Low-temperature, manganese oxide-based, thermochemical water splitting cycle

Abstract

Thermochemical cycles that split water into stoichiometric amounts of hydrogen and oxygen below 1,000 °C, and do not involve toxic or corrosive intermediates, are highly desirable because they can convert heat into chemical energy in the form of hydrogen. We report a manganese-based thermochemical cycle with a highest operating temperature of 850 °C that is completely recyclable and does not involve toxic or corrosive components. The thermochemical cycle utilizes redox reactions of Mn(II)/Mn(III) oxides. The shuttling of Na⁺ into and out of the manganese oxides in the hydrogen and oxygen evolution steps, respectively, provides the key thermodynamic driving forces and allows for the cycle to be closed at temperatures below 1,000 °C. The production of hydrogen and oxygen is fully reproducible for at least five cycles.

Thermochemical production of hydrogen and oxygen from water involves a series of chemical reactions that convert water into stoichiometric amounts of hydrogen and oxygen using heat as the only energy source. Thermochemical water splitting is of interest because it directly converts thermal energy into stored chemical energy (hydrogen and oxygen). Research on thermochemical water splitting cycles largely began in the 1960s and 1970s and involved nuclear reactors (1, 2) and solar collectors (3) as the energy sources. Numerous reviews of the thermochemical cycles proposed and experimentally investigated are available, e.g., (4). A large number of thermochemical cycles for splitting water has been proposed, and generally can be grouped into two broad categories: high-temperature two-step processes (5, 6) and low-temperature multistep processes (7, 8). One of us (M.E.D.) has conducted previous research on low-temperature multistep processes in the 1970s (9).

Low-temperature multistep processes, typically with the a highest operating temperature below 1,000 °C, allow for the use of a broader spectrum of heat sources, such as heat from nuclear power plants, and hence have attracted considerable attention. The majority of existing low-temperature processes produces intermediates that can be complex, corrosive halide mixtures. One of these processes, the sulfur-iodine cycle, has been studied extensively, and even piloted for implementation (8). This process produces strongly acidic mixtures of sulfuric and iodic acids that create significant corrosion issues, but requires only one high-temperature step at ca. 850 °C.

Two-step processes typically involve simpler reactions and intermediates, e.g., solid metal oxides, than the low-temperature multistep cycles. However, the temperatures required to close these types of cycles are well above 1,000 °C. Because of the requirement of high-temperature heat sources, these types of cycles have been investigated for use with solar concentrators (10, 11). These cycles typically consist of one step that involves the oxidation of a metal [such as zinc (12)] or a metal oxide [such as iron(II) oxide (6, 13)] by water to produce hydrogen, and a subsequent step to recover the starting material from its oxidized form (thermal reduction to produce oxygen).

The objective of our work is to create thermochemical water splitting cycles that involve non-corrosive solids and operate at below 1,000 °C. Essentially, we wish to create new cycles that take advantage of both the low-temperature multistep and high-temperature two-step cycles. Here, we show that more than two reactions will be necessary to perform themochemical water splitting below 1,000 °C, and then introduce a new thermochemical water splitting cycle that involves non-corrosive solids that can operate with a maximum temperature of 850 °C.

Results and Discussion

Thermodynamic Analysis of Two-Step Thermochemical Cycles Shows Temperatures Above 1,000 °C will be Necessary.

Consider a thermochemical cycle with the following two steps: (i) oxidation of a metal or metal oxide, referred to as Red, by water to its oxidized state, referred to as Ox, with the metal at a higher oxidation state; and (ii) thermal reduction of the Ox phase back to the Red phase, accompanied by the release of oxygen, as follows:

[1]

[2]

T₁ and T₂ are the reaction temperatures where ΔG = 0 for Eqs. 1 and 2, respectively. From the changes in the Gibbs free energy, ΔG_r = ΔH_r - TΔS_r, expressions relating T₁ to T₂ with either () or () held constant (isenthalpic or isentropic, respectively) can be obtained (Fig. 1 and SI Appendix).

Fig. 1.

Isenthalpic and isentropic lines relating the temperature of the hydrogen generating step and the thermal reduction step.

The thermodynamic data shown in Fig. 1 demonstrate that it is unlikely, if not impossible, to split water with a two-step cycle where there is complete conversion between the oxidized and reduced forms at below 1,000 °C. Typical differences in the formation enthalpy (ΔΔH) and entropy (ΔΔS) of the Red and Ox phases for metal/oxide systems (SI Appendix, Table S2) are found to be below 400 kJ/mol and 50 J/mol/K, respectively (conservative upper bounds). Therefore, only the region above both the lower black (entropic) and green (enthalpic) lines in Fig. 1 (blue shaded area) has practical relevance, as it is only in this region that both thermodynamic criteria are satisfied. The target region for low-temperature thermal water splitting (yellow shaded area, T₁, T₂ < 1,000 °C) has no overlap with the practically accessible region when using a two-step process. Additionally, existing two-step thermochemical water splitting cycles that have been reported previously all operate at above 1,000 °C (5, 6, 11, 12). The conclusion that can be derived from these results is that thermochemical water splitting cycles accomplished below 1,000 °C will require more than two steps, and is consistent with the previous thermodynamic analysis of Meredig and Wolverton (14).

Multistep, Thermochemical Cycle for Splitting Water Below 1,000 °C.

Thermochemical cycle overview.

We have developed a Mn-based, multistep, low-temperature water splitting cycle that has a highest operating temperature of 850 °C. The thermochemical cycle consists of four main steps (Fig. 2): (i) thermal treatment of a physical mixture of Na₂CO₃ and Mn₃O₄ to produce MnO, CO₂, and α-NaMnO₂ at 850 °C; (ii) oxidation of MnO in the presence of Na₂CO₃ by water to produce H₂, CO₂, and α-NaMnO₂ at 850 °C; (iii) Na⁺ extraction from α-NaMnO₂ by suspension in aqueous solutions in the presence of bubbling CO₂ at 80 °C; and (iv) recovery of Mn₃O₄ by thermally reducing the sodium ion extracted solid produced in step (iii) at 850 °C. The net reaction is the stoichiometric splitting of water to hydrogen and oxygen without any by-product. The incorporation and extraction of Na⁺ into and out of the manganese oxides are the critical steps in lowering the temperature required for both the hydrogen evolution and the thermal reduction steps (vide infra).

Fig. 2.

Schematic representation of the low-temperature, Mn-based thermochemical cycle.

Steps 1 and 2: Hydrogen evolution on Mn₃O₄/Na₂CO₃.

The presence of Na⁺ enables the oxidation of Mn²⁺ in Mn₃O₄ to Mn³⁺ by water, leading to the formation of α-NaMnO₂, CO₂, and H₂. In the absence of Na₂CO₃, oxidation of Mn₃O₄ to Mn₂O₃ is always thermodynamically unfavorable (ΔG > 0) (Fig. 3A, top dotted line) (15). The introduction of Na₂CO₃ drastically changes the thermodynamics of the oxidation reaction; the ΔG of reaction decreases with increasing temperature and becomes negative around 250 °C (Fig. 3A, solid line). These thermodynamic estimates are consistent with our experimental observations in that water does not react with Mn₃O₄ in the absence of Na₂CO₃ at 850 °C, and the amount of D₂ obtained by reacting D₂O (D₂O is used instead of H₂O to enhance the signal-to-noise ratio in the product detection and quantification) with the Mn₃O₄/Na₂CO₃ mixture at 850 °C is equivalent to the amount that would be expected when Mn²⁺ is totally converted into Mn³⁺ (Fig. 3B).

Fig. 3.

Addition of sodium carbonate is essential for low-temperature water decomposition on Mn₃O₄. (A) Thermodynamic estimates show that Na₂CO₃ enables oxidation of Mn₃O₄ by H₂O. ΔG for α-NaMnO₂ is adopted from Azad and Sreedharan (25), and the remainder of thermodynamic data are taken from ref. 26. (B) Production of D₂ and CO₂ in the oxidation of Mn₃O₄ to form α-NaMnO₂ by D₂O in the presence of Na₂CO₃ at 850 °C.

We hypothesize that Na₂CO₃ extracts the Mn³⁺ from Mn₃O₄ at 500–850 °C to form α-NaMnO₂, CO₂, and MnO [Fig. 2, step 1; note that CO₂ is observed while heating the Mn₃O₄/Na₂CO₃ mixture prior to exposure of water (Fig. 3B)]. Hydrogen is then formed from the water oxidation of MnO at 850 °C in the presence of Na₂CO₃ (Fig. 2, step 2). We confirmed that this step can occur independently at 850 °C (SI Appendix, Fig. S2). The ΔG for both steps decreases with increasing temperature and becomes energetically favorable above 400 °C (Fig. 3A). Upon the introduction of D₂O, a sharp peak indicating the release of CO₂ is observed. In contrast, the rate of D₂ evolution increases slowly after the D₂O introduction, and reaches a plateau after approximately 30 min (SI Appendix, Fig. S3 shows reproducibility). The drastically different kinetics for the evolution of CO₂ and D₂ suggest that step 2 is not an elementary step. The stoichiometry of the proposed reaction predicts that two-thirds of the α-NaMnO₂ and CO₂ should be formed via step 1 and the remaining third via step 2. These amounts are experimentally confirmed by the 2∶1 (± 15%) ratio of the amount of CO₂ evolved before and after the introduction of D₂O (Fig. 3B).

Further support for this reaction pathway is provided by the identification of the reaction intermediate MnO by powder X-ray diffraction (XRD) measurements. The XRD pattern of the solid obtained after the hydrogen evolution reaction (Fig. 2, steps 1 and 2) contains α-NaMnO₂ and a hydrated product α-Na_0.7MnO_2.14 [Fig. 4, i; α-NaMnO₂ can form α-Na_0.7MnO_2.14 when exposed to water (16)]. Importantly, the XRD pattern of the powder collected after step 1 clearly shows the presence of MnO in addition to the peaks attributed to α-NaMnO₂ and α-Na_0.7MnO_2.14 (Fig. 4, ii). The diffraction peaks corresponding to MnO persist after α-NaMnO₂ is fully hydrolyzed in the presence of CO₂ (Fig. 4, iii); the hydrolysis process of α-NaMnO₂ is discussed further below). Furthermore, after annealing the hydrolyzed sample at 180 °C in Ar, the only sharp peaks are those from MnO (Fig. 4, iv). The identification of the reaction intermediate MnO strongly supports our proposed reaction pathway.

Fig. 4.

Powder X-ray diffraction patterns used to identify solids phases in hydrogen evolution steps. (i) Solid collected after the hydrogen evolution step (after step 2); (ii) solid collected after reacting Mn₃O₄ with Na₂CO₃ at 850 °C (after step 1); (iii) sample from (ii) hydrolyzed in an aqueous suspension in the presence of CO₂ at 80 °C for 3 h; and (iv) sample from (iii) annealed at 180 °C in Ar for 1 h.

Step 3: Na⁺ extraction of α-NaMnO₂.

Na⁺ can be efficiently extracted from α-NaMnO₂ via hydrolysis in the presence of CO₂. Na⁺ extraction is a critical step in closing the low-temperature thermochemical cycle, since α-NaMnO₂ cannot be thermally reduced below 1,000 °C (17). The compound α-NaMnO₂ is layered, with Na⁺ sandwiched between MnO₆ octahedral sheets (18). Water can intercalate into these sheets, expanding the distance between adjacent layers to form sodium birnessite (Fig. 5) (18), as evidenced by the disappearance of the diffraction peak at 16.7° in α-NaMnO₂ and the appearance of the 12.5° peak in birnessite (Fig. 5, i). The mobility of Na⁺ is greatly enhanced in birnessite compared to that in α-NaMnO₂ because the MnO₆ sheets are pillared by water, and therefore can easily be exchanged by other cations including protons (18). Complete Na⁺ extraction from α-NaMnO₂ by hydrolysis in acidic conditions to form protonic birnessite (H⁺ birnessite) has been reported (19). Here, this conversion is achieved by bubbling CO₂ through an aqueous suspension of α-NaMnO₂ at 80 °C for 3 h. A disproportionation mechanism has been proposed to explain the oxidation state change of Mn in α-NaMn(III)O₂ when it converts into the birnessite phase with an average oxidation state of 3.5–3.8 (18, 19):

[3]

The Mn(IV) remains in the solid birnessite phase while Mn(II) is generally believed to dissolve in the aqueous phase (18, 19). An insoluble Mn(II) salt or other compounds with Mn(II) are expected to form in the CO₂-assisted hydrolysis of α-NaMnO₂. Very weak and broad diffraction peaks corresponding to the MnCO₃ and Mn₃O₄ phases are present in the sample collected after hydrolysis of α-NaMnO₂ with CO₂ under ambient condition (Fig. 5, ii). However, characteristic diffraction peaks for MnCO₃ and Mn₃O₄ phases are observed after hydrolysis of α-NaMnO₂ with CO₂ under hydrothermal conditions (Fig. 5, iii). Accelerated crystal growth under hydrothermal conditions is most likely responsible for MnCO₃ and Mn₃O₄ crystals large enough to be detected by the diffraction measurements. The presence of the MnCO₃ and Mn₃O₄ phases is strong evidence to support the disproportionation mechanism (Eq. 3; implies the average oxidation state of Mn in all Mn-containing solids is still +3).

Fig. 5.

Powder X-ray diffraction patterns used to identify intermediate phases in hydrolysis of α-NaMnO₂. (i) Hydrolysis of α-NaMnO₂ in an aqueous suspension at 80 °C for 3 h; (ii) hydrolysis of α-NaMnO₂ in an aqueous suspension at 80 °C for 3 h with bubbling CO₂; and (iii) hydrolysis of α-NaMnO₂ in water vapor and CO₂ under hydrothermal conditions at 140 °C for 5 h.

Step 4: Oxygen evolution by thermal reduction of solids from Na⁺ extraction of α-NaMnO₂.

Thermal reduction of the mixture formed after sodium extraction of α-NaMnO₂ (Fig. 6A, i) in Ar at 850 °C recovers Mn₃O₄. The layered structure of protonic birnessite collapses upon heating to 180 °C in Ar, presumably yielding amorphous Mn(III,IV)O_x (Fig. 6A, ii). The amorphous phase persisted up to 500 °C, where broad and weak diffraction peaks of Mn₃O₄ begin to appear. These results are consistent with the temperature-programmed desorption profile of O₂ (Fig. 6B, ii), with the onset of the O₂ desorption peak at approximately 450 °C. The first oxygen desorption peak at approximately 565 °C from the mixture is attributed to the thermal reduction of MnO₂ to Mn₂O₃ (Fig. 6B, i), indicating the reduction of amorphous MnO₂ to Mn₂O₃. The diffraction peaks of Mn₃O₄ for the Na⁺ extracted mixture gradually grow more intense and narrow as temperature increases; however, no diffraction peaks corresponding to MnO₂ or Mn₂O₃ are observed throughout the temperature range tested. The oxygen desorption peaks for the Na⁺ extracted phase above 565 °C do not correspond to the desorption peaks from the reduction of Mn₂O₃ to Mn₃O₄ at 810 °C (Fig. 6B, i); these desorption events are attributed to the solid state reaction between amorphous Mn₂O₃ and MnO present in the mixture. The CO₂ desorption peak from the mixture appears in a similar temperature range as the decomposition of MnCO₃ to MnO and CO₂ (Fig. 6B, iii and iv), confirming the presence of MnCO₃. The XRD pattern for the sample after thermal reduction is almost identical to that of commercial Mn₃O₄, except for a very weak peak at approximately 16 °C corresponding to a trace amount of α-Na_0.7MnO_2.14.

Fig. 6.

Thermal reduction of Na⁺ extracted α-NaMnO₂. (A) X-ray diffraction patterns of (i) Na⁺ extracted α-NaMnO₂, and (i) after annealing at (ii) 180 °C, (iii) 300 °C, (iv) 400 °C, (v) 500 °C, (vi) 600 °C, (vii) 700 °C, and (viii) 850 °C in Ar for 1 h. (B) Temperature programmed reaction of (i) MnO₂, (ii) and (iii) Na⁺ extracted α-NaMnO₂, and (iv) MnCO₃.

Recyclability of the complete cycle.

The Mn-based thermochemical system shown here exhibits > 90% yield for both hydrogen and oxygen evolution and shows no sign of deactivation during five cycles (Fig. 7). The amount of O₂ released from a thermal reduction of a commercial, crystalline Mn₂O₃ to form Mn₃O₄ (Fig. 7, solid black circle) is identical to that released from the thermal reduction presented above, consistent with the recovery of Mn₃O₄ in the thermochemical cycle. Furthermore, the XRD pattern of the Mn₃O₄ recovered after the oxygen evolution is identical among the five cycles (SI Appendix, Fig. S4) and match the reference diffraction pattern.

Fig. 7.

Multiple cycles of the Mn-based thermochemical water splitting system.

The key feature contributing to the recyclability of the Mn-based system is the complete shuttling of Na⁺ into and out of the manganese oxides. The Na⁺ incorporation takes advantage of thermodynamically favorable reactions (Fig. 2, steps 1 and 2) to form α-NaMnO₂. The Na⁺ extraction step exploits the mobility of Na⁺ in the layered structure when intercalated by water, and is further enhanced by the presence of CO₂, which drives the equilibrium towards the mixture of protonic birnessite, Mn₃O₄, and MnCO₃. This mixture can be thermally reduced to Mn₃O₄ at 850 °C, closing the thermochemical cycle. Importantly, the trace amount of by-product formed by the incomplete Na⁺ extraction is reintegrated into the α-NaMnO₂ phase in the next cycle, avoiding the accumulation of a permanent, inactive phase.

Comparisons to other multistep thermal water splitting systems.

The highest temperature required for the Mn-based cycle is near those temperatures used with other low-temperature multistep water splitting cycles. Both the hydrogen and oxygen evolution steps are conducted at 850 °C in the Mn-based cycle, similar to the temperature for the highest temperature step in the piloted sulfur-iodine system (8).

The Mn-based system does not involve any corrosive materials, while almost all previously reported multistep low-temperature thermal water splitting cycles have toxic and/or corrosive intermediates in one or more of the steps in the cycle. Most notably, in each step of the sulfur-iodine cycle, there is at least one of the following chemicals: SO₂, H₂SO₄, I₂, and HI, which pose great environmental and engineering challenges (8).

Hydrogen production via the water oxidation of metal oxides with a spinel structure in the presence of sodium carbonate was pioneered by Tamaura et al. (20, 21). However, their cycle was not closed due to the use of sacrificial Fe₂O₃ to extract Na⁺ from sodium manganese iron oxide produced in the hydrogen evolution step (22). Sodium sources other than Na₂CO₃, such as NaOH, have also been employed to facilitate the oxidation of Mn²⁺ to Mn³⁺ in the water splitting step (23). The volatility of NaOH at > 800 °C and incomplete Na⁺ extraction by water to recover NaOH poses challenges to its implementation (17). Based upon the work of Tamaura et al., Sturzenegger et al. first pointed out the possibility of creating a Mn-based cycle that utilizes Na₂CO₃ rather than NaOH (23). Here, we have shown that a closed cycle of this type can be created.

Implementation issues.

There are a number of technical aspects of the Mn-based cycle proposed in this work that require solution prior to effective implementation. The kinetics of hydrogen evolution (Fig. 3B) will need to be increased. Here, no efforts have been made to control parameters such as solid particle size. Thus, there are several ways that the kinetics of hydrogen evolution may be enhanced, e.g., creation of small particles. The need for cooling the α-NaMnO₂ produced in the hydrogen evolution step from 850 °C to 80 °C—to perform the Na⁺ extraction in the aqueous phase before heating the Na⁺ extracted phase back to 850 °C—may impose a considerable energy cost. Moreover, Na₂CO₃ needs to be recovered from the aqueous solution in the Na⁺ extraction step. Further research in alternative Na⁺ extraction methods is certainly warranted. In addition, the energy cost for the separation of CO₂ produced in the water oxidation step from H₂ should also be taken into account in the overall efficiency analysis. Sturzenegger and Nuesch have performed an efficiency analysis for a cycle that does require heating and cooling solids to the temperatures used here (24). Their results suggest a process of this type could be feasible. Of course, higher energy efficiency will be achieved by educated reactor design with efficient heat recovery and recycling among different stages of the cycle (24).

Materials and Methods

Materials Preparation.

Mn₂O₃ (99%), Mn₃O₄ (97%), and Na₂CO₃ (99.5%) were purchased from Aldrich and used without further treatment. The mixture of Mn₃O₄ and sodium carbonate (molar ratio of 2∶3) for the hydrogen evolution step was prepared by mixing these two powders in an agate mortar under ambient condition. Sodium was extracted from α-NaMnO₂ by bubbling CO₂ (99.997%, 10 cc/ min) through an aqueous suspension of the powder (approximately 5 wt % of solid) for 3 h at 80 °C. Hydrothermal treatment of α-NaMnO₂ was carried out in an autoclave at 140 °C for 5 h with the mass ratio of α-NaMnO₂/H₂O/CO₂ (dry ice) being roughly 1/10/10. The compound α-NaMnO₂ was placed on a small vessel in the autoclave to avoid direct contact with liquid water. The powder for the oxygen evolution step was obtained by separating the solid by centrifugation and drying at 100 °C in air.

Reaction Tests.

All powders (approximately 200 mg) were pelletized (20–35 mesh) before being introduced to a quartz reaction tube with an alumina sheath. The pellets were supported between two layers of alumina sand (16 mesh) to avoid contact with the quartz tube. Hydrogen and oxygen evolution steps were tested using an Altamira flow reaction system (AMI-200), and the effluent gas stream was monitored by an on-line mass spectrometer (Dymaxion 2000). In the hydrogen evolution step, the mixture (Mn₃O₄/Na₂CO₃) was heated from room temperature to 850 °C at a ramp rate of 20 °C/ min under Ar (99.999%, 50 cc/ min) before water (D₂O) vapor was introduced by flowing the carrier gas through a bubbler (50 cc/ min) with D₂O at room temperature. D₂O was used instead of H₂O to obtain a better signal-to-noise ratio of the signal in the water splitting step (m/z = 4 for D₂ instead of m/z = 2 for H₂). Oxides recovered from the CO₂ treatment were heated up at a ramp rate of 20 °C/ min from room temperature to 850 °C in Ar atmosphere.

Characterization.

Powder X-ray diffraction (XRD) patterns were collected using a Rigaku Miniflex II diffractometer using Cu Kα radiation.