Electroreduction via Gas Diffusion Electrode Reactor Designs

<p></p><p></p><p>In this work, the
effect of ion-selective membranes on the detailed carbon balance was systematically
analyzed for high-rate CO<sub>2</sub> reduction in flow electrolyzers. By using
different ion-selective membranes, we show nearly identical catalytic
selectivity for CO<sub>2</sub> reduction, which is primarily due to a similar
local reaction environment created at the cathode/electrolyte interface via the
introduction of a catholyte layer. In addition, based on a systematic exploration of gases released from electrolytes
and the dynamical change of electrolyte speciation, we demonstrate the explicit
discrepancy in carbon balance paths for the captured CO<sub>2</sub> at the
cathode/catholyte interface via reaction with OH<sup>-</sup> when using different
ion-selective membranes: (i) the captured CO<sub>2</sub> could transport
through an anion exchange membrane in the form of CO<sub>3</sub><sup>2-</sup>,
subsequently releasing CO<sub>2</sub> along with O<sub>2</sub> in<sub> </sub>the
anolyte, (ii) with a cation exchange membrane, the captured CO<sub>2</sub>
would be accumulated in the catholyte in the forms of CO<sub>3</sub><sup>2-</sup>,
(iii) whereas under the operation of a BPM, the captured CO<sub>2</sub> could
be released at the catholyte/membrane interface in the form of gaseous CO<sub>2</sub>.
The unique carbon balance path for each type of membrane is linked to ion
species transported through membranes.</p><p></p><p></p>


SUMMARY
In this work, the effect of ion-selective membranes on the detailed carbon balance was systematically analyzed for high-rate CO2 reduction in flow electrolyzers. By using different ionselective membranes, we show nearly identical catalytic selectivity for CO2 reduction, which is primarily due to a similar local reaction environment created at the cathode/electrolyte interface via the introduction of a catholyte layer. In addition, based on a systematic exploration of gases released from electrolytes and the dynamical change of electrolyte speciation, we demonstrate the explicit discrepancy in carbon balance paths for the captured CO2 at the cathode/catholyte interface via reaction with OHwhen using different ion-selective membranes: (i) the captured CO2 could transport through an anion exchange membrane in the form of CO3 2-, subsequently releasing CO2 along with O2 in the anolyte, (ii) with a cation exchange membrane, the captured CO2 would be accumulated in the catholyte in the forms of CO3 2-, (iii) whereas under the operation of a BPM, the captured CO2 could be released at the catholyte/membrane interface in the form of gaseous CO2. The unique carbon balance path for each type of membrane is linked to ion species transported through membranes.
However, CO2 reduction in H-type reactors only allow for relatively low current densities due to mass transport limitations in aqueous solutions. [17][18][19] Large-scale utilization of electrochemical conversion of CO2 requires high reaction rates (i.e. high current densities). In this context, flow electrolyzers with gas-diffusion electrodes (GDEs) has gained considerable attention for CO2 reduction, owing to that GDEs allow for a very thin mass-transfer boundary layer (~50 nm). 18,19 By using GDE-type flow electrolyzers, the mass-transport of CO2 and gaseous products on the surface of the catalysts can be accelerated, achieving commercially-relevant current densities (> 100 mA/cm 2 ) along with high selectivity toward a desired product. [20][21][22][23][24][25][26][27][28][29] To date, most of the high-rate CO2 reduction studies based on GDE-type flow electrolyzers have been performed using anion exchange membranes (AEM) to avoid product crossover. [20][21][22][23][24][25][26][27][28][29] However, our recent work demonstrated a substantial crossover of anionic CO2 reduction products such as acetate and formate through AEM in GDE-type flow electrolyzers. 29 More importantly, after the electrolytes reach steady state, it was found that about 70% of consumed CO2 is captured at the cathode/electrolyte interface via reaction with OH -, forming CO3 2-, which transports to the anolyte via AEM as a charge-carrier. 29 Subsequently, CO3 2coming from the catholyte reacts with H + in the vicinity of the anode, releasing gaseous CO2 from the anolyte along with the O2 stream, which means that most of the consumed CO2 (70%) is captured in catholyte and emitted from the anolyte.
In other words, only 30% of the CO2 consumption is involved for CO2 conversion into products.
This finding means that many of the current techno-economic analysis for high-rate electroreduction of CO2 must be reconsidered if the significant CO2 crossover occurs.
One approach to reduce the CO2 crossover would be to use a two-step cascade process, which consists of an initial CO2 reduction to CO and a subsequent CO conversion into highly valuable multi-carbon products that has no carbon source crossover. 30,31 However, even in this two-step procedure with 100% CO faradaic efficiency in the first step, 50% of all consumed CO2 could still be emitted out of the anolyte using an AEM. 29 Theoretically, utilization of a cation exchange membrane (CEM) or a bipolar membrane (BPM) can prevent the CO2 crossover in GDE-type flow electrolyzers. However, only a few works on high-rate CO2 reduction (> 100 mA/cm 2 ) have been carried out in GDE-type electrolyzers using a CEM [32][33][34] or BPM 35,36 to date.
This study describes a systematic exploration of the effect of ion-selective membranes on the detailed carbon balance including CO2 consumption, products and CO2 crossover, as well as CO2 emission in GDE-type flow electrolyzers. Herein, we demonstrate the comparison of catalytic selectivity, CO2 consumption rate (via the reaction with OH -), and the dynamical change of electrolyte speciation among three different types of ion-selective membranes. By a systematic exploration of gases released from catholyte or anolyte, ion species change in electrolyte and ion species transport via membranes, this work provides mechanistic insights into the role of ionselective membranes in carbon balancing for high-rate CO2 reduction.

Electrocatalytic CO2 reduction performance
In this work, Cu electrocatalyst layers (~70 nm) were prepared on top of microporous carbon layers of GDEs by magnetron sputtering at an argon pressure of 2 mTorr ( Figure S1). The detailed materials characterization of the Cu catalyst layers on GDEs has been reported in our previous work. 29 We conducted CO2 reduction electrolysis experiments in a three-compartment flow electrolyzer where a Cu catalyst coated on GDE was positioned between the gas and catholyte chambers, as shown in Figure 1a. An ion-selective membrane was used to separate the catholyte and anolyte flow chambers in which electrolytes continuously flow. During CO2 reduction, gaseous CO2 at a constant flow rate (45 ml min) was continuously fed into the gas chamber (Figure 1a), and a fraction of the CO2 diffused to the surface of the catalysts in an electrolyte and then converted into gas products such as C2H4 and liquid products such as ethanol ( Figure 1b). Gas products mixed with the unreacted CO2 were directly vented into the gas-sampling loop of a gas chromatograph (GC) for periodic quantification. The liquid products were diluted and circulated in the given catholyte and anolyte reservoirs, and was detected via high-performance liquid chromatography (HPLC) after completion of the CO2 reduction electrolysis experiment.
In order to get reliable catalytic selectivity for gas products in high-rate CO2 reduction, gas flow out of the reactor was monitored via a volumetric flowmeter according to our previous work ( Figure   S2). 29 Figure 1c shows nearly identical gas flowrates were observed out of the electrolyzer when using AEM, CEM and BPM in 1 M KHCO3 at 200 mA/cm 2 , indicating a similar CO2 consumption rate. This observation is primarily due to that the same OHgeneration rate via cathodic reactions (i.e. similar local pH created at the cathode/electrolyte interface). The faradaic efficiencies of gas products calculated using these corrected gas flowrates were plotted for different ion-selective membranes ( Figure 1c). As shown in Figure 1c, C2H4 is the primary gas product for all the different ion-selective membranes, along with small amounts of CO and H2 and only trace amounts of CH4. types of membranes were utilized (at nearly identical potentials, as shown in Figure 1c). This result indicates that catalytic selectivity of gaseous products is independent of the type of ion-selective membrane for high-rate CO2 reduction in the three-compartment electrolyzers.  In addition to the detected gas products, liquid-phase products in both catholyte and anolyte were all analyzed due to the potential crossover of liquid products from the catholyte to anolyte via membranes. As noted in Figure 2a, substantial anionic CO2 reduction products (such as formate and acetate) crossed over from the catholyte to the anolyte via the AEM by electromigration, with only minimal crossover for uncharged liquid products. In contrast, the CEM and BPM exhibited negligible crossover for both anionic liquid products and uncharged products ( Figure 2a). This observation entails both CEM and BPM are capable of inhibiting the crossover of anionic and neutral products.
For determining the total amounts of liquid products, liquid products evaporated from GDEs into the gas chamber of the reactor were also collected for analysis (using a setup shown in Figure   S3). 37 No matter which type of ion-selective membrane was used, alcohol products such as npropanol and ethanol experienced considerable evaporation through the gas diffusion layer of GDE ( Figure 2b), which is due to their high volatility. In addition, we found that acetaldehyde had the highest evaporation ratio among the evaporated liquid products ( Figure S4). This finding may be attributed to two reasons, (i) its relatively high vapor pressure and (ii) its further reduction to ethanol on the cathode where a substantial amount of acetaldehyde was produced initially and subsequently converted into ethanol. 38 Based on the quantification of liquid products in both catholyte and anolyte as well as liquid products evaporated from GDEs into the gas chamber (Equation S10), faradaic efficiencies of all liquid products were evaluated for all the different types of membranes ( Figure 2c). As shown in Figure 2c, ethanol was the dominant liquid product along with n-propanol, acetate and formate as minor products. There appears to be no significant variation in liquid product formation across all types of membranes. All the above results imply that the role of ion-selective membrane is almost negligible in affecting catalytic selectivity of high-rate CO2 reduction in the three-compartment electrolyzers, owing to the similar local reaction environment created on the cathode via the introduction of a catholyte layer.

Capture and emission of CO2 throughout the electrolyte
According to our recent carbon balance study, 29 the gases released from the anolyte were systematically explored for CO2 reduction via an AEM with 1 M KHCO3, elucidating a two-step procedure of CO2 capture at the cathode/electrolyte interface via reaction with OHand subsequent CO2 degassing from anolyte due to H + in the vicinity of the anode (Scheme 1a). With the nearly identical catalytic selectivity ( Figure 2c) and similar total CO2 consumption rate (similar gas outlet shown in Figure 1c), the same OHgeneration rate via cathodic reactions means that the capability of capturing CO2 for carbonate formation at the cathode/electrolyte interface using CEM and BPM should be similar to that of AEM. Thus, for a CEM and BPM, substantial additional carbonate anions produced in the reaction of CO2 and OHgenerated via the cathodic reactions must be balanced with extra cation species (total anion charge equals to total cation charge) or emitted from electrolyte as gaseous CO2. To uncover the role of different membrane types in the carbon balance for flow electrolyzers, gases released from the electrolyte were detected for CEM and BPM, respectively (using a closed-cycle anolyte with a vent for gases shown in Figure S5).
Theoretically, the composition ratio of CO2/O2 in the gas stream from the anolyte will be 4, 2 or 0 if the only anion species for neutralizing H + generated on the anode is HCO3 -, CO3 2or OH -. 28,29 In addition, under the consideration of that HCO3 -, CO3 2or OHis the only anion species of neutralization reaction with H + , the theoretically calculated CO2 flowrate will be 6.0, 3.0 or 0 ml/min at 200 mA/cm 2 with the geometric active area of 2 cm 2 (Table S2).  When an AEM was used, the CO2/O2 ratio decreased from ~3 to ~2 in the initial 4 h and then maintained at ~2 over the rest of electrolysis. This observation is due to that the CO2 evolution via the H + neutralization reaction changed rapidly from a mixture of HCO3and CO3 2to nearly pure CO3 2using an AEM (Figure 3a). In contrast, as noted in Figure 3b, the CEM experienced a consistent CO2/O2 ratio of ~4 and a constant CO2 flowrate with 6 ml/min for the duration of electrolysis at 200 mA/cm 2 , which implies that the CO2 formation was always derived from HCO3in the anolyte. This finding is ascribed to that the CO3 2formed via CO2 capture in the catholyte cannot transport to the anolyte via the CEM since the function groups (typically SO3groups) only allows cation species (such as K + ) to transport through (Scheme 1b). It should be noted that the CO2 reduction electrolysis via CEM was tested for just ~ 3 h, since that the anolyte conductivity rapidly decreased from ~70 mS/cm to ~3 mS/cm after ~ 3 h using the CEM ( Figure   S8b). All the above results with the CEM entail that almost no anionic species transported to the anolyte via the membrane, but cation species such as K + served as the main charge carrier via the CEM. Thus, the concentration of KHCO3 in the anolyte was significantly reduced over time as K + constantly transported to the catholyte and the remaining HCO3in the anolyte was consumed for CO2 evolution (Scheme 1b).
A bipolar membrane is comprised of a cation exchange layer (CEL) and an anion exchange layer (AEL) as well as a catalyst layer that is sandwiched between CEL and AEL. The catalyst layer in a BPM dissociates water (fed from both the catholyte and anolyte) into H + and OH -, which subsequently transports to the catholyte and anolyte via the CEL and AEL, respectively (Scheme 1c). 39 Under operation of a BPM (Figure 3c), the flowrate of CO2 released from the anolyte rapidly decreased from 1.4 ml/min to 0.5 ml/min in the initial 4 h, corresponding to a decline in the CO2/O2 ratio from ~1 to ~0.3. This observation indicates that OHserved as the major anion species that neutralizes with H + produced as a byproduct of O2 evolution. In addition, the almost constant conductivity in both catholyte and anolyte over the 10 h electrolysis ( Figure S7c) may imply that neither anionic species (CO3 2or HCO3 -) nor cationic species (K + ) had any apparent crossover. This result reveals that the additional anion species (CO3 2or HCO3 -) generated by CO2 capture could not be accumulated in the catholyte during CO2 reduction electrolysis due to charge balance issue (the total anion charge must equal to the total cation charge). Thus, the additional CO3 2or HCO3should be emitted from the catholyte as gaseous CO2. As expected, gas bubbles released from the catholyte were observed when a BPM was used (no gas evolution was observed in the catholyte using AEM or CEM), and this gas evolution immediately disappeared after stopping the electrolysis.
To verify that CO2 was generated in the catholyte, the gases released from the catholyte during the CO2 reduction electrolysis were analyzed by a setup shown in Figure 4a. Figure 4b shows CO2 degassing from the catholyte when using a BPM, owing to the neutralization reaction of CO3 2or HCO3with H + near the CEL of BPM (Scheme 1c), which is in line with previous BPM work. 36 In addition, the related flowrate of CO2 released from the catholyte slightly decreased from ~ 3.5 4b). This observation can be attributed to that the carbon source (anion species) for CO2 evolution abruptly transformed from a mixture of HCO3and CO3 2to almost pure CO3 2-. In addition, a fraction of CO2 released from catholyte chamber can transport back to the cathode surface to be reused for both CO2 reduction 36   While each type of ion-selective membrane had a different CO2 flowrate released from the anolyte, O2 was detected with a constant flowrate of ~1.5 ml/min during the electrolysis irrespective of membrane type (Figure 3a-c). This finding is consistent with the theoretical value of the O2 flowrate (1.5 ml/min shown in Table S2) at 200 mA/cm 2 for the geometric active area of 2 cm 2 .
To further understand the transformation of anionic species in the electrolyte, the pH of the electrolyte was also monitored over the course of the electrolysis for all the membranes. Figure 3e shows that for a CEM the catholyte pH was enhanced from 8.3 to nearly 9.8 after ~3 h. The catholyte pH with the AEM increased to 10.2 after ~3 h under identical conditions. Thus, the similar increasing trend in catholyte pH between the AEM and CEM over 3 h further confirms that the captured CO2 at the cathode/electrolyte interface (via reaction with OH -) mainly formed CO3 2using the CEM, 29 leading to CO3 2acting as the dominant anion species in the catholyte after 3h. The catholyte pH with the BPM was maintained below 9 over the entire electrolysis experiment ( Figure   3f) due to that a constant supply rate of H + from water dissociation in the BPM enables carbonate and bicarbonate concentrations in the catholyte to reach a steady sate. However, the observed CO2 flowrate (2.6 ml/min) from catholyte (after reaching steady state) also reveal that CO2 was captured and converted to CO3 2at the cathode/electrolyte interface, and then combined with the aforementioned H + at the BPM/ catholyte interface to release CO2. In addition, it should be noted that the theoretical calculations have shown that the pH near the cathode is ~ 13 in 1 M KHCO3 at 200 mA/cm 2 , 18 which means that the reaction of CO2 with OHat the cathode/electrolyte interface forms CO3 2instead of HCO3 -(equation S8 and S9). Thus, all these results reveal that the captured CO2 by electrolyte near the cathode formed CO3 2irrespective of membrane types.
We found that the anolyte quickly reached a near neutral pH for both the AEM and the CEM during the electrolysis (Figure 3d and e), which allows for CO2 degassing in the anolyte.
Specifically, the anolyte pH with the AEM was maintained at ~7.9 after 20 min (Figure 3d), owing to that the constant H + generation rate near the anode and continuous carbonate supply derived from the catholyte created a steady state for all the anion species in the anolyte via the neutralization reactions (Scheme 1a). In contrast, with the CEM, the anolyte pH rapidly deceased from 8.3 to 6.7 over 3 h (Figure 3e). This finding is due to that the CO2 degassing with the continuous consumption of KHCO3 in the anolyte created a CO2-saturated KHCO3 anolyte and its concentration gradually  Figure 3f. This observation may be linked to a slow variation in the anionic species concentrations (here, more CO3 2was likely created along with less existing HCO3 -) in the anolyte during the electrolysis. This alteration in anionic species concentrations is ascribed to that when OHis transported toward the anode as the major anion species being neutralized with H + generated near the anode, a fraction of OHmay also react with the existing HCO3in the anolyte to form CO3 2-.

Carbon balance via different types of membranes
For high-rate CO2 reduction in flow electrolyzers, the carbon source for CO2 fed from the inlet of reactor must be balanced with that of all CO2 reduction products, captured CO2 by electrolyte (carbonate formation) and residual CO2 out of reactor (i.e. unreacted CO2). As noted in Figure 5a, (i) the residual unreacted CO2 flowrate out of the reactor, (ii) the consumed CO2 flowrate for carbonate formation via the reaction with OH -(i.e. captured CO2 throughout the electrolyte) and (iii) the consumed CO2 flowrate that was converted into all the gaseous and liquid products add up to a total CO2 flow of ~ 45 ml/min for each type of the ion-selective membrane. Thus, the carbon element during the electrolysis is balanced with that of CO2 inlet flowrate (45 ml/min) in this work.
In addition, Figure 5b shows the nearly identical CO2 consumption rate for the formation of gaseous and liquid products using different ion-selective membranes, which is in line with the roughly same catalytic selectivity shown in Figure 2b.
It should be noted that there should be nearly the same carbonate formation rate (via CO2 reaction with OH -) near the cathode among all the different membranes due to the identical OHgeneration rate via cathodic reactions at identical current densities. While membrane types should have minimal effect in the total carbonate formation rate near the cathode, the BPM had a slightly lower consumption rate of CO2 from the gas chamber for carbonate formation compared to those of the AEM and CEM, as shown in Figure 5b. This finding is correlated with the discrepancy in carbon balance paths among the three different types of membranes. In other words, while the unavoidable CO2 capture near the cathode forms carbonate in the catholyte, the end results of where the carbonate goes is different in each type of membrane. For the CEM, the captured CO2 was accumulated in the form of carbonate in the catholyte without emission. In contrast, under the use of the AEM, the captured CO2 in the form of carbonate crossed to the anolyte and was emitted as gaseous CO2 with O2 stream in the anolyte. Notably, with the BPM, the captured CO2 could be released from the catholyte as gaseous CO2. Thus, a fraction of the generated CO2 in the catholyte may be involved in the reaction with OHfor carbonate formation, which corresponds to a relatively low consumption rate of CO2 in the gas chamber for carbonate formation (~65% of total CO2 consumption), as shown in Figure 5b. In addition, the reuse of a fraction of the released CO2 in the catholyte, derived from the captured CO2 in the form of carbonate, also results in a slightly higher CO2 utilization rate of the BPM (ratio of CO2 converted into products versus total CO2 consumption) in Figure 5b.

Implications of the CO2 degassing
From an economic and environmental perspective, the released CO2 from the electrolyte in flow electrolyzers would require to be captured and recycled. Under the use of the AEM, the released CO2 in the anolyte only can be recycled for CO2 reduction after removing O2 in the gas mixture (mole ratio of CO2/O2 is 2:1). Interestingly, the operation of the BPM could degas CO2 from the catholyte, which can be directly fed into the gas compartment for CO2 conversion due to its high purity (~100% CO2 by mole). Thus, compared to the necessary CO2 and O2 separation process for CO2 recycling with the AEM, the BPM has the potential to reduce the total cost of carbon source.
However, it should be noted that using BPM at high-rate CO2 reduction (current densities > 100 mA/cm 2 ) requires an additional potential (> ~0.8 V) for membranes that may reduce the energy efficiency of CO2 conversion reactors. [39][40][41] Thereby, how to balance the energy efficiency along with the easy recyclability of the produced CO2 in the catholyte (from inevitably captured CO2) under the use of BPMs will need a full techno-economic analysis in the future.

CONCLUSION
In conclusion, our results present that the role of ion-selective membranes is minimal in affecting the catalytic selectivity of high-rate CO2 reduction, owing to the nearly same local reaction environment created near the catalysts through having a catholyte layer. By rigorously analyzing gases released from electrolytes as well as monitoring electrolyte pH, we found that most of the consumed CO2 source (≥ ~65%) was captured via reaction with OHnear the cathode to form formed CO3 2-, which is almost independent of membrane types.
Importantly, each type of ion-selective membrane produces a unique carbon balance path for the captured CO2 source. Specifically, the captured CO2 in the form of CO3 2could cross an AEM from the catholyte to the anolyte and then emitted as gaseous CO2 mixed with O2 stream. In contrast, the captured CO2 could not transport to the anolyte when using a CEM or BPM. With a CEM, captured CO2 in the form of carbonate continuously accumulated in the catholyte, since there was no concomitant H + supply for CO2 evolution (mainly K + crossed the membrane). With the bipolar membrane, the captured CO2 was released from the catholyte as gaseous CO2, owing to the reaction of carbonate with H + transported from its cation exchange layer. In addition, while for an AEM CO2 was emitted together with O2, for a BPM the pure CO2 was released, which can be directly recycled back to the gas compartment for CO2 conversion, correspondingly decreasing the cost of the CO2 source. This study shows that while the catalytic selectivity is independent of the types of ion-selective membranes, membrane types play an important role in the corresponding carbon balance path at high-rate CO2 reduction, thus future work should focus on the membrane exploration for achieving the practical utilization of high-rate CO2 reduction.
Iridium dioxide (IrO2) purchased from Dioxide Materials was used as an anode in flow electrolyzers of high-rate CO2 reduction.

Catalysts preparation
For obtaining high purity Cu catalsyts, Cu layers were prepared on top of microporous layer of gas-diffusion electrodes by direct current magnetron sputtering (50 W) from a Cu target at an argon pressure of 2 mTorr. Figure 1S shows a typial schematic illustration for Cu depostion using magnetron sputtering under an argon ambient. The energetic Ar + ions are created in a glow discharge plasma, thus Ar + bombardment occurs on the cathode Cu target, which leads to the removal of Cu atoms. Subseqently, the sputtered Cu atoms condense on a substrate (i.e. GDE) to form a Cu layer. In this work, with ~4 nm/min Cu deposition rate, the thickness of the Cu layers were controlled by the deposition time.

Gas and liquid product analysis
Gas products mixing with unreacted CO2 flowed out of the electrolyzers, directly injecting into the gas-sampling loop of a gas chromatography (PerkinElmer, Clarus ® 590). Ar was used as a carrier gas with a contant flowrate of 10 sccm. The gas chromatography was equipped with a packed Molecular sieve 13x column and a packed Hayesep Q column to separate the gas products. Thus, exsiting the columns, H2, CO, CH4 and C2H4 could be identified at different reaction times using a thermal conductivity detector. In addition, the peak area of each gas product was compared to standards (calibration gases) to determine the corresponding concentration of gaseous products. Thus, we can get the faradaic efficiency of a certain gas product as follows: where n is the number of electrons required for producing one molecule of the related gas product, and Cproduct is the concentration of gas product measured by GC. ∅ and is the gas flowrate out of the electrolyzers and the electrolysis time, respectively. is the ambient pressure, is the ideal gas constant, T is the absolute temperature, F is Faraday constant, and I is the applied current.
The liquid-phase products are analyzed after the electrolysis using a high performance liquid chromatography (HPLC, Agilent 1200 series). Liquid-phase products were separated by an Aminex HPX-87H column (Bio-Rad) that was maintained at 50 °C for the duration of the detection.
The HPLC was equipped with a diode array detector (DAD) and a refractive index detector (RID), and the signal response of the DAD and RID was calibrated by known concentration solutions.
Thus, we can get the concentration of the detected liquid-phase product. The faradaic efficiency of liquid products can be calculated by equation: where n is the number of electrons required for producing one molecule of the related liquid product, and Cproduct is the molar concentration of gas product measured by HPLC. V is the volume of the electrolyte. To obtain accurate selectivity of liquid products, we measured the volume of catholyte and anolyte after electrolysis, respectively.

Electrolyte pH and conductivity measurements
electrolysis. In addition, the pH meter was also equipped with a temperature sensor for the temperature-compensation. The pH meter was calibrated by a standard pH 7 buffer and a standard pH 10 buffer before the measurement.
The conductivity of the catholyte and the anolyte was monitored by a conductivity meter (PCE-PHD 1-PH, PCE Instruments) during CO2 reduction electrolysis. Before the measurement, the conductivity meter was calibrated via conductivity standard of 1413 µS / cm (25 ° C; 0.01 M KCl) and 111.8 mS / cm (25 ° C; 1 M KCl) purchased from VWR. It should be noted that both of the calibration and the measurement were temperature-compensated due to that the solution conductivity is also temperature-dependent at a fixed solution concentration.

SUPPLEMENTAL INFORMATION
Supplemental Information can be found online.
M.M and B.S developed the conceptual idea, designed the experiments and wrote the original manuscript. S. K carried out a part of the electrolyte pH and conductivity measurement. All authors contributed to discussing the results and editing the manuscript.

DECLARATION OF INTERESTS
There are no conflicts to declare. Catalysts preparation by magnetron sputtering Figure S1. Schematic illustration of magnetron sputtering for Cu deposition from a Cu target.

High local pH near the cathode
During the process of electrochemical reduction of CO2, CO2 and H2O can be electrochemically converted into a variety of gaseous products such as CO, C2H4 and CH4 on the surface of the catalyst in electrolytes based on the below reactions 1-4 : In the electrocatalytic CO2 reduction process, the competing H2 evolution is an unavoidable reaction. The electroreduction of H2O to H2 on the surface of catalyst according to the reaction 1 : Due to the above cathodic reactions (Equation (3-9), a large amount of OHions can be created at the cathode/electrolyte interface at high current densities, which creates a significantly higher pH near the surface of cathode compared to that of the bulk solutions. 2,3

CO2 capture via high local pH
During CO2 reduction in flow electrolyzers using 1 M KHCO3, CO2 from the gas chamber of the electrolyzers reacts with OHions produced at the cathode/electrolyte interface accoding to the below reactions of CO2 and OH -: * This is at a CO2 partial pressure of 1 bar in 1 M HCO3 -. 5

CO2 reduction performance
The electrocatalytic reduction of CO2 was peformed in a three-chamber flow electrolyzer made from Teflon at ambient temperature and pressure. At the reactor, an ion-selective membrane was used to sperate catholyte and anolyte flow compartments. Catholyte and anolyte bottles were filled with 50 ml 1 M KHCO3, respectively ( Figure S2). In addition, a fixed geometric surface area (2 cm 2 ) of Cu layer was used for all the tests in this study.
CO2 was purged into gas compartment at a constant flowrate of 45 ml/min, and then a fraction of gaseous CO2 diffuses to the surface of the catalyst in electrolyte for CO2 conversion. In addtion, CO2 also can be captured to form carbonate (equation S8 and S9) via the reaction of CO2 with OHgenerated at the cathode/electrolye inteface. 5 Thus, the high CO2 conversion rate to gas (C2) and liquid products as well as high local pH can lead to a substantial CO2 consumption at high current densities, correspondingly varying the gas flow (gas mixture) out of the reactor. The volumetric flowrate of gas outlet (gas mixture) after reactor was monitored S4 by flow meter in CO2 reduction ( Figure S2), and then faradaic efficiencies of gas products were evaluated under the consideration of gas flow variation between inlet and outlet. It should be noted that the average catalytic selevitity of gas products during 2.5 h CO2 reduction electrolysis was used in this work. Figure S2. Schematic illustration of flow cell setup for electrocatalytic CO2 reduction.

Collection of liquid from electrolyte
It should be noted that the ion species carried with water molecules (hydrated ion) transports via membrane, which means the volume of catholyte and anolyte was vaired after electrolysis.
For AEM, a decrease in catholyte volume was observed with increased anolyte volume after several hours of CO2 reduction electrolysis, because of the transportion of the anion species hydrated with water molecules from catholyte to anolyte via AEM as charge carriers. In contract, the use of CEM experienced an inceased catholyte volume with correspondingly decreased anolyte over the course of electrolysis, due to that the cation species hydrated water molecules transported from anolyte to catholyte via AEM as charge carriers. Notably, no obvious variation in both cathlyte and anolyte when BPM was used, which is due to that water supplied almost equally from both catholye and anolyte was disociated into H + and OH -, transporting to catholyte and anolyte, respectively.

S5
Based on the aforementioned discussion, in order to obtain accurate selectivity of liquid products, volume of catholyte and anolyte was also measured for each test after electrolysis, respectively.

Collection of liquid products evaporated from GDEs
Some liquid products can be evaporated from the gas diffusion layer of GDE and then flow out of the gas compartment of the reactor with unreacted CO2 and gas products. To collect the evaporated liquid products from GDEs (i.e. gas chamber), gas outlet flow after the reactor was directly purged into a sealed bottle filled with 30 ml de-ionized water (the outlet flow tube was immersed into de-ionized water), as shown in Figure S3. After completion of CO2 reduction, the liquid products diluted with de-ionized water in that sealed bottle were analysed via highperformance liquid chromatography (HPLC). Figure 2b presents the faradaic efficiencies of liquid products evaporated from GDEs when using distinct ion-selective membranes, indicating that only alcohols products such as ethanol and propanol evaporate and escape from the cathode/electrolyte interface irrespective of membrane types, which is due to their high volatility.
In addition, both catholyte and anolyte in the given reservoirs were collected for quantification of liquid products, owing to liquid products crossover from catholyte to anolyte via membranes. 5 Thus, the total amount of one certain liquid product formed on cathode GEDs can be written as: where and ℎ are the amount of one certain liquid product detected in anolyte and catholyte, respectively. is the amount of one certain liquid product evaporated from GEDs. Here, the evaporation ratio of one certain liquid product formed on cathode GDEs can be calculated based on the below equation: Thus, the equation (S11) was used to calculate a ratio between the amount of one certain liquid product evaporated from GDEs and the total amount of corresponding liquid product formed on the cathode, as shown in Figure S4.
S6 Figure S3. The schematic illustration of flow cell setup for collecting liquid products evaporated from GDEs during CO2 reduction. Figure S4. Evaporation ratio of related liquid products escaped from GDEs (i.e. gas chamber) at 200 mA/cm 2 when using AEM, CEM and BPM, respectively.

Analysis of gas released from the anolyte
When the electrocatalytic CO2 reduction occurs on the surface of the cathode, water oxidation reaction (i.e. O2 evolution) takes place on the anode surface. By the water oxidation reaction, a large amount of H + can be created at the anode/electrolyte interface, which leads to a decrease of pH locally near the anode. Subsequently, H + produced at the anode/electrolyte interface can be neutralized with HCO3 -, CO3 2or OHin anolyte. The H + neutralization with HCO3or CO3 2forms CO2, leading to CO2 degassing from anolyte with the stream of O2. 5 For analysing the gases released from anolyte over the course of CO2 reduction, the flow electrolyser setup in Figure S5 was utilized. In that setup, N2 at a constant flowrate was used as a carrier, thus gases released from anolyte were diluted with N2, directly venting into the gas sampling-loop of the GC for periodical quantification. The volumetric gas flow released from anolyte was also monitored by in situ flow meter over the electrolysis, as shown in Figure S5. Figure S5. The schematic illustration of flow cell setup for detection of gases released from the anolyte over the course of CO2 reduction.

Analysis of gas released from the catholyte using BPMs
No any gas evolution in catholtye was observed when AEM or CEM was used. However, under the use of bipolar membrane in flow electrolyzers, we found that gas bubbles released from the catholyte, which is unique in comparison with the other two membranes. To analyze the gas released from catholyte over the course of CO2 reduction using BPM, a test setup in Figure S6 was utilized. Similar to gas analysis from anolyte, a constant N2 flow was also used as a carrier gas, which mixed with gases released from catholyte, venting into the gas sampling-loop of the GC for periodical quantification, followed with an in situ volumetric flow meter ( Figure S6).
We found CO2 gas releasd from the cathlyte, along with only trace amount of H2 in Figure 4. It should be noted that the mole ratio of CO2/H2 released from the catholyte is 5000, which means that the purity of released CO2 is about 99.98%. Figure S6. The schematic illustration of flow cell setup for detection of gases released from the catholyte over the course of CO2 reduction under the use of BPM. 1 M KHCO3 was used as initial catholyte (50 ml) and anolyte (50 ml).

Applied potentials on the cathode
Potentiostatic electrochemical impedance spectroscopy (PEIS) was conducted on Cu deposited GDE in the flow electrolyzer to determine the solution resistance (Rs). A detailed procedure was described in a previous work. 5 It should be noted that the distance between reference and cathode was less than 2 mm in order to reduce Rs in this work. Table S1 shows the solution resistance for the different ion-selective membranes. However, the cathodic reactions at high current densities can lead to a significant change of ion species and related concentration in the vicinity of the cathode, which indicates a difference in solution resistance near the cathode at high current densities compared to that of PEIS which was performed at relative low current densities. Thus, this difference in the solution resistance is closely correlated with the accuracy in IR-corrected potentials at a high current.

Theoretical estimation of O2 and CO2 flowrate generated from electrolyte
Assuming that all charge passed through the anode is just employed for oxidation of water into O2, thus theoretical O2 flowrate released from anolyte can be expressed as 5 : where n and Qtot are the number (here is 4) of electrons lost from 2 H2O for forming one O2 molecule and totoal charge passed through the anode, respectively. F is the faradaic constant, is ideal gas constant, T is absolute temperature, and is ambient pressure.
From our prevous work, the ratio of CO2 and O2 released from the anolyte will be 4, 2 and 0 if the only anion species for neutralization reaction with H + is HCO3 -, CO3 2or OH -. 5 Thus, after getting the O2 flowrate at 200 mA/cm 2 (the cathode with 2 cm 2 geometric active area was used for all the tests) based on equation S12, we can easily get the related flow of CO2, as shown in Table S1.

Calculation of the carbon balance
The residual unreacted CO2 flowrate in the gas outlet (gas mixture) out of gas compartment of flow electrolzyers can be written as: where ∅ is the monitored gas flowrate out of the ractor during CO2 reduction electrolysis using the setup shown in Figure 2S. Here, ∅ , ∅ 4 , ∅ 2 4 and ∅ 2 are the gas flowrate of CO, CH4, C2H4 and H2 produced from electrochecmial CO2 conversion in the gas outlet, respectively. Based on the equation S1-3, each molecule of CO, CH4 and C2H4 formation requires 1, 1 and 2 CO2 molecule, Thus, the consumed CO2 flowrate that is converted into all gas products (CO, C2H4 and CH4) in CO2 reduction can be expresed as below: Depending on the number of carbon atoms in liquid molecule produced in CO2 reduction, the consumed CO2 flowrate involved in all liquid products formation can be written as: where ∅ 1 , ∅ 2 , and ∅ 3 are the consumed CO2 flowrate for forming C1, C2 and C3 liquid products, respectively. For high-rate CO2 reduction, the inevitably caputred CO2 in forms of carbonate via reaction with OHcould consume substantial CO2 flow, significantly reducing the total gas flow out of the reator (Figure 1c). It is known that the carbon element from CO2 inlet flowrate should be eventually balanced with those of residual unreacted CO2, all products and carbonate formed via reaction between OHand CO2. Thus, the consumed CO2 flowrate via the reaction with OHgenerated on the cathode surface can be expressed as: where ∅ 2 is CO2 flowrate fed into the gas chamber of the reactor. In this work, a constant CO2 flow was used. It should be noted that mass flow controller used in this work was calibrated for CO2 flow by volumetric flow meter before and after each CO2 reduction test for high accuracy. Thus, we got the below carbon balance in Table S2. reduction (such as ethanol); ∅ − : the consumed CO2 flowrate via the reaction with OHgenerated in cathodic reactions; ∅ 2 : the residual unreacted CO2 flowrate in the gas outlet (gas mixture) out of gas compartment of flow electrolzyers)