Integration of thermochemical water splitting with CO₂ direct air capture

While the chemistry of all 3 steps in the Mn–Na–CO₂ cycle has been established (15), several major challenges remain in coupling the cycle with a solar thermal power source and the DAC of CO₂: 1) solar energy is inherently intermittent, restricting the amount of time high-temperature reactions can be performed, and 2) DAC of CO₂ requires the utilization of massive airflow. Here, we a provide a solution to these issues (summarized in Fig. 3). Heat for the cycle is supplied by a solar collector field that provides superheating steam at 1,000 °C. Concentrated solar thermal power is useful for generating high-quality heat for water-splitting cycles (22, 23). However, the diurnal cycle of available solar energy restricts the amount of time a water-splitting reactor can be held at high temperatures on a daily basis. Conveniently, the combined water splitting and CO₂ DAC process has 2 distinct operating regimes that pair well with this diurnal constraint. While high-quality heat is required for the high-temperature steps, sodium extraction is exothermic and time-consuming if dilute atmospheric CO₂ is used as a CO₂ source. Thus, cyclic quasibatch operation (SI Appendix, Fig. S2 shows a schedule of operation) where the high-temperature steps are performed during the day and Na⁺ extraction is performed at night enables effective utilization of reactor time despite the limitations imposed by a diurnal power source. CO₂ DAC requires large flowrates of air due to the low concentration (∼400 ppm) of CO₂. This leads to a correspondingly large power requirement. We address this issue by utilizing a solar updraft tower (24, 25). These systems have been proposed since the late 1990s for generating power from solar heating via thermal updraft through a power-generating turbine suspended in a large tower surrounded on ground level by a large solar collection basin (Fig. 3, Left). While this technology usually shows low efficiency in producing electricity (24), it is an excellent source of high flowrates of air without the corresponding electricity costs. Large thermal reservoirs are typically included in the base of updraft towers such that the airflow is relatively constant 24 h a day (24). This allows the tower to provide airflow for CO₂ capture at night, while also producing electricity during the day, when large air flowrates are not required. In general, diurnal systems suffer from low throughput and low efficiencies due to poor process intensification (26). We stress that this diurnal quasibatch operating schedule is only 1 possible implementation of this concept. With the implementation of a high-temperature heat storage system (27) and/or a system for the recirculation of solids (13) the cycle could be executed with higher frequencies or even continuously. We evaluate only our diurnal process in this work in order to simplify the necessary process assumptions.

The integration of these 2 passive solar technologies with the Mn–Na–CO₂ water-splitting cycle yields a process ideal for coupling DAC of CO₂ and water splitting. High-temperature steps that produce hydrogen and oxygen operate during the daytime when concentrated solar energy is available with steam as a heating medium (Fig. 3, Top Right). The heat exchange/separation system (Fig. 3, Bottom Right) receives either dilute streams of H₂ and CO₂ in steam (during the H₂ evolution step) or O₂ in steam (during the thermal reduction step) from the TWS reactor. These steps are separated by either the introduction of Na₂CO₃ or sodium extraction, such that these 2 gas streams are produced in distinct time periods throughout the cycle. This gas stream is first cooled to ∼600 °C in a heat exchanger that provides heat to dry Na₂CO₃ solutions produced during sodium extraction. This 600 °C stream is then introduced to a steam turbine that cools the stream further to 100 °C and produces electricity. A water knockout vessel is used to condense the majority of the steam and separate the final gas mixture (H₂ and CO₂ or O₂). The condensed water can then be reused as a source of steam such that each cycle consumes only the stoichiometric amount of water for water splitting. The use of steam as a reactant, heat-transfer medium, and carrier gas significantly simplifies the process as compared to other water-splitting cycles that require inert carrier gases (28, 29) and makes leveraging waste heat (via steam turbine) possible. At night, liquid water is supplied to the water-splitting reactor and a constant flow of air from the solar updraft tower (heat captured during the day by the solar tower allows operation at night; 24-h operation of solar towers is already proven in other applications) is introduced in order to extract sodium from α-NaMnO₂ via CO₂ absorption. At the end of the night the resulting Na₂CO₃ slurry is removed and dried during the day using excess heat from the TWS reactor effluent. The dried solid is then reintroduced to the TWS reactor between thermal reduction and H₂ evolution via solid–gas fluidization. This design yields a process that requires only water, air and sunlight to produce a stream of concentrated CO₂ and H₂ for further upgrading, a stream of pure oxygen, and electricity. It should be noted that pure H₂ and CO₂ streams can also be produced with this TWS if necessary.

We designed and fabricated an integrated reactor for the Mn-Na-CO₂ TWS cycle to demonstrate its feasibility. In our previous work, we elucidated the chemistry of each step in Mn–Na–CO₂ TWS in separate reactors on a test-tube scale (∼200-mg solid) (15). A key challenge in the implementation of this cycle is the drastically different operating conditions, i.e., solid–gas phase reactions at 850 °C for hydrogen and oxygen evolutions, and solid–liquid phase Na⁺ extraction and CO₂ capture at 90 °C. The fabricated reactor eliminates the necessity of movement of solids with a capacity of handling ∼10-g solid, which is a 50-fold increase from our initial work (SI Appendix, Fig. S3, details provided in Experimental Procedures). The reactor consists of an alumina vessel with multiple inlets and outlets to control the flow of liquids and gases into and out of the reaction zone. We demonstrate below that all 3 steps in the Mn–Na–CO₂ TWS cycle can be conducted in the fabricated reactor with close to stoichiometric H₂ and O₂ yields in multiple cycles.

During hydrogen evolution, a stoichiometric mixture of Mn₃O₄ and Na₂CO₃ (2:3 molar ratio) is initially heated to 850 °C under an inert gas flow and then under steam to form NaMnO₂, CO₂, and H₂. Prior to the introduction of steam (Fig. 4 A, i), the only gas-phase product formed as the solid mixture is heated to 850 °C is CO₂. This reaction leads to the formation of a 2:1 ratio of α-NaMnO₂ and MnO as we have shown in our previous work (15). No oxidation state change of manganese occurs in this step, as Mn^III and Mn^II in Mn₃O₄ form α-NaMnO₂ and MnO, respectively. Upon introduction of steam to the solid mixture at 850 °C at the end of 5 h (Fig. 4 A, ii), MnO is oxidized by water in the presence of unreacted Na₂CO₃ to form additional α-NaMnO₂, and a mixture of H₂ and CO₂ with a molar ratio of 1:1.

Na⁺ intercalated in the MnO_x layers of α-NaMnO₂ must be removed before Mn₃O₄ can be thermally reduced to complete the TWS cycle. Sodium extraction is performed by cooling the solids produced in the hydrogen evolution step to 90 °C, followed by the introduction of water to the reactor via the liquid inlet and bubbling of 5% CO₂ in N₂ via the gas inlet into the liquid–solid slurry. Initially complete CO₂ absorption is observed (Fig. 4B), demonstrating the high binding affinity of Na⁺ to CO₃²⁻ and facile sodium extraction from the birnessite phase formed upon contact of α-NaMnO₂ with liquid water (30). A detectable amount of CO₂ is observed after 2.3 h in the breakthrough curve, followed by a quick increase of the CO₂ concentration to the level in the feed. Periodic pH measurements of the slurry during extraction reveal that the slurry is very alkaline in nature, with an initial pH of 12.8, that decreases gradually as CO₂ in adsorbed into the liquid. This indicates that Na⁺ is gradually extracted from the birnessite phase by CO₂. The onset of incomplete CO₂ adsorption occurs when the pH of the slurry decreases to ∼11.4, that corresponds to ∼63% of Na⁺ extracted based on the amount of CO₂ absorbed. The final pH of the slurry is ∼9.6, which is roughly consistent with the formation of a Na₂CO₃–NaHCO₃ buffer (31) after the majority of Na⁺ has been extracted from the solid. Upon the completion of sodium extraction, the Na₂CO₃–NaHCO₃ solution is extracted from the liquid outlet equipped with an inline filter.

The solid component remaining in the reactor from the sodium extraction step is dried and heated to 850 °C to regenerate Mn₃O₄. O₂ is produced during the temperature ramp and hold (Fig. 4C) due to the thermal reduction of manganese oxides produced in the previous step to Mn₃O₄ (15). Additionally, a negligible amount of CO₂ is also produced during this step via decomposition of MnCO₃ [formed during sodium extraction (15)] during heating. Higher amounts of CO₂ formation were observed in our previous work (15). The difference may be due to the significantly higher α-NaMnO₂ weight percentage in the slurry during sodium extraction (∼15 wt %) employed in this work, as compared to our previous work (∼5 wt %) (15). Sodium extraction takes place via competing redox extraction (2 NaMnO₂ + 4 H⁺ → Mn^IVO₂ + Mn²⁺ + 2 Na⁺ + 2 H₂O) and ion exchange (NaMn^IIIO₂ + H⁺ → HMn^IIIO₂ + Na⁺) where the redox mechanism dominates at low pH (32). MnCO₃ formation is attributed to the reaction between the aqueous Mn²⁺ formed via the redox mechanism and CO₂ (15). Thus, higher weight loadings of α-NaMnO₂ in the slurry suppress the redox mechanism by maintaining a higher pH, which in turn reduces the MnCO₃ formation. It is desirable for CO₂ evolution to only take place during the H₂ production step as mixtures of CO₂ and H₂ are valuable while mixtures of CO₂ and O₂ would require costly separation. At the end of the thermal reduction step, the reactor is cooled to ∼100 °C, followed by the introduction of a Na₂CO₃ solution. Evaporating the water yields a solid mixture of Na₂CO₃ and Mn₃O₄ ready for the hydrogen evolution step in the subsequent cycle.

Near-theoretical yields of H₂ and O₂ are produced from the Mn–Na–CO₂ TWS cycle in the fabricated reactor over the course of 6 consecutive cycles (Fig. 5A). A stoichiometric mixture of Mn₃O₄/Na₂CO₃ is charged in the reactor prior to the hydrogen evolution step of the first cycle, and no additional manganese oxide is added to or removed from the reactor in all consecutive cycles. Roughly 90% of the theoretical amounts of H₂ and O₂ are produced in each cycle (Fig. 5A) with no discernible decrease in production with increasing cycles. The lack of deactivation is in part due to the fact that complete phase change is involved in every step, e.g., the intercalation of Na⁺ into and extraction of Na⁺ from MnO_x layers (15). Thus, each cycle has a fresh start without any “memory” from previous cycles. Further, roughly 90% of the stoichiometric amount of CO₂ is produced in the hydrogen evolution step in each cycle (Fig. 5B), with a negligible amount of CO₂ produced in the thermal reduction step. Powder X-ray diffraction patterns of solids after all 3 steps in the 6th cycle were collected (SI Appendix, Fig. S4A) and the data confirm that the expected phases are formed based on our previous work (15). Analysis of the X-ray diffraction patterns of produced Mn₃O₄ at several points in the cycle (SI Appendix, Fig. S4B) reveals that the average particle size decreases slightly during the 6-cycle test. Typically, CaO/CaCO₃ particles in Kraft-based cycles increase in size dramatically over the course of many cycles, limiting practical reuse (13). The fact that our Na–Mn–CO₂ system does not suffer from the same issue may allow substantially longer times between the replacement of solids. Results from these cycling experiments demonstrate that the fabricated reactor can be used to facilitate the Mn–Na–CO₂ TWS cycle.

The alkaline nature of aqueous suspensions of α-NaMnO₂ (Fig. 4B) suggests that it could be an effective adsorbent for CO₂ DAC. To evaluate the feasibility of this we employ ambient air with ∼400 ppm of CO₂ as a CO₂ source to extract sodium from α-NaMnO₂ rather than 5% CO₂ that was used in our cycling test. The breakthrough curves for CO₂ absorption in several α-NaMnO₂ slurries with varying NaMnO₂ loadings (Fig. 6A) show an excellent initial performance, absorbing more than 75% of CO₂ from air. The fraction of sodium extracted by CO₂ before the air effluent contains more than 100 ppm of CO₂ increases (up to 72%) with the amount of α-NaMnO₂ in the slurry (Fig. 6B). This is likely a consequence of the lower initial pH of slurries with low weight loading of α-NaMnO₂ (0.2–0.8 wt %) caused by near-complete extraction of Na⁺ from the NaMnO₂ layers upon initial introduction of water. In contrast, ∼15 wt % of NaMnO₂ slurry in water is used in the cycling experiments. The primary reason to use a relatively low NaMnO₂ wt % slurry in the proof of concept CO₂ DAC experiments is to measure these breakthrough curves in reasonable timescales. The amount of CO₂ captured when the CO₂ concentration in the effluent recovers to 400 ppm is in all cases consistent with ∼90% of intercalated Na⁺ being converted to Na₂CO₃ (similar to the case when 5% of CO₂ is used). As the weight fraction of NaMnO₂ in the slurry increases, the fraction of Na⁺ extracted before the effluent air contains more than 100 ppm of CO₂ is expected to increase.

A preliminary technoeconomic analysis is performed on a large-scale design based on the proposed flowsheet (Fig. 3) on a plant that removes 25,000 metric tons of CO₂ from the atmosphere per year. The major unit processes used are 1) a water-splitting reactor, 2) a field of heliostats focused on a central tower, 3) a solar updraft tower to generate the airflow during Na⁺ extraction at night and provide electricity during the day, and 4) a series of steam turbines to utilize waste heat in cooling the water-splitting effluent. The system is simulated as a quasibatch system where a water-splitting reactor is cycled through all of the required steps of the cycle once a day, performing the high-temperatures step during the day and the low-temperature step at night (SI Appendix, Fig. S2). In order to determine the extent to which integration of the updraft tower and steam turbines affects the feasibilities of the plant, we consider 4 alternate cases. Case 1 considers a plant that is not integrated with solar airflow or steam turbines for power recuperation. Case 2 considers a plant that is integrated with a solar updraft tower for airflow and power generation. Case 3 considers the construction of a full plant with the implementation of a series of turbines but no solar updraft tower. Case 4 takes into account the implementation of both a solar updraft tower and a series of steam turbines. A detailed account of all 4 cases and all relevant assumptions are provided in SI Appendix, Fig. S5. For each case we evaluate the economic viability using an optimistic CO₂ price of $200/ton. Cases 1–3 exhibit negative net present values (NPVs) even at this quite optimistic CO₂ price, highlighting the importance of proper integration of the water-splitting cycle with other technologies for power generation. The NPV of case 4 is dependent on financial assumptions but is positive, assuming a CO₂ price of $200/ton. Thus, we focus on the fully integrated system for further analysis.

The total costs of the fully integrated system investigated in case 4 are dominated by initial capital costs, mainly the solar energy harvesting facilities (Fig. 7A). The combination of concentrated solar power, water splitting, CO₂ removal, and turbines to produce electricity makes this process different from other DAC processes in its reliance on several revenue streams (Fig. 7B). The generation of H₂, O₂, CO₂, and electricity contributes to the plant’s revenue, although the vast majority of the revenue is derived from the generation of electricity. As a result, the economic viability of the proposed process depends on the value of all these outputs (Fig. 7C). For example, assuming H₂, O₂, and electricity prices of $2/kg (33), $0.04/kg (34), and $0.167/kWh [average cost of electricity in California in June 2019 (35)], respectively, a CO₂ price of $83/ton is needed for the proposed plant to break even. This hypothetical price point is lower than previous reports (4, 12–14). This relatively low projected break-even CO₂ price illustrates the economic feasibility of such an integrated process, despite its dependence on the assumed electricity price (SI Appendix, Fig. S6). It is noted that no carbon subsidy is assumed in the analysis. Due to the nature of the proposed cycle, the degree of uncertainty of the technoeconomic analysis will be greater than those on DAC schemes leveraging years of pilot plant data and vast industrial knowledge on similar processes. Our process is leveraging 2 reasonably mature solar-to-power technologies (updraft towers and solar thermal concentrators coupled to heat engines) to offset the costs of CO₂ DAC via electricity generation; thus, the comparison to processes that aim only to capture CO₂ may not be a fair comparison. Although the current analysis focuses on CO₂ DAC, the proposed cycle can also use more concentrated sources of CO₂, which is expected to yield improved economics as less inlet gas flow is needed for a given amount of CO₂ captured. As the majority of the capital cost of the process is derived from the solar-harvesting facilities, it is expected that as these technologies mature and improve, these costs will decrease. There is precedence in literature for dramatic cost reduction after the construction of a pioneer plant as a result of improvements in manufacturing/experience in both solar-harvesting technologies (36) and for direct air capture (13).

The experimental test reactor is constructed from readily available parts. The reactor tube is an 8.5-in.-long closed-ended Al₂O₃ tube with an outer diameter of 1 in. and an inner diameter of 0.75 in. The gas-handling and liquid-handling tubes are Al₂O₃ tubes with outer diameters of 1/4 and 1/8 in. and inner diameters of 3/16 and 1/16 in., respectively. All Al₂O₃ tubes are purchased from AdValue Technologies. An Inconel clad thermocouple (Omega) is threaded through the gas-handling tube in order to accurately measure temperatures in the reactor tube. All gas flows are controlled with Brooks mass-flow controllers. Steam is introduced via a Cole Parmer syringe pump with a 60-mL syringe. All other liquid additions or extractions are performed using the reactor syringe port using a Corning 0.20-μm nylon filter as an inline filter. The reactor is heated by a clamshell furnace (Lingberg/Blue M) and the gas distribution section is heated to 150 °C to prevent condensation of steam. All heaters and mass flow controllers are controlled via an integrated LabView program made in-house. An online differentially pumped quadrupole mass spectrometer (Stanford Research Instruments RGA100) is used to quantify gas products and a 100-mL condenser suspended in an ice bath is positioned upstream of the residual gas analyzer (RGA) to remove excess water.

Cycling experiments are performed by first preparing a physical mixture of Mn₃O₄ and Na₂CO₃ (Sigma-Aldrich) via grinding with a mortar and pestle. This mixture is then loaded into the reactor with care so as not to clog the gas-handling tube with solids. The water-splitting step is performed by first heating the physical mixture to 850 °C under a flowrate of 50 sccm of nitrogen at a ramping rate of 10 °C/min. After the set temperature is reached, steam is introduced to the heated inlet gas line by syringe pump at a rate of 2 mL/h of liquid. Once the reaction is completed (hydrogen production ceases) the steam flowrate is stopped, and the reactor is cooled to room temperature under nitrogen.

Once cool, sodium extraction is performed. First ∼40 mL of deionized (DI) water is introduced to the reactor using the syringe port on the side of the reactor and the reactor is heated to ∼80 °C while 50 sccm of N₂ is bubbled through the solution. Once the reactor has reached the set temperature a small flowrate of CO₂ is introduced into the nitrogen carrier gas such that a 5% CO₂ in N₂ stream at 50 sccm is fed to the reactor. The online RGA is used to measure CO₂ concentration during this time to establish a rough degree of extraction. Sodium extraction is performed in 3 steps starting from 5% CO₂ in N₂, 10% CO₂ in N₂, and finally pure CO₂ to insure complete sodium extraction. Between each step ∼40 mL of liquid is extracted, filtered, and replaced with the same volume of DI water using the inline filter attached to the syringe port. After the pH of the extracted solution is ∼7.5–8 the reactor is heated to and held at 100 °C for several hours to remove the remaining water.

Thermal reduction of the postextraction solid is performed by heating the reactor to 850 °C at a rate of 10 °C/min under a flow of 50 sccm of nitrogen. The RGA is used to quantify produced O₂ and the reactor is cooled once oxygen production ceases. After cooling to room temperature an aqueous solution of Na₂CO₃ is introduced to the reactor via the syringe port and the reactor is heated to 100 °C, removing water and leaving a physical mixture of Mn₃O₄ and Na₂CO₃ which is used in the next cycle.

Air-scrubbing tests are performed using a small 4-in. closed-end Al₂O₃ tube with an outer diameter of 1 in. and an inner diameter of 0.75 in. A 60-sccm flowrate of air is supplied with a small 5VDC diaphragm pump with a potentiometer to control the flowrate. The same RGA as above is used to track CO₂ concentration during Na extraction. A ppm level CO₂ calibration is obtained for the RGA by diluting atmospheric air with nitrogen in various ratios. NaMnO₂ used in tests was synthesized via solid-state synthesis using stoichiometric amounts of Mn₂O₃ (Sigma-Aldrich) and Na₂CO₃ heated in flowing air at 700 °C for 6 h.