0.5Sr_0.5MnO₃ and Ln_0.5Ca_0.5MnO₃ (Ln = La, Nd, Sm, Gd, Dy, and Y), have been used for the two-step thermochemical process recently (50). As shown in Fig. S4, O₂ evolution increases on decreasing the size of rare earth ions from La to Y (T_red = 1,400 °C, 99.9995% Ar). Yttrium derivative shows the highest O₂ production, the O₂ released by YCM50 and YSM50 being 573 (74% reduction) and 481 μmol/g (77% reduction), respectively (50). The ability of oxygen vacancy formation is correlated with the tolerance factor (τ) of perovskites in Fig. 8. Decreasing the radius of the rare earth ions reduces τ, which in turn increases the lattice distortion with large tilting of the MnO₆ octahedra. A decrease in τ reduces the Mn–O–Mn bond angle as well as the spatial overlap of Mn e_g and O 2pσ orbitals, favoring oxide ions removal (51). Fig. 8A shows that YSM50, with a τ of 0.965, generates oxygen nonstoichiometry of 2.81, much higher compared with LaSM50 (τ = 0.996). YCM50 (τ = 0.948) exhibits a further improvement in oxygen nonstoichiometry, reaching 2.80 (Fig. 8B) (50). The O₂ evolution temperature decreases gradually by reducing the size of rare earth ions from La to Y in case of LnSM50 (Fig. S4A, Inset). Size mismatch between the Ln and A cations (termed as size variance factor, σ²) induces local disorder in the perovskite structure (52), which helps in the greater displacement of the oxygens from the mean position and can act as the driving force behind easy removal of oxygens at lower temperature. Notably, σ² is higher for LnSM50 derivatives than that of LnCM50 and highest for YSM50. Thus, in case of YSM50 (σ² = 15.6 $\times$ 10⁻³) reduction starts at ∼860 °C, whereas YCM50 (σ² = 5.6 $\times$ 10⁻³) starts to get reduced at ∼970 °C (50).

Fig. S4.

Thermogravimetric reduction profiles of (A) Ln_0.5Sr_0.5MnO₃ and (B) Ln_0.5Ca_0.5MnO₃ where Ln represents lanthanides: La (1), Nd (2), Sm (3), Gd (4), Dy (5), and Y (6). Reduction temperature is 1,400 °C (99.9995% Ar, flow rate 40 sccm). Histograms quantify the amount of O₂ produced by (C) Ln_0.5Sr_0.5MnO₃ and (D) Ln_0.5Ca_0.5MnO₃. A, Inset shows the mass loss profiles of Ln_0.5Sr_0.5MnO₃ with reduction temperature in which the blue arrow specifies the gradual decrease in reduction temperature while moving from La to Y. Adapted with permission from ref. 50, copyright 2015, Wiley-VCH.

Fig. 8.

Variation of O₂ production quantity and oxygen nonstoichiometry generated (3-δ) at 1,400 °C reduction temperature as a function of tolerance factor (τ) of (A) Ln_0.5Sr_0.5MnO₃ and (B) Ln_0.5Ca_0.5MnO₃ perovskites where Ln = La, Nd, Sm, Gd, Dy, and Y. Adapted with permission from ref. 50, copyright 2015, Wiley-VCH. Black arrows indicate variation of production of O₂ with tolerance factor of perovskites; blue arrows indicate variation of oxygen nonstoichiometry with tolerance factor of perovskites.

CO produced on splitting CO₂ (T_oxd = 1,100 °C, ${\text{p}}_{{\text{CO}}_2}=0.4\hspace{0.17em}\text{atm}$) is reported to increase with the decrease in the size of the rare earth ion, the highest amount being found with yttrium derivatives. The maximum amount of CO produced by YCM50 (671 μmol/g) at 1,100 °C is even higher than YSM50 (571 μmol/g). Interestingly the oxidation yield of YSM50 increases from 59 to 79% by decreasing the oxidation temperature from 1,100 °C to 900 °C (50). The fuel production activity of YSM50 has been tested during three different reduction/oxidation temperatures (Fig. S5). Production of 624 µmol/g and 418 µmol/g of CO occurred on T_red/T_oxd of 1,300 °C/900 °C and 1,200 °C/900 °C, respectively, highlighting YSM50 as a potential candidate for energy application (50).

Fig. S5.

(A) Thermochemical CO₂ splitting profile and (B) the corresponding histogram of Y_0.5Sr_0.5MnO₃ (YSM50) derived from TGA profile. T_red/T_oxd pairs are (1) 1,200 °C/900 °C, (2) 1,300 °C /900 °C, and (3) 1,400 °C /900 °C. Reduction and oxidation are performed under 99.9995% Ar and ${\text{p}}_{{\text{CO}}_2}=0.4\hspace{0.17em}\text{atm}$, respectively. Green arrow indicates the point of CO₂ injection. Adapted with permission from ref. 50, copyright 2015, Wiley-VCH. Black arrow indicates change in temperature with time.

The H₂O splitting activity tested in a tubular furnace also shows that the O₂ evolution of YSM50 begins at a lower temperature than YCM50 (Fig. 9A). Satisfactory H₂ production is also obtained with YSM50 and YCM50 with amounts of 320 μmol/g and 310 μmol/g, respectively, in a span of 140 min (Fig. 9B). The yield of H₂ during the first 140 min of the reaction arises from the fast kinetic regime, followed by a slower kinetic regime controlled by diffusion (29, 50).

Fig. 9.

(A) Oxygen (T_red = 1,400 °C) and (B) hydrogen (T_oxd = 1,100 °C) evolution profiles of Y_0.5A_0.5MnO₃ (A = Sr, Ca). A laboratory-fabricated furnace was used for this particular measurement. Adapted with permission from ref. 50, copyright 2015, Wiley-VCH. Black arrow indicates change in temperature with time.

Carbonate Formation.

Alkaline earth metals such as Sr and Ca are prone to carbonation during CO₂ exposure. SrCO₃ and CaCO₃ decompose at ≤940 °C, which indicates that weight gain of nonstoichiometric perovskites on passing CO₂ at lower temperatures (<900 °C) could be partly due to carbonate formation instead of oxidation alone (45). Segregated SrCO₃ and CaCO₃ on oxide surfaces are found in microscope images (45). In situ X-ray diffraction would monitor the real-time changes of oxides under CO₂ exposure (47). Measuring the produced gases with a gas sensor would be more conclusive. Al³⁺ substitution on La_1-xA_xMnO₃ (A = Sr, Ca) suppresses the carbonation significantly, as revealed by thermogravimetric analysis (Fig. S6) (45). Sc³⁺ substitution is also effective for this purpose, because it suppresses SrO segregation (42, 53).

Fig. S6.

Thermogravimetric (TG) curve shows the initial weight loss under Ar followed by exposure to CO₂ at 900 °C and finally ramped up to 1,350 °C. The weight gain during 900 °C oxidation and subsequent immediate weight loss is partly due to the carbonation and decarbonation process. Clearly it is higher for LSM40. La_0.6Sr_0.4MnO₃ and La_0.6Sr_0.4Al_0.4Mn_0.6O₃ are denoted by LSM40 and LSMA, respectively. Reproduced with permission from ref. 45, copyright 2015, Royal Society of Chemistry.

Reaction Thermodynamics.

From the above discussion it can be surmised that perovskite manganites and Zr-doped CeO₂ exhibit high oxygen yields (${\text{n}}_{{\text{O}}_2}$) at low T_red. In a practical scenario, η_{solar-to-fuel} depends extensively on ${\text{n}}_{{\text{H}}_2\text{O}}/{\text{n}}_{{\text{H}}_2},$ reduction enthalpy and entropy of oxide $\left({\mathrm{\Delta }\text{H}}_{\text{red}},{\mathrm{\Delta }\text{S}}_{\text{red}}\right)$, ${\text{C}}_{\text{p},\text{redox}}$, and $\mathrm{\Delta }\text{T}$ (Eqs. S9–S12). Besides ${\mathrm{\Delta }\text{G}}_{\text{f},{\text{T}}_{\text{oxd}}}^{{\text{H}}_2\text{O}}$, these parameters depend on ${\mathrm{\Delta }\text{H}}_{\text{red}}\hspace{0.17em}\text{and}\hspace{0.17em}{\mathrm{\Delta }\text{S}}_{\text{red}},$ which are inherent properties of the material. We find that (i) increasing Zr⁴⁺ content decreases both the ${\mathrm{\Delta }\text{H}}_{\text{red}}$ and ${\mathrm{\Delta }\text{S}}_{\text{red}}$ of CeO₂. ${\mathrm{\Delta }\text{H}}_{\text{red}}$ linearly decreases as δ reaches nearly 0.075–0.1, but for larger reduction extents it become constant (31, 54) (ii) La_1-xSr_xMnO₃ perovskites show lower ${\mathrm{\Delta }\text{H}}_{\text{red}}$ than CeO₂ for all compositions. (iii) Increasing Sr²⁺ content decreases the ${\mathrm{\Delta }\text{H}}_{\text{red}}$ and ${\mathrm{\Delta }\text{S}}_{\text{red}}$ of LaMnO₃ (35). (iv) Al³⁺ doping decreases the ${\mathrm{\Delta }\text{H}}_{\text{red}}$ of La_1-xA_xMnO₃ (A = Sr, Ca), in the higher δ range of 0.075–0.15 (41). On the one hand, decreasing ${\mathrm{\Delta }\text{H}}_{\text{red}}$ favors ${\text{n}}_{{\text{O}}_2}$, decreases T_red (decreases η_abs), and favors high η_{solar-to-fuel}. On the other hand, decreasing ${\mathrm{\Delta }\text{H}}_{\text{red}}\hspace{0.17em}\text{also}$ disfavors $\mathrm{\Delta }{\text{G}}_{\text{oxd}}$ (Eqs. S9–S12) and increases the need of ${\text{n}}_{{\text{H}}_2\text{O}}/{\text{n}}_{{\text{H}}_2}$ and $\mathrm{\Delta }\text{T}$ (low T_oxd), thereby lowering η_{solar-to-fuel}. Manganites favor the reduction thermodynamics whereas CeO₂ favors the oxidation thermodynamics. To design a material that satisfies both the reduction and oxidation thermodynamics seems difficult.

Reaction Kinetics.

Generally, the reduction profile is initially rapid followed by a long extended tail, independent of the oxide. This is determined solely by the heat transfer rate rather than by the kinetics of oxygen diffusion or surface reactions (26). In contrast, H₂O splitting kinetics varies with the oxidation thermodynamics $\left(\mathrm{\Delta }{\text{H}}_{\text{oxd}}^0\right)$ and the oxide composition. η_{solar-to-fuel} depends on the upper limit of thermodynamic fuel production, considering infinite amount of gas supply and reaction time, but in practice, however, t_red and t_oxd need to be optimized (55). (i) For CeO₂, t_oxd<< t_red and oxidation completes in 1–2 min whereas the H₂ production time of La_1-xSr_xMnO₃ increases monotonically with increasing the Sr content and can take even more than 1 h for x = 0.4, which results in deviation of H₂:O₂ from 2 (35). Near stoichiometric or slight excess ${\text{n}}_{{\text{H}}_2\text{O}}$ is enough to reach the target amount of H₂ for CeO₂, although it can be a few orders higher in the case of La_1-xSr_xMnO₃ and increases with x (Fig. S7) (35). Thus, CeO₂ has a high theoretical η_{solar-to-fuel} compared with La_1-xSr_xMnO₃ (x > 0.2) at all temperature ranges. Scheffe et al. (33) found that with excess ${\text{n}}_{{\text{H}}_2\text{O}}$ input and 100% heat recuperation manganites can surpass the efficiency of CeO₂. (ii) The CO production rate is sluggish on increasing Ca content, as in La_0.35Ca_0.65MnO₃ (36). (iii) Decreasing the size of the lanthanide retards the oxidation kinetics further. Slow oxidation of Y_0.5Ca_0.5MnO₃ decays the CO:O₂ significantly in comparison with La_0.5Ca_0.5MnO₃ (50). (iv) Substitution with +3 ions (Al, Ga, and Sc) slows down the oxidation kinetics, resulting in a CO:O₂ ratio less than 2 (42). It seems difficult to ascertain that slow oxidation is solely due to kinetic or thermodynamic limitations.

Fig. S7.

The molar quantity of H₂O vapor required (800 °C, ${\text{p}}_{{\text{CO}}_2}=0.2\hspace{0.17em}\text{atm}$) to fill up the oxygen nonstoichiometry created during reduction at 1,400 °C (10 ppm O₂) as computed for CeO₂ and La_1-xSr_xMnO₃ (x = 0–0.4) oxides. Reproduced with permission from ref. 35, copyright 2014, Royal Society of Chemistry.

Bulk oxygen diffusion and surface reaction are factors that control the reaction kinetics of mixed ionic–electronic conductors. The chemical diffusivity (D_chem, Eq. S13) and chemical surface exchange constant (k_s, Eq. S14) are measured by conductivity relaxation or potentiostatic step methods (56, 57). D_chem of CeO₂ is nearly $2\times {10}^{-5}$ cm²/s (1,073 K). Yasuda and Hishinuma (56) found that D_chem of La_0.8Sr_0.2MnO₃ ($\sim 8\times {10}^{-6}$ cm²/s at 1,173 K and $\sim 3\times {10}^{-5}$ cm²/s at 1,273 K) is lower than CeO₂. Belzner et al. (57) found an order of magnitude decrease in D_chem with increasing Sr content from 0.2 to 0.5 due to the oxygen activity variation. D_chem of La_0.5Sr_0.5MnO₃ is around ${10}^{-7}$ cm²/s (1,073 K), which renders the characteristic diffusion time of <1 s for the diffusion length (l) of 3–5 μm (diffusion time, $\mathrm{\tau }=l^2/4{\mathrm{D}}_{\mathrm{chem}}$). The oxidation rate is thus limited by surface reaction rather than bulk oxygen diffusivity. The variation k_s with Sr content remains inconclusive, k_s being sensitive to the gaseous atmosphere. Thus, k_s of Sm_0.15Ce_0.85O_1.925-δ in a CO–CO₂–Ar mixture is 40 times higher than in an H₂–H₂O–Ar mixture due to kinetic benefits of CO₂ splitting (58). Deposition of Rh on the surface or increasing the surface area enhances the surface reaction rate (59, 60).

Light and Heat Penetration.

Microporous structures (felt and monolith) show rapid oxidation due to high surface area. Average and peak η_{solar-to-fuel} of 0.4% and 0.7% have been obtained by monolith, porous CeO₂ for H₂O splitting (26), whereas CeO₂ felt gives 0.15% and 0.31% for simultaneous splitting of H₂O and CO₂ (27). These structures are opaque to incident radiation, causing a temperature gradient across the thickness of CeO₂ and retards the reduction rate. In contrast, macroporous structures (foam and honeycomb) have pores in the millimeter range through which deeper, volumetric absorption of solar radiation results in homogenous temperature distribution, although their low surface area retards the oxidation rate. Three-dimensional ordered CeO₂ macroporous structures with interconnected pores show a higher CO₂ splitting rate (>1.5 times) than a nonordered entity (61). To accommodate dual-scale porosity (millimeters and micrometers) reticulated porous ceramic (RPC) CeO₂ has been fabricated, resulting in the highest average and peak η_{solar-to-fuel} of 1.73% and 3.53% for CO₂ splitting (62, 63). The optical thickness of RPC (ε = 280 m⁻¹) is two orders lower than monolith structure (ε = 40,000 m⁻¹) (62). The higher mass loading leads to 17 times more CO production of RPC than the felt in spite of its slow oxidation rate (27, 62). Perovskite oxides of dual-scale porosity have been used in solid oxide fuel cells and have yet to be used for thermochemical splitting of H₂O/CO₂.

Isothermal Cycle

Single-step splitting of H₂O would seem to be the best pathway but the reaction is thermodynamically unfavorable below 4,300 K (1 bar) (9). Thermodynamic considerations demand T_red> T_oxd for a two-step process (7). Such cyclic rotation between T_red and T_oxd is accompanied by irreversible heat and time losses and creates thermal stresses on the system. In this context, it is desirable to use isothermal splitting of H₂O (ITWS) which depends on the large pressure swing in the gas composition between reduction and oxidation processes. Fig. S8A, based on Eq. 5, shows that an ideal isothermal cycle must operate either at high T_red, low reduction ${\text{p}}_{{\text{O}}_2}$ or at high ${\text{p}}_{{\text{H}}_2\text{O}}/{\text{p}}_{{\text{H}}_2}$ (64):

$${\text{T}}_{\text{iso}}=\frac{-\mathrm{\Delta }{\text{H}}^0}{\text{R}\left(\ln {\text{p}}_{{\text{O}}_2}^{1/2}-\ln \frac{{\text{p}}_{{\text{H}}_2\text{O}}}{{\text{p}}_{{\text{H}}_2}}\right)-\mathrm{\Delta }{\text{S}}^0}.$$ [5]

First, altering any one of these three parameters results in a penalty from others for maintaining η_{solar-to-fuel}, irrespective of the material or the reactor. Second, water thermodynamics dictates that the fuel production per cycle of ITWS is less than two-step splitting of H₂O (TSWS) (Fig. S8B). Third, ITWS requires a much higher ratio of fluid input (sweep gas and water vapor) to output products (O₂ and H₂) than TSWS, causing an energy penalty (65). A pumping process has been suggested as an alternative to sweep gas flow but experimental investigations are pending (64, 66). The calculated efficiency of ITWS remains lower than the TSWS (59, 64). ITWS has its own list of advantages. Due to the absence of sweeping between T_red and T_oxd, the efficiency of ITWS does not depend on solid-state heat recovery (SSHR), which not only makes the reactor concept simple but also releases mechanical and thermal stresses. Thermodynamic analysis of ITWS of CeO₂ suggests a η_solar-fuel of 10% (65). Time is the savior in ITWS considering the global reaction rates (55, 59). In this context, the H₂ production rate of CeO₂ at 1,500 °C (9.2 mL⋅g⁻¹⋅h⁻¹) is not less than the rate predicted for TSWS (59). Finally, high temperature operation of ITWS should minimize the kinetic barrier of oxidation (15). Recent findings of Davenport et al. (67) show that isothermal cycling of CeO₂ at 1,500 °C is limited only by thermodynamic constraints.

Fig. S8.

(A) Relationship between T_iso, molar ratio of H₂O to H₂ with ${\text{p}}_{{\text{O}}_2}$ for ideal isothermal cycle. Reproduced with permission from ref. 64, copyright 2014, Royal Society of Chemistry. (B) $\delta -T-{\text{p}}_{{\text{O}}_2}$ diagram of CeO₂. The gray region highlights the equilibrium nonstoichiometry not accessible during isothermal operation at respective temperatures and pressure. It shows the oxygen nonstoichiometry inaccessible in ITWS cycle of CeO₂ that is otherwise accessible in TSWS. Reproduced with permission from ref. 65, copyright 2013, American Chemical Society.

Poor reducibility restricts the application of CeO₂ at <1,500 °C. Sublimation of CeO₂ and radiative power loss hinder the application even at high isotherm (>1,500 °C) (55). La_1-xSr_xMnO₃ (x = 0–0.5) and YSM50 have been used for isothermal decomposition of CO₂ (ITCS) (Fig. 10) and their performance is superior to CeO₂ even at 1,300 °C (reduction under ${\text{p}}_{{\text{O}}_2}={10}^{-5}\text{atm}$; ${\text{p}}_{{\text{CO}}_2}=1.0\hspace{0.17em}\text{atm}$) (68). The first cycle establishes the equilibrium nonstoichiometry under CO₂ and from the second cycle onward the CO:O₂ ratio is nearly 2 (Fig. 10A). The low-temperature performance of manganites requires higher ${\text{p}}_{{\text{H}}_2\text{O}}/{\text{p}}_{{\text{H}}_2}$ and a greater need of fluid state heat recovery (FSHR) than CeO₂ (Eq. 5). Manganites are going to get more benefit from high temperature ITWS due to the more sluggish oxidation kinetics than CeO₂. Above 1,100 K, ITCS yields more fuel than ITWS (Ellingham diagram). Simultaneous variation of both ${\text{p}}_{{\text{O}}_2}$ and temperature is recommended independent of the materials. Neither TSWS nor ITWS near isothermal is estimated to score the highest η_solar-fuel (64).

Fig. 10.

(A) Isothermal CO₂ splitting profile of Ln_0.5Sr_0.5MnO₃ where Ln are Y (solid line) and La (dotted line) for isotherms of 1,400 °C (blue) and 1,300 °C (red). Reduction and oxidation are performed under pO₂ = 10⁻⁵ atm and pCO₂ = 0.4 atm, respectively. (B) Histogram quantifies the production of gases by Ln_0.5Sr_0.5MnO₃ in comparison with CeO₂ standard (1,400 °C). Adapted with permission from ref. 68, copyright 2016, American Chemical Society.

Solar Reactor Technologies

Use of the solar thermal route for H₂O/CO₂ splitting crucially depends on the availability of suitable solar reactors. Construction of solar reactor to satisfy the operational need of different kind of cycles, materials, and temperature regimes has been discussed elaborately in a few reviews (9, 69, 70). In the case of solar reactors implemented for nonstoichiometric oxide-driven, nonvolatile, two-step cycles, the solar configuration must consist of three basic parts: concentrators, receiver, and reactor. The use of an indirect and direct receiver reactor is discussed in detail in Supporting Information. The oxides can be placed either as regular monoliths (structured reactors) or assembled in a packed/fluidized bed (nonstructured reactors) to promote uniform heat and mass transfer as discussed in Supporting Information (2, 66, 71, 72). Most of the reactor concepts are devoid of SSHR facilities. Ferrites have been investigated extensively using solar irradiation and there are a few attempts to use CeO₂ in solar reactors. Perovskite oxides have not been executed so far in solar reactors for two-step H₂O/CO₂ splitting purposes.

Future Outlook and Conclusions

The preceding discussion of the solar thermochemical route of generating hydrogen by splitting water should suffice to convince one that it is indeed a viable method for this important process. There are many challenges and opportunities in this area. Thus, it would be necessary to carry out a large-scale experiment on generating hydrogen by the low-temperature, multistep thermochemical cycle based on Mn(III)/Mn(II) oxides. Investigation of an energy-saving pathway for NaMnO₂ conversion to Mn₃O₄ is desirable. This is because of the attractive temperature at which H₂ is obtained.

It would be desirable to set up a pilot plant for generating CO and H₂ from CO₂ and H₂O, respectively. Syngas so produced can be useful for generating organics. The applicability of the two-step or the isothermal cycle depends on SSHR and FSHR, respectively. A reactor that minimizes heat loss through radiation or conduction and avoids a quartz window is desirable. Rare earth manganites would be ideal for the low-temperature reduction process. However, the oxidation thermodynamics of ceria is much superior. It may be useful to explore other oxide materials for the two-step process for a tradeoff between the reduction and oxidation reactions. Thermodynamic analysis of various oxides should be carried out for preliminary materials screening. In practice, efficiency should be calculated taking the reaction kinetics into account. The use of reactant gases should be minimized or their separation from product gases should have to be energy-efficient. Successful integration of vacuum pumping of the reactor should be undertaken. Clever design of solar reactors and use of porous materials can overcome heat transfer-limited reduction. Rapid oxidation kinetics and longevity of materials need to be demonstrated for a large number of cycles. Near isothermal performance with the optimization of temperature difference and of the partial pressures of gases should be carried out, especially with manganite-based perovskites.

Instead of NaOH, the use of Na₂CO₃ for the production of H₂ from redox active ${\text{MnFe}}_2{\text{O}}_4$ has been suggested by Tamaura et al. (11) as shown in Eqs. S1 and S2:

$$2{\text{MnFe}}_2{\text{O}}_4+3{\text{Na}}_2{\text{CO}}_3+{\text{H}}_2\text{O}\left(\text{g}\right)\to 6\text{Na}\left({\text{Mn}}_{1/3}{\text{Fe}}_{2/3}\right){\text{O}}_2\left(\text{s}\right)+3{\text{CO}}_2\left(\text{g}\right)+{\text{H}}_2\left(\text{g}\right)$$ [S1]

$$6\text{Na}\left({\text{Mn}}_{1/3}{\text{Fe}}_{2/3}\right){\text{O}}_2\left(\text{s}\right)+3{\text{CO}}_2\left(\text{g}\right)\to 3{\text{Na}}_2{\text{CO}}_3\left(\text{s}\right)+2{\text{MnFe}}_2{\text{O}}_4+0.5{\text{O}}_2\left(\text{g}\right).$$ [S2]

The thermochemical cycle-based Mn₃O₄/MnO oxides are stated in Eqs. S3–S6:

$$2{\text{Mn}}_3{\text{O}}_4+3{\text{Na}}_2{\text{CO}}_3\to 4{\text{NaMnO}}_2\left(\text{s}\right)+2\text{MnO}+{\text{Na}}_2{\text{CO}}_3+2{\text{CO}}_2\left(\text{g}\right)$$ [S3]

$$2\text{MnO}+{\text{Na}}_2{\text{CO}}_3+{\text{H}}_2\text{O}\left(\text{g}\right)\to 2{\text{NaMnO}}_2\left(\text{s}\right)+{\text{CO}}_2\left(\text{g}\right)+{\text{H}}_2\left(\text{g}\right)$$ [S4]

$$6{\text{NaMnO}}_2\left(\text{s}\right)+{\text{ayH}}_2\text{O}\left(\text{l}\right)+\left(3+\text{b}\right){\text{CO}}_2\left(\text{g}\right)\to 3{\text{Na}}_2{\text{CO}}_3\left(\text{aq}\right)+{\text{aH}}_{\text{x}}{\text{MnO}}_2\cdot {\text{yH}}_2\text{O}\left(\text{s}\right)+{\text{bMnCO}}_3+{\text{cMn}}_3{\text{O}}_4$$ [S5]

$${\text{aH}}_{\text{x}}{\text{MnO}}_2\cdot {\text{yH}}_2\text{O}\left(\text{s}\right)+{\text{bMnCO}}_3\to \left(2-\text{c}\right){\text{Mn}}_3{\text{O}}_4\left(\text{s}\right)+{\text{ayH}}_2\text{O}\left(\text{g}\right)+{\text{bCO}}_2\left(\text{g}\right)+0.5{\text{O}}_2\left(\text{g}\right),$$ [S6]

where a + b + 3c = 6 and (4 − x)a + 2b + 8c = 18.

The metal oxide (MO_oxd) reduces to the metal or to a lower valent metal oxide (MO_red) (Eq. S7) with the release of O₂(g) during the endothermic step, whereas in the next step it gets reoxidized on reaction with H₂O (Eq. S8), releasing stoichiometric amount of H₂(g):

$$\text{Endothermal}\hspace{0.17em}\text{Step}:{\text{MO}}_{\text{oxd}}\to {\text{MO}}_{\text{red}}+\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$2$}\right.{\text{O}}_2$$ [S7]

$$\text{Exothermal}\hspace{0.17em}\text{step}:{\text{MO}}_{\text{red}}+{\text{H}}_2\text{O}\to {\text{MO}}_{\text{oxd}}+{\text{H}}_2.$$ [S8]

Knowledge of the partial molar enthalpy $\mathrm{\Delta }{\text{H}}_{\text{redox}}\left(\mathrm{\delta }\right)$ and the partial molar entropy $\mathrm{\Delta }{\text{S}}_{\text{redox}}\left(\mathrm{\delta }\right)$ of oxygen vacancy formation as a function of δ allows one to deduce oxygen nonstoichiometry under different ${\text{p}}_{{\text{O}}_2}$ and T by Eq. S9:

$$\mathrm{\Delta }\text{G}\left(\mathrm{\delta },\text{T}\right)=\mathrm{\Delta }{\text{H}}_{\text{redox}}\left(\mathrm{\delta }\right)-\text{T}\mathrm{\Delta }{\text{S}}_{\text{redox}}\left(\mathrm{\delta }\right)=-\frac{1}{2}\mathrm{RT}\text{ln}({\text{p}}_{{\text{O}}_2}^{\text{eq}}\left(\mathrm{\delta },\text{T}\right).$$ [S9]

The Gibbs free energy of oxidation ($\mathrm{\Delta }{\text{G}}_{\text{oxd}}$) of the reduced oxides is obtained by using the partial oxygen molar free energy ($\mathrm{\Delta }\text{G}$) as shown in Eq. S10. Oxidation is thermodynamically favorable only when $\mathrm{\Delta }{\text{G}}_{\text{oxd}}-\mathrm{\Delta }{\text{G}}_{\text{f},{\text{T}}_{\text{oxd}}}^{{\text{H}}_2\text{O}}$ < 0. The equilibrium H₂ yields (${\text{n}}_{{\text{H}}_2}$) and the molar ratio of the oxidant (${\text{n}}_{{\text{H}}_2\text{O}}$) needed for that purpose are obtained from a knowledge of $\mathrm{\Delta }\text{G}$ and ${\mathrm{\Delta }\text{G}}_{\text{f},{\text{T}}_{\text{oxd}}}^{{\text{H}}_2\text{O}}$ (Eq. S11):

$$\mathrm{\Delta }{\text{G}}_{\text{oxd}}=\frac{{\displaystyle {\int }_{{\mathrm{\delta }}_{\text{i}}}^{{\mathrm{\delta }}_{\text{f}}}\mathrm{\Delta }\text{Gd}\mathrm{\delta }}}{{\mathrm{\delta }}_{\text{i}}-{\mathrm{\delta }}_{\text{f}}}$$ [S10]

$$\mathrm{\Delta }\text{G}\left({\mathrm{\delta }}_{\text{f}},\text{T}\right)=\mathrm{RT}\ln \left(\frac{{\text{n}}_{{\text{H}}_2\text{O}}{\text{K}}_{\text{w}}}{{\text{n}}_{{\text{H}}_2}}\right).$$ [S11]

The solar to fuel conversion efficiency is defined by Eq. 2 (main text) where HHV stands for higher heating value of the fuel produced. Furthermore, the solar input energy (${\text{Q}}_{\text{solar}}$) is mainly expenses for heating water (first term), heating the redox material from T_oxd to T_red (second term), creating oxygen nonstoichiometry (third term) and depends on solar energy absorption efficiency of receiver reactor (${\mathrm{\eta }}_{\text{abs}}$) as shown in Eq. S12:

$${\text{Q}}_{\text{solar}}=\frac{\mathrm{\Delta }{\text{H}}_{{\text{H}}_2\text{O},298\text{K}\to {\text{T}}_{\text{oxd}}}{\text{n}}_{{\text{H}}_2\text{O}}+{\text{n}}_{\text{redox}}{\displaystyle {\int }_{{\text{T}}_{\text{oxd}}}^{{\text{T}}_{\text{red}}}{\text{C}}_{\text{p},\text{redox}}.\text{dT}}+\mathrm{\Delta }{\text{H}}_{\text{red}}\mathrm{\delta }}{{\mathrm{\eta }}_{\text{abs}}}.$$ [S12]

Taking the ideal solution behavior and charge neutral condition, the ambipolar diffusion coefficient can be expressed as

$$\stackrel{\sim }{\text{D}}=\frac{\left({\text{c}}_{\text{ion}}+{\text{c}}_{\text{el}}\right){\text{D}}_{\text{ion}}{\text{D}}_{\text{el}}}{{\text{c}}_{\text{ion}}{\text{D}}_{\text{ion}}+{\text{c}}_{\text{el}}{\text{D}}_{\text{el}}}.$$ [S13]

Surface exchange constant (k_s) can be expressed as function of oxygen flux ($J_O$) across the reaction surface and the difference in concentration of oxygen across the solid-gas interface:

$${\text{J}}_{\text{O}}=-{\text{k}}_{\text{s}}\mathrm{\Delta }{\text{C}}_{\text{O}}.$$ [S14]

Various Solar Rectors

Concentrators of tower and dish optical configurations are mostly used for high-temperature applications. The receiver system exchanges heat with the thermal fluid (indirect process) or acts as a single unit with the reactor to volumetrically absorb the solar radiation on the oxides (direct process). The oxides can be placed either as regular monoliths (structured reactors) or assembled in a packed/fluidized bed (nonstructured reactors) to promote uniform heat and mass transfer.

Structured Reactors.

Adjustment of solar radiative power and switching of reactant gases happens depending on the reduction or oxidation process in a single-chamber reactor. A modular dual chamber has been implemented further for simultaneous production of O₂ and H₂ (69). Metal oxide can be coated on monolithic supports but the use of support decreases the effective mass loading and encounters the side reaction, which leads to the use of nonsupported monolithic structure.

The monolithic solar cavity reactor is the simplest configuration that allows the pinging of solar irradiation through a quartz window into an insulated cavity coated with monolith while performing both the steps in a single chamber by switching the reactant gases. The cavity reactor replicates black-body behavior by multiple reflections in inner reactor walls. Porous monolith CeO₂ cylinder splits H₂O and CO₂ in cavity receiver (26, 27). The cross-sectional view of the reactor shows the entrance route of radiation from solar simulator through a quartz windowed aperture and the outlet for product gases is axial to the inlet of reactant gases. Chueh et al. (26) performed 500 H₂O splitting cycles, obtaining efficiencies of net and peak 0.4% and 0.7%, respectively, for the first 23 cycles. During reduction the heating gradient builds from the outer to inner shell of the cavity, rate limits the reduction. Heat losses through conductive and radiative transfers contribute to lower efficiency (27, 63). No SSHR technique is associated with this type of setup. To achieve SSHR, a counter rotating-ring receiver–reactor–recuperator (CR5) was investigated. The reactor consists of a set of rings coated with active oxides. Simultaneous reduction (exposed to solar irradiation) and H₂O splitting occur in two opposite quarters of each ring, while at the same time the remaining two quarters promote the heat exchange with their neighboring rings, which rotate in the opposite direction. The ring traveling toward the oxidation zone transfers its heat to its adjacent neighboring ring entering the reduction zone (71). CeO₂ has been investigated in the CR5 reactor, although the reactor for scale-up operation fails shortly (71).

Nonstructured Reactors.

The simplest of this type of reactor is a packed bed catalyst residing in a quartz tube placed in focus of a solar concentrator and heated under sweeping of reactant gases during a two-step cycle. In a simple aerosol-based design the sample fed into the vertical tube reactor, allowing it to reach the down-hot reaction zone gravimetrically, while the inert gas flow helped to reduce the mass transfer limitations (69). The tube wall is directly irradiated by solar energy and transfer heat to the reactive site via conduction and radiation processes. This concept lacks inbuilt material recirculation techniques and SSHR implementation.

An internally circulating fluidized bed reactor contains a center draft tube through which the redox particles flow from bottom to top by the force exerted by the sweep gas and falls back through the annulus. Although solar energy focuses on the top part of the oxide bed, circulation of the oxide particles transfers heat to the lower section as well. Switching from inert gas to H₂O takes place during shuttling from the reduction cycle to the fuel production cycle (72).

Another reactor concept is the “moving particle packed bed solar reactor.” The H₂O splitting here occurs at the bottom chamber of reactors, followed by the stepwise lifting of the oxidized particles to the top of the upper reduction chamber using elevator and rotating cases successively to perform the reduction under solar illumination entering through a quartz window (66). The reduced particles then move downward to the oxidation zone via a stationary ceramic screw, which also simultaneously transfers heat to the upward-moving oxidized particles and facilitates SSHR. The spatial separation of O₂ from H₂, temperature, and pressure are achieved using this setup. Further modification of this setup by integrating vacuum pumping and cascaded pressure reduction for removing O₂ has been suggested (69).