Molecular electrocatalysts can mediate fast, selective CO2 reduction in a flow cell

See allHide authors and affiliations

Science  26 Jul 2019:
Vol. 365, Issue 6451, pp. 367-369
DOI: 10.1126/science.aax4608

Flowing CO2 boosts a molecular catalyst

Molecular electrocatalysts for CO2 reduction have often appeared to lack sufficient activity or stability for practical application. Ren et al. now show that design of the surrounding electrochemical cell can substantially boost both features. They directly exposed a known molecular catalyst, cobalt phthalocyanine, to gaseous CO2 in a flow cell architecture, rather than an aqueous electrolyte. The configuration accommodated current densities exceeding 150 milliamperes per square centimeter, with longevity limited by local proton concentration rather than catalyst stability.

Science, this issue p. 367


Practical electrochemical carbon dioxide (CO2) conversion requires a catalyst capable of mediating the efficient formation of a single product with high selectivity at high current densities. Solid-state electrocatalysts achieve the CO2 reduction reaction (CO2RR) at current densities ≥ 150 milliamperes per square centimeter (mA/cm2), but maintaining high selectivities at high current densities and efficiencies remains a challenge. Molecular CO2RR catalysts can be designed to achieve high selectivities and low overpotentials but only at current densities irrelevant to commercial operation. We show here that cobalt phthalocyanine, a widely available molecular catalyst, can mediate CO2 to CO formation in a zero-gap membrane flow reactor with selectivities > 95% at 150 mA/cm2. The revelation that molecular catalysts can work efficiently under these operating conditions illuminates a distinct approach for optimizing CO2RR catalysts and electrolyzers.

The electrolytic reduction of CO2 is an appealing method for making a range of carbon-based products, including CO, methane, formate, methanol, ethylene, and longer alkyl chains (14). Among these CO2 reduction reaction (CO2RR) products, the formation of CO is arguably one of the easier reactions to negotiate during catalysis on the basis that fewer electrons and protons are needed than for all other carbon products aside from formate. Notwithstanding, the best technology known today can produce CO with a Faradaic efficiency (FE) of >90% at current densities of 200 mA/cm2 for merely 1000 hours (5, 6), which remains inadequate for commercially relevant electrolysis.

CO2RR electrolyzers designed to produce CO typically rely on heterogeneous catalysts, such as gold, silver, or copper (1, 2, 710). Modifications to these solid-state electrocatalysts have not generally led to meaningful improvements in efficiencies and selectivities at high current densities (11, 12). This situation prompted us to consider whether molecular catalysts could convert CO2 to CO in a flow cell, a pilot reactor capable of operating at conditions more relevant to an industrial electrolyzer (3, 6, 1315).

The chemical and electronic environment about the active site of a molecular catalyst is often more substantially and acutely tunable than solid-state catalysts (1618). The ligand environment about a transition metal can be modified to yield competent homogeneous catalysts capable of mediating CO2RR with selectivities as high as 99% (Fig. 1) (1922). However, with one exception (23), electrocatalytic testing on these molecular systems was performed in organic media or at low current densities (<40 mA/cm2) to accommodate the mass-transport limitations of batch-type (e.g., “H-cell”) electrolysis experiments. Immobilizing high concentrations of molecular catalysts on high–surface-area conductive electrodes has allowed for a current density of 33 mA/cm2 to be reached while attaining 90% selectivity for CO production (23).

Fig. 1 Selectivity and activity for CO production as a function of current density and cell voltage for molecular and heterogeneous CO2RR electrocatalysts.

(A) FECO as a function of current density, J. (B) CO partial current density, JCO, as a function of overall cell voltage, Ecell, for high-performing molecular catalysts (17, 19, 23, 27, 3644) (blue; details in table S1) and the state-of-the-art heterogeneous Au (45) and Ag (6) catalysts (gray). The data for CoPc featured in this study are indicated in orange.

We report here that a molecular CO2RR electrocatalyst can form CO with >95% selectivity at current densities of 150 mA/cm2 in a flow cell, and a partial current density for CO production (JCO) of 175 mA/cm2 with an overall two-electrode cell voltage (Ecell) of 2.5 V. This value is 0.4 V lower (or a 4% gain in energy efficiency) relative to that reported for a silver solid-state CO2 catalyst at a similar CO partial current density (6). These performance metrics were realized in a zero-gap membrane reactor by using a commercially available cobalt phthalocyanine (CoPc) CO2RR catalyst immobilized on a gas diffusion layer in tandem with a nickel foam oxygen evolution reaction catalyst. This flow reactor provides access to higher current densities by (i) overcoming the mass-transport issues inherent to batch-type electrolysis and (ii) supplying CO2 to the cathode in the gas phase to overcome the inherently low diffusion and solubility of CO2 in aqueous media (3, 13, 2426). The catalyst was selected because it is inexpensive and selective for CO2-to-CO conversion in batch-type testing (27, 28).

In our custom-designed flow cell (Fig. 2B), a catalyst ink, consisting of carbon powder–supported CoPc and Nafion, was spray coated on hydrophobic carbon paper (that measured 2 cm by 2 cm) with a densely packed layer of microporous carbon to form the gas diffusion electrode (GDE) (15). The serpentine cathode and anode flow plates that manipulate reactant delivery to the GDE and the nickel foam were made of titanium and stainless steel, respectively. The membrane electrode assembly consists of a nickel foam anode (which acts as both a diffusion layer and the oxygen evolution reaction catalyst), an anion exchange membrane with high ion conductivity (29), and the GDE. The active areas of the cathode and anode were each 4 cm2. The cathode was fed with a humidified CO2 gas stream at a flow rate of 100 standard cubic centimeters per minute (SCCM), whereas the anode was fed with recirculated 1 M KOH at a flow rate of 20 ml/min.

Fig. 2 Membrane flow reactor for efficient CO2RR with a cobalt-based molecular electrocatalyst.

(A) Structure of the molecular catalyst, CoPc. (B) Exploded diagram of the zero-gap membrane reactor used for CO2 electroreduction testing. The membrane electrode assembly comprises the cathode and anode GDEs on either side of the anion exchange membrane (AEM). (C) Image of the flow cell.

CO2RR electrolysis in the flow cell was measured under constant-current conditions. We used in-line gas chromatography to detect and quantify the evolution of the CO and H2 electrolytic products over the current density range of 25 to 200 mA/cm2 in 25-mA increments (Fig. 3). The measurements show that FE values of >95% were maintained at current densities up to 150 mA/cm2 (Fig. 3). The selectivity measured at these high current densities compares favorably to a previously tested molecular catalyst immobilized on carbon nanotubes in a CO2RR flow reactor that showed an FECO of 56% (23). The higher selectivity in our case is likely a result of the delivery of gaseous CO2 to the electrode (15). This conjecture is corroborated by experiments in which FECO was tracked at various CO2 flow rates (fig. S5): As the flow rate of CO2 entering the flow cell increased from 2 to 100 SCCM, the selectivity toward CO increased from 90 to 99%. However, the single-pass conversion efficiencies decreased at progressively higher CO2 flow rates (table S3). The Ecell for our flow cell containing CoPc and a nickel anode remained below 2.7 V at current densities up to 200 mA/cm2. The Ecell of 2.61 V measured at 200 mA/cm2 compares favorably to the 2.72 V measured with a solid-state silver catalyst in the same flow cell and the >2.9 V required for related reactor architectures containing solid-state silver and IrO2 catalysts at the cathode and anode, respectively (6, 30).

Fig. 3 CO2RR selectivity for CO formation and applied voltage as a function of current density.

The Ecell and the corresponding FECO were measured for the CoPc-mediated conversion, with (red) and without (blue) phenol additive, of CO2 to CO at the indicated current densities. CC and turnover frequency at corresponding current densities are provided in table S2 (anode: nickel foam; cathode: carbon-supported CoPc with or without phenol (PhOH); anolyte: 20 ml/min, 1 M KOH; gas catholyte: 100 SCCM CO2; membrane: Sustainion AEM).

The FECO dropped from >90 to ~60% when the current density was increased from 150 to 200 mA/cm2. We hypothesized that this drop in selectivity resulted from a suppression of proton concentration at higher current densities. Our search for solutions to this problem converged on reports that phenol can help CoPc-mediated CO2RR (31, 32). We therefore modified our catalyst ink to also include phenol (denoted in Fig. 3 as “CoPc + phenol”). Experiments performed with this modified ink maintained a high FECO of 88% at 200 mA/cm2 while also providing a 90-mV voltage saving (Ecell = 2.52 V at 200 mA/cm2) (Fig. 3). This lower cell potential tracks molecular studies showing that phenol can lower the CO2RR overpotential (32). We conjecture that phenol acts as a local pH buffer that slows the formation of catalytically inactive bicarbonate at the electrode interface.

Molecular CO2RR catalysts typically degrade over a period of minutes, but they can be stabilized through molecular design (17, 22) or changing the reaction medium (33). Lu et al. reported that the immobilization of molecular CO2RR catalysts on carbon nanotubes can improve the stability of molecular electrocatalysts, but this composite system lasts for merely a few hours when measured at <40 mA/cm2 (23). We tested the stability of CoPc-mediated CO2RR electrolysis in our flow cell by quantifying product formation every ~1200 s at a constant current. The FECO could be held at >90% over 8 hours of electrolysis at 50 mA/cm2 (Fig. 4A), which corresponds to >4000 catalytic cycles (CCs) and a turnover frequency of 3.6 min−1 for each active site. At 100 mA/cm2, the FECO could be held at >90% over 3 hours of electrolysis before dropping to 60% after 5 hours of operation (fig. S4).

Fig. 4 Temporal stability of CoPc under flow at 50 mA/cm2.

FECO as a function of time for CO2RR electrolysis at a constant current density of 50 mA/cm2 in (A) a membrane reactor (anode: nickel foam; cathode: carbon powder–supported CoPc with phenol on carbon GDE; anolyte: 20 ml/min, 1 M KOH; gas catholyte: 100 SCCM CO2 humidified at 25°C; membrane: Sustainion AEM) and (B) a microfluidic reactor (anode: Ti/Pt; cathode: carbon powder–supported CoPc on a carbon GDE; anolyte: 20 ml/min, 0.5 M KHCO3; catholyte: 20 ml/min, 0.5 M KHCO3; gas catholyte: 20 SCCM CO2 humidified at 25°C; membrane: Sustainion AEM). The Ecell was measured at the same time as when gas chromatography measurement was taken.

It was unexpected that CoPc in the hydrodynamic environment of a flow cell at 50 mA/cm2 could match the longevity of a batch-type electrolysis experiment recorded at merely 10 mA/cm2 (34). The primary reason for the drop in performance in the flow cell is not CoPc degradation. Independent experiments performed in a microfluidic flow cell, which allows for the voltage reading at the cathode to be measured in isolation, maintained an >80% FECO for >100 hours at 50 mA/cm2 (Fig. 4B), corresponding to approximately 16,000 CCs. We hypothesize the rapid drop in output from the membrane reactor is largely due to a progressively lower proton inventory at the catalyst-membrane interface. This contention is supported by the fact that phenol helps stabilize the activity and selectivity of the reactor. The formation of KHCO3 crystals on the cathode flow plate during electrolysis also obstructs CO2 flow and reactor output. This crystal formation is a known phenomenon that arises from the crossover of potassium and hydroxyl ions (35). The original performance of the reactor can be regenerated by washing away the crystals after disassembly of the reactor. We continue to explore different strategies to realize robust CO2RR electrolysis in a flow cell reactor by circumventing crystal formation and developing methods to anchor the molecular catalysts to the supports. Regardless of the failure mode(s), there are few cases of immobilized molecular catalysts capable of sustaining electrolysis for longer time periods than the data reported here.

These results show that a widely available and abundant molecular catalyst is capable of operating in a flow cell with a high selectivity for CO production at commercially relevant current densities (≥150 mA/cm2). This finding challenges the accepted dogma that molecular catalysts are not capable of performing electrolysis at commercially relevant rates of product formation. The fact that CoPc lasts merely 10 hours at 10 mA/cm2 in a batch-type electrolysis cell (34), yet can sustain electrolysis at 50 mA/cm2 for >100 hours in a flow cell, provides the impetus to study other molecular catalysts under flow conditions. The identification that molecular catalysts can be used in a flow cell also offers the opportunity to lower the high cell potentials required to drive CO2RR chemistry at meaningful rates of production.

Supplementary Materials

Materials and Methods

Figs. S1 to S7

Table S1 to S3

References (46, 47)

References and Notes

Acknowledgments: Funding: We are grateful to the France-Canada Research Fund (New Scientific Collaboration Support Program), Canadian Natural Science and Engineering Research Council (RGPIN 337345-13), Canadian Foundation for Innovation (229288), Canadian Institute for Advanced Research (BSE-BERL-162173), and Canada Research Chairs for financial support. Partial funding from Air Liquide and the Institut Universitaire de France is also gratefully acknowledged. S.R. was supported by the University of British Columbia with an International Doctoral Fellowship. M.W. thanks the China Scholarship Council (CSC) for her doctoral fellowship (CSC student no. 201606220034). This research was funded in part by the Canada First Research Excellence Fund, Quantum Materials and Future Technologies Program. Author contributions: C.P.B. and M.R. conceived the idea and supervised the project. D.J. and S.R. performed electrolysis experiments and measurements and carried out data analysis. D.S. supervised the cell assembly and electrolysis measurements. K.T. and M.W. supervised the microfluidic flow cell experiments and the GDE fabrication. All authors discussed the results and assisted with manuscript preparation. Competing interests: The authors declare no competing interests. Data and materials availability: The data supporting the findings of the study are available in the paper and its supplementary materials.

Stay Connected to Science

Navigate This Article