Dynamic transformation of cubic copper catalysts during CO2 electroreduction and its impact on catalytic selectivity

To rationally design effective and stable catalysts for energy conversion applications, we need to understand how they transform under reaction conditions and reveal their underlying structure-property relationships. This is especially important for catalysts used in the electroreduction of carbon dioxide where product selectivity is sensitive to catalyst structure. Here, we present real-time electrochemical liquid cell transmission electron microscopy studies showing the restructuring of copper(I) oxide cubes during reaction. Fragmentation of the solid cubes, re-deposition of new nanoparticles, catalyst detachment and catalyst aggregation are observed as a function of the applied potential and time. Using cubes with different initial sizes and loading, we further correlate this dynamic morphology with the catalytic selectivity through time-resolved scanning electron microscopy measurements and product analysis. These comparative studies reveal the impact of nanoparticle re-deposition and detachment on the catalyst reactivity, and how the increased surface metal loading created by re-deposited nanoparticles can lead to enhanced C2+ selectivity and stability.


Supplementary
. Comparison of catalysts on the working electrode before and after a linear potential sweep to -1.1 VRHE. Slow-scan STEM images taken from a larger field (a) before and (b) after sweeping the potential. The yellow dashed boxes indicate the area imaged during the in situ experiment shown in Figure 1 of the main text. Note that the better contrast and resolution of the area near the window edge (thinner liquid layer) indicate the absence of small redeposited particles before the potential is applied. The observation of re-deposited particles extending beyond the in situ imaged area after the experiment also confirms that the re-deposition of particles is caused by the applied potential, and not due to beam-induced effects. STEM images showing an experiment where (a) re-deposited NPs were initially seen but were sweep away after (b) a bubble formed and (c) was subsequently pushed out with a higher injection rate on the syringe pump. Images comparing of (d) the end of an in situ experiment where the redeposited NPs could be seen at but were gone after (e) the liquid cell was dis-assembled and the chip was rinsed for ex situ measurements. Notice that the cube in the upper right corner of (d) also detached. Figure 6. Representative plots of currents measured during the EC-TEM chronoamperometric experiments. In general, currents are consistent with the catalyst density observed during the experiments, but there were some differences between similar experiments, which we attribute to variation in flow conditions between different liquid cells as reported previously 1 . The spikes in the currents are due to the stalling of the syringe pump caused by an empty syringe. Generally, the Cu cubes on the chip (a) have a broader size distribution and lower loading that the cubes on the glassy carbon plate (b). Figure 11. Chronoamperometric traces obtained during the benchtop timeresolved product analysis experiments. The plot shows the chronoamperometric traces for 80 nm, 170 nm, and 390 nm Cu cubes on glassy carbon plates measured for 12 h at -1.1 VRHE. Each trace is the average of three independent measurements. The error bars give the standard deviation of three measurements. Figure 12. Comparison of the time-dependent evolution of the morphology of 170 nm cubes and gaseous products of CO2RR at different reduction potentials. (a) Ex situ SEM images of Cu2O cubes deposited on glassy carbon plates before and after 12 hours of the reaction. Comparison of Cu cubes deposited on a glassy carbon plate before (pristine) and after CO2RR at -0.9 VRHE, -1.1 VRHE and -1.3 VRHE. It should be mentioned that the after-reaction images at -1.3 VRHE also indicate that there was detachment and aggregation of these larger cubes, similar to what we observed in the 80 nm cubes at -0.9 VRHE, as described in the main text. (b) Partial current density of 170 nm Cu cubes measured at -0.9 V, -1.1 V, and -1.3 VRHE for CO, H2, CH4, and C2H4. The biggest changes were observed within the first hour of reaction, with secondary long term stability changes. The error bars give the standard deviation of three measurements.  Figure 3 and a glassy carbon plate. The sample labelled as "lower loading" has ~50% less cubes than the samples described in the main text. The signals of the liquid products for the lower loading samples were below the detection limit. The hydrocarbon production significantly decreased with the reduced loading. The error bars give the standard deviation of three measurements.

Supplementary
Supplementary Figure 17. Ex situ TEM images of 50 nm Cu2O cubes synthesized with colloidal chemistry. (a) Image of the as-synthesized cubes. Images of cubes after 1 hour of CO2RR that are found in two areas of the silicon nitride window: (b) the carbon WE of the EC-TEM chip and (c) the bare silicon nitride membrane. Images on the carbon WE show significant restructuring and formation of small NPs whereas the cubes on the silicon nitride show only small changes in their shape (presumably un-reacted due to the absence of electrical contact). Figure 18. Dynamics of the 170 nm of Cu2O cubes obtained at high electrolyte flow velocities. Image sequences acquired under higher electrolyte velocity (5 ml/min). The scan rate during the linear sweep was 15 mV/s and the applied constant potential during chronoamperometry was -0.9 VRHE. The electron flux used was 3.5 e -Å -2 s -1 . Figure 19: Identical location SEM images comparing the same 170 nm cubes before and after 12 hours of KHCO3 exposure. The middle column presents an overlay of the SEM images acquired before (blue), and after (red) KHCO3 exposure There were no significant changes observed in the morphology and coverage due to the extended KHCO3 exposure in the absence of an applied potential, suggesting that the dissolution is only limited to the cube surface and the initial electrolyte exposure. Figure 20. Schematic of the experimental geometry. The in situ imaging is achieved by encapsulating the electrolyte in between two chips with electron-transparent silicon nitride membrane windows. The electrolyte is flowed into the cell using a syringe pump, whereas the electrochemical biasing is achieved through a micropatterned carbon film that is connected to the working electrode of a potentiostat. Image of the TEM holder is courtesy of Hummingbird Scientific. Figure 21. A representative EELS spectrum acquired during our in situ TEM experiments. The profile of the spectrum indicates significant scattering of the primary electrons as they propagate through the electrochemical cell. Note the absence of the zero-loss peak, which indicates that the liquid thickness exceeds at least a few inelastic mean free paths of the 300 kV electrons. The features of the spectrum indicate that we have a liquid layer thickness of more than 1 µm. 3

Comparison of the Electrochemical Performance of the EC-TEM Cell and a Standard H-Type Cell
To determine if the electrochemical data obtained from the EC-TEM cell accurately reproduced the behavior found in the benchtop measurements, we compared the current densities that were obtained after correcting for the surface area of the Cu catalysts. The exposed surface area of Cu (areaCu) was estimated from the size and distribution of the as-synthesized cubes in the electron microscopy images. Here, we assume that each cube has 5 exposed facets of similar size, with the sixth one covered in contact with the support. The contribution to the current from the pure glassy carbon plate is negligible compared to the catalysts. For the ex situ experiments, SEM images of at least three different locations have been taken on the sample and the number of cubes per µm 2 as well as their average size have been determined. For the in situ experiments, the same has been done using the STEM images. The linear sweep voltammetry currents and chronoamperometric traces have been normalized by dividing the measured currents by the total surface area of Cu.
We are aware that such normalization is not ideal, since the total Cu surface area is changing under reaction conditions due to the morphological changes in the catalysts, as shown our in situ experiments. Nevertheless, we consider that this approach is a reasonable approximation as it captures the differences that are caused by the initially distinct cube size and loading, which would be otherwise ignored if the support geometrical surface area is used for the normalization.
In addition, we performed further measurements to compare the onset potentials of CO2RR and Cu dissolution in the holder and in the flask to check that the applied potential in the EC-TEM microfluidic cell can be translated to that found in our standard-type cell. First, we applied a structured cyclic potential recipe to electrodeposit Cu2O particles on the EC-TEM chips and the glassy carbon plates. 4 Then, these samples were subjected to a series of cyclic voltammetry scans at 25 mV/s in 0.1 M KHCO3 (Supplementary Figure 22). Such an approach is necessary because a sustained oxidative potential quickly drives the complete dissolution of Cu from the carbon surfaces. From these measurements, we determined that the cyclic voltammograms obtained from these specially fabricated reference electrodes are reproducibly shifted by ~-0.4 V from the cyclic voltammograms obtained from the H-type cell. We attribute the measured offset to the unique design of the EC-TEM cell/holder and the longer (fixed) separation between the reference and working electrodes. To mitigate any discrepancy in the applied potential, we adopted an additional calibration step during our experiments to ensure that we are reproducibly applying the right potential. Every experiment was started with two linear sweep voltammograms and then, the profiles were compared to the CO2RR onset potential to our benchtop reference measurement of a sample with similar loading and size (Supplementary Figure 23 compares LSVs taken on different days). These measurements also indicate that the reference potential remained reasonably stable over long periods of time. In general, we only found relatively small day-to-day shifts (< 50 mV) between different experiments when using the same reference electrode. The day-to-day shifts can be explained by degradation of the KCl solution or the AgCl in the reference electrode over time, both of which needs to be much smaller to fit within the TEM holder.
An example where the initial linear sweep voltammograms are superimposed after correcting for the reference potential offset and normalization over the exposed surface area of Cu for the samples in the two setups is shown in Supplementary Figure 24. It is clear from the profiles that the current densities also compare well against each other. We emphasize here that our experiments have shown that the structural changes in range of -0.9 VRHE to -1.3 VRHE is not strongly potential dependent and only the rate of change is affected by applied potential. Hence, both time-resolved imaging and electrochemical data from both in situ TEM and benchtop setups indicate a robust correspondence between the two systems, allowing us to extrapolate the in situ findings and associate them with the real changes in catalytic performance measured from the identically prepared samples on the bulk electrodes. The currents have been normalized over the estimated exposed surface area of Cu on the working electrode, which is a glass carbon plate in the H-type cell and a carbon thin film in the EC-TEM experiments respectively.

Time-Resolved Ex Situ SEM Imaging of Cubes Reacted in a Standard H-Type Cell
To confirm that the EC-TEM observations can be extrapolated to the behavior in a standard electrochemical cell, we further followed ex situ the morphological evolution of the Cu2O cubes deposited on the glassy carbon plates during CO2RR. The samples were extracted periodically over one hour and investigated by SEM. Supplementary Figure 25 shows a sequence of SEM images from a sample with 170 nm cubes acquired after different CO2RR times at -0.9 VRHE. Indeed, the SEM images show the restructuring of the cubes and NP re-deposition similar to that observed in the in situ experiments. As shown in the histograms in Supplementary Figure 25(b) and 25(c), the cubes in our standard cell also did not change significantly in size and number density during the experiments, whereas the re-deposited NPs increased in size and number.
We mention here that the increase in re-deposited NP size (not resulting from aggregation) and number is artifact of these time-resolved ex situ experiments. The repeated removal of the glassy carbon plates from the electrolyte and applied potential, leads to surface re-oxidation and a new cycle of re-deposition when the samples are re-introduced into the reaction environment, as shown in the chronoamperometric plots shown in Supplementary Figure 25(d). This behavior, in turn, results in a perceived increase in re-deposited NPs density as seen in the time-resolved series. In addition, we reiterate that these time-resolved measurements are also not sensitive to catalyst detachment events as shown in Supplementary Figure 5. Therefore, while they can be sufficient for supporting the appearance of structural change during electrochemistry, they should not be used to make any quantitative conclusions regarding the catalyst dynamics.
Lastly, we show in Supplementary Figure 26 the images of the catalysts found on the EC-TEM chips after 1 hour of reaction, which indicates that the morphological changes are consistent with those found on the glassy carbon plates. The high initial reductive currents following each re-immersion of the glassy carbon plate suggested that the catalysts re-oxidized when they were removed from the electrolyte and the applied potential. Hence, each re-immersion also introduced a new cycle of dissolution and redeposition.

Supplementary Figure 26. Ex situ SEM images of Cu2O on the EC-TEM chips before and
after the CO2RR reaction. The reaction time was 60 minutes at a potential of -0.9 VRHE. All images are at the same magnification.

Converting the Applied Potential to the Reversible Hydrogen Electrode Potential
The conversion of the applied potential to the potential of the reversible hydrogen electrode can be done with the Nernst equation:

Log10(H + ) = -pH
where Log10(H + ) is simply the negative of the electrolyte pH value, which is 6.8 for of CO2 saturated KHCO3.

Formula used for the product analysis
Calculation of Faradaic efficiency of gas products Calculation of Faradaic efficiency of liquid products Calculation of partial current density itotal: total electrolysis current in A ia: partial current density for the respective product "a" in A

Electron Beam-Induced Effects in In Situ Experiments
Electron beam-related effects are always a concern for liquid cell TEM experiments, and so we have adopted a low electron dose imaging protocol to minimize these artifacts. In our previous work, 5 we determined the electron flux limit for observing noticeable beam-induced effects in our TEM to be ~7 e -Å -2 s -1 and we stayed under this threshold flux in the current experiments. In Supplementary Figure 27, we present another control experiment where the electron beam was blanked until only after potential application. While it was unfortunate that a bubble formed during the potential sweep, it was clear that we could see the fragmentation of the cubes and the formation of re-deposited NPs. Supplementary Figure 28 further shows the after-reaction morphologies acquired after the in situ experiments from the areas not exposed to the electron beam. The cubes located in the two areas do not exhibit significant differences in morphology, indicating that those extended imaging did not significantly alter the cubes. Hence, we can conclude that we were able to avoid significant electron beam-induced artifacts with our low electron dose protocol, even with extended imaging times of about an hour.
There appears, however, to be subtle effects caused by the extended electron beam exposure since there seems to be more secondary NPs in the imaged area as compared to non-imaged areas (Supplementary Figure 29). Nevertheless, this effect does not affect the conclusions of our work the morphologies are similar between the two areas and are also consistent with the behavior found in our ex situ SEM experiments described in Supplementary Note 2. frames were averaged to create one frame of the movie. The movie playback rate is ×30 times real time. The electron flux was 1.7 e -Å -2 s -1 .

File Name: Supplementary Movie 4
Description: EC-TEM movie describing the motion of particles driven by bubble formation and their removal process during the chronoamperometry at -1.1 VRHE. The movie play back rate is ×5 times real time. The electron flux was 0.11 e -Å -2 s -1 .

File Name: Supplementary Movie 5
Description: EC-TEM movie describing the structural changes in the 390 nm Cu2O cubes during frame per second. 5 frames were averaged to create one frame of the movie. The movie playback rate is ×100 times real time. The electron flux was 3.5 e -Å -2 s -1 .

File Name: Supplementary Movie 6
Description: EC-TEM movie describing the structural changes in the 170 nm Cu2O cubes during 1 h of chronoamperometry at -0.9 VRHE in 0.1 M KHCO3. The recording rate of the movie was 1 frame per second. 5 frames were averaged to create one frame of the movie. The movie playback rate is ×100 times real time. The electron flux was 3.5 e -Å -2 s -1 .

File Name: Supplementary Movie 7
Description: EC-TEM movie describing the structural changes in the 80 nm Cu2O cubes during 50 min of chronoamperometry at -0.9 VRHE in 0.1 M KHCO3 from an area that has mostly cubes. The recording rate of the movie was 1 frame per second. 5 frames were averaged to create one frame of the movie. The movie playback rate is ×100 times real time. The electron flux was 3.5 e -Å -2 s -1 .

File Name: Supplementary Movie 8
Description: EC-TEM movie describing the structural changes in the 80 nm Cu2O cubes during 45 min of chronoamperometry at -0.9 VRHE in 0.1 M KHCO3 from an area that has mostly partial cube fragments and re-deposited NPs. The recording rate of the movie was 1 frame per second. 5 frames were averaged to create one frame of the movie. The movie playback rate is ×100 times real time.

File Name: Supplementary Movie 9
EC-TEM movie describing the structural changes in the 390 nm Cu2O cubes during 25 min of chronoamperometry at -0.9 VRHE in 0.1 M KHCO3. The recording rate of the movie was 1 frame per second. 10 frames were averaged to create one frame of the movie. The movie playback rate is ×200 times real time. The electron flux was 1.7 e -Å -2 s -1 .

File Name: Supplementary Movie 10
EC-TEM movie describing the structural changes in the 170 nm Cu2O cubes during 25 min of chronoamperometry at -0.9 VRHE in 0.1 M KHCO3. The recording rate of the movie was 1 frame per second. 10 frames were averaged to create one frame of the movie. The movie playback rate is ×200 times real time. The electron flux was 1.7 e -Å -2 s -1 .

File Name: Supplementary Movie 11
EC-TEM movie describing the structural changes in the 80 nm Cu2O cubes during 25 min of chronoamperometry at -0.9 VRHE in 0.1 M KHCO3. The recording rate of the movie was 1 frame per second. 10 frames were averaged to create one frame of the movie. The movie playback rate is ×200 times real time. The electron flux was 1.7 e -Å -2 s -1 .

File Name: Supplementary Movie 12
Description: EC-TEM movie describing the structural changes in the 390 nm Cu2O cubes with higher loading than Movie 5 during 1 h of chronoamperometry at -0.9 VRHE in 0.1 M KHCO3. The recording rate of the movie was 1 frame per second. 10 frames were averaged to create one frame of the movie. The movie playback rate is ×200 times real time. The electron flux was 1.7 e -Å -2 s -1 .

File Name: Supplementary Movie 13
Description: EC-TEM movie describing the structural changes in the 170 nm Cu2O cubes with lower loading than Movie 6 during 1 h of chronoamperometry at -0.9 VRHE in 0.1 M KHCO3. The recording rate of the movie was 1 frame per second. 10 frames were averaged to create one frame of the movie. The movie playback rate is ×200 times real time. The electron flux was 1.7 e -Å -2 s -1 .

File Name: Supplementary Movie 14
Description: EC-TEM Movie describing the structural changes in the 30 nm Cu2O cubes synthesized by colloidal chemistry during 30 min of chronoamperometry at -0.9 VRHE in 0.1 M KHCO3. The recording rate of the movie was 1 frame per second. 5 frames were averaged to create one frame of the movie. The movie playback rate is ×100 times real time. The electron flux was 13 e -Å -2 s -1 . Due to the small size of these Cu cubes, a higher electron flux had to be used for in these experiments. Nevertheless, comparisons between electron irradiated and non-irradiated areas of the sample reveal the similar morphologies of the catalysts after reaction.