Characterizing CO2 Reduction Catalysts on Gas Diffusion Electrodes: Comparing Activity, Selectivity, and Stability of Transition Metal Catalysts

Continued advancements in the electrochemical reduction of CO2 (CO2RR) have emphasized that reactivity, selectivity, and stability are not explicit material properties but combined effects of the catalyst, double-layer, reaction environment, and system configuration. These realizations have steadily built upon the foundational work performed for a broad array of transition metals performed at 5 mA cm–2, which historically guided the research field. To encompass the changing advancements and mindset within the research field, an updated baseline at elevated current densities could then be of value. Here we seek to re-characterize the activity, selectivity, and stability of the five most utilized transition metal catalysts for CO2RR (Ag, Au, Pd, Sn, and Cu) at elevated reaction rates through electrochemical operation, physical characterization, and varied operating parameters to provide a renewed resource and point of comparison. As a basis, we have employed a common cell architecture, highly controlled catalyst layer morphologies and thicknesses, and fixed current densities. Through a dataset of 88 separate experiments, we provide comparisons between CO-producing catalysts (Ag, Au, and Pd), highlighting CO-limiting current densities on Au and Pd at 72 and 50 mA cm–2, respectively. We further show the instability of Sn in highly alkaline environments, and the convergence of product selectivity at elevated current densities for a Cu catalyst in neutral and alkaline media. Lastly, we reflect upon the use and limits of reaction rates as a baseline metric by comparing catalytic selectivity at 10 versus 200 mA cm–2. We hope the collective work provides a resource for researchers setting up CO2RR experiments for the first time.


Material Characterization
Most material characterization techniques can be performed by one person, which makes executing the experiments and producing comparable results much easier. Nevertheless everyone should understand how and why we perform each technique and discuss the results of each method.

(High Resolution) Scanning Electron Microscopy -(HR)SEM
SEM ( fig. S2) and HR-SEM are used to obtain information about surface morphology and with that track whether its structure changes after 1h operation. Found abnormalities like clusters, deformation, crystals or exposed MPL could indicate poor stability. To be able to compare materials (HR-)SEM images are used to show similar morphologies across different metals as a result of sputtering. The HR-SEM is located in a cleanroom. Images are made at the following magnifications:  S3) is another surface probing technique in addition to the SEM. This technique will give information on surface topography and phase separation. For AFM we make use of TESPA-v2 probes. Height and phase images of the catalyst surface at a magnification of 500 x 500 nm and 1000 x 1000 nm are taken. Some parameters of the AFM can be sample specific and need to be optimized during operation.

Profilometry
Profilometry ( fig. S4) is a simple technique that measures the thickness of the samples by comparing the height to a reference point. The coarseness of the microporous layer is ±1 µm and the sputtered thickness we aim for is 100 nm. Because of this coarseness, profilometry will be performed on a piece of Menzel glass, added to each sputtering procedure. By partially Scotchtaping the glass surface and peeling the tape away after sputtering, an abrupt interface between sputtered and non-sputtered glass is created. This  S-4 interface is easily measured by profilometry, giving the thickness with a ±5 nm error. Although the porosity of catalysts on GDEs is higher than on glass, which results in a thicker layer on GDE, it is assumed that the translation of catalyst layer thickness between glass and GDE is constant over the range of materials. In the end the importance of this characterization is not the exact thickness, but to ensure similar thicknesses are present amongst all samples to obtain comparable mass transport properties.

X-ray Photoelectron Spectroscopy -XPS
XPS ( fig. S5) gives clear information about the purity and oxidation state of the samples. This way we can exclude or take into account the effects of impurities and oxidation of the catalyst. For each catalyst a survey scan is performed first. Elements of interest are scanned separately to obtain more accurate data. A C 1s scan shows the presence of substrate carbon (and some omnipresent surface carbon). This peak will hint at if the substrate became more or less exposed after operation. Very close to the C 1s peak is the K 2p peak used to identify potassium, which can deposit from the electrolyte on the surface during operation. A respective metal scan (e.g. Au 4f for Au) is used to see change in the catalysts abundance and oxidation. Finally, an O 1s scan is used to measure the degree of oxidation.

Sample handling and rinsing
While producing, installing and ex-situ measuring GDE samples it is important to avoid any direct contact with the MPL/catalyst layer. Due to the powder like structure of the Sigracet 38 BC MPL this layer gives off its catalyst and carbon black easily. In such a case a new sample needed to be deployed, since this could have affected the homogeneity, loading and pore size distribution. Therefore it is important to have proper handling strategies. During the experiments samples were only touched on the sturdy GDL backside or along the edges of the catalyst side using tweezers (outside of the active area). After placing a sample between two gaskets a mask was placed over to prevent any contaminations. During ex-situ experiments this care was also taken and it was made sure only the active area was studied.
During the reaction with the potassium containing electrolytes it is likely that at the cathode secondary reactions with the potassium ions in solution would occur. In order to prevent post-reaction salt-crusting or residual oxidations a rigorous rinsing protocol was deployed. After each experiment the catholyte and anolyte streams were switched with a 250 mL DI water solution that was continuously recirculated through the cell for approx. 5 minutes to remove any residual KHCO 3 /KOH electrolyte species. Afterwards the cell was opened and the sample was taken out with tweezers and sprayed again with bottled DI water for 30 seconds and subsequently dried using an abundant nitrogen stream. After this the samples were stored until ex-situ characterization was performed.

Performance Characterization
A major component of this comparison is the reproducibility of the experimental procedure. This section documents the details of the setup.

Cell construction and parameters
The experimental setup and the internal electrochemical cell will be built according to fig.  S6 and S7: The external setup consists of 2 external electrolyte compartments (80 mL electrolyte each) from which the pump (Cole-Parmer Masterflex L/S, φ pump = 10 mL min -1 ) transported the anolyte and catholyte to the lower cell inlets. The corresponding upper outlets go back to the electrolyte compartments from the outlets. The cell is tilted slightly to aid in the transport of anode formed oxygen out from the top of the anolyte compartment, hereby reducing potential fluctuations. The CO 2 MFC (Bronkhorst EL-flow, φ MFC = 30 sccm) is connected to the gas inlet of the cell and the outlet is connected directly to the GC once measurements are taken. During circulation of both liquid and gas an overpressure of 80-100 mbar is witnessed on the MFC pressure gauge. GC injections close the gas pathway temporarily (2-3 seconds) and cause overpressure to shortly spike to ~200 mbar, leading to minor gas crossover into the catholyte and subsequently leaving through the external compartment. The loss of gas has a minor effect on the gaseous product collection, but since this only occurs after every injection the system has a 4 minute time window to equilibrate before the following injection takes place. Figure S6. Practical experimental setup. a. External electrolyte compartments. b. Peristaltic pump (10 mL min -1 ). c. MFC for CO 2 (30 sccm). d.+ e. Potentiostat,BPR and GC control. f. PTFE flowcell. g. Liquid trap. h. Back pressure regulator. d.+ e. S-6 Between the anolyte and catholyte compartments the anode and membrane are positioned in a sandwich of gaskets. This sandwich consists of 5 parts: a gasket, the anode (bent away from the membrane and taped to the first gasket for electrical connection), a second gasket, the Nafion-212 cation exchange membrane and a third gasket. The sputtered GDE sample is also squished between two gaskets alongside a current collector. A more detailed description can be found in a paper by Liu et al. on assembly and operation of GDE cells [1].

Potentiostat
All potentiostatic measurements are performed with the ParStat 4000 ( fig S8) or ParStat MC. pH is measured before and after experiments. By combining pH and EIS results, the measured potential can be converted to RHE: Where E cath is the measured negative potential between the reference and the cathode, E 0 Ag/AgCl is the standard reference potential (E 0 Ag/AgCl = 0.1976V vs RHE @ 25 °C) and iR drop the negative current times measured resistance. Unfortunately it was not possible to accurately determine the iR drop through EIS at elevated current densities due to the long path of electrolyte between cathode and reference. More details on this can be found in SI B. Electrochemical impedance spectroscopy. Instead, all reported potentials are iR-uncorrected as described by the following formula: E (uncorr. vs RHE) = E cath + 0.0591 pH + E 0

Ag/AgCl
Chronopotentiometry experiments are the core part of this research. By keeping the current density (C.D.) constant and combining this with gas and liquid analysis we are able to determine the product selectivity at a certain production rate as well as the stability over time. The planar active surface area is 2.25 cm 2 , so the potentiostat input are adapted as shown in brackets. Chronopotentiometry measurements are taken for 1 hour at the following currents: After an experiment catholyte samples are taken and analyzed by HPLC for liquid products.

Gas Chromatography (GC)
GC ( fig. S9) measurements are taken every 4 minutes from the gas phase outlet during all chronopotentiometry experiments to measure the concentration of gaseous products. Before each measurement 3 injections without applied potential are performed as a second check of the baseline after flushing the system. It is important to perform regular calibration checks to ensure measurements are accurate. The gas products CO, H 2 and C 2 H 4 are calibrated at 3 levels (10, 100, 1000 ppm).

High Performance Liquid Chromatography (HPLC)
After each chronopotentiometry experiment a sample of catholyte will be taken to the Agilent Technologies 1260 Infinity II HPLC (fig. S10) equipped with VWD (dual wavelength: 210 nm and 280 nm) and RID (T = 40 °C) to measure the concentration of liquid carbon containing species on basis of their retention times in an Hi-Plex H column (T = 50 °C). Main products of interest here are formic acid, acetic acid, acetaldehyde, ethanol and propanol. 10 levels of calibration were performed (10-10000 ppm, R 2 > 99.9 %)

SI B. Characterization data
This section displays the results of the performed experiments. Data is collected via chronopotentiometry, electrochemical impedance spectroscopy (EIS), gas and liquid product analysis (GC/HPLC), (high resolution-) scanning electron microscopy (SEM & HR-SEM imaging), X-ray photoelectron spectroscopy (XPS) and atomic force microscopy (AFM).

Electrochemical impedance spectroscopy -EIS
Electrochemical impedance spectroscopy can be used to determine the resistance between the reference and working electrode and consecutively correct the measured potential for any resistive losses. During this series of experiments an EIS measurement was performed before and after each experiment. It was found that the configuration of our system had a significant drawback: Due to the relatively large distance between the reference electrode and the GDE cathode (~8 mm) the measured resistance was 3-4 Ω for 1 M KOH and 7-8 Ω for KHCO 3 . This extremely large correction factor caused the iR corrections to lower the voltage significantly. To illustrate: 8 Ω * 450 mA = 3.6 V of ohmic drop. Furthermore, it was found that correcting potentials at 100/200 mA cm -2 with said resistance caused overcompensation. For instance, the corrected potential at 200 mA cm -2 in 1 M KHCO 3 was lower than the corrected potential at 50 mA cm -2 . In some cases the iR-drop even exceeded the applied potential, indicating that the resistance during operation at elevated currents was actually lower than we were able to measure during 'offline', no-current EIS.
Concluding, it was decided to not correct the measured potential for its resistive losses. This was done to avoid confusion about the measured potentials due to changing electrolyte conductivities throughout the length of the experiment, and to evade overcompensation of unrealistic resistances.

S-9
Chronopotentiometry Below V cat -t chronopotentiometry diagrams of all metal-electrolyte combinations are shown as measured against an Ag/AgCl reference electrode without corrections for iR-drop (fig. S11 -S15). Most of the high current experiments show diverging from the patterns. These values indicate the instabilities of the system at elevated activities, due to bubble formation in the electrolytes resulting in issues such as GDE flooding and gas crossover.

Time-dependent Faradaic efficiencies -FE
During all experiments GC samples were taken every 4 minutes. The measured signal is converted into a concentration. Liquid analysis by HPLC is only performed when the experiment is completed after which the production is averaged over the duration. The following formula is used to calculate and plot the time-dependent Faradaic efficiencies (fig. S16 -S20). FE n is the Faradaic efficiency of product n, z n the number of electrons per formed molecule of product n, F the Faraday constant, c n the concentration of n measured by the GC, φ CO2 the molar flowrate of CO 2 and I tot the total current going through the system.

Scanning electron microscopy -SEM
In figure S21 below HR-SEM images of the as-deposited catalysts are displayed. All catalysts show similar porosity and catalyst coverage on the GDE. Sn shows slightly increased agglomeration of the catalysts due to the more volatile nature of the metal (Note: the deposition power of Sn was 20 W as compared to 50 W for the other metals.) Figure S21. HR-SEM images of all 5 as-deposited catalysts. Throughout a similar porosity and particle size can be observed.
The remainder of this section contains scanning electron microscopy images taken from a bare Sigracet 38 BC GDL (fig. S22) as well as fresh and used samples ( fig. S23 -S32). Initially 500 mA cm -2 (=1.125 A) experiments were also attempted, but due to the limiting compliance voltage of our potentiostat (ΔU max = 12 V cell ) alongside the high resistance of the wide cell, it was not possible to achieve the requested current. Instead, the physical effects of 12 V total cell potential can be witnessed, at which current densities between 300-500 mA cm -2 were achieved. SEM images of these experiments do show interesting features of potassium depositions, catalyst reconstruction and a continuation of the earlier observed surface changes to a greater extent.

Magnifications and total dimensional image sizes
Magnification: x 50 Size: 2.6 x 1.7 mm Magnification: x 100 Size:

x 85 µm
Magnification: x 5000 Size: 26 x 17 µm Bare gas diffusion layer (GDL) Figure S22. SEM images of a bare gas diffusion layer (Sigracet 38 BC) Figure S23. SEM images of 100 nm Ag on GDE after 1 hour reaction in 1M KOH at various current densities. Overall this system seems stable up to 200 mA cm -2 . At 500 mA cm -2 the surface displays branched coverage, which was identified to mainly contain increased amounts of potassium and oxygen. It is likely the high overpotentials initiated a growing deposition of potassium from the electrolyte.

High-resolution scanning electron microscopy -HR-SEM
This section contains three magnifications of high resolution scanning electron microscopy (HR-SEM) images of the 5 metals as deposited and after electroreduction for 1 hour in 1 M KOH and 1 M KHCO 3 at V cell = 12 V ( fig. S33 -37. Above the images the used magnification can be found. Between the sputtered and used samples differences such as agglomeration, clustering, smoothening, deposition and reformation can be found.

X-ray photoelectron spectroscopy -XPS
This section contains all XPS scans of the 5 metals as deposited and after electroreduction at 200 mA/cm 2 for 1 hour in both KOH and KHCO 3 ( fig. S38 -S52). During XPS the following scans were performed for each of these samples: Corresponding to the numbers above the following format is used:

6.
The survey and respective metal elements give us information on the stability of the metal, for instance, in the Sn fresh vs Sn KOH/KHCO 3 case there is a significant signal drop-off of the Sn3d peak (~99% for KOH, ~50% for KHCO 3 ) after testing, indicating its instability.
To obtain additional information on the presence and stability of the GDE the carbon (and fluorene) peaks are looked into. Each scan is performed on and averaged over 2 separate locations (random spots, not on substrate native crevices).

Atomic force microscopy -AFM
Atomic Force Microscopy (AFM) is a powerful tool to look into the smallest details of the reactive surface. Below μm size topography heightmaps of a bare GDE sample can be found as well as of before and after catalysis samples of Ag, Au, Sn and Cu (Fig. S53 -S57). The Z-axis of in the 1 μm x 1 μm heightmaps has an aspect ratio of 1:1:0.3 (the 2 μm x 2 μm 'Bare GDE' sample has a normal 1:1:1 ratio). Although it is hard to compare AFM images, it gives an idea of what the surface looks like on the nanoand microscale.