Role of Mass Transport in Electrochemical CO2 Reduction to Methanol Using Immobilized Cobalt Phthalocyanine

Electrochemical CO2 reduction (CO2R) using heterogenized molecular catalysts usually yields 2-electron reduction products (CO, formate). Recently, it has been reported that certain preparations of immobilized cobalt phthalocyanine (CoPc) produce methanol (MeOH), a 6-electron reduction product. Here, we demonstrate the significant role of intermediate mass transport in CoPc selectivity to methanol. We first developed a simple, physically mixed, polymer (and polyfluoroalkyl, PFAS)-free preparation of CoPc on multiwalled carbon nanotubes (MWCNTs) which can be integrated onto Au electrodes using a poly(3,4-ethylenedioxythiophene) polystyrenesulfonate (PEDOT:PSS) adhesion layer. After optimization of catalyst preparation and loading, methanol Faradaic efficiencies and partial current densities of 36% (±3%) and 3.8 (±0.5) mA cm–2, respectively, are achieved in the CO2-saturated aqueous electrolyte. The electrolyte flow rate has a large effect. A linear flow velocity of 8.5 cm/min produces the highest MeOH selectivity, with higher flow rates increasing CO selectivity and lower flow rates increasing the hydrogen evolution reaction, suggesting that CO is an unbound intermediate. Using a continuum multiphysics model assuming CO is the intermediate, we show qualitative agreement with the optimal inlet flow rate. Polymer binders were not required to achieve a high Faradaic efficiency for methanol using CoPc and MWCNTs. We also investigated the role of formaldehyde as an intermediate and the role of strain, but definitive conclusions could not be established.


Cell Fabrication
The fabrication of electrolysis cells with a 1 cm 2 active cathode and anode area is described here.
Two approximately 2 x 2 x 0.5 in PEEK blocks (McMaster Carr) were cut using a band saw.Each block was then placed in a mill (Bantam Tools, Othermill) where the liquid chambers, pilot holes where the screws are to be inserted were drilled, and the cell contours were cut.1/8" bits were used for all cutting operations.The cell was then flipped over and face cuts were used to cut the cell to thickness (1 cm).The cells were then sanded either manually or with a belt sander to remove machine marks and to smooth the surface.
The pilot holes were then drilled completely using a 0.209" drill bit.Using a 0.043" drill bit, holes for the gas inlet and outlet were drilled at the ends of the flow lines.
For the cathode/anode chamber, a hole punch was used to make pilot holes above and below the chamber.A 0.128" drill bit was then used again to drill a hole for liquid flow. 2 cm segments of 1/8" OD PEEK tubing (Trajan) were then cut and inserted into the holes with epoxy as the adhesive.
On the catholyte chamber, on the center of another face, using a 0.063" drill bit, a thin hole was drilled for the reference electrode.Then a shallow hole was drilled in accordance to tap a ¼-28 hole, and the hole was then tapped using the ¼-28 tap.This is to fit a 1/8" flangeless PEEK fitting (Cole-Parmer), and a leakless Ag/AgCl reference electrode (eDAQ) was inserted into this fitting.
The back plates were prepared in the same manner, except using a 0.25" thick PC plate.Nature rubber gaskets (40A Durometer, McMaster Carr) were cut using a laser cutter.Long screws and wingnuts were used to hold the cell together.
See Figure S1 for the geometry of the machined areas for the 1 cm 2 active area cell and for an overall schematic.
Step files for the CAD drawings that were used to generate machine paths have been uploaded as additional supplementary materials.

Electrochemical Testing
The cell was assembled according to the schematic shown in Figure S1 using a gasket to seal the layers (4 total).Cu tape was attached to the back of a carbon fiber paper anode, and the measured electrode.A cotton swab is then used to remove some of the catalyst coating from the top of the electrode such that Cu tape can better adhere to it.The flow cell stack was order from bottom to top as followed: polycarbonate back plate, carbon fiber paper anode, rubber gasket, PEEK anode chamber, rubber gasket, SELEMION™ membrane, rubber gasket, PEEK cathode chamber, rubber gasket, tested cathode, and polycarbonate top plate.Upon assembly completion, the four screws were tightened with nuts to finger tightness.The Ag/AgCl reference is inserted last and screwed into side of the cathode chamber.10 mL of 0.05 M K2CO3 were then poured into the anolyte and catholyte reservoirs.These reservoirs each had 4 inlets/outlets (gas in, gas out, liquid in, liquid out), though only the liquid in and out were used in the anolyte reservoir.Flexible tubing (Cole Parmer Masterflex 13) was used to connect the cell, electrolyte reservoirs, gas source, peristaltic pump (Cole Parmer), and outlet to GC as shown in Figure S2.The gas outlet was connected to a 5 mL vial which was connected to the GC.This vial served as a mixing volume for the gases in order to time average the gas production since bubble formation and bursting can lead to inhomogeneities in the apparent gas flow.
The catholyte chamber was sparged and saturated using CO2 gas at a flow rate of 50 sccm for 15 minutes before electrolysis.After 15 minutes the flow rate was switched to 5 sccm, which will continue for the remainder of the experiment.The peristaltic pump was then set to the desired flow rate (given the cell dimensions, volumetric flow in mL/min = lateral flow speed in cm/min) The cell was connected to a potentiostat (Biologic SP-300), and an EIS measurement was taken to determine the solution resistance.The solution resistance was then inputted into a manual IR compensation method and a chronoamperometry (CA) experiment was performed for 80 minutes with 85% IR compensation.

Detection of Gas Products
The outlet of the catholyte chamber is connected to the inlet of the GC (SRI Multi Gas 3) to allow for gas detection.The GC is equipped with a 2 mL sampling loop which will inject the gas every 16.5 minutes.Two columns (Hayesep-D) and Mol Sieve 13x were used to separate the gases, and an electronic valve was used to redirect the bulk of the CO2 band as to not overwhelm the detectors.
A methanizer + FID detector was used for CO detection, and a TCD was used for H2 detection.
Argon was used as the carrier gas.

Detection of Liquid Products
Liquid products were determined using 1 H NMR (Bruker 500 MHz) following procedures similar to those described by Chatterjee et al. 1 An internal standard of 50 mM DMSO, and 50 mM phenol in DI water was prepared.Each NMR tube was prepared with 400 uL analyte, 50 uL internal standard, and 50 uL D2O.The 90° pulse width was obtained for each solvent system and the center of the water peak.A solvent suppression method was then used to obtain the spectrum using a 25 second delay time.The integration ratio between the MeOH peak and DMSO peak was used to calculate the MeOH concentration.Calibration of the internal standard was done using different standard solutions of MeOH and potassium formate.
Detection of formaldehyde requires further sample preparation.900 μL of electrolyte sample was mixed with 900 μL of a 100 mg/L solution of PFBHA and then heated at 60 °C for 10 minutes.
The solution was then injected into an Agilent 7890A GC-MS.A fritted glass liner was used to hold all the salts would precipitate out upon heating.Method details are shown in Table S1.
Calibration was done by doing the same steps as above but with standard solutions of 0.5, 1, 3, 6, and 10 mg/L of formaldehyde.

Calculation of Faradaic Efficiency for Gas and Liquid Products
The following formula was used to calculate the gas product faradaic efficiency: where C is the concentration of the product as determined by the GC, J is the molar flow rate of CO2 into the reservoir, n is the number of electrons to the product (2 for both H2 and CO), F is Faraday's constant, and i is the average current at the time of injection.The current used was the time averaged current in a 3-minute window, which corresponds approximately to the total mixing volume of the system.To calculate the error, 6 gas measurements were taken over the course of the 80-minute experiment for each experimental run.The first gas measurement is excluded from analysis as it is taken at the start (t= 0 min).Each experimental condition was repeated in triplicate separate run with fresh samples.The error of the gas measurements was calculated from the standard deviation of the all the gas measurements from the experimental condition.
The following formula was used to calculate the liquid product faradaic efficiency: where Cref is the concentration of the reference, IP and IR are the integrations of the product and reference respectively, HR and HP are the number of corresponding protons on the reference and product respectively (6 for DMSO, 3 for MeOH), n is the number of electrons to product (6 for MeOH), F is Faraday's constant, V is the volume of electrolyte used for the experiment (10 mL typically), Q is total charge passed, and the factor of 5/4 is due to 400 uL of analyte being diluted to 500 uL for the NMR analysis.
The error of the liquid product Faradaic efficiencies were calculated from the standard deviation of the triplicate experimental runs.

Calibration of Reference Electrodes
An aqueous solution of 0.05 M K3Fe(CN)6 was made in a glass scintillation vial.A glassy carbon disk was used as the working electrode (0.4 mm diameter), and a graphite rod was used as a counter.
A CV was done from -0.1 V to 0.5 V, starting at 0.2 V at a sweep rate of 50 mV/s.The wellbehaved redox feature was compared against literature and was monitored for any electrochemical drift or shifts.

CO2 Labeled Experiments
The experiment for using labeled gas was exactly the same except for as follows.0.1 M KHCO3 was used, 13 CO2 was used as the gas (Sigma-Aldrich 99.0 atom % 13 C, < 3 atom % 18 O), the electrolyte was first sparged with Ar for 20 minutes at 50 sccm, then 5 sccm of 13 CO2 was flowed for 12 minutes before starting the experiment.The electrolysis was run for 45 minutes at -1.2 V vs RHE at 8.5 mL/min flow rate.

Details of Continuum Modeling
Steady-state mass conservation and Fick's law of diffusion is used to solve for the concentration profiles of CO2 and CO within the electrolyte boundary layer adjacent to the cathode surface (model is illustrated in Figure S13).
where   and   are the diffusivity and concentration of species  , respectively, and  is the electrolyte velocity.The first term captures the transport of species by diffusion, and the second term describes transport by convection. 2 The diffusivities and other model parameters are provided in Table S2.
The velocity profile is calculated by the continuity equation and the Navier-Stokes equation for an incompressible, Newtonian fluid. 3∇ •  = 0 (S4) where  and  are the density and dynamic viscosity of water, respectively (used material property value for water within COMSOL at 20 o C),  is pressure, and  is the identity matrix.Equations S3-S5 were solved simultaneously within COMSOL Multiphysics 6.0 software with a triangular mesh that coarsens going from the catalyst surface to the bulk of the electrolyte.A sensitivity analysis on the mesh identified a minimum mesh size of 11842 domain elements and 642 boundary elements.
The boundary conditions are presented in Figure S13.In short, a no slip condition is used at the catalyst surface, the bulk electrolyte boundary is treated as an open boundary, and inflow and outflow conditions were used at the entrance and exit, respectively.The surface reactions were modeled with simple first-order rate expressions where the rate constants (k1 and k2) were fit so that the simulated methanol current density versus inlet flow rate peaks around the experimentally measured optimal inlet flow rate (~ 8.5 cm min −1 ).In general, changing k1 and k2 changes the inlet flow rate that maximizes methanol production.       is the H 13 CO3 -which comes from sparging with the isotopically labeled 13 CO2.
Figure S10: 1 H NMR of solution after CA using CO as the gas feed.Electrolyte was 0.1 M KHCO3, electrolyte flow rate of 8.5 cm 3 min -1 with an applied potential of -1.8 V vs Ag/AgCl.

Figure S1 :
Figure S1: Schematic of electrochemical cell, and the geometry of the anode and cathode chambers.

Figure S2 :
Figure S2: Schematic of liquid flow and gas flow through the electrochemical reactor and a schematic of the electrode stack along with an image of the electrode.

Figure S3 :
Figure S3: (a) Ratio of CoPc to MWCNT to the Faradaic efficiency to methanol.(b) partial current density to methanol.All tests were done in CO2 saturated 0.1 M KHCO3 with a linear electrolyte flow speed of 8.5 cm/min at -1.2 V vs RHE.

Figure S4 :
Figure S4: CAs of control experiments of (a) a bare, planar Au, (b) Au with PEDOT:PSS, (c) Au with PEDOT:PSS and MWCNTs.(d) FEs of products in the control experiments.All tests were

Figure S5 :
Figure S5: An example of the gas Faradaic efficiency over the course of a chronoamperometry experiment using CoPc/MWCNT at a 0.6:1 CoPc:MWCNT ratio.The gas Faradaic efficiency evolves during the first part of the experiment before reaching a steady state average.This electrode was held at -1.3 V vs. RHE in CO2 saturated 0.1 M KHCO3 with a linear electrolyte flow speed of 8.5 cm/min.

Figure S8: 1 H
Figure S8:1 H NMR of a control experiment with similar FE methanol as a reference for comparison with the 13 CO2 labeled experiment.The 1 H NMR shows that the predominate methanol species formed is 12 C based.

Figure S9 :
Figure S9: 13 C NMR of a control experiment with similar FE methanol as a reference for comparison with the 13 CO2 labeled experiment.This 13 C NMR shows that only 13 C labeled species

Figure S11 :
Figure S11: CA using CO as the gas feed.Electrolyte was 0.1 M KHCO3, electrolyte flow rate of 8.5 cm 3 min -1 with an applied potential of -1.8 V vs Ag/AgCl.

Figure S12 :
Figure S12: (a) Current density with varying potential on CoPc/MWCNT catalyst, (b) Current density with varying flow rate on CoPc/MWCNT catalyst.(c) Faradaic efficiency of the different products with flow rate, plotted individually.These corresponded with the main Figures1 and 2and were done with the 0.6:1 CoPc:MWCNT ratio at 0.15 mg cm -2 loading.Flow rate was kept constant (8.5 cm/min) when varying potential and potential was kept constant when varying flow rate (-1.2 V vs RHE).

Figure S13 :
Figure S13: Illustration of model domain and all the mathematical forms of the boundary conditions.

Figure S14 :
Figure S14: Surface plot of CO concentration throughout the model domain at an inlet velocity of (a) 1 cm min −1 , showing a case when lateral outward diffusion of CO limits efficient conversion (b) 8.5 cm min −1 , our optimized case where convection helps CO transport but not so fast as to sweep it away.(c) 30 cm min −1 ; a case where convection dominates such that that CO is swept away before it can be further converted to methanol.

Figure S15 :
Figure S15: Surface plot of CO2 concentration throughout the model domain at an inlet velocity of (a) 1 cm min −1 , showing a case when CO2 depletion allows for stronger HER competition (b) 8.5 cm min −1 , our optimized case.(c) 30 cm min −1

Figure S16 :Figure S17 :
Figure S16: XPS core level spectra for N 1s for CoPc electrodes with different preparations:

Figure S18 :
Figure S18: (a) Example chromatogram of a sample containing formaldehyde after derivatizing with PFBHA.The peak right after 6 minutes is the formaldehyde-oxime peak.The two peaks after are from the unreacted PFBHA.They elute as separate peaks due to equilibrium of the PFBHA with water.(b) Mass spectrum of the formaldehyde-oxime peak showing the selected ion masses.