Design and construction of seawater flow reactor for CO2 capture and in-situ conversion
Simulated seawater was prepared by dissolving 35.5 g of simulated seawater salt (Instant Ocean®, see table S1 for detailed chemical compositions) in 1 L 18 mega ohm deionized water. The chemical composition of this simulated seawater solution was designed to simulate the chemical, electronic, and ionic properties of natural seawater (Table S1). Although Na+ and Cl− ions are the primary carriers of the ionic current in this solution, it is essential for protons generated at the upstream photoanodes to be transported downstream photocathodes for consumption. To address these issues, we developed an original vortex reactor design and 3D-printed the flow reactor (Fig. 1a, photographs in Fig. S1a and S1b).
The 3D-printed vortex reactor design features a periodic array of photoanode-photocathode pairs, each configured back-to-back and aligned in parallel to each other. In each pair, a Si photocathode and a BiVO4 photoanode are positioned back-to-back, maintaining a distance of 2 cm apart. The BiVO4 photoanode is 2 cm in length and 1 cm in width, while the Si photocathode is 0.5 cm in length and 0.5 cm in width. Ag-Au/CrOx catalysts are deposited on the Si photocathode for CO2R (fabrication details in the Methods section). Each BiVO4-Si pair is oriented at an angle towards the sun and arranged horizontally in parallel to optimize light capture. This design, where parallel pairs of photoelectrodes are utilized, can be indefinitely repeated for manufacturing scale-up. Electrical contacts are made between the back-to-back photoanode and photocathode pairs, while the ionic currents are between the upstream photoanode and the downstream photocathode. The seawater flow over each photoelectrode pair is identical.
Seawater flow was designed to sweep across the surface of BiVO4 photoanode, carrying H+, HCO3−, and CO2(aq) from the photoanode surface to downstream photocathodes that facilitate CO2(aq) capture and in-situ CO2 reduction. The seawater flow field, confirmed through computational simulations, is characterized by laminar seawater flow across the BiVO4 photoanode surface, which oxidizes seawater to O2 (Fig. 1b). Consideration was also given to the time needed to generate CO2(aq) and its limited lifetime when its concentration exceeded its equilibrium concentration in the bulk solution. To manage this, a vortex flow is created for CO2(aq) to pass over Si photocathode surfaces (Fig. 1c) mounted and electrically wired to the back of a BiVO4 photoanode (shown in supporting movie S1). This design enables the direct transport of the generated CO2(aq) to the Si photocathode without hindering light absorption. Moreover, this flow tunes the residence time of CO2(aq) on the surface of the Si photocathodes: an excess of CO2(aq) can undergo subsequent reduction into valuable chemical products before it diffuses into the bulk seawater solution or reacts with alkaline species (OH−, CO32−) near the Si photocathode. Gas bubbles are continuously emitted from photoelectrode surfaces (Fig. 1d, see supporting movie S2), underscore the importance of well-defined convective flow for solar-driven ocean-based CO2 capture and conversion.
The effect of proton convective transport on the photoelectrochemical CO2 reduction rate and the overall solar-to-CO conversion efficiency was investigated by systematically varying the flow velocities during PEC device operation under AM 1.5G illumination (100 mW/cm2). During the process of seawater oxidation at the BiVO4 photoanodes, protons are generated in quantities exceeding their equilibrium concentration in the bulk seawater solution. This leads to a reduction in the local pH and drives the conversion of HCO3− to CO2(aq) as it moves along the path of convective transport. The enhancement effects of convective proton flow were experimentally measured and then supplemented with analysis from numerical simulations.
In this study, flow rate was measured using the average volumetric rate at which seawater solution left the inlet nozzle, which was 0.3 cm away from the closest photoelectrode pair (Fig. S1c). To quantify the flow velocity, the volumetric flow rate was divided by the cross-sectional area of the tubing. The fuels produced in this system include CO and H2 gases, with liquid products falling below the detection limit, as confirmed by nuclear magnetic resonance (NMR) spectroscopy (Fig. S29). Analysis of the gas product composition at a flow velocity of 0 m/s, revealed that the Faradaic efficiency for CO and H2 are 3% and 97%, respectively. The low CO selectivity suggests that there is a preference for electrons to reduce of H+ to form H2 under static flow conditions, where [CO2(aq)] is expected to be around 2-µmol. Increasing the flow velocity to 0.77 m/s, resulting in a CO Faradaic efficiency increase from 3–19%. This significant increase in CO production underscores the crucial role that flow plays in both the in-situ generation and mass transfer of CO2(aq) from the upstream photoanode to the downstream Si photocathode for in-situ CO2 conversion (Fig. 2a). Under static flow conditions, CO generation over a period of 4.5 hours was determined to be 2.97 µmol per cm2 illumination area. However, CO generation over the same period significantly increased to 14 µmol/cm2, when the flow velocity was increased to 0.77 m/s (Fig. 2c). Other flow velocities also increased the CO production rate and Faradaic efficiency, as shown in Fig. S2. The increase in CO production can be attributed to the elevated [CO2(aq)] at the cathodic surfaces.
A similar trend was observed when measuring gaseous CO2 generation (Fig. 2b). Under static flow conditions, CO2 generation was determined to be 1.85 µmol per cm2 illumination area over a period of 4.5 h. When the flow velocity was increased to 0.77 m/s (see other flow velocities shown in Fig. S2), the amount of CO2 extracted over the same timeframe increased to 3.5 µmol/cm2. Excess CO2(aq) at the headspace-seawater interface may eventually be released as CO2 gas. Correspondingly, the STF efficiencies increased from 0.4–0.71% with the increase in flow velocity. This level of performance serves as our benchmark STF for BiVO4-based devices for unbiased CO2R (Table S6). The STF efficiency reached a maximum at a flow velocity of 0.77 m/s (Fig. 2d). This increase in efficiency highlights the importance of convective flow in improving ionic transport and minimizing concentration overpotentials, which are in systems without flow. The observed increase in CO2 generation supports our hypothesis that in-situ generated CO2(aq) has a sufficient lifetime to participate in CO2R downstream. To summarize, this BiVO4/Si vortex reactor in seawater demonstrates the first instance of utilizing an unbiased PEC tandem device to simultaneously capture and convert utilize dissolved inorganic carbon in seawater.
Construction of BiVO4 photoanode for extracting CO2 from seawater
BiVO4 was chosen to serve as the photoanode material for its ability to provide the high oxidation potential per charge transfer required for seawater oxidation. Thin-film BiVO4 photoanodes (Fig. 3a) were fabricated on fluorine-doped tin oxide (FTO) glass via a metal-organic decomposition method (see Methods with the photograph in Fig. S3). The thickness of the BiVO4 photoanode was approximately 200 nm (Fig. S4). X-ray diffraction (XRD) (Fig. S5) measurements were consistent with what has been previously observed in relevant studies13,26. Scanning electron microscopy (SEM) images (Fig. 3b) revealed the nanoporous morphology of the BiVO4 photoanode, which can increase the photoanode’s contact area with seawater, improve charge transfer efficiency, and induce surface pH variations. The presence of Bi and V was confirmed through SEM element mapping (Fig. 3c and 3d). To enhance oxygen evolution kinetics in seawater, NiFe(OH)x catalysts were deposited on the photoanode’s surface by dip coating27. A thin CrOx layer was then photodeposited onto the NiFe(OH)x catalysts to prevent chlorine oxidation and the subsequent corrosion of the BiVO4 photoanodes in seawater.
The PEC oxygen evolution measurements were performed under AM 1.5G illumination (100 mW/cm2) in seawater with a of pH 8.3 at ambient conditions. The BiVO4 photoanode with NiFe(OH)x/CrOx catalysts biased at 0.6 vs RHE exhibited the same onset potential of 0.2 V vs RHE but at a higher current density of 1.8 mA/cm2 than the 0.5 mA/cm2 without cocatalysts (Fig. 3e). The NiFe(OH)x catalyst increased the oxidation reaction’s rate, allowing for applied bias photon-to-current efficiencies (ABPE) of 0.51% to be achieved (Fig. S6). BiVO4 photoanodes with CrOx exhibited the same onset potential and saturated current density (Fig. 3f), indicating that the thin CrOx layer does not impede photogenerated hole charge transfer. Under AM 1.5G illumination, an O2 Faradaic efficiency of 94.5% at 0.6 V vs RHE over a 2 h period was achieved, suggesting that the majority of the photogenerated holes were utilized for oxygen evolution, with significant oxygen bubble formation also observed (Fig. S7). Additionally, the BiVO4 photoanode with NiFe(OH)x catalysts exhibited a robust photocurrent density after 10 h (Fig. 3g). The generation of hypochlorite (ClO−) was negligible, as confirmed with hypochlorite detection.28,29 Furthermore, the color of the seawater remained clear after 10 hours of operation (Fig. S8). The high O2 FE and the excellent stability were attributed to the CrOx layer, which may behave as a Lewis acid to effectively modulate the local reaction microenvironment of the NiFe(OH)x’s water-oxidation sites to effectively prevent chloride attack28,29.
pH map above photoelectrodes reveals spatial confinement effects of flowing seawater
The concentration of reactants is often used as a quantitative indicator for the limiting electrochemical currents. Applying this concept to our system, if diffusion were the predominate transport mechanism, bulk [CO2(aq)] in seawater would act as our indicator for the CO2(aq) flux, and consequently the CO2R partial current. In this study, it is important to note that the convective flux of in-situ generated CO2(aq) substantially contributes to the downstream CO2R currents. Calculations reveal that the bulk seawater [CO2(aq)] remains at the 0.11 mmol level near the Si photocathodes. In other words, the diffusive transport of mmol-level CO2(aq) does not sufficiently account for the measured CO2R partial current density of ~ 0.2 mA cm− 2. To elucidate the coupled processes of light-generated excess proton generation and acid-base reactive transport, we employed experimental pH mapping and numerical calculations. We took this approach to quantify and visualize both the concentration and flux of CO2(aq) that is otherwise difficult to measure.
To measure the pH spatial distribution in a seawater flow over the BiVO4 photoanode surfaces, we developed an original in-situ photo-fluorescence technique using confocal fluorescence spectroscopy (Fig. 4a). Carboxy SNARF-1 was selected as the fluorescent pH indicator due to its emission spectrum exhibiting a pH-dependent wavelength shift. pH at a given point was measured by calculating the fluorescence intensity at two different emission wavelengths (580 and 650 nm)30,31. To mitigate the potential influence of illumination (415 nm laser) on the pH indicator dye’s peak intensities and position, back-illumination was employed (photographs of the apparatus used in the pH measurement are shown in Fig. S9, and S10). The pH indicator was first calibrated using various standard solutions (Fig. S11). An immersion water lens (60 x) was used for high resolution spectra acquisition, and the Raman laser was focused on the top surface of BiVO4 photoanodes (Fig. 4b and Fig. S12). Spatial mapping revealed uniform pH of ~ 8.3 across the entire BiVO4 photoanode at open circuit potentials (Fig. 4c). This observation suggests that the introduction of flow confines the acidification process that the BiVO4 photoanode surface uniformly generate H+, whereas the pH of the seawater at > 100 µm above the photoanode surface plane remains identical to the inflow seawater pH. Each pH data point was determined from the pH-indicator emission spectra, with an emission peak at 650 nm indicating a pH value of ~ 8.35 (Fig. S11), which is consistent with the seawater’s pH of 8.30 measured by a pH meter. When a 0.6 V vs RHE potential was applied to the BiVO4 ohmic back contact, flowing seawater was oxidized at a current density of 0.5 mA/cm2. Under these conditions, the near-surface fluorescence spectra revealed an increase at the 580 nm emission peak, indicating a decrease in pH during seawater oxidation (Fig. S13).
To illustrate the effect of flow velocities on spatial pH differences, we mapped spatial pH variations from 0–100 µm above the BiVO4 photoanode surface during PEC seawater oxidation at a current density of 0.5 mA/cm2. At a flow velocity of 0 m/s (Fig. 4d), protons generated by water oxidation diffuse from the BiVO4 photoanode surface to the bulk solution, driven by a concentration gradient. Protons accumulate near BiVO4 surface for ~ 5 minutes and diffuse outwards. The pH is lowered to ~ 4.3 at 0–100 µm above the BiVO4 photoanode surface, indicating that the water oxidation reaction at the BiVO4 photoanode surface significantly alters the local pH. When the flow velocity is increased to 0.16 m/s, pH mapping revealed four distinct regions colored blue, green, yellow, and red. Near the BiVO4 photoanode surface was the blue region, corresponding to lower pH and attributed to the higher proton concentration at the photoanode surface (Fig. 4e). The random pH variation within the blue region (Fig. 4e and 4f) was attributed to random voids in BiVO4 thin film’s nanoporous structure (shown in Fig. 3d).
At 60-µm above the photoanode surface, there was a gradual increase in the pH, indicated by the yellow region that continued until it approached the red region. As the flow velocity increased from 0.16 to 0.77 m/s, there was a corresponding increase to red region’s area on the pH maps (Fig. 4e – 4h). This suggests that convective flow serves to effectively confine generated protons to the photoanode surface, to prevent their diffusion into the bulk seawater solution, and to enhance CO2 extraction from seawater (Fig. S14). Modeling indicates that the downstream pH increases are due to proton consumption at the cathode. Furthermore, the pH variations downstream are significantly less under flow conditions than when there is no flow. The overall effect of convective flow is to reduce ionic conductivity, pH gradient, and CO2(aq) concentration overpotential losses. Subsequent analysis will focus on the effects of the flow field on non-equilibrium CO2(aq) transport flux, and the rate and selectivity of CO2R at the photocathodes.
Modeling time-dependent CO2(aq) convective-diffusive transport coupled with the non-equilibrium generation and neutralization of protons
We modeled our PEC reactor as a flatbed device, with an upstream photoanode and downstream photocathode arranged side-by-side. The modeling was conducted in 2D as the electrode behavior across the width dimension is uniform. Both the anode and cathode were 1 cm long and 2 cm apart at the bottom of the seawater chamber, with the electrolyte above the electrodes 1 mm in height (Fig. 5a). These dimensions match the dimensions of the 3D printed vortex reactor (Fig. 1a). Electrolyte flow velocity in our model was determined from our experimental data, with the chamber walls modeled with no-slip boundary conditions. To develop an understanding of the flow within our system, we first determined the flow field at varying velocities: 0.16 m/s, 0.34 m/s, 0.56 m/s, and 0.77 m/s. The resulting Reynolds numbers for these flow velocities were 320, 680, 1120, and 1540 respectively, all below 2000, confirming that our vortex reactor operates under laminar flow conditions.
The simulated PEC reaction was conducted in 2 mM HCO3− with pH 8.3. The H+ and OH− generation rates, alongside the CO2(aq) consumption rate were based on the partial current density of CO2R on the Si photocathode (Fig. 2d). This model only considers convective and diffusive transport across anode and cathode near the surface of electrodes and in the bulk solution, because the flux contribution from electric field-induced migration is negligible: the distance between the photoanode and photocathode is 2 cm; and in seawater, ions such as K+, Na+, Mg2+, and Ca2+ drive the conductive ion flux. The model tracked the path of the protons generated at the photoanode as they transported toward the photocathode via convection and diffusion. During transport, the protons may encounter OH–, HCO3–, and CO32– ionic species. The flow field also allows for “in situ”-generated CO2 to be transported to the cathodic sites for CO2R. The concentration of a specific chemical species i, denoted as \({C}_{i}\),were calculated using multi-physics simulations. Equations (2) and (3) were used to describe the reactive transport of reactive species i, such as CO2(aq), H+, OH–, HCO3–, and CO32–:
$$\frac{\partial {C}_{i}}{\partial t}=-\nabla \bullet {N}_{i}+{R}_{i}$$
2
$${N}_{i}=-{D}_{i}\nabla {C}_{i}+\varvec{v}{C}_{i}-\frac{{Z}_{i}F}{RT}{D}_{i}{C}_{i}\nabla \varphi$$
3
, where \({C}_{i}\) represents the concentration of species i, \({N}_{i}\) is the flux, \(-\nabla \bullet {N}_{i}\) is the net influx (with the divergence of the flux is taken as a negative value), and \({R}_{i}\) is the net rate of production from chemical reactions per unit time, \({D}_{i}\) is the diffusion coefficient of species i, and \(\varvec{v}\) represents the local velocity. The reactive transport of seawater DIC species (e.g., CO2(aq), HCO3−, and CO32−), protons, and hydroxides were analyzed to determine whether a species is being produced (\({R}_{i}>0\)) or consumed (\({R}_{i}<0\)) in a chemical reaction. Using \({R}_{CO2}\) as an example, details of \({R}_{i}\) for are provided in the supporting information. According to Eq. (3), the Nernst-Planck equation, \({N}_{i}\) considers contributions from diffusion, convection, and a negligible migration flux. The boundary conditions, kinetic rate constants, diffusion coefficients, and electric charge of species are shown in Fig. S15b, tables S2 and S4.
A seawater flow confines photoanode-generated protons within boundary layers
The pH mapping at the BiVO4 photoanodes revealed that varying the seawater flow velocity leads to significant changes in pH spatial distribution (Fig. 4e - f), and consequently affects [CO2(aq)] generated through acidification. We employed an analytical model that considers the convection and diffusion of H+ in a semi-quantitative manner to analyze the principle of flow-induced confinement of H+ produced at the photoanode. We then utilized COMSOL Multiphysics to couple mass transfer with acid-base reactions. Our COMSOL model considered the diffusion of all chemical species dissolved in seawater, while accounting for the real-time acid-base speciation reactions of all dissolved carbon species. H+ and OH− were actively generated at the anodes and cathodes, respectively. The in-situ generation of CO2(aq) and the measurement of the CO2 flux are performed under non-equilibrium conditions.
In our analytical model, for simplicity, we start by assuming [H+] in bulk seawater solution is zero. Drawing inspiration from Pohlhausen's integral approach to solve for boundary layer thickness, we proposed a linear concentration distribution for the CO2 flux as a function of distance from electrode surfaces32. This approach, combined with a series of boundary conditions elaborated on later, allows us to estimate the boundary layer thickness. We theorize that at a given point along the x-axis, the concentration exhibits a linear distribution relative to the distance from the cathode, z, as described in Eq. (4). This linear distribution matches closely with the results obtained from COMSOL Multiphysics (Fig. S16). Within this framework, the boundary layer is defined as the envelope where c > 0, and the thickness of the boundary layer is denoted as δ(x).
$$c\left(x,z\right)={c}_{0}\left(x\right)\left(1-\frac{z}{\delta \left(x\right)}\right)$$
4
This model is constrained by two boundary conditions: proton generation and their transport by convection and diffusion. We assume a uniform generation rate of H+ from seawater oxidation as described in Eq. (5). The number of protons flowing through the cross-section at x = x0 is equal to the total number of protons generated at the anode at x < x0, as stated in Eq. (6). Furthermore, the electrolyte flow between the two parallel plates conforms to Poiseuille flow, as described in Equations (7), where u0, k, z, and d are the average flow velocity among the reactor, velocity gradient, the height above the anode, and thickness of the reactor, respectively.
$$-D\frac{\partial c}{\partial z}{|}_{z=0}=\frac{i}{F}$$
5
$${\int }_{0}^{\delta \left(x\right)}c\left({x}_{0},z\right)u\left(z\right)\hspace{0.25em}\text{d}z=\frac{i}{F}{x}_{0}$$
6
$$u\left(z\right)=6{u}_{0}\cdot \left(-{\left(\frac{z}{d}\right)}^{2}+\frac{z}{d}\right)\approx kz \left(k=6\frac{{u}_{0}}{d}\right)$$
7
Equations (4) to (7) are combined to solve the thicknesses of the H+ boundary layer as a function of the flow velocity u0 and flow distance x0 (Fig. 5a).
$${\delta }\approx {{\left(\frac{6D{x}_{0}}{k}\right)}^{\frac{1}{3}}=\left(\frac{{x}_{0}Dd}{{u}_{0}}\right)}^{\frac{1}{3}}$$
8
Equation (8) provides a semi-quantitative expression for calculating the thickness of the concentration boundary layer. As the flow velocity increases, the CO2(aq) boundary layer thickness decreases (Fig. 5a) and the CO2 flux produced by acidification becomes concentrated within the narrowed region. At the same flow velocity, the boundary layer downstream is thicker than upstream (Fig. 5a). This can be attributed to higher downstream CO2(aq) and proton concentrations. At a fixed position in the x0 direction (take x0 = 0.8 cm as an example), the boundary layer thickness decreases from 77 µm at a flow velocity of 0.16 m/s to 40 µm at flow velocity of 0.77 m/s (Fig. 5a). This reduction is due to an increase in the convective mass transport flux in the z-direction as the velocity increases, resulting in a thinner boundary layer.
Quantitative analysis of CO2(aq) transport flux to Si photocathodes
The COMSOL numerical model allows us to precisely define boundary layer thicknesses (definition shown in supplementary materials section). Our COMSOL model considers reversible CO2-speciation reactions that can occur during transport. The dashed line in Fig. 5b shows the linear distribution assumption for CO2(aq) flux, which is consistent with the linear flux-position relationship that exists outside the boundary layer according to our model calculation. This allows for our numerical model to be compared against our analytical boundary layer analysis: based on Eq. (8), the boundary layer thicknesses at x = 0.8 cm along the anode at flow velocities of 0.16, 0.34, 0.56, and 0.77 m/s are 80, 60, 50, and 40 µm, respectively (Fig. 4e - h). In comparison, the numerical boundary layer thickness at the same position and flow rates are 77, 60, 51, and 45 µm. Furthermore, the pH map contours match the quantitative flow boundary (see supplementary materials). These strong correlations not only indicate that the assumptions proposed during our numerical calculations are reasonable but also confirm the accuracy and applicability of in-situ spectroscopy for monitoring local pH during PEC water splitting and CO2R.
Across all flow velocities, the observations made from our pH mapping using in-situ fluorescence measurements (Fig. 4e – 4h) closely matched the trends of the COMSOL-simulated boundary layers (Fig. 5a). This validates our model and allows us to visualize CO2(aq) generation, transport flux, and consumption behavior, along the photoanode-photocathode flow path. The cross-section located 1 cm from the anode’s far end (x = 1) was taken to illustrate the distribution of CO2(aq) flux (Fig. 5b). At this specific cross-section, CO2(aq) flux peaks within a 40-µm range from the photoelectrode surface in the z-direction. The peaking of the CO2(aq) flux was observed at all flow velocities. As the flow velocity increased from 0.16 to 0.77 m/s, the CO2(aq) flux increased from 11.7 to 18.8 mmol·m− 2·s− 1, demonstrating the ability of the flow field to enhance the CO2(aq) flux. The CO2(aq) flux is primarily driven by convective mass transfer. CO2(aq) generated through acidification was concentrated within a specific height range, closely matching what was determined with our semi-quantitative model. During CO2(aq) transport at flow velocities of 0.77 and 0.16 m/s from the anode to the cathode, the CO2(aq) fluxes at the front end of the cathode were 13.53 and 7.5 mmol·m− 2·s− 1 were achieved respectively (Fig. S20). This indicates that higher flow velocities correspond to higher CO2(aq) flux near the cathode surface, which is consistent with our findings that the CO2 selectivity and STF efficiency increased at higher flow velocities (Fig. 2a). Moreover, the vector profile of CO2(aq) flux at 0.77 m/s provides an overview of CO2 generation, transportation, and conversion (Fig. 5c). The arrows indicate a downward CO2(aq) flux on the cathode surface, indicating the presence of diffusive mass transfer when CO2(aq) is transported to the cathode surface (other flow rates shown in Fig. S21), driven by the CO2R reaction occurring on the cathode surface that lowers the local CO2 concentration. As a result, downward diffusion contributes to CO2(aq) transport within the boundary layer.
Chemical-species confinement via a boundary layer flow allows photoanodes to generate CO2(aq) flux effectively
By applying a seawater flow, the H+ produced at the anode and the CO2(aq) generated from these H+ can be confined within the boundary layer. We further experimentally benchmarked the effect of CO2 compensation on CO2R Faradaic efficiency using a model flat-plate flow reactor. Electrocatalytic electrodes are side-by-side in a flow-by configuration. Instead of photoelectrodes, we have the opportunity to arbitrarily turn on and turn off proton generation upstream.
An Ag-Au/CrOx cathode and an FTO/NiFe(OH)x/CrOx anode were employed (fabrication details in the Methods), and the catalysts used are the same as photoelectrodes. To compare the current density and selectivity of cathodes with and without acidification effects from upstream anodes, the terminology used is as follows: “with CO2 compensation” means that the anode was placed side-by-side with the cathode within the same reactor, while “without CO2 compensation” indicates that the anode was situated outside the cell, separate from the cathode. The flow rate-dependent chronoamperometry data reveals that under static conditions, both with and without CO2 compensation configurations exhibit comparable current densities (Fig. 5d). However, as the experiment progresses, the current densities steadily decline. This trend suggests an insufficient mass transfer of reactants, i.e., H+ and CO2, to adequately sustain the reductive current. The scenario takes a different turn when introducing the flow into the cell. In this case, the current densities remain stable and exhibit an increase with the ascending flow rate within the range of 0 to 0.77 m/s, implying that the mass transfer is enhanced. Notably, the reductive current densities observed with adjacent anodes (with CO2 compensation) are greater than those observed when no acidifying CO2 is transferred to the cathode (without CO2 compensation). This promotion in current can be attributed to the elevated CO2 concentration on the cathode surface in our configuration. The gas composition analysis indicates that, under a flow rate of 0 m/s, the faradaic efficiency for CO and H2 is 3% and 97%, respectively, which suggests that electrons primarily favor the reduction of H+ to form H2 under these conditions. The CO Faradaic efficiency increases from 3–21% as the flow rate rises, highlighting the crucial role played by mass transfer and CO2 compensation for CO2 in-situ conversion (Fig. 5e). The partial current for CO production increases proportionally with the flow rate, providing further evidence that the precisely controlled seawater flow serves to accelerate both the overall reaction rate (current density) and enhance CO selectivity (Fig. 5f). Figure 5e shows a higher CO2R FE than Fig. 5f. The flat plate reactor (Fig. 5e and 5f) and vortex flow reactor (Fig. 2b – 2d) exhibit comparable CO2R performance, indicating that vortex flow can effectively enhance convective mass transfer to confine the CO2 in the boundary.