Protecting effect of mass transport during electrochemical reduction of oxygenated carbon dioxide feedstocks

Electrochemical CO 2 reduction is a promising path toward mitigating carbon emissions while also monetizing waste gas through chemicals production and storage of surplus renewable energy. However, deploying such a technology for use on industrial CO 2 sources requires an understanding of the e ﬀ ects that gas feed impurities have upon CO 2 reduction reaction (CO 2 RR). In this work, we elucidate the impact of molecular oxygen on the network of reactions occurring in a CO 2 reduction system. Our ﬁ ndings indicate that for a planar, polycrystalline Au electrode in an aqueous environment, oxygen reduction current is limited by the transport characteristics speci ﬁ c to the cell geometry and solvent; as a result, mass transport confers a protective e ﬀ ect by mitigating the otherwise thermodynamically and kinetically favorable reduction of oxygen. The presence of oxygen does not appear to have a signi ﬁ cant impact on either CO 2 RR or hydrogen evolution partial currents, indicating that the mechanisms of reduction reactions involving oxygen are independent of CO 2 RR and hydrogen evolution. Further, an electrokinetic mechanistic analysis indicates many feasible candidates for the rate-determining step of CO 2 RR; there is no indication that the CO 2 RR mechanism at P CO 2 ¼ 0.5 atm is altered by the presence of oxygen, as the Tafel slopes (59 mV dec (cid:2) 1 ) and reaction orders with respect to bicarbonate (0), CO 2 ( (cid:3) 1.5), and protons (0 from lack of KIE) are consistent between systems with P O 2 ¼ 0 atm and those with P O 2 ¼ 0.5 atm. While this is promising for the robustness of CO 2 RR to oxygen impurities in gas feeds, the ultimate design tradeo ﬀ when utilizing CO 2 sources containing oxygen is between the cost of separation processes and the corresponding cost of power ine ﬃ ciency as a result of electrons lost to oxygen reduction. This represents a ﬁ rst step in understanding kinetic and transport considerations in the design of gas-impurity-tolerant CO 2 reduction systems. ORR


Introduction
With the rise of concern regarding carbon dioxide (CO 2 ) emissions, CO 2 valorization as a means of energy storage or chemicals production is an attractive pursuit. While round-trip efficiencies for energy storage technologies such as lithium ion batteries and pumped hydro power are greater than those for CO 2 electroreduction to fuels, 1 the generation of energy-dense chemicals allows for exibility of energy recovery strategy as well as for product utilization in synthetic organic applications. CO 2 reduction reactions (CO 2 RR) have been employed to reach products ranging from carbon monoxide (CO) and formate to more deeply-reduced products such as ethylene, methanol, and ethanol. [2][3][4] Generating such products from point sources of CO 2 may prove to be an attractive means of both curtailing and monetizing CO 2 emissions in the future.
A major limitation of existing studies of CO 2 electroreduction, however, is that almost all experiments conducted to date have utilized gas feeds of 99.9% or purer CO 2 . Those investigations which have focused on dilute streams of CO 2 have so far mostly done so in the context of low-conversion molecular catalysts and inert gas impurities such as nitrogen (N 2 ). 5 If CO 2 reduction is ever to be practically employed, it may not be feasiblelogistically or economicallyto expect endusers of CO 2 reduction technology to supply feedstocks of high CO 2 purity. Some estimates have indicated that the puri-cation of CO 2 from air (which currently contains $400 ppm or 0.04% CO 2 ) 6 would cost roughly $1000 per metric ton CO 2 captured, while similar capture from the effluents of coal-red power plants (containing 10-15% CO 2 ) would cost about $25-100 per metric ton CO 2 . 7 A more promising study on the design of a system for direct CO 2 capture from air was able to decrease the cost of air capture to $94-232 per ton CO 2 . 8 However, even this cost may prove prohibitive for the commercialization of CO 2 reduction technology, as many technoeconomic analyses investigating CO 2 reduction viability rely on CO 2 capture costs being no more than $30-60 per ton CO 2 . 9,10 Therefore, it would be valuable for developed CO 2 reduction technologies to exhibit some degree of tolerance to gas feed impurities.
Previous studies involving non-inert impurities in CO 2 reduction streams have focused on the effects of NO x , SO x , and O 2 on CO 2 RR at Cu cathodes. 11,12 In these contexts, O 2 was treated either as a species with the propensity to interfere with other impurities (e.g. by oxidizing poisoning sulde ions), 11 or as a catalyst oxidant. 12 Ex situ catalyst characterization and changes in bulk product distribution were the primary metrics of the impurities' effects, whereas impacts on the electroreduction mechanisms occurring in the cell and the physical phenomena governing the feed impurities' behavior in these contexts were not explicitly treated. Thus, there is a need to build upon these efforts in order to ensure the catalysts being developed for CO 2 reduction today will be viable for implementation in the future.
In particular, an abundant and reactive impurity in most anticipated CO 2 feed streams is molecular oxygen (O 2 ). Oxygentolerant catalytic networks are increasingly being studied for hydrogen evolution; methods for minimizing O 2 competition include the development of bio-inspired catalytic pockets in order to exclude, photo-reduce, or otherwise steer molecular oxygen away from poisoning catalytic systems for proton reduction. 13 Platinum catalysts for the hydrogen evolution reaction (HER) seem to maintain current density to HER in the presence of 0.21 atm O 2 ; 13 however, the exact means by which this is possible have not been investigated. It is equally as important, if not more so, for the impact of O 2 on CO 2 reduction to be studied and mitigated, since O 2 may very well be a constituent of CO 2 reduction gas feeds.
The goal of the present work is to demonstrate the effects of the presence of oxygen on the network of reactions occurring during electrochemical CO 2 reduction on gold catalysts. Gold was chosen as a test catalyst due to its high activity toward CO 2 reduction and for ease of study due to its high selectivity toward only gas-phase products (CO and H 2 ). 2,4,[14][15][16][17][18] Cultivating an understanding of oxygen's effect on such an electrocatalytic system will pave the way for the development of systems with tolerance toward a broader variety of gas-phase impurities characteristic of many CO 2 point sources, which may be required for industrially viable CO 2 reduction.

Results and discussion
Materials and methods Electrochemical cell. A two-compartment cell design was used for CO 2 reduction experiments (Fig. S1 †). 19 A full treatment of cell design and assembly can be found in the ESI (Sections S.2 and S.7 †).
Catalyst, electrolyte, reference, and gas feed preparation. Gold foils were mechanically wet-polished using new 400-grit sandpaper for 3 minutes, sonicated in Milli-Q® water for 3-5 minutes, and dried with compressed nitrogen gas prior to insertion in the electrochemical cell. Unpolished platinum foil was used as the counter-electrode, and the two electrode compartments were separated by a Selemion™ AMV anionexchange membrane. Sodium bicarbonate electrolytes were prepared by dissolving sodium carbonate into Milli-Q® water to yield solutions with carbonate molarities of 1/2 the desired bicarbonate molarity, followed by bubbling overnight using instrument-grade (99.99%) CO 2 to convert the carbonate to bicarbonate. A leak-free Ag/AgCl reference electrode was used, and was calibrated to a master Ag/AgCl reference in saturated KCl before each experiment. Gas feeds during experiments were mixed by combining research-grade (99.999%) CO 2 , industrialgrade (99.7%) O 2 , and house-supplied N 2 (the inert feed) in the appropriate ratios. Total gas ow was set to 10 sccm unless otherwise noted, and the pressure in the electrochemical cell remained at roughly 1 atm. Further details on materials and experiments can be found in ESI (Section S.7 †).
Product characterization. While minor electroreduction products were detected through a combination of NMR spectroscopy and colorimetric assays (Section S.8 †), the majority of conrmed products were detected through in-line gas chromatography (GC). An SRI MG #5 Model 8610C GC was set up to auto-sample gas effluent from the electrochemical cell during operation. Products were detected using both a ame ionization detector (FID) and a thermal conductivity detector (TCD). Full details of GC setup and calibration can be found in ESI (Section S.8 †). GC samples taken at t ¼ 20, 30, and 40 minutes during each electrolysis experiment were averaged for the data reported herein. Oxygen reduction reaction (ORR) current densities for most experiments were calculated from the balance of current aer accounting for all detected products. The validity of this approach is conrmed both by lack of detectable liquid-phase products (Section S.8.2 †) and by FE closure to 100% in the absence of oxygen.
Oxygen's effect on reactions in a CO 2 RR cell Initial experiments were conducted holding the partial pressure of CO 2 constant at 0.5 atm, while the pressure of O 2 was varied between 0 and 0.5 atm using N 2 as the balance. The effect of the oxygen partial pressure on partial current densities toward CO 2 RR, HER, and ORR was quantied (Fig. 1). Results of this experiment conducted at different voltages can be found in ESI (Section S.9 †).
As increasing amounts of oxygen were added to the gas feed, current toward ORR increased linearly while partial currents toward CO 2 RR and HER remained roughly constant. This implies that ORR chemistry did not affect CO 2 RR or HER catalysis. Further, it suggests the nature of the catalyst itself may have been unaffected by the presence of oxygen. It should be noted that there is somewhat larger uncertainty in the partial current measurements toward CO 2 RR and HER, consistent with previous studies using foil electrodes. 20,21 This is because foil preparation by polishing introduces surface irregularities that cannot be easily replicated from run to run. However, the uncertainties expressed in the error bars of this gure are noticeably small for the ORR data points. This is indicative that the quality of the foil preparation has no effect on ORR currents, and reinforces our condence in the data.
At less reductive potentials and lower current densities, an apparent negative trend in both CO 2 RR and HER current is observed (Fig. S11 †). The fact that this trend is evident only at low overpotentials suggests either that oxygen interferes with these mechanisms in a distinct way at lower overpotentials, or that the effect of O 2 in the system is constant, but kept small on an absolute scale, therefore only showing up at low current densities.
To investigate why these trends hold true, the factors governing each reactionnamely, the kinetic versus transport control of ORR, as well as mechanistic aspects of CO 2 RR and HERwere studied in the context of P CO 2 ¼ 0.5 atm.
ORR under CO 2 reduction conditions on gold CO 2 RR, HER, and ORR can be described through the following respective equilibria on gold at pH 7: 2,22 Thermodynamics would dictate that at CO 2 reduction conditions that allow for the formation of CO, reducing oxygen is extremely favorable. Since the equilibrium potential for CO 2 reduction is roughly À0.52 V vs. SHE in neutral media, the cathode must be held at more reductive potentials than À0.52 V vs. SHE in order to achieve signicant CO 2 reduction currents. It is reasonable to imagine that at greater than 1.3 V of overpotential, the rate of ORR may not be strictly dictated by kinetics. However, it was prudent to conrm this experimentally.
In this vein, cyclic voltammograms taken at P CO 2 ¼ 0.5 atm were compared with and without a half-atmosphere of oxygen ( Fig. 2A). It can be observed that at around 0.25 V vs. SHE, the onset of ORR occurs. This reduction current density plateaus at À0.26 mA cm À2 around À0.2 V vs. SHE. HER and CO 2 RR onset, meanwhile, do not occur until the potential is swept down to almost À0.8 V vs. SHE. It is therefore concluded from this plot that ORR is transport-limited at the CO 2 RR potentials of interest. This explains the phenomenon wherein ORR current is not impacted by changes introduced to the catalyst surface by polishingthe current is limited by diffusion.
For a planar electrode, assuming a stagnant boundary layer model (Fig. 2B), the formation of a linear species concentration prole at the catalyst surface leads to the expression: 23 where i lim,j represents the diffusion-limited current density for a reduction reaction involving species j, n is the number of electrons involved in the reduction, F is Faraday's constant, D j is the diffusivity of species j, c bulk,j is the concentration of species j in the bulk of the solution (in turn dictated by Henry's law in the dilute limit), and d j is the mass transfer boundary layer thickness for species j. Because of the relative insolubility of oxygen in waterthe Henry's law constant for O 2 is 0.0013 mol per kg per bar, compared to 0.034 mol per kg per bar for CO 2 24 the relative transport-limited current density of oxygen compared to carbon dioxide reduction in solution is quite low.  To understand more about the mass-transport boundary layer in the cell, its thickness was determined by measuring the diffusion-limited current density of the ferricyanide ion and back-calculating its mass-transport boundary layer thickness using eqn (1). 25 A conversion was then applied to account for the difference in the diffusion coefficients of ferricyanide and gasphase species such as CO 2 and O 2 (see ESI Section S.7.5 †). For the transport of dissolved O 2 gas, the calculated boundary layer thickness was 200 AE 7 mm. Using the values n ¼ 4 (see below), D O 2 ¼ 2.10 Â 10 À5 cm 2 s À1 , 26 and C bulk,O 2 ¼ 0.65 mM at 0.5 atm, 24 the resulting prediction for the ORR diffusion-limited current density is 0.26 AE 0.01 mA cm À2 , in remarkable agreement with our voltammogram.
In addition, it was necessary to conrm which type of ORR was occurring under CO 2 reduction conditions. Gold is a known peroxide-forming ORR catalyst, 27 more typically implicated in selective H 2 O 2 formation under acidic conditions, 28,29 whereas within the bicarbonate mass transport boundary layer, conditions should be slightly basic (up to pH 9-10, depending on the current density drawn). 30 Long-term experiments passing over 50C of charge were conducted in order to study the accumulation of ORR liquid-phase products; the colorimetric assay corresponding to the quantication of hydrogen peroxide in this electrolyte returned negative results, meaning that less than 10% of the oxygen being reduced is forming hydrogen peroxide (Section S.8.3 †). We can therefore posit not only that ORR is transport-limited at relevant testing potentials, but also that it is primarily forming water. As water is the chief component of the electrolyte, the ORR products are not expected to interfere with any other catalytic cycles occurring in the cell.

Changes to mechanism of CO 2 reduction
To follow up on these observations, it was pertinent to interrogate the effect that oxygen's presence in the CO 2 reduction environment had on the HER and CO 2 RR mechanisms, especially in the low-current density regime. As the kinetics of HER were difficult to probea kinetically interpretable Tafel slope for HER was never attained, consistent with previous observations 16 the focus of this investigation was on the mechanism of CO 2 RR. Probing the effect of oxygen in this context necessitates not only the study of the CO 2 reduction mechanism under oxygenated conditions, but also the study of the same mechanism at dilute concentrations and with no oxygen present, as most investigations of CO 2 reduction to this point have been conducted at 1 atm of CO 2 . [15][16][17][18]31 To this end, a mechanistic study of CO 2 reduction was conducted.
Kinetic data were captured through Tafel analysis, bicarbonate dependence, CO 2 dependence, and kinetic isotope effect (KIE) experiments. The results are displayed and discussed below (Fig. 3).
Tafel slope. Tafel slopes were measured under 0.5 atm of CO 2 on Au, both with and without 0.5 atm O 2 (Fig. 3A). The Tafel slope for CO 2 RR under the testing conditions without oxygen is 72 AE 4 mV dec À1 , while in the presence of oxygen the slope is 68 AE 6 mV dec À1 . The results presented are consistent with a theoretical 59 mV dec À1 Tafel slope. The resulting data were interpreted as being linear at potentials less reductive than À0.845 V vs. SHE. At potentials more reductive than this value, nonlinearities are observed which may be due to transport effects, mechanistic changes, or non-exponential kinetic response at high overpotentials.
The Tafel slope of 59 mV dec À1 indicates that a chemical step past the initial electron transfer to CO 2 is the rate-determining step (RDS) (see ESI Section S.11 †). Subsequent experiments were designed to probe what this chemical step might be. Experiments were intended to probe purely the kinetics of the reaction and therefore were primarily conducted in the linear Tafel regionhere selected as 300 mV overpotential, or À0.41 V vs. RHE in 0.1 M NaHCO 3 (À0.82 V vs. SHE).
Bicarbonate order dependence. One possibility for the RDS is the transfer of a proton from bicarbonate to the adsorbed CO 2 anion radical q CO2 À to form q COOH (here q is used to denote a surface adsorption site). The most straightforward way to test this RDS possibility is to conduct a bicarbonate order dependence test. This was done by varying the concentration of sodium bicarbonate in the electrolyte while holding the total solution ionic strength constant with sodium perchlorate. Perchlorate was regarded as an innocent bystander ion, while buffers such as phosphate were avoided due to their ability to poison the catalyst surface. 17 Further, the potential with respect to SHE rather than RHE was held constant, as this voltage is absolute in nature, and therefore represents the activity of the electron in the reactiona factor which should be held constant when trying to probe the effects of a different reagent on the reaction rate. A full treatment of this topic can be found in the ESI (Section S.11 †). These tests were performed at P CO 2 ¼ 0.5 atm both with and without oxygen (Fig. 3B). In processing this data, partial current density results were normalized by the solubility of CO 2this controls for the secondary "salting-out" effect (see Section S.7.2 †). 18,32 While most conceivable CO 2 RR mechanisms consistent with the Tafel data might have either a zeroth-or rst-order dependence on bicarbonate, the slope in the bicarbonate plot suggests that CO 2 RR is negatively impacted by increased bicarbonate concentrations. The slope without O 2 is À0.1 AE 0.2, while the slope with O 2 is À0.23 AE 0.05. This trend is especially true as we approach the solubility limit of sodium bicarbonate at roughly 1.2 M (see Section S.7.2 †). 33 Results of the same experiment repeated with potassium bicarbonate, which is of a different purity than and nearly three times as soluble as the sodium salt, are shown in the ESI (Section S.7.2.2 †). Because the potassium bicarbonate data yields a fairly clear zero-order dependence on bicarbonate, it is anticipated that the negative slope in the sodium case is the result of a secondary effect and that the reaction order in bicarbonate is 0. CO 2 order dependence. Experiments probing the effect of P CO 2 in the inlet gas on CO 2 RR current were conducted (Fig. 3C). It should be noted that changes in P CO 2 corresponded to bulk electrolyte pH changes; however, since bicarbonate order was found to be 0, this was regarded as having no convoluting effect on the CO 2 order dependence measurement. With slopes of 1.6 AE 0.2 and 1.2 AE 0.3 respectively for the cases with and without oxygen, the reaction order with respect to CO 2 can neither be said to be 1 nor 2. In fact, it is possible that we are operating in a regime of mixed control, where the rates of different paths toward CO 2 reductionone of which involves a solution-phase CO 2 accepting an oxygen atom from a surface-bound species to promote reduction of the bound speciesare comparable. From these experiments alone, we cannot entirely rule out the possibility that the activity of CO 2 in solution may have a superlinear effect on current densities toward CO 2 reduction, although the slope here could also be due to changes in adsorbate coverage and other second-order effects.
Kinetic isotope effect. A proton involved in the RDS of CO 2 RR need not come from the bicarbonate ion; protons are also plentiful in the aqueous solvent. To examine the role of protons in the RDS, it is useful to note that a proton transfer occurring in an RDS is normally subject to a kinetic isotope effect; that is, because the proton is so light, and because its vibrational mode is the coordinate along which the reaction proceeds, affecting the frequency of vibration by swapping out an H for a D can elucidate the participation of the H in the reaction. Thus, experiments were conducted in which the H 2 O solvent was replaced by D 2 O, with proper accounting for additional required drying, membrane soaking, and bubbler replacement in order to minimize adventitious H 2 O sources (Fig. 3D) (Section S.7.3 †).
It was shown that at the testing conditions, the KIE is negligible (KIE ¼ k H /k D ¼ 1 within error) for CO 2 reduction, while for HER the value of the KIE is greater than 10. This is suggestive that while the KIE could be considerable in this experiment given the participation of a proton in the RDS, the KIE is not observed, and therefore it is unlikely that a proton transfer is involved. Notably, KIE experiments have been known to lead to false negative conclusions, but only in very specic and rare instances. 34 Mechanistic interpretation of kinetic data. Notably, the kinetic data obtained here in the context of P CO 2 ¼ 0.5 atm differ from the outcomes presented in existing literature on CO 2 RR mechanisms with P CO 2 ¼ 1 atm. 17, 18 We therefore present a list of possible mechanisms, with a focus on those candidates which t the Tafel slope data ( Table 1). Derivations of rate laws and further discussion may be found in ESI (Section S.11 †).
In these mechanisms, CO 2 RR is regarded as proceeding through inner-sphere transformations. The mechanistic step X.1, representing an adsorptive electron transfer to CO 2 , is regarded as the rst step in each of the proposed mechanisms. While the list of mechanistic possibilities compiled here is not exhaustive, it encompasses many of the commonly reported mechanisms of CO 2 reduction to CO on Au, as well as a number of speculative possibilities. 31 In particular, excluded from this list are possibilities involving concerted proton-electron transfer (CPET). However, all mechanisms with CPET corresponding to initial CO 2 adsorption can also be collapsed down into the starting points represented by mechanistic steps C.2 and F.2, and therefore are also encompassed by this treatment.
The more commonly accepted intermediates for CO 2 RR are implicated in mechanisms A and B. However, RDS candidates A.2 and B.2 are ruled out due to lack of bicarbonate dependence and KIE, respectively.
If we allow for the participation of surface sites or adsorbed species which are uncommon in the existing literature, then many RDS candidates remain (bolded options, Table 1). These include steps involving the formation of an Au-O bond, which we deem unlikely on the basis of gold's low oxophilicity; examples of this are q COOH dissociation into adsorbed CO and OH (mechanism C, excluded on the grounds of pH dependence) and the dissociation of adsorbed CO 2 c À into CO and an oxygen atom anion radical adsorbate (D.2). Mechanisms forming an Au-O bond can be avoided by instead invoking a cationic q CO + adsorbate species, which seems only to have been described in metal complexes rather than on surfaces; 35,36 this is exemplied in mechanisms which exhibit the participation of CO 2 in the RDS (E.2) and the dissociation of the q COOH intermediate into charged species (mechanism F, excluded on the grounds of pH dependence). Further, under certain mathematical assumptions and allowing for the q COOH À anion, the desorption of CO

RDS
Tafel slope form Tafel slope at 298 K, b ¼ 0.5 Acidic proton order P CO 2 order KIE?
Outcomes of experiments to probe the kinetics are listed next to each step in every postulated mechanism, under the circumstances that the step in question is the respective mechanism's rate-determining step (RDS). q represents a catalytic active site on the Au electrode. b is the symmetry factor, interpreted as being 0.5. X.1 represents the rst step in each of the proposed mechanisms. Bolded steps remain plausible RDS candidates aer the kinetic investigation conducted here. *C.2 and F.2 are written such that any proton donor could in principle serve as the proton source, including water and bicarbonate. † G.1 is a separate mechanistic step also involved in the HER pathway. ‡ Mechanism G requires the assumption that q H is dictated by equilibrium described by the Tafel step of HER. See derivation (S.11.4). > H.n steps could in principle be almost anything, including previously listed mechanisms. H.2 is simply the rearrangement of CO 2 c À on the surface to generate an intermediate which then reacts to form products. It does not necessarily require that g differs from q.
(G.3) is an RDS that is consistent with the data. A conformation change of adsorbed CO 2 c À (H.2) may also serve as the RDS. Accumulation of surface intermediates involved in these mechanisms could be assessed by use of FT-IR in order to further narrow down the mechanistic possibilities or develop new, more plausible hypotheses. For instance, a large IR band associated with q CO may indicate that this species has accumulated on the surface and is therefore the bottleneck in CO 2 RR, supporting a CO desorption step as an RDS. The observation of such intermediates is currently a subject of debate. 16,18 More relevant to the discussion of oxygen's effects, however, is the understanding of whether the mechanism of CO 2 reduction changes under oxygenated conditions in the electrochemical cell. The ndings presented above suggest that so far, we have no reason to believe the mechanism of CO 2 reduction changes in the presence of oxygen. This is a promising result if we hope to develop oxygen-tolerant CO 2 reduction systems in the future.

Synthesizing the data into a picture of oxygen's interactions
Our understanding of the effect of ORR on the product distribution allows us to envision the interactions at the catalyst surface wherein the electrocatalytic steps involving CO 2 and HER are distinct from and unaffected by ORR intermediates and products (Scheme 1). In this conception of the relevant catalytic interactions, while the mechanisms of none of the individual reactions can be isolated, the chemistries of CO 2 RR and HER vs. ORR are nonetheless distinct from one another.
Because oxygen is less soluble than CO 2 in aqueous electrolyte, transport provides a protecting effect against ORR. However, that does not mean that ORR even in this context is entirely benign. It is important to note that for practical design purposes, a high-rate CO 2 reduction reactor may need to operate under congurations such as a gas diffusion electrode (GDE) or a owing-electrolyte design, wherein the length scales over which interfacial transport occurs are much smaller, and protecting effects of solubility are lessened. In such a scenario, the mechanistic analysis conducted here may not apply; further, power inefficiencies as a result of ORR may dominate economic considerations. Even in the context of the aqueous cell with a planar electrode reported here, assuming CO 2 : O 2 ratios typical of ue gas streams from natural gas power plants (roughly 1.67 : 1), 37 and assuming activity and selectivity to CO 2 matching those reported here at À1.06 V vs. SHE, roughly 12% of applied power would be lost to ORR. In the case that this value in a practical system is smaller than the equivalent power loss required to purify a CO 2 stream of oxygen, direct reduction of a mixed O 2 /CO 2 stream may be a viable design option. The strategy to accept loss of power in a CO 2 electrolyzer could be viewed as complementary to other techniques which have been suggested for thermochemically reducing dilute oxygen impurities in CO 2 -rich feedstocks using methane. 38

Conclusions
The presence of oxygen was shown to not have a signicant impact on the current toward CO 2 reduction at high current densities, largely due to the suppression of ORR by the transport characteristics of a ooded aqueous electrolysis system. The presence of oxygen has not to this point been implicated in affecting the mechanism of CO 2 reduction. While these are promising results for CO 2 reduction system design and scaleup, it is likely that ORR could become an economic consideration, with the design trade-off being between cost of CO 2 separation and power inefficiencies introduced by the presence of oxygen.
Future work should be dedicated toward developing a better understanding of the RDS of CO 2 RR on gold, as well as conducting similar studies on more oxophilic metals. Further, for power efficiency purposes, it may prove economical to invest effort toward the design of an inherently oxygen-tolerant CO 2 RR system through modulating transport parameters at the electrode surface.

Conflicts of interest
There are no conicts to declare. Scheme 1 Interactions between CO 2 RR, HER, and ORR during cathodic polarization on a bulk Au surface.