# A direct coupled electrochemical system for capture and conversion of CO2 from oceanwater

Sep 4, 2020

### Design and fabrication of the BPMED cell for CO2 capture from oceanwater

Figure 1a shows the schematic illustration of the BPMED cell for CO2 capture from oceanwater. The BPMED cell contained two oceanwater compartments separated by a bipolar membrane (BPM), two reversible redox-couple compartments, each separated from the oceanwater compartment by a cation exchange membrane (CEM), and two electrodes for electrochemical reactions. The electrochemical reactions at the electrodes, ionic transport across the membranes, and water dissociation at the BPM interface are illustrated in Fig. 1a. At the middle of the BPMED cell, a BPM that generates proton (H+) and hydroxide ion (OH) fluxes via water dissociation reactions at the BPM interface was used to convert the input oceanwater into output streams of acidified and basified oceanwater. The electrode solution, i.e., catholyte and anolyte, contained a reversible redox-couple solution, potassium ferro/ferricyanide (K3/K4[Fe(CN)6)]), and was re-circulated to minimize any polarization losses associated with concentration overpotentials at the electrodes. Two CEMs were then employed to charge balance the acidified or basified streams of oceanwater by selectively transporting cations from the anolyte or toward the catholyte, respectively. The electrode reactions in the BPMED cell were one electron, reversible redox reaction as the following:

$${mathrm{Cathode}}:;left[ {{mathrm{Fe}}left( {{mathrm{CN}}} right)_6} right]^{3 – } + e^ – to ;left[ {{mathrm{Fe}}left( {{mathrm{CN}}} right)_6} right]^{4 – }.$$

(1)

$${mathrm{Anode}}:;left[ {{mathrm{Fe}}left( {{mathrm{CN}}} right)_6} right]^{4 – } to ;left[ {{mathrm{Fe}}left( {{mathrm{CN}}} right)_6} right]^{3 – } + e^ -.$$

(2)

One unique advantage of this new BPMED configuration is that it can be employed and scaled up both in a single stack configuration or a multi-stack configuration without introduction of any unintended chemical reactions or any additional voltage losses. By contrast, the BPMED configuration in Supplementary Fig. 1a17 can only be employed in a single stack configuration with a untunable ratio of the CO2 capture rate and the H2 generation rate, while BPMED configuration as shown in Supplementary Fig. 1b15 can only be employed in a multi-stack configuration to minimize the voltage penalty associated with water-splitting reactions.

Figure 1b shows the experimental flow diagram of the electrochemical capture and conversion of CO2 from oceanwater. Dissolved gasses in the input oceanwater stream, e.g., O2 and N2, were vacuum stripped using three commercial membrane contactors (3M™ Liqui-Cel™ MM-0.5 × 1, each with a maximum operating liquid flow rate of 30 ml min−1) connected in a series prior to entering the acidification compartment. The acidified oceanwater was directed toward another series of three membrane contactors for removal of dissolved CO2 by a vacuum pump. A cold trap surrounded by dry ice was used to condense moisture from the gas output. The acidified oceanwater was then fed to the base compartment where the pH was retrieved close to the initial value and the effluent was disposed of as a waste in a collection tank.

### Polarization losses in the BPMED cell

The electrochemical performance of the BPM is the key to the operation of BPMED cell, and to understand the voltage loss across the BPM, a multi-physics model was used to simulate the voltage–current density characteristics, electrochemical potentials, and partial current densities carried by different ions in the system (Supplementary Note 1). Figure 2a indicates that the simulated and the experimental data showed good agreement throughout the entire current density range. The total current density (jtotal) equals to the sum of the current density carried by the hydrogen ion (left( {j_{{mathrm{H}}^ + }} right)), the hydroxide ion (left( {j_{{mathrm{OH}}^ – }} right)), and the sodium and chloride co-ions (left( {j_{{mathrm{Na}}^ + };mathrm{and};j_{{mathrm{Cl}}^ – }} right)) in the solution. Other counter ions also co-exist in the oceanwater but their concentration was too small to impact the total current density. The complete list of anions and cations in the synthetic oceanwater is provided in Supplementary Table 4. At low current densities, the leak current of Na+ and Cl was substantial due to the imperfect permselectivity32,33 of the cation exchange layer (CEL) and anion exchange layer (AEL) of the BPM. When the BPM voltage exceeded 0.4 V, the water dissociation reaction started to take place due to the increased electric field across the BPM interface, and at large BPM voltages, the water dissociation rate, i.e., (j_{{mathrm{H}}^ + }) and (j_{{mathrm{OH}}^ – }), became the dominating partial current density in the system. In this study, the operating current density of the device was set to >3.3 mA cm−2 and from Fig. 2b, >93% of the ionic transport was carried by (j_{{mathrm{H}}^ + }) and (j_{{mathrm{OH}}^ – }) with minimal contribution from co-ion crossovers.

Figure 2c shows the total cell voltage as a function of the operating current density using different electrode solutions and flow conditions. The thermodynamic limit of the total cell voltage (Vcell, ideal) in the BPMED cell can be expressed in the following equation:

$$V_{{mathrm{cell,}};{mathrm{ideal}}} = frac{{RT}}{F}left( {{mathrm{pH}}_{{mathrm{basified}}} – {mathrm{pH}}_{{mathrm{acidified}}}} right),$$

(3)

where R is the universal gas constant (8.3144 J K−1 mol−1), T is the temperature, F is the Faraday constant (9.6485 × 104 C mol−1), pHbasified is the pH of the solution in the basified compartment and pHacidified the pH of the solution in the acidified compartment. In comparison, the practical total cell voltage (Vcell, practical) can be expressed as the following equation:

$$V_{{mathrm{cell,}};{mathrm{practical}}} = , frac{{RT}}{F}left( {{mathrm{pH}}_{{mathrm{basified}}} – {mathrm{pH}}_{{mathrm{acidified}}}} right) + V_{{mathrm{BPM}};{mathrm{loss}}} + V_{{mathrm{CEMs}}} \ + V_{{mathrm{oceanwater}}} + V_{{mathrm{electrolyte}}} + V_{{mathrm{electrode}}},$$

(4)

where VBPM loss is the voltage loss across the BPM, VCEMs, is the voltage loss across the CEMs, Voceanwater and Velectrolyte are the voltage loss across the oceanwater and electrolyte compartment, respectively, and Velectrode is the voltage loss at the two electrodes. The dominating voltage penalty in the BPM-based electrodialysis cell originated from the water dissociation kinetics and polarization loss across the BPM. As shown in Fig. 2c, the voltage of the electrodialysis cell with the traditional 0.5 M Na2SO4 electrode solution was significantly higher than any of the redox-couple-based cell configurations in all current density ranges due to the required thermodynamic voltage window (1.23 V) for water-splitting as well as kinetic overpotentials for OER and HER. Eliminating the water-splitting reaction in the BPM-based electrodialysis cell by replacing the traditional electrode electrolyte with K3/K4[Fe(CN)6] redox-couple solutions significantly reduced Velectrode and hence reduced the total operating cell voltage.

Figure 2c also indicates that polarization losses associated with concentration overpotentials in the redox-couple compartments can be minimized by increasing the concentration and flow rates of the redox-couple solutions. The total cell voltage was very close to the voltage difference across the BPM at a concentration of 0.4 M and a flow rate of 40 ml min−1, particularly at the low current density regime. The main difference between (V_{{mathrm{cell,}};{mathrm{practical}}}) and (V_{{mathrm{cell,}};{mathrm{ideal}}}) was (V_{{mathrm{BPM}};{mathrm{loss}}}) when the rest of the voltage losses were minimized by the optimized cell design. In the linear region of voltage–current density curves, the discrepancy between the BPM voltage and cell voltage was primarily due to the resistance of the CEM and the 0.4 M K3/K4[Fe(CN)6] solution. The dominating voltage penalty in the BPM-based electrodialysis cell originated from the water dissociation kinetics and polarization loss across within the BPM. Similar or higher voltage losses across the BPM during operation were observed in previous reports32,33,34,35.

The calculated and the experimentally measured pH (Fig. 2d) as a function of the current density in the BPM electrodialysis cell showed excellent agreement throughout the whole range of applied current density. The contributions of co-ions transport, e.g., (j_{{mathrm{Na}}^ + }) and (j_{{mathrm{Cl}}^ – }), to the total current density were significant when the operating current density was lower than 0.4 mA cm−2. The water dissociation reaction, i.e., (j_{{mathrm{H}}^ + }) and (j_{{mathrm{OH}}^ – }), was the major charge carrier during the process at higher current densities. As shown in Fig. 2d, to attain the desired pHs at the acidified stream, the required current densities were higher for oceanwater that flowed at higher rates. The solution pH has a significant impact on the concentration of the dissolved CO2 and hence on the capture efficiency of the BPM electrodialysis cell. In the synthetic oceanwater, the concentration of the dissolved CO2 increases from 0.016 to 3.08 mM when the solution pH decreases from 8.1 to 4 (Supplementary Fig. 4). Therefore, to efficiently capture CO2 from oceanwater, the solution pH needed to be kept close to 4. Note that the pH of the acidified stream was dictated solely by the operating current normalized with the volumetric flow rate of the oceanwater (Supplementary Fig. 6). All the reported electrodialysis cells at near optimal operating conditions yielded a normalized operating current of 5.71 mA min ml−1.

During the experiment, the net ion movement between the electrolyte and the oceanwater replaced the K+ ions with the Na+ ions in the electrolyte, resulting Na3–/Na4–Fe(CN)6 solution in the electrode compartment that is free of K+ after a period of operation. The exchange of the cations in the redox-couple species did not impact the performance of BPMED. Because of the much higher concentration of Na+ (~0.4 M) in oceanwater relative to H+ at mild pH ~5 (10−5 M), the transference number was close to unity for Na+ across the CEM, and the transport of H+ to the electrolyte was negligible. In addition, the catholyte and anolyte were circulated during the operation. As a result, the change of the catholyte or anolyte pH was not observed under operating conditions. However, it is important to maintain the acidified oceanwater compartment at mild pHs so that minimal H+ transfer took place between the oceanwater and the catholyte.

### Performances of the BPM-based electrodialysis cell for CO2 capture from oceanwater

One critical metric for evaluating the performance of the BPMED cell for CO2 capture from oceanwater is the electrochemical energy consumption via electrodialysis. The electrochemical energy consumption is defined as the amount of electrical energy required (in kilowatt hour, kWh) for electrodialysis divided by the amount of captured CO2 (in mass, kg). Figure 3a shows the calculated electrochemical energy consumption as a function of the applied current density and the oceanwater flow rate. The experimentally measured voltage–current density characteristics of the cell with the optimized K3/K4[Fe(CN)6] electrode solution (Fig. 2c), the calculated pH-current density relations (Fig. 2d), and the CO2 concentration-pH equilibrium (Supplementary Note 2) were used to determine the electrical power consumption and the resulting dissolved CO2 in the oceanwater that can be captured using a traditional liquid–gas membrane contactor. Detailed calculation and flowchart outlining calculation steps are provided in Supplementary Note 3 and Supplementary Fig. 7, respectively. In this calculation, current densities greater than 0.4 mA cm−2 were used, where the water dissociation at the BPM interface dominated the ionic transports (Fig. 2b). As shown in Fig. 3a, at any given oceanwater flow rate, there is an optimal operating current density of the cell that would yield the lowest electrochemical energy consumption for CO2 capture. Improving the water dissociation kinetics at the BPM interface as well as lowering the series resistance at high current densities by improving the cell design would further lower electrochemical energy consumption for CO2 capture from oceanwater.

Another important parameter for the BPM-based electrodialysis device is the output rate of the captured CO2. Figure 3b shows the calculated rate of ideal CO2 output as a function of the applied current density and the oceanwater flow rate. The ideal CO2 output rate assumes that all the dissolved CO2 can be captured from the acidified oceanwater using membrane contactors. At any given oceanwater flow rate, the rate of CO2 output increased as the operating current density of the cell increased until a maximum rate for CO2 capture was reached. At a higher oceanwater flow rate, a higher operating current density was required to acidify the oceanwater to the pH at which most carbonate and bicarbonate ions were converted to dissolved CO2. Figure 3c indicates that, at a given oceanwater flow rate of 37 ml min−1, there is an optimal regime of operating current density between ~1.5 and ~3.5 mA cm−2, where the lowest electrochemical energy consumption was achieved. At the lower operating current density, the pH of oceanwater stream was not sufficiently low to convert the majority of DIC into dissolved CO2, which resulted in low capture efficiency and high electrochemical energy consumption. At higher operating current densities, while the oceanwater stream was sufficiently acidified, the increased voltage across the BPMED led to increased electrochemical energy consumption. The output CO2 flow rate increased linearly as a function of the operating current density until a turning point was reached, where the vast majority of DIC was converted into dissolved CO2. As a result, at an oceanwater flow rate of 37 ml min−1, the operating current density was set to 3.3 mA cm−2 to minimize the electrochemical energy consumption, while maximizing the output CO2 flow rate.

Figure 4a shows the experimentally measured CO2 capture rate in the BPMED cell, at an applied current density of 3.3 mA cm−2 (or an absolute current of 211.2 mA) and at an oceanwater flow rate of 37 ml min−1. At these operating conditions, the pH of the acidified stream was 4.7, the measured total voltage across the cell was at ~1 V (Supplementary Fig. 8), the DIC rate was calculated to be 2.8 sccm and the ideal CO2 output rate was 2.6 sccm (Supplementary Note 3). The initial total gas output was ~3 sccm, and stabilized at 2.1 sccm after 1 h of operation when the vacuum pressure was sufficiently low and the remaining air in the gas stream line (e.g., membrane contactors, cold trap, tubing) had been completely evacuated.

To accurately quantify the CO2 concentration, the output gas stream was diluted with pure N2 gas in a mixing chamber before introducing to a gas chromatograph (GC). Figure 4b indicates that during the first hour of the experiment, the output CO2 gradually increased to a constant concentration of 93%, while O2 and N2 decreased to a stable concentration of 1.5% and 5.5%, respectively. The initially low concentration of CO2 was attributed to the incomplete removal of dissolved air from the input oceanwater and the incomplete evacuation of air from the gas stream line. After 1 h of operation, the remaining O2 and N2 measured in the output gas were likely from the slight leak in the membrane contactors because the ratio between O2 and N2 was roughly 1:4. In the absence of a vacuum stripping stage prior to the BPMED unit, the captured gas contained more than 30% of N2 and O2 gas mixtures (Supplementary Fig. 9).

Figure 4c shows the CO2 capture efficiency and the membrane contactor efficiency over the course of 2 h. The CO2 capture efficiency is defined as the measured rate of the gaseous CO2 output divided by the rate of total DIC in the oceanwater that was introduced into the BPMED cell. The total DIC in the synthetic oceanwater was 3.12 mM (Supplementary Note 2). The membrane contactor efficiency is defined as the measured rate of the gaseous CO2 divided by the rate of dissolved CO2 gas present in the acidified stream at the corresponding pH and the oceanwater flow rate. After the output gas reached a stable rate and composition, the CO2 capture efficiency and the membrane contactor efficiency were 71% and 76%, respectively. The small discrepancy between the CO2 capture efficiency and the membrane contactor efficiency suggests that at these operating conditions (at an applied current density of 3.3 mA cm−2, or an absolute current of 211.2 mA and at an oceanwater flow rate of 37 ml min−1), most of the DIC in the oceanwater was converted to dissolved CO2. It is important to note that three membrane contactors in series (3M™ Liqui-Cel™ MM-0.5×1) were used at the inlet (for dissolved gas removal, such as O2 and N2, from fresh oceanwater) and the outlet (for CO2 capture) of the acid compartment of the BPMED cell, each with a maximum operating liquid flow rate of 30 ml min−1, to separate dissolved CO2 from oceanwater at efficiencies reported herein. Using only 1 membrane contactor also allowed for the separation and removal of CO2, but at a lower efficiency (Supplementary Fig. 10), and two series of membrane contactors were at least required to capture CO2 at efficiencies of more than 70% (Supplementary Fig. 11). The highest experimentally recorded CO2 capture efficiency in this study was 77% at an oceanwater pH of 3.7, where ~99% of the DIC was converted to dissolved CO2 (Supplementary Fig. 12). Adding more membrane contactors may increase the air leakage into the membrane contactors and will not significantly improve the capture efficiency because the maximum removal of dissolved CO2 by vacuum stripping through this type of contactor is typically ~80%15, which is also consistent with the specification provided by the manufacturer. Figure 4d shows the electrochemical energy consumption for CO2 capture in kWh kg−1 and kJ mol−1. After 1 h of operation, the electrochemical energy consumption was stabilized to 0.98 kWh kg−1 or 155.4 kJ mol−1 of CO2. The low electrochemical energy consumption was achieved by eliminating the voltage loss at the electrodes.

In order to sustain the CO2 capture from oceanwater at scale, the oceanwater waste from the BPMED should be returned to the ocean at restored alkalinity to allow for continuous uptake of atmospheric CO2. The removal of CO2 in the oceanwater allows the carbonate species to re-equilibrate and prompts the oceanwater pH to adjust according to the new equilibrium condition. At an oceanwater flow rate of 37 ml min−1, a current density of 3.3 mA cm−2, the pH of the acidified stream was 4.7, and increased to 5.3 after CO2 capture at an efficiency of 71%. According to the rate of hydroxide generation in the base compartment, which is equal to the rate of proton generation in the acid compartment, the output of oceanwater in the basified stream should have a pH of 10.46. However, experimental measurements showed a pH of 8.5, close to the original oceanwater pH. This discrepancy was attributed to the presence of non-negligible amounts of Mg2+ and Ca2+ ions in the oceanwater (Supplementary Table 4), which preferentially reacted with OH and formed white precipitates of divalent hydroxides and carbonates, as observed during the experiments. In the absence of Mg2+ and Ca2+, the basified stream would reach a pH of ~10.46 (Supplementary Note 4). Using simplified oceanwater without Mg2+ or Ca2+, the measured pHs from the acidified stream and basified stream showed good agreement with the calculated values. By contrast, when synthetic oceanwater that contained Mg2+ and Ca2+ was used, the basified stream exhibited smaller pH increase than the calculated value due to the preferential reaction between OH and Mg2+/Ca2+ (Supplementary Fig. 13). Softening the oceanwater feed prior to BPMED will solve this issue and restore the oceanwater alkalinity but the environmental impacts of returning decarbonized oceanwater at pH >10 with the same salt level is not well understood presently. Subsequent processes will need to be developed and implemented to levitate any impact on oceanic life.

The operating parameters of the proof-of-concept device was constrained by the laboratory hardware. As a result, at an oceanwater flow rate of 37 ml min−1, the operating current density was set to 3.3 mA cm−2 to minimize the electrochemical energy consumption, while maximizing the CO2 output rate. In a scaled-up device, a much higher oceanwater feed rate would require a higher operating current density to achieve the optimal pH in the acidified compartment and to capture the majority of the CO2. The higher operating current density in a practical device reduces the BPM cost per kilogram of CO2 captured in the system but increases the electrochemical energy consumption due to the increased polarization loss. The trade-off between the electrochemical energy cost and the membrane cost in the overall capture cost of CO2 in the BPMED system at different current densities was analyzed in Supplementary Note 5.

### Electrochemical conversion of CO2 in a vapor-fed device

Vapor-fed CO2R cells have several advantages for CO2R such as the ability to overcome mass transport limitations of CO2 solubility in aqueous electrolytes36. Examples of vapor-fed CO2R cells have exhibited large current densities, increased selectivity, and high single pass conversion rates for CO2R on Cu-based electrodes37,38,39,40,41,42. To test the proof-of-concept design, a vapor-fed CO2R cell similar to other previously reported cells was employed38. The outlet CO2 stream from the BPMED cell was directly fed through tandem vapor-fed cells, the first for oxygen reduction reaction (ORR) and second for CO2R. The first ORR pre-electrolysis cell was used to eliminate any residue O2 from flowing into the vapor-fed CO2R cell (Supplementary Fig. 17), as any O2 will be selectively reduced and lower the FE of the CO2R cell. Two types of CO2R catalysts were used in this study. The Cu catalyst was deposited on a gas diffusion layer (GDL, Sterlitech PTFE) by magnetron sputtering, and the Ag catalyst was deposited on carbon-based GDL (Ion Power, Sigracet 29 BC) by drop casting Ag nanoparticles43. A traditional three-electrode configuration with the CO2R catalysts as the working electrode, Pt mesh as the counter electrode, and Ag/AgCl (1 M KCl) as the reference electrode in the anolyte reservoir, was used44. Bulk electrolysis was performed at an applied potential of −1.14 V vs. the reversible hydrogen electrode (RHE) for 2 h with an average CO2 flow rate of 2.2 sccm from the BPMED system and the anolyte flow rate of 5 ml min−1. The resulting current densities ranged from 53 to 77 mA cm−2 (Supplementary Fig. 18). During the bulk electrolysis, 0.5 ml aliquots of electrolyte was taken every 10 min to monitor the liquid product distribution over time. As shown in Fig. 5a, using the Cu catalyst, 73% of the electrons were selective toward CO2R products, while ~20% went toward HER. The remaining 7% of electrons were lost due to the re-oxidation of liquid products to CO2 in the anode chamber, or from being absorbed in the anion exchange membrane (AEM) and GDL44. The product distributions for both liquid and gas products are relatively stable during the 2 h bulk electrolysis with an average single pass conversion rate of ~6.7% for Cu. A range of reduction products were obtained using the Cu catalyst, and producing a single desired product either by improving the selectivity of the reaction, or by downstream product separation, is a key area of CO2R, which is currently under intense research and development. Ag electrodes have been used for the selective reduction of CO2 to CO at high FE in both aqueous and gas diffusion configurations43,45,46,47,48. As shown in Fig. 5b, at −0.6 V vs. RHE, the FE toward CO increased to >90% for a large portion of the bulk electrolysis, with HER as low as 5%, and an average single pass conversion rate of ~9.7%. The operating current density of the Ag-based vapor-fed CO2R cell ranged from 7.6 to 11.7 mA cm−2 (Supplementary Fig. 19). The selectivity of CO dropped, while HER increased with time due to flooding of the GDL. In a separate experiment, we have tested a CO2 feed directly from the outlet of the BPMED with O2 present using a Ag-based CO2R cell and found that more than 80% of the electrons went toward ORR (Supplementary Fig. 20). As a result, CO2 gas feeds free of O2 impurities are important for high CO2R selectivity in lieu of the parasitic loses from ORR.

To operationally match the CO2 flux in the BPMED and the vapor-fed CO2R cell, the ratio between the CO2 reduction current density and the CO2 capture current density was estimated to be 4, assuming and a 6-electron process with 100% CO2 utilization and selectivity (Supplementary Note 6). For example, if the vapor-fed CO2 cell operates at 300 mA cm−2, the areal-matched BPMED should operate at ~75 mA cm−2. Our cost analysis indicated that there is a trade-off between the cost of the electrochemical energy consumption and the membrane cost, and the BPMED is cost-effective when operates at current density of >60 mA cm−2 (Supplementary Note 5 and Supplementary Fig. 14). The successful coupling of the electrochemical CO2R cell and the CO2 capture unit showed that CO2 captured from oceanwater could be a potential carbon feedstock for renewable generation of fuels or chemicals.