Improving the Stability, Selectivity, and Cell Voltage of a Bipolar Membrane Zero‐Gap Electrolyzer for Low‐Loss CO2 Reduction

Electrolyzers for CO2 reduction containing bipolar membranes (BPM) are promising due to low loss of CO2 as carbonates and low product crossover, but improvements in product selectivity, stability, and cell voltage are required. In particular, direct contact with the acidic cation exchange layer leads to high levels of H2 evolution with many common cathode catalysts. Here, Co phthalocyanine (CoPc) is reported as a suitable catalyst for a zero‐gap BPM device, reaching 53% Faradaic efficiency to CO at 100 mA cm−2 using only pure water and CO2 as the input feeds. It is also shown that the cell voltage can be lowered by constructing a customized BPM using TiO2 water dissociation catalyst, however this is at the cost of decreased selectivity. Switching the pure‐water anolyte to KOH improved both the cell voltage and CO selectivity (62% at 200 mA cm−2), but cation crossover could cause complications. The results demonstrate viable strategies for improving a BPM CO2 electrolyzer toward practical‐scale CO2‐to‐chemicals conversion.


Introduction
Electrochemical CO 2 reduction, combined with electricity from renewable sources, is a key technology to achieve net zero by converting waste CO 2 into valuable chemicals, thus enabling a closed-loop carbon cycle. [1] To achieve practical-scale CO 2 reduction, we need to consider not only metrics of the electrolyzer itself (current density, Faradaic efficiency), but also other components in the process such as product separation. In a well-Ag catalysts, and the device operated with only CO 2 and pure water as feeds. The intrinsic catalyst selectivity avoids the use of a pH buffering layer and high alkali salt concentrations which is potentially beneficial as salt precipitation is completely avoided. However, the full cell voltages were relatively high, and Ni cyclam was affected by product inhibition and eventual degradation which led to low stability (on a scale of <1 h).
In this work, we report strategies to improve both the stability and cell voltages of this electrolyzer configuration. Cobased molecular catalysts for CO 2 reduction have been reported, mainly comprising a phthalocyanine or porphyrin ligand or their derivatives. [15] While the majority of studies were conducted in alkaline conditions (either alkaline electrolyte or an AEM zero-gap cell), some studies have reported these catalysts operating in acidic conditions, albeit at lower current densities and only in an H-cell configuration. [16] Here we chose Co phthalocyanine (CoPc), supported on carbon, as the cathode catalyst and demonstrated that it can be used at high current densities (up to 200 mA cm −2 ) with good selectivities in an acidic environment. The introduction of water dissociation (WD) catalysts into the CEL/AEL junction can enhance WD and thus lower the cell voltage, and here we report the effects of using a customized BPM, as well as the analysis of various voltage loss pathways by electrochemical impedance spectroscopy. Figure 1 shows a diagram of the cathode and BPM, and the structure of CoPc.

Results and Discussion
First, we show the overall performance of the zero-gap BPM cell (commercial Fumasep FBM membrane) with CoPc as the cathode catalyst. The cathode was constructed by spraying CoPc supported by carbon powder (17 wt% CoPc, total loading 1.2 mg cm −2 , for full experimental details see methods) onto carbon paper (5 cm 2 ). The CoPc/carbon cathode was cold-pressed in the electrolyzer cell, together with the BPM and a RuO 2 anode, with the CEL of the BPM toward the cathode (reverse-bias configuration). We used a pure H 2 O feed at the anode, to avoid possible co-ion transport and crossover of ions which could convolute the performance at the cathode. [17] In reverse biased BPM studies an alkaline electrolyte (e.g., 1 m KOH) is sometimes used at the anode and this should lower the overall cell voltage further, but crossover of K + can influence the cathode behavior (see below). Figure 2A shows the initial faradaic efficiencies to H 2 and CO, obtained during 2-electrode chronopotentiometric measurements. The FE for CO reached 69 ± 4% at a total current density of 25 mA cm −2 , decreasing to 53 ± 4% at 100 mA cm −2 and 34 ± 2% at 200 mA cm −2 . The selectivity reached here is a large improvement over our previous report with Ni cyclam and its derivative in the same cell configuration (≈20-30% at 100 mA cm −2 ), [14] mainly because CoPc does not show noticeable product inhibition by CO and subsequent reductive deactivation. The CoPc also significantly outperforms the previously reported Ag nanoparticle benchmark catalyst in this cell configuration (20±2% at 100 mA cm −2 ). [14] Figure 2B shows the full cell voltages and the CO partial current densities. A breakdown of the cell voltage and its improvement is discussed further below. Linear sweep voltammograms of the cell, comparing CoPc/carbon and blank carbon cathodes, are shown in Figure S1, Supporting Information. Under Ar, the blank carbon cathode showed activity toward hydrogen evolution reaction (HER), which is suppressed under CO 2 , as also observed by Burdyny and coworkers. [18] Interestingly, in contrast to Ag in their study, which showed increased HER throughout the potential range, CoPc under Ar showed suppressed HER at lower currents compared to the blank, indicating that CoPc has intrinsic higher overpotential for HER under our zero-gap acidic cathode environment.
The CO partial current density increased toward a plateau with total current density, reached 68±3 mA cm −2 at 200 mA cm −2 total current density. The plateau in CO partial current density is presumably due to the intrinsic turnover frequency of CoPc and total electroactive content (also observed in previous work using Ni-based molecular catalysts [14] ), and not due to mass transport limitations (see below for CO 2 flow rate dependence). We measured the electroactive coverage by cyclic voltammograms of a fresh CoPc/carbon on carbon paper electrode (see Figure S2, Supporting Information), we observed   [19] and obtained a coverage of ≈3.8 × 10 −8 mol cm −2 for a fresh cathode. Using this electroactive coverage, we calculate the maximum turnover frequency (TOF) achieved (from the maximum CO partial current density) to be 12 s −1 , which is comparable to literature. [15b] (We note that there is a large spread in the reported coverage and TOF values, as this is heavily influenced by preparation methods and cell configurations). As a minor product, CH 4 was also observed, but at the levels of <0.1% Faradaic efficiency. A small amount of CH 4 was previously reported on a Co-protoporphyrin complex [20] in acidic electrolyte (≈0.1% FE at ambient conditions, rising to 2% with 10 atm CO 2 ). The CH 4 is generated from further protonation and reduction of the adsorbed CO, and while this is promoted under acidic conditions, the contribution of this pathway is still small relative to CO generation and desorption.
We conducted longer measurements at 25 and 100 mA cm −2 (Figure 3). At 25 mA cm −2 , there was an initial decline in CO FE but it stabilized at ≈40% which was sustained for the duration of the experiment with no further decreases (up to 4 h tested). For the run at 100 mA cm −2 , the selectivity reached ≈30% after 3 h. As for cell voltage, we note that this commercial BPM (Fumasep) is not designed for operation at high current density at reverse bias (>100 mA cm −2 ) for long periods, and the voltage increase with time at 100 mA cm −2 is partly attributable to membrane degradation. The decrease in cell voltage at 25 mA cm −2 is not due to the shift toward H 2 production, since the selectivity was stable after the initial drop, but rather due to change in hydration of the membrane or the electrodes after commencing operation (see below for impedance analysis of cell voltage). Pausing the applied current (while keeping the CO 2 and H 2 O feed on) had no effect in restoring the selectivity, indicating that the partial loss in selectivity is not due to product inhibition or catalyst desorption, since in both cases we would expect some recovery in CO selectivity, either due to CO being flushed out, or catalyst readsorption, during the pause period. Overall the stability of the CoPc cathode is a large improvement on the past Ni cyclam catalyst where activity at both 25 and 100 mA cm −2 was limited to <1 h.
One possible cause of selectivity decrease is the partial flooding of the gas diffusion layer (GDL) of the cathode, which could result in lower CO 2 supply to the catalyst. Flooding can occur irrespective of carbonate formation, due to water transport through the BPM as well as water formation as part of the reaction of CO 2 conversion to CO. Periodic drying has been shown to help sustain the activity of CoPc on GDEs at higher pH's. [21] In a separate experiment, we measured the double layer capacitance of the cathode (as a quantification of water penetration into the GDE) [22] by cyclic voltammetry. Initially, after preconditioning, the cathode capacitance was 51 µF cm −2 , which increased to 514 µF cm −2 after chronopotentiometry at 100 mA cm −2 for 2 h. During this period, the CO Faradaic efficiency decreased from 42% to 18%. The cell was then disassembled and the cathode taken out to dry in ambient air overnight, and after reassembly the capacitance decreased to 306 µF cm −2 . When the chronopotentiometry was restarted at 100 mA cm −2 , there was a concomitant recovery of CO FE to 41%, which then declined again during operation, thus confirming that at least part of the selectivity decline is due to flooding.
Another possible cause of selectivity decrease is the demetallation of Co from CoPc. [23] To observe the chemical nature of the CoPc catalyst, we conducted X-ray photoelectron spectroscopy (XPS) measurements of the fresh and used (at 100 mA cm −2 , "short" = 0.5 h, "long" = 4 h) cathodes, and for reference, XPS of CoPc powder was also measured using the same parameters (Figure 4). Signals assigned to Co 2+ in CoPc were clearly observed in both fresh and used samples. The peaks at 781.1 eV and 796.6 eV correspond to Co 2p 3/2 and Co 2p 1/2 , respectively. The positions of the peaks and spin-orbit separation of 15.5 eV between them matches well with the literature. [24] The peak at 783.0 eV is assigned to multiplet splitting of Co which occur due to the unpaired electrons in the valence level. The peak could not be the satellite contribution because for it to be satellite, there should have been a satellite contribution from Co 2p 1/2 also, which is absent. [25] The accompanying C 1s and N 1s spectra are shown in Figure S3, Supporting Information, and these are consistent with CoPc on a carbon paper substrate, with contributions of C-F signals due to the added Nafion in the cathode ink. Performance of CoPc/carbon in a reverse-bias zero-gap BPM electrolyzer. A) Initial Faradaic efficiency toward CO and H 2 , and B) full cell voltage (red, left axis) and CO partial current density (blue, right axis) as a function of total current density. Conditions: Cathode area 5 cm 2 , cathode feed: CO 2 saturated by a water bubbler, 20 sccm, anode RuO 2 9 cm 2 , anolyte pure water, recirculated at 15 mL min −1 , membrane Fumasep FBM, room temperature. Error bars correspond to 1 standard deviation, from three independent samples.

www.advmatinterfaces.de
We also conducted ex situ scanning electron microscopy and energy dispersive X-ray (SEM/EDX) of the fresh and used (100 mA cm −2 , 4 h) cathode ( Figure S4 and Table S2, Supporting Information). The Co signals were clearly observed and relatively well-distributed on the µm scale, for both fresh and used samples. As a semiquantitative measure, we used the signal of Co normalized by that of F, which is assumed to be similar across the samples (due to the background Nafion and PTFE present in the carbon paper), and found that the Co:F atomic ratio was slightly decreased from 0.090 (fresh) to 0.072 (used 30 min and 4 h), indicating that a small amount of Co was lost during operation. This is also observed in the decreased electroactive coverage of CoPc in the used sample ( Figure S2, Supporting Information).
Taken together with the sustained CO selectivity (Figure 3), the characterization (XPS and EDX) indicates that although some loss of Co did occur, a substantial portion of CoPc still remained active on the cathode after operation at 100 mA cm −2 for 4 h. From the capacitance measurements and post-run characterization, we conclude that the decline in selectivity is due to both GDL flooding (leading to decreased CO 2 supply to CoPc), and CoPc demetallation. Water management and GDL flooding are critical issues in zero-gap electrolyzers, [4a,26] and future work on membrane water transport and cathode hydrophobicity is needed.
A key performance metric is the CO 2 utilization efficiency, or single-pass yield of carbon products (defined as the proportion of inlet CO 2 that was converted into the desired carbon products, CO in this case) (note that this has also been referred to as single-pass conversion in the literature, although this is a potentially confusing usage if CO 2 can also be converted into undesired products, i.e., carbonates). At constant current (100 mA cm −2 ), we varied the inlet CO 2 flow rate and measured the CO Faradaic efficiency and the CO single pass yield   Figure 5A). This was conducted on the same sample, measured after the initial drop in selectivity (after ≈1 h) such that the performance is relatively stable. The CO Faradaic efficiency only decreased slightly with lower inlet flow rates, resulting in a maximum of 51% CO yield at 3 sccm inlet flow. This is comparable to the carbon product yield of a recently reported BPM system with a Cu cathode (≈20-60% at similar inlet flow rates). [13] There is a trade-off between CO productivity (measured in amount of CO produced per unit time) and the CO yield, which was analyzed by Hawks and coworkers. [27] Higher inlet flow results in higher productivity (defined as moles or mass of CO produced per unit time), because higher CO 2 flow lessens the effect of CO 2 depletion but this comes at a cost of lower percentage yield due to dilution of the product stream. We illustrate this trade-off in Figure 5B, for 25 and 100 mA cm −2 , together with trendlines obtained from least-squares fitting to the relationship reported in Hawks et al. The more scattered fit at the higher current is due to degradation during the measurement (conducted in the order 10, 3, 40, 5, 20 sccm).
Up to this point, we have been using a commercially-available BPM (Fumasep FBM). However, the full cell voltages remain high (≈4.7 V at 100 mA cm −2 ). In addition, this commercial Fumasep BPM is not suited for long-term operation at >100 mA cm −2 , according to the supplier. This has resulted in variation in cell voltage after long operation, as well as batchto-batch differences. The voltage requirement of a BPM electrolyzer depends on the ion transport characteristics of the CEL and AEL, and the rate of water dissociation (WD) at the CEL/AEL junction. The addition of catalysts (e.g., metal oxides, graphene oxide) has been shown to increase the WD rate at the junction, leading to higher currents during reverse-bias operation. [28] Here, we constructed a custom BPM composed of a Nafion 117 membrane as the CEL, a Sustainion X37-50 membrane as the AEL, and TiO 2 (anatase) particles as WD catalysts at the CEL/AEL junction (at a loading of 30 µg cm −2 by airbrushing). Figure 6A compares the CO FE obtained from this custom Figure 5. A) Dependence of CO Faradaic efficiency (black, left axis) and CO single-pass yield (orange, right axis) on the CO 2 inlet flow rate. B) Trade-off between CO single-pass yield and CO productivity, at 25 mA cm −2 (blue) and 100 mA cm −2 (black). Conditions: Cathode CoPc/carbon area 5 cm 2 , cathode feed: CO 2 saturated by a water bubbler, anode RuO 2 9 cm 2 , anolyte pure water, recirculated at 15 mL min −1 , membrane Fumasep FBM. Conditions: Cathode CoPc/carbon area 5 cm 2 , cathode feed: CO 2 saturated by a water bubbler, 20 sccm, anode RuO 2 9 cm 2 , anolyte pure water, recirculated at 15 mL min −1 .

www.advmatinterfaces.de
BPM to the commercial Fumasep BPM (from Figure 2A). The cathode and anode remained the same CoPc/carbon and RuO 2 , respectively. The CO selectivity using the custom BPM was significantly lower than the Fumasep BPM throughout the current density range measured here, we believe this is because it generated a more acidic local environment at the cathode compared to the Fumasep BPM. Whilst we can reasonably conclude that the increased hydrogen evolution is due to a change in pH, as noted alongside Figure S5, Supporting Information, the local proton concentration is due to a number of factors meaning the cause of lower selectivity cannot be readily ascertained.
Future improvements to the selectivity can be expected, for example by modifying the acidity of the CEL, or by inserting a thin intervening layer between the CEL and the cathode. The comparison of cell voltages is shown in Figure 6B. A custom BPM with no WD catalyst showed very high cell voltages (>7 V at 30 mA cm −2 ), demonstrating that TiO 2 was effective in promoting water dissociation. Overall, the cell voltage from the custom BPM with a TiO 2 WD catalyst was ≈0.5-0.8 V lower than that of the commercial Fumasep membrane.
To identify the contributions of various processes, we conducted electrochemical impedance spectroscopy (EIS) measurements at various operating current densities. Previous work by Chen et al. [28a] deconvoluted two processes, and attributed the higher-frequency and lower-frequency semicircles to the water dissociation at the BPM (R WD ), and charge-transfer at the electrodes (R CT ), respectively. Here we also observe 2 semicircles, and fitted each of the process to a constant-phase element (CPE), for a non-ideal capacitive element, in parallel to a resistance element ( Figure 7A). We term the high-frequency intercept the cell resistance (R cell ), and this comprises the cell plates and contacts, the ionic transport through the membranes, and the resistance of the WD catalyst layer itself. Figure 7A shows the Nyquist plot of the resistance-area product (the real and imaginary parts of the impedance normalized by multiplying the cathode area) of the cell with either the commercial Fumasep or the custom BPM (Nafion/TiO 2 /Sustainion) under operation at 25 mA cm −2 , which gives a reasonable fitting with the circuit in Figure 7A. The EIS measurement was conducted at operating current density 25-200 mA cm −2 and Figure 7B,C shows these resistances (R cell , R WD , R CT ) for the Fumasep and the custom BPM. The deconvolution of these three resistances allowed a breakdown of the overall cell voltage ( Figure 7D,E).
Since the cell materials, cathode, and anode are the same for all measurements, the change in R cell is attributable mainly to the ionic transport properties of the membrane. For the Fumasep BPM, overall the R cell was higher than that of the custom BPM, indicating that the Nafion and Sustainion membranes, taken together, showed better ionic transport properties than the Fumasep FBM. The R cell of Fumasep FBM showed a slight decrease with increasing current, unlike the custom BPM which was relatively unchanged, and we attribute this to a change in hydration with more H 2 O pulled through at higher currents. We also note that our results were obtained at room temperature (20-22 °C), and the Fumasep FBM's stated upper limit is 40 °C, but industrial electrolyzers are operated at elevated temperatures (>60 °C) which would be beneficial for improving the membrane ionic transport. The R CT (which contains contributions from both the cathode and the anode) was lower in the custom BPM, in line with a lower local pH at the cathode. The R WD was overall slightly higher in the custom BPM, indicating that future optimization of the WD catalyst (material, loading amount, loading methods) should be able to reduce the cell voltage down even further.
We measured the changes in each of these resistances, as well as the cathode capacitance, at hourly intervals during operation for 3 h at 100 mA cm −2 using the Fumasep membrane ( Figure S6, Supporting Information). The cell resistance remained relatively constant, and the water dissociation resistance increased slightly, while the charge transfer resistance increased noticeably, indicating possible instability at the cathode or anode. The capacitance also increased significantly, which is associated with flooding as noted above.
Finally, we investigate the effect of changing the anolyte from pure H 2 O to a highly alkaline (1 m KOH) electrolyte, and the results are shown in Figure 8, using the Fumasep BPM for comparison with pure H 2 O anolyte results (from Figure 2A). As expected, the cell voltage decreased by ≈0.7-1.0 V, due to the improved anode kinetics as well as lower solution resistance. Although the cathode is separated from the anolyte, we found that CO selectivity was increased at high current density compared to using pure water anolyte (62% versus 34% at 200 mA cm −2 ). A selectivity enhancement effect has been previously observed in a zero-gap BPM system with a Ag cathode when a K + based anolyte was used, [17] and the improvement was attributed to the crossover of cations (K + ) from the anolyte to the cathode through the BPM. The beneficial effect of cations was confirmed in a separate experiment, adding KCl directly to the CoPc/carbon cathode, while retaining pure H 2 O as the anolyte ( Figure S7, Supporting Information). The extent of CO selectivity enhancement was almost identical to Figure 8A with KOH anolyte.
The effect of cations in enhancing the selectivity in CO 2 reduction is well-studied on metal cathodes, though there is still no consensus on the main cause (modification of interfacial electric field, buffering the local pH, or stabilization of an intermediate [29] ). Here we also find evidence for K + crossover to the cathode from the anolyte ( Figure S8, Supporting Information). The cation effect is less studied on molecular catalysts, but there are reports on selectivity enhancement over CoPc [30] and Fe porphyrins, [31] mainly attributed to stabilization of the M-CO 2 adduct (M is the transition metal center). Whilst this paper was in preparation a preliminary report of K + stabilization of CO 2 binding at a similar Co catalyst has also been reported supporting that a cation effect may also be occurring here. [32] Future work will explore the beneficial effect of direct cation addition to the cathode, which we caution may become convoluted with potential detrimental changes associated with increased flooding and salt precipitation as a result of the K + crossover, Figure S9, Supporting Information (extended 44 h experiment).
Here we have shown that high Faradic efficiencies for CO production can be achieved in a reverse biased BPM electrolyzer operating at up to 200 mA cm −2 , despite the acidic environment, by using a CoPc molecular electrocatalyst. We have also demonstrated that a reduction in cell voltage of a BPM electrolyzer is achievable via BPM design and optimizing reaction conditions. Although the overall voltages are still higher than typical monopolar membrane (AEM) devices the costs associated with carbonate regeneration and CO 2 separation from crossover into the anode stream need to be included. CO 2 crossover losses are minimized by the use of the BPM struc-ture and here we achieve a single-pass CO yield of 51%. Various approaches for evaluating electrolyzers have been proposed, from simpler models considering only the energetic costs [5,33] to more complex models involving other economic parameters such as separator cost, electrolyzer cost, plant lifetime, etc. [34] Although these economic models vary with different assumptions and parameters, some of which still have large  , green), on the operating current density. D,E) Cell voltage breakdown, the voltages are summed so the green line provides the total cell voltage. Conditions: cathode CoPc/carbon area 5 cm 2 , cathode feed: CO 2 saturated by a water bubbler, anode RuO 2 9 cm 2 , anolyte pure water, recirculated at 15 mL min −1 .

www.advmatinterfaces.de
uncertainties involved due to the technology being in an early stage, systems-level analysis that take downstream processes into account suggest that BPM electrolyzers are competitive due to high CO 2 utilization efficiencies, even when considering potentially higher cell voltages.

Conclusion
Bipolar membrane zero-gap electrolyzers are promising CO 2 reduction devices, due to low carbonate formation and crossover, which translates to high CO 2 utilization and product yield. However, challenges remain in the low selectivity, due to the acidic CEL being in direct contact with the cathode, and the additional voltage requirements for water dissociation within the BPM. Here, using only CO 2 and pure water feeds, we demonstrate that selection of CoPc as the cathode catalyst afforded a higher CO selectivity compared to previously reported benchmark Ni cyclam-based molecular catalysts and Ag metal catalysts. CoPc cathodes show stable selectivities for CO over 4 h experiments at 25 mA cm −2 . At higher current densities (100 mA cm −2 ) there is some loss of selectivity which we propose is primarily due to partial flooding of the gas-diffusion layer. Electrochemical impedance spectroscopy led to a detailed analysis of voltage losses, which is crucial in future development of such BPMs and BPM-containing devices. Switching the anolyte from pure water to 1 m KOH improved both CO selectivity and cell voltage, proposed to be due to cation crossover however further studies are required to assess the mechanism of how KOH enhances the selectivity of this molecular catalyst cathode. Overall, this study has demonstrated strategies to improve key components of a BPM electrolyzer toward practical-scale electrochemical CO 2 conversion.

Experimental Section
Materials: Co(II) phthalocyanine (Sigma Aldrich, 97%), Ensaco 350G carbon powder (Imerys), TiO 2 (anatase, <25 nm, Sigma Aldrich, 99.7%), isopropanol (Sigma Aldrich, 99.5%), carbon paper (Sigracet 39BB), RuO 2 (FuelCellStore, nanoparticles ≈5-10 nm), Fumasep FBM bipolar membrane (stored in 1 m NaCl) were used. Nafion solution (Sigma Aldrich, 5% in a mixture of a lower aliphatic alcohols and water), CO 2 (BOC, CP grade), Nafion 117 membrane (FuelCellStore), Sustainion membrane (X37-50, Grade RT) were also used. Electrode and Membrane Fabrication: For the cathode, first the Ensaco carbon powder was dispersed in isopropanol by sonication for 1 h, then CoPc was added then sonicated for a further 30 min. The Co/carbon suspension was left to stir overnight (≈16 h), Nafion solution was added then stirred for 3 h, then sprayed (Harder & Steenbeck Evolution with a N 2 stream) onto carbon paper over a hot plate set to 40 °C. For the anode, the RuO 2 powder was dispersed in isopropanol by sonication for 1 h, then sprayed onto carbon paper over a hot plate set to 80 °C. The electrodes were left to dry in ambient air.
For the custom BPM, a procedure was adapted from the literature.
[28a] TiO 2 (anatase) was suspended in a mixture of isopropanol and water then sonicated for 1 h. This suspension was sprayed onto a sheet of Nafion 117 (at a loading of 30 µg cm −2 ), held on a glass holder on a hot plate set at 90 °C. After spraying, the TiO 2 -loaded Nafion was returned to and stored in H 2 O, then it was interfaced with Sustainion (with the TiO 2 -loaded side in the middle) directly before assembly of the electrochemical cell.
Electrochemical Measurement: Electrochemical measurements were carried out using an Ivium Vertex potentiostat. The membrane electrode assembly was assembled in the electrolyzer cell (dioxide materials, cathode area 5 cm 2 ) at ambient conditions ("cold pressing"). The Fumasep membrane was soaked in H 2 O for 1 h before use. The cation exchange layer of the BPM was toward the cathode ("reverse bias"). The cell was tightened to 3 Nm using a torque wrench. The CO 2 inlet stream was passed through a water saturator at room temperature, at a flow rate of 20 sccm unless specified. The anolyte was pure H 2 O, typically 100 mL, recirculated at a rate of 15 mL min −1 . All measurements were taken at room temperature (20-22 °C). After assembly, the cell was preconditioned at open circuit, with CO 2 and H 2 O flowing, for 1 h, before starting electrochemical measurements. The CO 2 inlet flow rate was controlled by a mass flow controller (Alicat), and the outlet stream flow rate was measured using a digital flow meter (Agilent). . Conditions: Cathode CoPc/ carbon area 5 cm 2 , cathode feed: CO 2 saturated by a water bubbler, anode RuO 2 9 cm 2 , recirculated at 15 mL min −1 . Error bars correspond to 1 standard deviation, from 2 independent samples for KOH run.

www.advmatinterfaces.de
Measurement of electroactive coverage was conducted in a onecompartment glass cell with 0.1 m KHCO 3 electrolyte, Ag/AgCl reference electrode, Pt wire counter electrode, and Ar purging.
The turnover frequency (TOF) was calculated from the maximum CO partial current density, J CO , and the electroactive coverage, Γ, as follows T = TOF J nF CO (1) where n is 2, and F is Faraday's constant. The electroactive coverage was calculated from integrating the peak area during cyclic voltammetry where n is 1, A is the electrode area, and ν is the scan rate. Electrochemical impedance spectroscopy (EIS) was conducted galvanostatically, with a 20 mA amplitude, with frequencies from 100 kHz to 1 Hz. The cell was held at the operating current for 3 min for equilibration prior to starting the measurement. The equivalent circuit fitting was conducted using the IviumSoft potentiostat control program.
The equilibrium (thermodynamic) voltage was calculated from yielding E cell,equiv. = 1.34 V, assuming the same pH at both electrodes.
The voltage requirements for cell resistance and water dissociation resistance were calculated by summing the partial voltages Product Detection: The outlet stream of the electrolyzer was connected to a gas chromatograph (Varian CP-4900 MicroGC) with a Molsieve 5Å column with Ar carrier gas for H 2 and CO detection by a thermal conductivity detector. The initial injection was taken 2 min after the start of the chronopotentiometry.
Faradaic efficiency calculation ( )= = × xv P RT JA F CO FE % Current toward CO production Total current / /2 100 out (6) where F is Faraday's constant, J is current density, A is electrode area, ν out is the total volumetric outlet flow rate, x is the outlet molar fraction of CO, P is the pressure, R is the gas constant, and T is the temperature.
where x is the outlet molar fraction of CO, ν out is the total volumetric outlet flow rate, and ν in is the inlet volumetric CO 2 flow rate. Characterization: SEM and EDX measurements were carried out with a Hitachi SEM S4800 at 20 kV. XPS examinations were carried out using Thermo Scientific K-Alpha X-ray photoelectron spectrometer fitted with Al Kα X-ray source (1486.7 eV). The samples were analyzed without further surface cleaning on a spot of 400 × 400 µm 2 area. The survey measurements were recorded in 0-1350 eV range at 200 eV pass energy and the high-resolution scans for elements of interest were obtained in the appropriate range at 50 eV pass energy. The spectra were calibrated by using C 1s peak at 285.0 eV as reference.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.