Hierarchically conductive electrodes unlock stable and scalable CO2 electrolysis | Nature Communications

Mar 04, 2025

Nature Communications volume 15, Article number: 9429 (2024) Cite this article

9143 Accesses

173 Altmetric

Metrics details

Electrochemical CO2 reduction has emerged as a promising CO2 utilization technology, with Gas Diffusion Electrodes becoming the predominant architecture to maximize performance. Such electrodes must maintain robust hydrophobicity to prevent flooding, while also ensuring high conductivity to minimize ohmic losses. Intrinsic material tradeoffs have led to two main architectures: carbon paper is highly conductive but floods easily; while expanded Polytetrafluoroethylene is flooding resistant but non-conductive, limiting electrode sizes to just 5 cm2. Here we demonstrate a hierarchically conductive electrode architecture which overcomes these scaling limitations by employing inter-woven microscale conductors within a hydrophobic expanded Polytetrafluoroethylene membrane. We develop a model which captures the spatial variability in voltage and product distribution on electrodes due to ohmic losses and use it to rationally design the hierarchical architecture which can be applied independent of catalyst chemistry or morphology. We demonstrate C2+ Faradaic efficiencies of ~75% and reduce cell voltage by as much as 0.9 V for electrodes as large as 50 cm2 by employing our hierarchically conductive electrode architecture.

Electrochemical CO2 reduction has seen an increase in attention for its promise to convert captured CO2 into a variety of useful fuels, chemicals, and products otherwise produced from petrochemicals1,2,3. When paired with renewable electricity, CO2 reduction can serve as a means for energy storage, or enable net-negative sequestration of CO2 in durable products4. Over the last decade, substantial progress in developing novel catalysts, reaction pathways, and electrode architectures have collectively pushed electrochemical CO2 reduction toward the possibility of commercial relevance5,6,7. Many of the broadly accepted benchmarks for commercial viability have been reached individually at the lab scale with small electrodes of several cm2 size. Operation at many 100’s of mA/cm2 has been shown concurrently with C2+ selectivity over 80%8,9,10,11,12. Separately, CO2 utilizations as high as 80–90% and carbonate crossover as low as 3% have been shown13,14,15,16. Stability tests have shown stable performance for >1000 h17. Some systems have demonstrated cathodic electrical efficiencies as high as 45%12. However, integration of all these characteristics into a single, scaled system has proved challenging. Promisingly, technoeconomic analyses indicate that the production of carbon monoxide would be profitable if lab performance can be scaled to large sizes, while optimistic predictions indicate ethylene production costs could meet market costs in the future18,19,20. However, maintaining lab-scale performance when scaling to commercial sizes poses significant challenges, such as the large-scale production of catalysts and membranes, maintenance of favorable mass and electron transport in larger cells, and thermal management. As CO2 conversion technologies continue to mature and approach commercial relevance, the scalability of electrodes and their supporting systems to large sizes must be increasingly taken into consideration.

A generalized equation for electrochemical reduction of CO2 can be written:

Due to the poor diffusivity of CO2(aq), the immediate accessibility of all three reactants is critical: gaseous CO2, electrolyte, and the catalyst must be in close proximity to ensure high reaction rates6,21,22. Gas Diffusion Electrode (GDE) architectures establish such three-phase contact by locating the catalyst layer atop a non-wettable gas diffusion layer (GDL), and have thus become the predominant and highest performing architecture for gaseous phase CO2 reduction6,23,24. Though the catalyst has been the subject of intense study, the GDL also plays an important role in optimizing high-rate reactions. A well-designed GDL must: i) physically support the catalyst, yet must be sufficiently porous for gas transport, ii) have adequate electronic conductivity to facilitate electron transport to the catalyst layer with minimal ohmic loss, and iii) maintain robust hydrophobicity to ensure that the triple phase contact is sustained close to the catalyst. The simultaneous requirement of robust hydrophobicity and good conductivity presents a challenging design tradeoff. Generally, there exists a negative correlation between hydrophobicity and conductivity for engineering materials commonly used in GDEs: highly conductive materials tend to be hydrophilic, and the most hydrophobic materials tend to be nonconductive. The ramifications of this fundamental tradeoff are readily apparent in the GDLs employed in today’s state of the art electrochemical CO2 reduction systems.

One of the most common GDLs is carbon paper, consisting of carbon fibers coated with a thin PTFE layer. Carbon paper has good conductivity, allowing electrons to conduct directly through the thickness of the GDL to reach the catalyst as shown in Fig. 1A. However, carbon paper electrodes are known to lose their hydrophobicity during operation due to coating degradation25 (see Supplementary Fig. S25) among many other physical mechanisms26,27,28,29,30. Carbon paper is therefore susceptible to flooding events, in which the electrolyte impregnates the GDL, increasing the diffusion distance of CO2(aq), and reducing CO2 reduction rates and specificity27. Recent works have also elucidated that salt formation in the GDL plays a major role in stability and selectivity31, and that decreased hydrophobicity exacerbates flooding associated with salt formation30. We experimentally demonstrate the flooding instability of carbon paper in Fig. 1B, in which the ethylene selectivity decreases and hydrogen evolution increases concurrently with a decrease in cell voltage, closely matching previously observed flooding oscillatory patterns30. We also note that liquids were observed on the backside of the carbon paper GDL after cell disassembly as further evidence of liquid impregnation. The water contact angle of the GDL backside before and after operation is shown in Supplementary Fig. S24, demonstrating that the hydrophobicity of carbon paper degrades dramatically during operation. Our long-term stability experiments (see Supplementary Fig. S23 and Supplementary Section 7) indicate that carbon paper is unstable via at least two distinct mechanisms over a range of current densities and electrode combinations, losing most of its C2H4 FE over 40 h. Significant research has focused on preventing flooding of carbon paper, including constant re-hydrophobization of the GDE8, meticulously balancing the relative pressures, modifying the catalyst or catalyst layer32,33,34, and controlling pore size distributions35. Such approaches, however, are difficult to sustain over long periods and significantly constrain the operational freedom of these systems.

A Electrons conduct from the current collector through the thickness of a conductive Carbon paper Gas Diffusion Layer (GDL) to reach the catalyst layer. B Carbon paper GDEs lose their hydrophobicity, leading to flooding events which result in increased Hydrogen generation and reduced C2+ FE. Complete data in Supplementary Fig. S23B. All voltages are reported without iR correction. C ePTFE GDLs are insulating, therefore electrons must be conducted in-plane through the thin catalyst layer. D As unmodified ePTFE GDEs are scaled to larger sizes, the cell voltage increases and the CO2 reduction selectivity decreases. Error bars (standard deviation for 3 trials) are known for the 4.2 cm2 electrode, but are smaller than the symbols. All voltages are reported without iR correction; single experiments are reported for 0.5 and 2.5 cm2. E A hierarchically conductive GDE with micro-scale conductors piercing through the ePTFE membrane reduces the in-plane conduction distance within the catalyst, increasing the conductivity. F The intrinsic material tradeoff between conductivity and hydrophobicity. An optimal GDE design should possess both excellent hydrophobicity and excellent conductivity. Hydrophobicity data represents the Contact Angle (CA) and is described in more detail in Supplementary Fig. S24. Source data for B and D are provided in the Source Data file.

Another GDL which has gained significant popularity in recent years for its ability to resist flooding is expanded Polytetrafluoroethylene (ePTFE). It is composed of a porous network of pure PTFE fibers and thus has very robust hydrophobicity. We show in Supplementary Fig. S24 that even after long-term (70 h) operation, its wettability is largely unaffected. Its superior stability25,36 and hydrophobicity has led to its frequent use in high performing electrodes10,11,12,13,25,37 (see also Supplementary Table S2). However, PTFE is no exception to the aforementioned intrinsic material tradeoff between hydrophobicity and conductivity: it is a poor electrical conductor, with a conductivity on the order of 10−23 S/m. Electrons cannot be transported through ePTFE GDLs, forcing the catalyst layer itself to facilitate in-plane electron transport as shown in Fig. 1C. To assess the adequacy of the catalyst layer as the primary electron conductor, we form a scaling relationship for the ohmic losses \(\Delta V\) at a current density \(i\) for a square electrode with sidelength \(L\) and area \(A={L}^{2}\). The catalyst has a thickness \({t}_{{cat}}\) and a resistivity \({\rho }_{{cat}}\).

We observe that the ohmic losses through the catalyst layer scale as a function of the electrode area, \({L}^{2}\). Critically, this indicates that ohmic losses will become increasingly problematic as the technology progresses to commercialization and the electrode size increases. For carbon paper, we find that the overpotential scales as \(\Delta {V}_{{carbon}}\, \sim {i}{\rho }_{{carbon}}{t}_{{carbon}}\) due to its ability to conduct electrons through its thickness. Importantly, ohmic losses here are independent of electrode scale, enabling scaling of carbon paper electrodes to 50–100’s of cm2 for C2+ products13,38. This difference between the ePTFE and carbon paper architectures arises from the fundamentally different directions of electron transport illustrated in Fig. 1A, C. The poor conductivity of ePTFE electrodes impacts their scalability: their ohmic losses at commercial scales are as much as six orders of magnitude larger than the ohmic losses in a carbon paper architecture (see Supplementary Section 1). This poor conductivity has also been associated with catalyst layer instability and degradation by Montfort et al., who demonstrated layered current collectors on ePTFE electrodes to improve conductivity, but did not scale the approach39,40.

Previous works have added carbon and graphite atop the catalyst layer in order to increase conductivity25,41, but these effects do not fundamentally overcome the direction of electron transport and thus the adverse scaling behavior. Previous attempts to scale this ePTFE architecture by an order of magnitude were thus not pursued due to the lack of in-plane conductivity13. For this reason, implementations of ePTFE electrodes to date have been constrained to electrode areas of \(\le \)10 cm2 (typically 1–5 cm2) to the best of our knowledge (see Supplementary Section 6).

Electrodes of such limited size are not feasible for commercial implementation, as the immense number of cells would incur extensive balance of plant costs. We look to other more mature electrochemical applications to contextualize a reasonable electrode size which reconciles balance of plant, fabrication costs, and operational costs associated with electrode performance. Proton exchange membrane (PEM) electrolysis installations ranging from several kW to several megawatts have electrode sizes from 300 cm2 to 3000 cm2 42. Even more advanced technologies like fuel cells and chlor-alkali electrolyzers have electrode areas of 10,000 cm2 or greater43. We thus conclude that ePTFE electrodes of several cm2 are infeasible to scale. The resolution of the ePTFE electrode conductivity bottleneck is thus of critical importance to ensure that this highly performing electrode architecture can be scaled to commercial relevance.

The performance of ePTFE electrodes at increasing electrode sizes has not been explored in depth, leaving the length scales at which such ohmic losses become relevant unknown. Moreover, there is a distinct opportunity to design a new family of GDE architectures consisting of composite GDLs which possess both excellent hydrophobicity and conductivity, as shown conceptually in Fig. 1E, F. In this work, we explore the fundamental underpinnings of ohmic losses in ePTFE-based electrodes, and subsequently design and validate a hierarchically conductive composite electrode architecture. We demonstrate ~75% C2+ Faradaic Efficiency (FE) at a current density of 200 mA/cm2 for a 50 cm2 electrode, over an order of magnitude larger than typical ePTFE electrodes (see Supplementary Section 6 and Supplementary Table S2).

We use cathodes with a 250-nm thick copper catalyst deposited on an ePTFE membrane with 450 nm pores for our model system, as optimized in other works41 (see Methods). The cathodes are assembled in a Membrane Electrode Assembly (MEA) along with an iridium oxide carbon paper anode, separated by a Dioxide Materials anion exchange membrane for ion migration. All experiments and data in this work are performed without iR correction.

To assess the effects of ePTFE electrode size on performance, three square electrodes of sizes 0.5 cm2, 2.5 cm2, and 4.2 cm2 were evaluated for their ethylene productivity, with results shown in Fig. 1D. The 0.5 cm2 electrode performs well, achieving 47% FE towards ethylene and an estimated 84% FE towards C2+ products (see Supplementary Section 4 and Supplementary Fig. S17 for full product analysis) at an industrially relevant current density of 200 mA/cm2 and at voltages comparable to literature. However, the 4.2 cm2 electrode exhibits a substantially reduced peak ethylene FE of 29% and a full cell voltage which is 0.81 V greater than the 0.5 cm2 electrode at 200 mA/cm2. Clearly, the effects of scale on ePTFE electrode performance are drastic even at single cm2, lending credence to the size limitation in literature and further motivating the development of novel GDE architectures with simultaneous hydrophobicity and conductivity.

To this end, our aim in this work is to develop a novel GDE architecture by supplementing ePTFE electrodes with additional hierarchical conducting elements as shown in Fig. 1E. This approach makes good use of the inherent hydrophobicity of ePTFE membranes, requiring only the improvement of conductivity. To rationally approach the conductivity challenge, we first sought to further understand the fundamental effects driving the voltage increase and selectivity decrease observed in our experiments.

The increasing cell voltage with increasing electrode size is readily accounted for by the ohmic losses through the catalyst layer expressed in Eq. (2). On the other hand, the mechanism behind decreasing product selectivities was not as immediately clear. Noting that different products begin to be produced at different overpotentials (CO, then C2+, then H2, from low to high overpotential), we hypothesized that when ohmic losses reach a similar order as these overpotentials, there can exist significant spatial gradients in overpotential and thus product generation, leading to decreased overall selectivity44. This is shown conceptually in Fig. 2A, wherein the left side of the electrode closest to the electrical connection has the highest absolute overpotential and is expected to exhibit a greater degree of Hydrogen evolution. Conversely, the right side has a lower absolute overpotential due to significant ohmic voltage drops and is expected to produce predominantly CO. To experimentally validate this phenomenon, we designed and fabricated a new reactor with three independent serpentine flow paths, each one aligned with a third of the electrode as denoted I, II, and III in Fig. 2A. (see Supplementary Fig. S8) The local product distributions of each of the three sectors shown in Fig. 2B match our hypothesis, with sector I producing more Hydrogen and sector III producing more CO.

A A conceptual schematic showing the front-view of an electrode which is electrically connected from the left side only, causing a voltage gradient across the electrode. Three separate gas pathways (I, II, III) isolate the products produced on the left, middle, and right sides of the electrode. B Shows the experimental variation in product distribution across the three aforementioned product streams (see Supplementary Fig. S8 for more details). C Simulated voltage field on a 4.2 cm2 square electrode at 250 mA/cm2 with 0 V defined as the lowest voltage on the surface. D Simulated ethylene FE at the above condition is spatially variable due to voltage gradients. E Pointwise integration of the model predicts the decreased peak FE and relatively increased voltage for the 4.2 cm2 electrode compared to the 0.5 cm2 electrode with single-side connection as shown in A. A slight over-prediction of FE is expected, as the model does not consider mass transport and other effects. F Simulation model prediction of i-V curve matches experimental data for an unmodified ePTFE electrode with single-side connection as shown in A. 4.2 cm2 experimental data is repeated three times with standard deviation smaller than symbol size. All voltages are reported without iR correction. Source data for B, E, and F are provided in the Source Data file.

We then developed a model to confirm the role of ohmic effects, knowing the model could subsequently be used as a design tool to predict the performance of proposed electrode designs. Briefly, a 2-D finite element model was constructed which iteratively computes the non-Laplacian voltage field on the electrode by coupling Ohm’s law and current conservation given the out-of-plane current which contributes to the electrochemical reaction. The point-wise electrochemical current is estimated as a function of the local voltage by using the experimental data of the 0.5 cm2 electrode in which ohmic losses are minimal and thus the entire electrode can be considered equipotential. A representative simulated voltage field is shown in Fig. 2C. A full description of the model and assumptions are available in the Methods and Supplementary Section 3.

To validate the accuracy of the model, we simulated a 4.2 cm2 electrode using only the reference performance at 0.5 cm2. As expected, the voltage and ethylene FE are shown to have significant spatial variations as shown in Fig. 2C, D. The modeled performance and experimental results present a strong match as shown in Fig. 2E, F. The model captures the increased cell voltages at 4.2 cm2, as well as the decreased peak FE and the shift in peak FE voltage. Further model validation is performed for a 50 cm2 unmodified ePTFE electrode as shown in Supplementary Fig. S17. The ability of this model to predict performance of larger electrodes while only accounting for the ohmic losses in the catalyst layer confirms the central role which insufficient conductivity plays in the limited performance of ePTFE electrodes.

To design an electrode architecture with improved electron transport, we draw inspiration from natural and engineered systems which leverage hierarchy to optimize transport. Efficient transport systems of all kinds, from the cardiovascular system to public transport networks, hold in common the use of large, high volume transport passages for long distance transport, with progressively smaller branching passages enabling medium-distance and then short-distance transport to the destination. By observing the journey of an electron in a conventional ePTFE electrode, we readily recognize the shortcomings of its design with respect to hierarchy. An electron first travels several centimeters through a current collector of millimetric thickness. It must then travel several centimeters through the catalyst layer only hundreds of nanometers thick. As evidenced by the ohmic losses, the sub-micron catalyst layer does not have sufficient conductivity to facilitate centimetric electron transport. To alleviate this bottleneck, we introduced a micrometric conductor into the GDL to bridge the conduction from the underlying current collector to the sub-micron catalyst layer. The micrometric conductor facilitates medium-range distribution of electrons within the catalyst layer, limiting the transport through the sub-micron layer to distances small enough to prevent significant ohmic losses. We term this GDE architecture with conducting elements across length-scales the Hierarchically Conductive GDE: HCGDE.

The HCGDE embodiment shown in Fig. 3A, B features a micrometric copper wire which pierces the ePTFE membrane and runs along the catalyst layer. The wire connects to the current collector on the backside of the GDL. To determine the wire spacing for which ohmic losses in the catalyst layer become negligible, we turn to the previously validated model as a design tool. In Fig. 3C, we show a representative simulation result of a wire design unit cell, with equipotential current source regions representing the wires at the top and bottom edges. The wire spacing s is swept from one to one hundred millimeters, and the voltage at which the average FE is maximized along with the maximized FE is presented in Fig. 3D. As the wire spacing increases, we see both a decrease in the maximum FE and an increase in the voltage at which this maximum occurs, consistent with our experimental and simulation findings in Figs. 1 and 2. Figure 3E, F illustrates the spatial variability of FE and voltage in the wire architecture. From these results, we reason that our HCGDEs should maintain sufficient conductivity and minimize ohmic losses through the catalyst layer when wire spacing is held below ~10 mm.

A Conceptual schematic of a wire-HCGDE with wires which pierce the membrane to reach the catalyst, effectively parallelizing the electrode and reducing the average electron conduction distance within the catalyst. The wires run along the catalyst for a length l, and are repeated at spacing distance s. B Image of a 14 cm2 wire HCGDE electrode showing the sewn wires. C Representative unit cell simulation with two equipotential current sources which replicate the effect of conductive wires at a spacing of s. D Maximum average ethylene FE and voltage at peak FE show decreasing performance with increasing spacing s beyond ~10 mm. Traces of (E) ethylene FE and (F) voltage at non-dimensionalized positions between the two equipotential current sources show how spatial gradients develop. Source data for D–F are provided in the Source Data file.

For the HCDGE design to remain practical, the micrometric conductive elements should not otherwise negatively affect the electrode or overall cell architecture. We chose copper as the wire material for its high conductivity and because it matches the catalyst material, reducing the likelihood of parasitic electrochemical reactions. Moreover, a 75-μm diameter copper wire at 4 mm spacing will occupy less than 2% of the electrode surface, negligibly decreasing the active surface area (see Supplementary Section 5). Though some small degree of liquid penetration may occur in the microscopic holes required for the penetration of the wire, the flooding will be limited to only this region by the robust hydrophobicity of the surrounding ePTFE. Thus we expect that the HCGDE should maintain at least the same degree of stability as ePTFE electrodes. The piercing wire embodiment is also simple and highly scalable, as it can be fabricated roll-to-roll with classic sewing techniques.

To evaluate the HCGDE performance, we fabricated 14 cm2 and 50 cm2 electrodes with the previously described ePTFE electrodes as a base, additionally sewing 75-μm diameter wires into the GDEs at 4 mm spacing as shown in Fig. 3B.

We constructed a new 50 cm2 MEA pilot scale reactor (Fig. 4A) to execute performance tests. Upon assembly and disassembly of the wire HCGDE, no differences or damage to the membrane, electrodes, or electrolyzer were observed due to the flexibility and small profile of the 75-μm wires. Figure 4B–D demonstrate how the HCGDE overcomes the conductivity limitations of ePTFE electrodes, enabling high performance at large scales. As the electrode size is increased a full two orders of magnitude from 0.5 cm2 to 50 cm2, the full cell voltage is unchanged: the 0.5 cm2 electrode requires a full cell potential of 3.58 V to reach 200 mA/cm2, while the 50 cm2 electrode requires 3.52 V (Figs. 1D and 4D). Moreover, C2+ product FEs remain at about 75% for large electrode sizes up to 50.0 cm2, leading to a sustained Electrical Efficiency (EE) of ~25% up to the 50 cm2 size (Fig. 4B). The HCGDE strongly outperforms a 50 cm2 ePTFE electrode without conductivity enhancements in Fig. 4C, D, producing more ethylene and attaining higher current densities at substantially lower voltages. Note that for a fair comparison, the 50 cm2 unmodified ePTFE GDE was electrically connected on all sides, hence the better performance than the unmodified ePTFE 4.2 cm2 case shown in Fig. 2E, F where the electrode was connected at only one side. The performance of the unmodified 50 cm2 electrode also matches our model prediction; see Supplementary Fig. S17.

A Pilot-scale 50 cm2 reactor flowfields (right) and original 5 cm2 flowfield (left), with a quarter for scale. B the HCGDE holds ~75% C2+ FE and ~25% C2+ EE up to 50 cm2. C When compared to an ePTFE GDE of the same size without conductivity enhancements, the HCGDE reaches higher current densities at lower cell voltages due to the improved conductivity. D The HCGDE architecture enables higher ethylene productivity at lower cell voltages. All voltages are reported without iR correction, and each experiment was performed once. Source data for B–D are provided in the Source Data file.

We expect that the performance increase conveyed by the HCGDE architecture will only become more significant as electrode sizes continues to increase beyond 50 cm2. We also reason that the HCGDE approach remains practical for electrodes operating at current densities in the A/cm2 range with improved catalysts, where minimum wire spacings would be around 2.5 mm with all other conditions held constant (see Supplementary Section 5). Our HCGDE is an order of magnitude larger than typical ePTFE electrodes as shown in Supplementary Table S2.

The reduction of ohmic losses afforded by our HCGDE may further contribute to the enhanced stability of CO2RR electrodes39. The stability of the HCGDE configuration was thus explored in Fig. 5 over 75 h. The C2H4 FE is not fully stable and initially decreases from ~40% to ~25%, then holds constant for 50 h. The non-increasing voltage over this time span implies that the decreasing FE is unrelated to the conductivity of our HCGDE electrode and may be due to morphological changes in the copper catalyst45 or to salt precipitation in the GDL which blocks gaseous mass transport30,31. Salt precipitation was observed after disassembly, and a DI water wash of the cathode compartment was performed every ~20 h. The rapidly rising voltage after hour 60 is ascribed to loss of the weakly-bound Iridium Oxide anode catalyst. Overall higher cell voltages compared to Fig. 4C are due to the use of a titanium felt-based anode necessary to avoid degradation of carbon-based anodes, and a reinforced anion exchange membrane necessary for preventing the much stiffer titanium anodes from shorting across the membrane (see Supplementary Figs. S22, S23). The use of titanium felt anodes necessitated a switch from 10 mM to 100 mM KHCO3, which may explain the increased C2H4 FE at 100 mA/cm2 in Fig. 5 when compared to Figure 1D46.

Stability of a 4.2 cm2 HCGDE at 100 mA/cm2 over 75 h. Sudden jumps in cell voltage at 15, 26, and 37 h coincide with accidental jostling of the cell and electrical connection leads, which were not tightly connected. Cell voltage is reported without iR correction. Source data are provided in the Source Data file.

As expected, the HCGDE demonstrates enhanced stability compared to a carbon paper electrode which had just <5% FE towards C2H4 after 40 h under the same conditions as shown in Supplementary Fig. S23. The significant challenges encountered in maintaining stability highlight the importance of managing salt precipitation and catalyst morphology after the flooding instability mode has been overcome.

The HCGDE platform employs hierarchical charge conduction to eliminate the ohmic losses which afflict conventional ePTFE electrodes, enabling low voltages and high CO2RR Faradaic efficiencies at pilot scale. The improvement of conductivity is achieved with a minimal footprint (<2% area) and without sacrificing the bulk hydrophobicity of the ePTFE. The HCGDE simultaneously achieves flooding resistance and good conductivity, reaching the upper right corner of Fig. 1F. The HCGDE was demonstrated to be more stable and have similar or lower cell voltages than carbon paper cathodes (see Supplementary Fig. S21). The purely physical HCGDE approach can be implemented in any existing GDE and can enable scaling of electrodes independent of their catalyst composition or morphology. Moreover, by decoupling the electron conduction and hydrophobicity requirements of the GDL, this approach unlocks electrode architecture design without placing additional constraints on catalyst or electrolyzer design. Integration of the HCGDE into flow cell types could be achieved by using a porous, flow-through spacer to ensure compression of the wires against the electrode.

Scaling lab-scale performance to large electrode stacks will be a critical aspect of fulfilling the promise of electrochemical CO2 reduction technology. In this work, we demonstrate with models and experiments how conventionally employed ePTFE electrodes fail to maintain their performance at larger electrode sizes due to their poor in-plane conductivity. We then leveraged a hierarchical design framework to design an HCGDE which introduces micrometric wires within the GDL to effectively distribute electrons across the electrode and alleviate ohmic losses. The ability of the HCGDE technique to maintain performance at scale is demonstrated with a 50 cm2 reactor. This approach is simple, low-impact, and flexible, and can readily be applied to existing high performing catalysts and electrolyzer architectures to enable their immediate scaling. The HCGDE platform achieves simultaneously excellent bulk hydrophobicity and good conductivity, thereby unlocking the design space of GDLs and relieving constraints on system operational parameters.

To fabricate the ePTFE cathodes, magnetron sputtering (AJA International) was used to deposit 250 nm of Copper on a 450 nm pore size Aspire Laminated PTFE membrane (Sterlitech) at a rate of 0.83 Å/s under Argon, calibrated via Quartz Crystal Monitor (QCM). To fabricate the carbon paper cathodes, the same sputtering procedure was performed on the MPL side Sigracet 36BB carbon paper (Fuel Cell Store). HCGDEs were fabricated by punching holes on the edges of the base ePTFE cathodes with a needle and threading a 75 μm diameter copper wire through the holes (see Supplementary Figs. S4, S7). For ePTFE cathodes without the HCGDE augmentation, a piece of copper foil (25-μm thick) was wrapped around a single edge of ePTFE electrodes to establish electrical connection to the flowfield current collector. Iridium oxide on carbon paper anodes were purchased from Dioxide Materials and cut to size. To avoid degradation of carbon-based anodes in long-term stability experiments, Iridium Oxide anodes were synthesized using a titanium felt substrate47. Briefly, the titanium felt was cleaned via sonication in Acetone and DI water for 2 min each. The titanium felt was then etched in 6 M HCl (SigmaAldrich, 37%) held at 80–100 °C for 20–30 min. The felt was then submerged in a solution of 9 mL isopropanol and 1 mL concentrated HCl with 30 mg IrCl3 (SigmaAldrich, reagent grade) dissolved. The solvent was evaporated on a hotplate at 100 °C, leaving behind Iridium residues. Finally, the titanium felt electrode was calcined at 500 °C for 10 min.

All electrochemical tests were performed in Membrane Electrode Assemblies (MEA) with a graphite cathodic serpentine flowfield and a titanium anodic serpentine flowfield. Experiments with electrodes <5 cm2 are performed in an MEA with 5 cm2 flowfield area, experiments with electrodes >5 cm2 are performed in an MEA with 50 cm2 flowfield area (see Supplementary Figs. S6 and S7). CO2 (99.995%, Airgas) was humidified and flowed through the cathode flowfield at 200 SCCM (50 cm2 electrode) or 100 SCCM (all other experiments). Note that cell outflow rates (and therefore GC inflow rates) are not necessarily equal to the cell inflow rate, and this discrepancy is addressed in Supplementary Fig. S11. Cell temperature was measured continuously with a thermocouple and was confirmed to remain in the range of 25–30 °C for all experiments. All experiments were chronopotentiometric, progressing from low to high current density (10, 50, 100, 150, 200, 250, 300 mA/cm2) and holding for 14 min at each current density. For electrodes <5 cm2, the current was supplied and full cell voltage was measured with a BioLogic SP-150 Potentiostat. For electrodes >5 cm2, the current was supplied by a power supply (TDK-Lambda GEN300-17) due to the maximum current limitations of the potentiostat. The full cell voltage was measured with the potentiostat. In no case was iR compensation used. When applicable, error bars are calculated as the standard deviation of three experiments.

For long-term stability tests, titanium felt IrOx anodes and Sustainion X37-50 Grade T anion exchange membranes were used. The reinforced anion exchange membrane was necessary to eliminate shorting across the membrane when using metallic anode substrates. For all other tests, Iridium oxide on carbon paper purchased from Dioxide Materials was used as the anode and a Sustainion X37-50 anion exchange membrane was used as the separator. For long-term stability tests, 100 mM KHCO3 (SigmaAldrich, 99.7%) was recirculated in the anode at 60 mL/min from a 200 mL reservoir with a peristaltic pump to reduce ohmic losses (see Supplementary Fig. S22). For all other tests, 10 mM KHCO3 was used. The increased electrolyte concentration in the long-term stability test setup may have contributed to increased salt precipitation46. All KHCO3 electrolytes were prepared within 10 min of the experiment start, and were left uncovered during the duration of the experiment.

The anode was surrounded by a 0.25-mm thick PTFE gasket. Carbon paper cathodes were similarly surrounded by a 0.25-mm thick PTFE gasket. ePTFE and HCGDE cathodes were placed atop a 25-μm thick PTFE gasket cut with a square open area of the appropriate active area.

Gas products were quantified with a Gas Chromatograph (Agilent 7890B) calibrated with a custom gas precision mixture (Gasco and Shopcross). For all experiments with electrodes smaller than 5 cm2, the gas flow rate was controlled and measured upstream the electrochemical cell, and periodic measurements were made downstream to validate that the downstream and upstream flowrates matched within 5%. For larger electrode experiments, the gas flowrate was measured directly downstream of the cell, and significant deviations were found and accounted for as shown in Supplementary Fig. S11. Liquid products are collected from both the anode electrolyte and a cold trap placed after the cathode, and are measured with Nuclear Magnetic Resonance (NMR) (Brucker 402) with DMSO and Phenol in D2O standard (see Supplementary Section 4 and Supplementary Fig. S19). The Faradaic Efficiency is calculated as:

Where \(z\) is the number of electrons transferred per mole of product, \(\dot{n}\) is the measured molar production rate, \(F\) is Faraday’s constant, and \({I}_{{total}}\) is the total current applied to the cell.

Contact Angle Measurements were performed with a Ramé-Hart goniometer.

Simulation Model description can be found in the Supplementary Information.

Source data are provided with this paper.

Hori, Y., Kikuchi, K. & Suzuki, S. Production of co and ch4 in electrochemical reduction of co2 at metal electrodes in aqueous hydrogencarbonate solution. Chem. Lett. 14, 1695–1698 (1985).

Article Google Scholar

Hori, Y., Murata, A. & Takahashi, R. Formation of hydrocarbons in the electrochemical reduction of carbon dioxide at a copper electrode in aqueous solution. J. Chem. Soc. Faraday Trans. 85, 2309–2326 (1989).

Article CAS Google Scholar

Kuhl, K. P., Cave, E. R., Abram, D. N. & Jaramillo, T. F. New insights into the electrochemical reduction of carbon dioxide on metallic copper surfaces. Energy Environ. Sci. 5, 7050–7059 (2012).

Article CAS Google Scholar

De Luna, P. et al. What would it take for renewably powered electrosynthesis to displace petrochemical processes? Science 364, eaav3506 (2019).

Article ADS PubMed Google Scholar

Nitopi, S. et al. Progress and Perspectives of Electrochemical CO2 Reduction on Copper in Aqueous Electrolyte. Chem. Rev. 119, 7610–7672 (2019).

Article PubMed CAS Google Scholar

Nguyen, T. N. & Dinh, C.-T. Gas diffusion electrode design for electrochemical carbon dioxide reduction. Chem. Soc. Rev. 49, 7488–7504 (2020).

Article PubMed CAS Google Scholar

Woldu, A. R., Huang, Z., Zhao, P., Hu, L. & Astruc, D. Electrochemical CO2 reduction (CO2RR) to multi-carbon products over copper-based catalysts. Coord. Chem. Rev. 454, 214340 (2022).

Article CAS Google Scholar

Ma, W. et al. Electrocatalytic reduction of CO2 to ethylene and ethanol through hydrogen-assisted C–C coupling over fluorine-modified copper. Nat. Catal. 3, 478–487 (2020).

Article CAS Google Scholar

Wang, X. et al. Efficient electrically powered CO2-to-ethanol via suppression of deoxygenation. Nat. Energy 5, 478–486 (2020).

Article ADS CAS Google Scholar

Zhong, M. et al. Accelerated discovery of CO2 electrocatalysts using active machine learning. Nature 581, 178–183 (2020).

Article ADS PubMed CAS Google Scholar

Cao, Y. et al. Surface hydroxide promotes CO2 electrolysis to ethylene in acidic conditions. Nat. Commun. 14, 2387 (2023).

Article ADS PubMed PubMed Central CAS Google Scholar

García de Arquer, F. P. et al. CO2 electrolysis to multicarbon products at activities greater than 1 A cm−2. Science 367, 661–666 (2020).

Article ADS PubMed Google Scholar

Xu, Y. et al. A microchanneled solid electrolyte for carbon-efficient CO2 electrolysis. Joule 6, 1333–1343 (2022).

Article CAS Google Scholar

Huang, J. E. et al. CO2 electrolysis to multicarbon products in strong acid. Science 372, 1074–1078 (2021).

Article ADS PubMed CAS Google Scholar

Kim, J. Y. ‘Timothy’ et al. Recovering carbon losses in CO2 electrolysis using a solid electrolyte reactor. Nat. Catal. 5, 288–299 (2022).

Article CAS Google Scholar

O’Brien, C. P. et al. Single Pass CO2 Conversion Exceeding 85% in the Electrosynthesis of Multicarbon Products via Local CO2 Regeneration. ACS Energy Lett. 6, 2952–2959 (2021).

Article Google Scholar

Kutz, R. B. et al. Sustainion Imidazolium‐Functionalized Polymers for Carbon Dioxide Electrolysis. Energy Technol. 5, 929–936 (2017).

Jouny, M., Luc, W. & Jiao, F. General Techno-Economic Analysis of CO2 Electrolysis Systems. Ind. Eng. Chem. Res. 57, 2165–2177 (2018).

Article CAS Google Scholar

Sisler, J. et al. Ethylene Electrosynthesis: A Comparative Techno-economic Analysis of Alkaline vs Membrane Electrode Assembly vs CO2–CO–C2H4 Tandems. ACS Energy Lett. 6, 997–1002 (2021).

Article ADS CAS Google Scholar

Masel, R. I. et al. An industrial perspective on catalysts for low-temperature CO2 electrolysis. Nat. Nanotechnol. 16, 118–128 (2021).

Article ADS PubMed CAS Google Scholar

Nesbitt, N. T. et al. Liquid–Solid Boundaries Dominate Activity of CO2 Reduction on Gas-Diffusion Electrodes. ACS Catal. 10, 14093–14106 (2020).

Article CAS Google Scholar

Soto, Á. M., Lake, J. R. & Varanasi, K. K. Transient Effects Caused by Gas Depletion during Carbon Dioxide Electroreduction. Langmuir 38, 1020–1033 (2022).

Article PubMed CAS Google Scholar

Wakerley, D. et al. Gas diffusion electrodes, reactor designs and key metrics of low-temperature CO2 electrolysers. Nat. Energy 7, 130–143 (2022).

Article ADS CAS Google Scholar

Lees, E. W., Mowbray, B. A. W., Parlane, F. G. L. & Berlinguette, C. P. Gas diffusion electrodes and membranes for CO2 reduction electrolysers. Nat. Rev. Mater. 7, 55–64 (2022).

Article ADS CAS Google Scholar

Dinh, C.-T. et al. CO2 electroreduction to ethylene via hydroxide-mediated copper catalysis at an abrupt interface. Science 360, 783–787 (2018).

Article PubMed CAS Google Scholar

Xu, Y. et al. Self-Cleaning CO2 Reduction Systems: Unsteady Electrochemical Forcing Enables Stability. ACS Energy Lett. 6, 809–815 (2021).

Article CAS Google Scholar

Yang, K., Kas, R., Smith, W. A. & Burdyny, T. Role of the Carbon-Based Gas Diffusion Layer on Flooding in a Gas Diffusion Electrode Cell for Electrochemical CO2 Reduction. ACS Energy Lett. 6, 33–40 (2021).

Article CAS Google Scholar

Leonard, M. E. et al. Editors’ Choice—Flooded by Success: On the Role of Electrode Wettability in CO2 Electrolyzers that Generate Liquid Products. J. Electrochem. Soc. 167, 124521 (2020).

Article ADS CAS Google Scholar

Jeanty, P. et al. Upscaling and continuous operation of electrochemical CO2 to CO conversion in aqueous solutions on silver gas diffusion electrodes. J. CO2 Utilization 24, 454–462 (2018).

Article CAS Google Scholar

Moss, A. B. et al. In operando investigations of oscillatory water and carbonate effects in MEA-based CO2 electrolysis devices. Joule 7, 350–365 (2023).

Article CAS Google Scholar

Disch, J. et al. High-resolution neutron imaging of salt precipitation and water transport in zero-gap CO2 electrolysis. Nat. Commun. 13, 6099 (2022).

Article ADS PubMed PubMed Central CAS Google Scholar

Niu, Z.-Z. et al. Hierarchical Copper with Inherent Hydrophobicity Mitigates Electrode Flooding for High-Rate CO2 Electroreduction to Multicarbon Products. J. Am. Chem. Soc. 143, 8011–8021 (2021).

Article PubMed CAS Google Scholar

Wu, Y. et al. Effects of microporous layer on electrolyte flooding in gas diffusion electrodes and selectivity of CO2 electrolysis to CO. J. Power Sources 522, 230998 (2022).

Article CAS Google Scholar

Wu, Y. et al. Mitigating Electrolyte Flooding for Electrochemical CO2 Reduction via Infiltration of Hydrophobic Particles in a Gas Diffusion Layer. ACS Energy Lett. 7, 2884–2892 (2022).

Article CAS Google Scholar

Baumgartner, L. M., Koopman, C. I., Forner-Cuenca, A. & Vermaas, D. A. When Flooding Is Not Catastrophic─Woven Gas Diffusion Electrodes Enable Stable CO2 Electrolysis. ACS Appl. Energy Mater. 5, 15125–15135 (2022).

Article PubMed PubMed Central CAS Google Scholar

Tiwari, P., Tsekouras, G., Swiegers, G. F. & Wallace, G. G. Gortex-Based Gas Diffusion Electrodes with Unprecedented Resistance to Flooding and Leaking. ACS Appl. Mater. Interfaces 10, 28176–28186 (2018).

Article PubMed CAS Google Scholar

Wang, Z. et al. Localized Alkaline Environment via In Situ Electrostatic Confinement for Enhanced CO2-to-Ethylene Conversion in Neutral Medium. J. Am. Chem. Soc. 145, 6339–6348 (2023).

Article PubMed CAS Google Scholar

Edwards, J. P. et al. Pilot-Scale CO2 Electrolysis Enables a Semi-empirical Electrolyzer Model. ACS Energy Lett. 8, 2576–2584 (2023).

Article CAS Google Scholar

Iglesias van Montfort, H.-P. et al. Non-invasive current collectors for improved current-density distribution during CO2 electrolysis on super-hydrophobic electrodes. Nat. Commun. 14, 6579 (2023).

Article ADS PubMed PubMed Central CAS Google Scholar

Rufer, S., Nitzsche, M., Garimella, S., Lake, J. & Varanasi, K. Hierarchically Conductive Electrodes Unlock Stable and Scalable CO2 Electrolysis | Chemical Engineering and Industrial Chemistry | ChemRxiv | Cambridge Open Engage. https://chemrxiv.org/engage/chemrxiv/article-details/65366f20c3693ca993f107d7 (2023).

Gabardo, C. M. et al. Continuous Carbon Dioxide Electroreduction to Concentrated Multi-carbon Products Using a Membrane Electrode Assembly. Joule 3, 2777–2791 (2019).

Article CAS Google Scholar

Harrison, K. MW-Scale PEM-Based Electrolyzers for RES Applications (CRADA CRD-18-00742 Final Report). https://www.osti.gov/biblio/1766848https://doi.org/10.2172/1766848 (2021).

Stinn, C. & Allanore, A. Estimating the Capital Costs of Electrowinning Processes. Electrochem. Soc. Interface 29, 44 (2020).

Article Google Scholar

Liu, K., Smith, W. A. & Burdyny, T. Introductory Guide to Assembling and Operating Gas Diffusion Electrodes for Electrochemical CO2 Reduction. ACS Energy Lett. 4, 639–643 (2019).

Article PubMed PubMed Central CAS Google Scholar

Nguyen, T. N. et al. Catalyst Regeneration via Chemical Oxidation Enables Long-Term Electrochemical Carbon Dioxide Reduction. J. Am. Chem. Soc. 144, 13254–13265 (2022).

Article PubMed CAS Google Scholar

El-Nagar, G. A., Haun, F., Gupta, S., Stojkovikj, S. & Mayer, M. T. Unintended cation crossover influences CO2 reduction selectivity in Cu-based zero-gap electrolysers. Nat. Commun. 14, 2062 (2023).

Article ADS PubMed PubMed Central CAS Google Scholar

Luc, W., Rosen, J. & Jiao, F. An Ir-based anode for a practical CO2 electrolyzer. Catal. Today 288, 79–84 (2017).

Article CAS Google Scholar

Download references

This work was supported by Shell through the MIT Energy Initiative (CW155703, K.K.V.). S.R. acknowledges financial support from the National Science Foundation (2021322124) We thank Flavia Cassiola, Sumit Verma, and Reza Mirshekari for the useful discussions. This work was carried out in part through the use of MIT.nano facilities. We thank Tal Joseph for assistance with the GDE renderings and Zara Aamer for the fabrication assistance.

Department of Mechanical Engineering, Massachusetts Institute of Technology 77 Massachusetts Avenue, Cambridge, MA, 02139, USA

Simon Rufer, Michael P. Nitzsche, Sanjay Garimella, Jack R. Lake & Kripa K. Varanasi

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

S.R. and K.K.V conceptualized the work. S.R., K.K.V., J.R.L. determined the methodology. S.R., S.G., M.P.N. carried out the investigation. M.P.N. performed the modeling. K.K.V. acquired the project funding. S.R., K.K.V managed the project. K.K.V. supervised the work. S.R., K.K.V., M.P.N. wrote and edited the manuscript.

Correspondence to Kripa K. Varanasi.

S.R. and K.K.V. have filed a provisional patent on the use of hierarchical conductors within GDEs. The remaining authors declare no competing interests.

Nature Communications thanks Adam Weber, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

Rufer, S., Nitzsche, M.P., Garimella, S. et al. Hierarchically conductive electrodes unlock stable and scalable CO2 electrolysis. Nat Commun 15, 9429 (2024). https://doi.org/10.1038/s41467-024-53523-8

Download citation

Received: 24 April 2024

Accepted: 15 October 2024

Published: 13 November 2024

DOI: https://doi.org/10.1038/s41467-024-53523-8

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative