Editors' Choice—Flooded by Success: On the Role of Electrode Wettability in CO₂ Electrolyzers that Generate Liquid Products

To better understand the wettability of aqueous-organic mixtures in the context of CO2R, we selected formic acid (FA; reagent grade, ≥95% purity, Sigma-Aldrich), methanol (MeOH; HPLC grade, ≥99.9%, Sigma-Aldrich), ethanol (EtOH; anhydrous, 200 Proof, KOPTEC), and 1-propanol (PrOH; ACS reagent, ≥99.5%, Sigma-Aldrich) for analysis. We prepared solutions across a range of dilutions from 0 to 100% by mass with deionized (DI) water (18.2 MΩ, Milli-Q). Salt-free solutions were used to isolate the interaction between each test liquid and water. Subsequent studies may elucidate the impacts of the chemistry and concentration of dissolved salts on relevant physical properties. FA mixtures were used in place of formate salt solutions because this study focuses on the effect of introducing organic solvent components into aqueous solutions. Although formate salt product mixtures are typically reported in the literature, some electrolyzer variants utilizing porous solid electrolytes can generate salt-free, concentrated acid product streams, making a focus on FA applicable.^22,49 Additionally, organic acid solutions are known to have lower surface tension than electrolyte solutions, so FA has utility for a bounding study focused on negative surface tension deviations from water.^36,37 PTFE (FP303050, Goodfellow) and graphite (99.997%, 867-421-20, Goodfellow) sheets were used as the primary solid substrates for droplet studies. PTFE sheets were cleaned with DI water and isopropyl alcohol (IPA; ≥99.5% ACS, VWR Chemicals BDH®) and dried using compressed air prior to analysis. Graphite sheets were prepared by removing the top layer of material with Scotch® tape (MFR#: 810, 3 M). 5-μL droplets⁵⁰ were dispensed onto substrates using an automatic pipetting unit. Measurements were taken in ambient air where the temperature and relative humidity remained between 20–24 °C and 10–40%, respectively. Videos of 30–60 s duration were captured at 30 frames per second using a contact angle goniometer system (Model 200, ramé-hart) and processed using DropPy V1.0.0a0, a Python ≥3.6-based goniometer software.⁵¹ Substrates were spot-cleaned before dispensing and imaging new droplets. Contact angles were determined by fitting edges with a two-parameter Bashforth-Adams model that accounts for the effects of gravity on droplet shape. To minimize the influence of evaporation on measurements, only the first 10 s of each recorded video were used for contact angle fitting. As a result of these practical constraints, the reported contact angles may not strictly reflect the equilibrium state. Additional descriptions of experimental procedures (Sections S.1 and S.2) as well as the data collected for each trial (Tables SI and SII) are provided in the Supplementary Information.

To determine the qualitative impact of mixed organic-aqueous product streams on electrode wettability, we measured the contact angles for the solutions described above as a function of water content, as shown in Fig. 1. The markers for each product represent the average contact angle from 5 trials at each concentration and the error bars are one standard deviation of the same measurements. The water contact angle on PTFE is found to be 112° ± 1.5°, which is consistent with previous work.⁵² The water contact angle on the graphite sheet is measured at 131° ± 1.8°. As expected, the contact angles of the mixtures on both surfaces decrease with increasing mass fraction of organic species due to the reductions in surface tension. The tendency to wet the solids is directly proportional to the carbon chain length of the product which is associated with decreased polarity and surface tension (PrOH < EtOH < MeOH < FA < water), especially for the primary alcohols.⁵³ The ability for each solid to prevent spontaneous liquid imbibition by a porous electrode can be assessed, at a high level, by comparing the point at which the test fluid would be neutrally wetting in the context of a cylindrical capillary (i.e., has a contact angle of 90°). When studying graphite, solutions with more than 10% alcohol fall below the 90°-threshold; however, the alcohols can be mixed in higher proportions before neutrally wetting conditions are reached on PTFE. In both cases, the FA mixtures reach neutrally wetting conditions at much greater mass fractions than the alcohols, suggesting that such product streams will not lead to significant changes in capillarity relative to pure aqueous solutions in PTFE-containing GDEs. As such, CO2R to FA appears to have a wide range of feasible operating compositions, exceeding the highest reported concentrations to date (ca. 15–30% by mass).^22,49 In contrast, the transition concentrations for alcohols are significantly lower and we anticipate that such compositions will be readily-achievable in practical CO2R-to-liquids electrochemical processes, posing a stability challenge for PTFE-containing GDEs.

Zoom In Zoom Out Reset image size

Beyond experimental measurements of the apparent contact angles of test fluids, γ can be used as a common predictor for the wettability of different fluid mixtures.³³ With the previous θ measurements, we can construct Zisman plots (Fig. 2) to predict the critical surface tension, γ_C0, for complete wetting (θ = 0°) and the surface tension at the cylindrical capillary transition composition, γ_C90, (θ = 90°) for both graphite (Fig. 2a) and PTFE (Fig. 2b). We fit the data (black open circles) for each surface with quadratic functions (red lines), which is reasonable based on previous analyses that used similar empirical fits.³⁴ We then predict γ_C0 values of 34.8 mN m⁻¹ for graphite (R² = 0.93, RMSE = 4.1 mN m⁻¹) and 21.9 mN m⁻¹ for PTFE (R² = 0.97, RMSE = 2.4 mN m⁻¹). We performed a similar analysis using a secondary set of well-defined test fluids and determined γ_C0 to be 14.8 mN m⁻¹ for PTFE. These data (Table SIII) along with an additional Zisman plot and linear empirical fit (Fig. S3) can be found in the Supplementary Information. As can be seen for graphite, the data below 34.8 mN m⁻¹ represent the product compositions that completely spread when contacting the solid. Note that none of the liquids tested had sufficiently low surface tensions to completely wet PTFE, so the empirical fit is needed to estimate γ_C0. The γ_C90 for graphite and PTFE are predicted to be 45.2 mN m⁻¹ and 47.2 mN m⁻¹, respectively. These values are useful for predicting sign changes in P_C, as will be considered in the next section. For comparison, Zisman reported a γ_C90 of ca. 40 mN m⁻¹ for PTFE,³⁴ but did not report a value for graphite, which is reasonable given that wettability of carbon surfaces vary widely depending on allotrope.⁵⁴ As such, microscopically smooth bulk materials that can serve as proxies for carbon particles or fibers remain elusive.⁵⁵

Zoom In Zoom Out Reset image size

While ex situ contact angle data only provide qualitative insights on wettability for porous electrodes, such understanding informs materials selection for different classes of reactions. Here, we use wettability data in combination with a simple mass balance model around the cathode reaction zone to estimate ranges of feasible operating conditions before liquid product enrichment near the gas-liquid-solid interface would be expected to induce electrode flooding. A mass balance model represented by the schematic in Fig. 3 accounts for the mass flow rates of water and organic products to/from a well-mixed liquid phase control volume. The results and possible implications of changing electrolyzer set points are primarily discussed in the context of the widely studied flowing liquid electrolyte configuration.^{8,10,11,25,26,29,56–67} However, some recently reported cell configurations can generate salt-free aqueous-organic mixtures by integrating polymer electrolyte components (dense polymer electrolyte membrane,²³ porous polymer electrolyte,⁴⁹ or ionomer-coated packed beads²²). Although the model framework is inspired by cells that use a flowing electrolyte, the zero-dimensional mass balance approach can serve to bound the operating space of CO2R systems without assuming device-specific geometry. It should also be noted that this model cannot predict location-specific flooding susceptibility based on operating conditions and cell geometry.

Zoom In Zoom Out Reset image size

Faraday's law of electrolysis connects the mass flow rates for product generation, _P,rxn, (Eq. 2) and water consumption, _W,rxn, (Eq. 3) at the cathode to the current (I) and one of the two stoichiometric constants, z_P and z_W, which correspond to the number of electrons per mole of product generated and water consumed, respectively.

$$\begin{eqnarray}&&{\dot{m}}_{{\rm{P}},{\rm{rxn}}}\,=\,\displaystyle \frac{I}{{z}_{{\rm{P}}}F}{M}_{{\rm{P}}}\end{eqnarray} \tag{ 2 }$$

$$\begin{eqnarray}&&{\dot{m}}_{{\rm{W}},{\rm{rxn}}}\,=\,\displaystyle \frac{I}{{z}_{{\rm{W}}}F}{M}_{{\rm{W}}}\end{eqnarray} \tag{ 3 }$$

In these equations, F is the Faraday constant, M_P is the molar mass (kg mol⁻¹) of a product species, and M_W is the molar mass of water. The mass flow rate of feed water, _W,in, is defined (Eq. 4) as a function of the inlet volumetric sweep rate, Q, and the density of water, ρ_W.

$$\begin{eqnarray}&&{\dot{m}}_{{\rm{W}},{\rm{in}}}\,=\,{\rho }_{{\rm{W}}}Q\end{eqnarray} \tag{ 4 }$$

Generally, the sweep rate of liquid electrolyte impacts product flux away from the catalyst layer to the bulk electrolyte and, by extension, the distribution of product concentrations along the reactor length and at the exit. In this treatment, we select Q directly (mL min^‒1) to regulate product dilution for a given current (mA), but this ability to independently control product removal and tune dilution would be hampered in polymer-electrolyte-based cells as alternative flux mechanisms, like evaporation and membrane transport, are less readily controllable.²³

We implement material balances around electrons, water, and liquid reaction products to directly calculate the total mass flow rate exiting the reactor, _out (Eq. 5), while ignoring dissolved gases (e.g., CO₂, CO, hydrogen, etc.) and dissociated ions such as hydroxide (OH^–) produced from the cathodic half-reactions as well as bicarbonate (HCO₃^‒) and carbonate (CO₃^2‒) that form as a result of carbonation reactions.⁶⁸ We also choose to set the product feed rate, _P,in, to zero in this study.

$$\begin{eqnarray}&&{\dot{m}}_{{\rm{out}}}\,=\,\left({\dot{m}}_{{\rm{W}},{\rm{in}}}-{\dot{m}}_{{\rm{W}},{\rm{rxn}}}\right)+\left({\dot{m}}_{{\rm{P}},{\rm{in}}}+{\dot{m}}_{{\rm{P}},{\rm{rxn}}}\right)\end{eqnarray} \tag{ 5 }$$

Through substitution, we define the product mass fraction, x_P (Eq. 6), as the total product mass divided by the total mass exiting the reaction zone as a function of total current, I, and inlet water volumetric sweep rate, Q.

$$\begin{eqnarray}&&\begin{array}{l}{x}_{{\rm{P}}}\,=\,\displaystyle \frac{{\dot{m}}_{{\rm{P}},{\rm{in}}}+{\dot{m}}_{{\rm{P}},{\rm{rxn}}}}{{\dot{m}}_{{\rm{out}}}}\\ \,=\,\displaystyle \frac{{\dot{m}}_{{\rm{P}},{\rm{in}}}+\tfrac{I}{{z}_{{\rm{P}}}F}{M}_{{\rm{P}}}}{\left({\rho }_{{\rm{W}}}Q-\tfrac{I}{{z}_{{\rm{W}}}F}{M}_{{\rm{W}}}\right)+\left({\dot{m}}_{{\rm{P}},{\rm{in}}}+\tfrac{I}{{z}_{{\rm{P}}}F}{M}_{{\rm{P}}}\right)}\end{array}\end{eqnarray} \tag{ 6 }$$

Here, I can either represent a partial current towards a target product or, assuming 100% faradaic efficiency, a total current. The water mass fraction, x_W (Eq. 7), is readily determined from x_P because we assume a binary mixture in the liquid phase.

$$\begin{eqnarray}&&{x}_{{\rm{W}}}\,=\,1-{x}_{{\rm{P}}}\end{eqnarray} \tag{ 7 }$$

Each cathodic half reaction consumes CO₂, H₂O, and electrons and produces hydrogenated products and OH^‒ as shown in Table I. Included are the relevant stoichiometric constants—n_P (the number of moles of CO₂ per mole of product), z_P, and z_W—as well as M_P for each product. We convert from a molar basis to a mass basis because it can be more convenient to work in a laboratory setting with mass (or weight) fractions at high solute concentrations.

This simple mass balance analysis enables consideration of the cumulative impact of water consumption and organic product generation on the physical properties of the solution and the wettability of the electrode. Note that the stoichiometric constants used in this model only account for the water consumption in cathode half-reactions, as the microenvironment local to the electrode-electrolyte interface will determine flooding. However, in a full cell water is generated at the anode during the oxygen evolution reaction. Depending on the cathode reaction stoichiometry, this source could offset some or all the water consumption (Table SIV, Supplementary Information). For example, there is no net water consumed for the conversion of CO₂ to formate/FA, but CO2R-to-alcohols reactions still result in net water consumption. By focusing on the cathode water consumption, this model estimates a conservative upper bound for organic product concentrations anticipated for a given chemistry, current, and liquid sweep rate.

We have constructed composition contour plots for FA, MeOH, EtOH, and PrOH (Fig. 4) by calculating x_P across many currents and flow rates. The y-axes are reported on a base-10 log scale for clarity across several magnitudes of flow rates. Composition isoclines, reported in product content mass (%, solid lines), start at 0.1, 1, and 10% and then continue from 10–100% in increments of 10%. Sweeping the current from 0–1000 mA at fixed Q results in a linear increase in the production rate (Eq. 2), while increasing Q from 0.001–1 ml min^‒1 at fixed I decreases x_P due to their inverse relation. The product composition contours generally shift downward with increasing number of electrons transferred (i.e., deeper reduction products) and increasing molar mass. The exception is FA, which has a similar molar mass to EtOH. FA composition is less sensitive to Q at fixed I, whereas the alcohols are more likely to reach high concentration through modest changes to Q.

Zoom In Zoom Out Reset image size

Determining 90°-threshold compositions from ex situ contact angle data allows us to estimate a band of operating conditions that may lead to an unfavorable P_C sign change (i.e., from positive to negative). We use wettability metrics for PTFE to represent GDE stability because they are assumed to be invariant to mild voltage biases within the electrode. In contrast, graphite is the more polarizable GDE component, so we may anticipate that its wettability will increase as a function of electrode voltage in accordance with electrowetting phenomena.^69,70 The measured transition composition, here, corresponding to a measured 90° contact angle on planar PTFE, is indicated with a black dot-dash (– –) line for each of the product subpanels in Fig. 4. These compositions were determined by interpolating between measured data points (Fig. S4, Supplementary Information). If making predictions using a Zisman rule, all liquids with γ below that of a threshold value, which is either 47 mN m⁻¹ (this work, ) or 40 mN m⁻¹ (Zisman, – – –),³⁴ should wet PTFE with a contact angle less than 90°. We estimate the transition composition for each CO2R product by finding the water composition at which the γ curves (Fig. S2b, Supplementary Information) reach the 90°-threshold. While there are discrepancies between the measurements and predicted isoclines, the differences between the operating conditions needed to achieve each composition are relatively minor. At flow rates above each transition line, the sweep stream provides enough water to the reaction zone at a given current to keep the product composition below the critical imbibition point. Put another way, for a given sweep rate, the electrochemical conversion rate is slow enough that enrichment of organic species in the reaction zone is not so great as to lead to flooding.

In agreement with the contact angle measurements, the ordering and position of the transition composition isoclines in Q-I space align with the γ and polarity of the organic species (Fig. S2b, Supplementary Information). Plotting the isoclines for different liquid species together (Fig. 5) is an effective way for determining if electrolyzer operating conditions need to be tailored according to product identity. For example, although FA mixtures reach the threshold at much higher concentrations as compared to the alcohol mixtures, the operating conditions required to reach zero capillary pressure are similar for species of equivalent polarity. At the extremes of species wettability (i.e., PrOH vs FA), however, the Q required to induce a contact angle transition varies by nearly an order of magnitude at the same I.

Zoom In Zoom Out Reset image size

Now with x_P mapped to different operating conditions, we can connect the wettability of the various liquid mixtures to a simple prediction of equilibrium P_C using Eq. 1, which is helpful for understanding the pressure differentials required to maintain a stable gas-liquid-solid interface in a gas fed CO₂ electrolyzer. Again, here we do not initially consider complex physical and geometric features evident in real GDE materials (e.g., thickness, pore size distribution, fiber spacing, particle sizes, mixed wettability)^71,72 to determine P_C or flooding dynamics because simplified models suffice for capturing general trends in capillarity. However, further analyses explicitly considering saturation or wetting dynamics in electrodes with finite volume could expand from these zeroth-order analyses of interfacial P_C and refine predictions of stable operating envelopes.

We compute P_C at various levels of water content, x_W, (Fig. 6) in order to translate product composition to equilibrium interfacial pressure along the contours in Fig. 4. The P_C data associated with this figure are reported in Table SV in the Supplementary Information. We employ interpolated PTFE contact angle values (Fig. S4) to calculate P_C with diameters of 30 μm and 0.1 μm as representative of the effective pore dimensions for macroporous carbon fiber substrates and microporous layers, respectively.⁷² As the pore diameters differ by a factor of 300, the P_C scales accordingly. Results can be interpreted for each pore by using the left (30 μm) and right (0.1 μm) vertical axes of Fig. 6. Generally, the smaller pores characteristic of a microporous layer—assuming it is crack-free—exhibit greater capillary pressure values as compared to the larger pores characteristic of a carbon fiber substrate.⁷³ For a given set of intrinsic solid-liquid affinities (specified by γ and θ), the pore diameter can serve to modulate the driving force for imbibition. If we were to overlay the P_C outputs onto the corresponding composition contours of Fig. 4, these new isoclines would serve to approximate the magnitude of the maximum liquid-gas pressure differential that a GDE could withstand while still maintaining interfacial stability. For example, once P_C becomes negative, a porous electrode may spontaneously imbibe the liquid phase and become flooded. This coarse approach allows for the insertion of P_C models that are uniquely suited to specific electrode microstructure, layer composition, and surface functionalization. Comparing the operating envelope for each liquid-PTFE combination as a function of current and flow rate is useful for predicting if any notable physical changes to the system pressure equilibrium emerge when targeting different CO2R products. The critical composition lines generally shift upward from FA to PrOH, according to chain length, depth of electroreduction, and decreasing polarity, which taken together indicate that the allowable operating space will narrow as the deeper CO2R products considered in this subset are pursued.

Zoom In Zoom Out Reset image size

These results suggest that the operating envelope for FA is likely to be wider than for alcohols for PTFE-containing GDEs. However, when considering that many existing commercial GDEs are composite materials (conductive metal and hydrophobic PTFE components) with mixed wettability properties (vide supra, oxygen depolarized cathodes), these contours may constitute an optimistic set of conditions correlating to P_C transition. Using composite GDEs may ultimately prove necessary when scaling to larger cell areas due to enhanced through-plane conductivity as compared to the PTFE-supported electrodes. Despite the greater flooding risk inherent to this category of GDEs, which are necessarily composed of high energy conductive constituents, there are still opportunities to tune wet-proofing content to achieve favorable P_C envelopes⁷⁰ as well as high CO2R activity and faradaic efficiency.^8,11

This mass balance analysis serves to estimate P_C thresholds for porous electrodes in contact with low surface tension liquid mixtures. However, what is not evident from the results until now is that pore geometry and surface wettability together determine P_C in real porous media. Therefore, in the next two sections we briefly discuss the potential for leveraging microstructure and surface chemistry to engineer more robust porous electrodes for CO2R-to-liquids electrolyzers.