Limitations of identical location SEM as a method of degradation studies on surfactant capped nanoparticle electrocatalysts

https://doi.org/10.1016/j.jcat.2020.12.006 0021-9517/ 2020 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). ⇑ Corresponding authors at: University of Bern, Department of Chemistry and Biochemistry, Freiestrasse 3, 3012 Bern, Switzerland (Y. Hou); Eötvös Loránd University, Department of Physical Chemistry, Pázmány Péter sétány 1/A, 1117 Budapest, Hungary (S. Vesztergom). E-mail addresses: yuhui.hou@dcb.unibe.ch (Y. Hou), vesztergom@chem.elte.hu (S. Vesztergom), peter.broekmann@dcb.unibe.ch (P. Broekmann). 1 These authors contributed equally. Yuhui Hou a,1,⇑, Noémi Kovács , Heng Xu , Changzhe Sun , Rolf Erni , María de Jesús Gálvez-Vázquez , Alain Rieder , Huifang Hu , Ying Kong , Menglong Liu , Benjamin J. Wiley , Soma Vesztergoma,b,⇑, Peter Broekmann a


Introduction
Due to the ever-increasing consumption of fossil fuels, gigatons of CO 2 are released yearly to the atmosphere, expediting global warming [1]. A possible way of mitigating the effects of atmospheric CO 2 is to reduce it electrochemically. Electrochemical reduction does not only allow CO 2 to be regarded as a valuable raw material instead of an environmentally dangerous waste, but it may also provide a solution for the storage of excess renewable (hydro-, solar or wind) energy [2].
Mostly due to this, electrochemical CO 2 reduction -a process that was first described more than 150 years ago [3]-has recently become the forefront of electrochemical research [4]. Searching for the term ''electrochemical CO 2 reduction" on the website of ACS Publications yields 3334 research papers about this topic, only from the past year; Google Scholar, when searched for the same term and for the same period of time, gives > 17000 matches. A majority of these publications are original research papers that describe new catalyst materials, which -somewhat remarkablyall exhibit excellent qualities when applied for CO 2 reduction. This means that by covering electrodes with the newly invented catalysts, and carrying out electrolyses of solutions that contain CO 2 dissolved in some form, high current densities of CO 2 reduction can be achieved at relatively low overpotentials, and the process may in an ideal case yield only one or just a few desired products [4].
Compared to the tremendous amount of research invested in the design of new electrocatalyst materials for CO 2 electroreduction, technologies that operate on an industrial scale are still rare. Undoubtedly, the most important obstacle that hinders the application of newly developed catalysts on an industrial level is an issue of stability: catalysts that may show remarkable features in lab experiments tend to degrade and lose their performance over prolonged use. This may especially be true for catalysts owing their activity to a fine structure, such as colloidally synthesized nanoparticles that are especially prone to degradation over long-time operation. In case of these catalysts, studying (electro-)mechanical degradation and its effects on the catalytic performance has to be the first step of technological up-scaling.
Although many operando techniques (e.g., X-ray diffraction, scattering or absorption, as well as Raman spectroscopies [5,6]) can provide an insight to nanoparticle transformations occurring during CO 2 reduction, it is still more common to use ex situ electron microscopic (EM) techniques to observe, in particular, the structural changes that electrocatalysts suffer during CO 2 reduction.
In order to apply EM in an electrocatalysis study, the catalyst has to be sampled before and after it is made subject to electrochemical treatment. When comparing images taken before and after electrolysis, we usually work under two implicit assumptions: (i.) that the areas scanned before and after the electrolysis are either physically the same, or are both representative of the sample as a whole; and (ii.) that any changes we observe are indeed caused by the electrochemical treatment and not by other operations, e.g., the pre-electrolysis scanning of the sample, careless sample transportation, exposition to air or to chemicals, etc.
The former of the above two assumptions can readily be made explicit, for example, if identical location scanning or transmission electron microscopies (IL-SEM or IL-TEM) are employed. IL-TEM was first described by a work of Mayrhofer et al. in 2008 [7], and the first report on the application of IL-SEM by Hodnik et al. [8] followed not much later, in 2012. In early studies, the catalyst material was loaded on a TEM finder grid (made of gold) to facilitate identical location imaging [7]. Later it was found that it is enough to apply a small incision (a cross-like scratch) on other (e.g., graphite) holders to relocate the scanned site after electrolysis, which rendered the use of finder grids unnecessary. Due to the fact that IL-EM is able to visualize changes of a catalyst surface, often down to the details of individual nanoparticles, IL-EM found immediate application in catalyst degradation studies on a variety of target reactions [9,10].
In the field of CO 2 electrolysis, IL-EM became a prominent method of studying catalyst degradation [11][12][13][14][15][16][17][18][19][20][21], mainly because it is considered (and, starting from its discovery, often advertised as) a non-destructive method. It is usually assumed that if a given catalyst preserves good performance characteristics over longer periods of electrolysis, and neither IL-SEM nor IL-TEM reveal any structural degradation, the catalyst is stable and can be considered a potential candidate for up-scaled (e.g., flow cell) studies [15].
Unfortunately, however, the situation is not this simple, especially because, in some cases, the pre-electrolysis EM imaging does affect the future catalytic performance of the sampled catalyst areas. For example, in the literature of IL-TEM studies of electrocatalysts, there are reports on the electron beam induced shrinkage (as well as some ripening) of Pt nanoparticles used in fuel cells [22]. Based on these results, Arenz and Zana strongly recommend that in order to check if the electron beam changes the sample, TEM analysis following the electrochemical measurements should also be performed at pristine locations; i.e., locations which have not been previously exposed to the electron beam [23].
For IL-SEM, probably based on the assumption that the electron dose is much lower than in the case of TEM, no such warning was given, and it is indeed not likely that the beam used under SEM conditions could induce similar sintering effects observed in TEM. The sintering of nanoparticles may however not be the only way an electron beam can alter a catalyst surface: another, equally important phenomenon -namely, the under-beam formation of a passive layer-should also deserve attention.
That electron bombardment of a conducting sample in vacuo, where only slightest traces of organic vapours occur, can result in the coverage of the sample with a non-conducting layer of polymerized carbon compounds was first noticed by Lariviere Stewart [24] in 1934 -that is, four years before von Ardenne built the first SEM [25]. That electron bombardment, especially during focusing, can also cause changes to the surface of a sample inside an SEM was first noticed as early as 1946 by Marton et al. [26]. Recently, two reviews from Postek et al. [27,28] discussed some issues of interpreting SEM images: the second part [28] was entirely devoted to the issue of electron beam-induced specimen contamination.
Postek et al. [28] pointed out that the origin of beam-induced contaminations can both be the sample itself and the vacuum system of the SEM. While the cleanliness of the latter can be significantly improved (for example, by the replacement of diffusion pumps with turbomolecular ones backed by dry backing pumps in modern instruments), the history of the specimen prior to entering the vacuum system still remains important [28]. In case of samples with significant organic content, organic molecules remaining on the sample surface can break, undergo polymerization, and get ''pinned" to the sample by the beam during scanning [28]. Depending on the electron dose, the formed carbonaceous layer can grow at a rate of a few nanometers/seconds over the sample surface, even if only low accelerating voltages are used.
It is interesting to note that although under-beam contamination is a well-studied subject in the literature of SEM (see [27,28], as well as the references cited therein), studies on the effect of under-beam contamination/passivation on the future electrochemical behaviour of the sample are scarce, and are mostly focused on corrosion and not on electrocatalytic properties [29]. Yet, as we are going to demonstrate in this paper, under-beam passivation can practically disable the sampled part of a catalyst, especially if it contains organic remnants (capping agents) from the synthesis process. While other parts of the catalyst (not affected by the electron beam before electrolysis) remain active and very often degrade significantly during the catalysed process, the part of the sample affected by pre-electrolysis scanning remains intact, and probably entirely passive, due to the carbonaceous film formed on it under the beam.
Here we demonstrate, by IL-SEM studies on polyvinylpyrrolidone (PVP) functionalized Ag nanocubes used as electrocatalysts for CO 2 reduction, a catalytic activity disabling effect of a passive carbonaceous layer that is known to be formed under the electron beam during pre-electrolysis SEM scans [30]. The aim of this paper is to emphasize the necessity of extreme care being taken not to misinterpret IL-SEM studies that seemingly demonstrate excellent catalyst stability.

Experimental
Catalyst preparation. Ag nanocubes (Ag NCs) were prepared by an upscaled synthesis route described elsewhere [31]. As support, a glassy carbon plate (2 mm thickness, Alfa Aesar, type 1) was mirror-polished (0.5 lm alumina suspension, Buehler), was thoroughly rinsed with ultrapure water and ethanol, dried, and masked with an inert PTFE tape to leave an 0.8 cm Â 1 cm geometric surface area open for catalyst coating.
In order to form a carbon-supported Ag NC catalyst, 5.6 mg of the as-prepared Ag nanocubes [31] (in the form of powder) was dispersed in 6 cm 3 isopropanol (VLSI Selectipur, BASF) by a 1hour sonication. 1.5 mg of technical carbon powder (Vulcan XC 72R, Cabot, USA) was also dispersed in 3 cm 3 isopropanol by 1hour of sonication, and the two suspensions were subsequently mixed by sonicating for 30 min. The resulting suspension was dried overnight under vacuum conditions, yielding a C-supported Ag NC catalyst powder. This powder was re-dispersed in 1.5 cm 3 of isopropanol containing 75 l' of a Nafion solution (Aldrich, 5 wt% dissolved in a mixture of lower aliphatic alcohols and water). The obtained dispersion was subjected to sonication for 30 min, and for each electrode, 25 l' of the resulted ink was drop-cast onto the glassy carbon plate and dried in a vacuum oven.
An Ag NC catalyst without carbon support was prepared by dispersing 22 mg of the as-prepared Ag NCs in 6 cm 3 isopropanol by 1-hour sonication and spin-coating 75 l' of this suspension onto a glassy carbon support in three steps over 1 minute, using 1000 min -1 rotation rate on an Ossila spin coater.
Both the C-supported and the unsupported Ag NC catalysts were exposed to a UV-ozone atmosphere (PSD Series, Novascan, operated with air at atmospheric pressure) for 12 min.
For studies on a gas diffusion electrode (GDE, experimental details were described elsewhere [15]) the suspension of carbonsupported Ag NCs was drop-cast on the hydrophobic surface of a Sigracet 39 BC (Fuel Cell Store) GDE, and the nanocubes were percolated through the porous body of the GDE by a vacuum filtration system placed on the rear side of the electrode, followed by airdrying at ambient conditions lasting 30 min. No UV-ozone treatment was applied to the thus prepared, Ag NC-modified GDE. The GDE was used as part of the gas flow cell described in [15], combined with a Sustainion alkaline membrane (X37-50 RT, Dioxide materials) and an anode compartment containing 2 mol dm À3 KOH solution.
XPS Characterization. X-ray photoelectron spectroscopy (XPS) studies were carried out using a Thermo ESCALAB 250 XI instrument at a pass energy of 30 eV using monochromated Al K-a line (hm ¼ 1486:7 eV). Charge correction was based on the position of the C1s peak (284.8 eV). The XPS spectra were subjected to a Shirley background subtraction and were analysed using the CasaXPS software.
Electrocatalysis studies. For all electrochemical experiments, a potentiostat/galvanostat (Metrohm Autolab 302N, The Netherlands) was used to control the potential, current density, and transferred charge. The electrolysis experiments were carried out using a custom-built, air-tight, H-type glass cell. Apart from the working electrode that was prepared as described above, the threeelectrode arrangement consisted of a ''leakless" Ag j AgCl j 3 mol dm À3 KCl reference (Pine) and a Pt-foil (1.5 cm Â 0.5 cm, Goodfellow) counter electrode. For electrolyses, 0:5 mol dm À3 KHCO 3 (ACS grade, Sigma-Aldrich) electrolyte solutions were prepared with ultrapure water (Milli-Q by Merck Millipore) and were saturated with CO 2 (99.999%, Carbagas, Switzerland). During the experiments, continuous gas flow was maintained through the electrolyte solution. To avoid possible fluctuations in CO 2 solubility caused by a change in the ambient temperature, all electrochemical experiments were performed at 20°C, by immersing the H-type cell into a thermostated water bath. Automatic IR compensation was applied following the determination of the cell resistance by positive feedback. For the sake of comparability, all potentials given herein were converted to the reversible hydrogen electrode (RHE) scale. The reported current densities were normalized to the geometric surface area.
Gaseous products generated in the cell were detected by connecting the purging gas outlet to a GC analyzer (SRI Instruments Multigas Analyzer N o 3). The continuous flow of the carrier CO 2 gas through the electrolysis cell carried volatile reaction products from the head-space into the sampling loops of the gas chromatograph. The partial current I i , corresponding to the formation of a gaseous product i, can be calculated [32] as where x i denotes the mole fraction of the products, determined by GC using an independent calibration standard gas (Carbagas); n i is the number of electrons involved in the reduction reaction to form a particular product (n ¼ 2 for both CO and H 2 formation); F ¼ 96485:3 C mol À1 is Faraday's constant; and v m is the molar CO 2 gas flow rate measured by a universal flowmeter (7000 GC flowmeter, Ellutia) at the exit of the electrochemical cell.
The Faradaic efficiency ðFEÞ of a given reaction product can be determined by dividing the respective partial current, determined from Eq. (1), by the total current measured electrochemically. A thermal conductivity detector (TCD, for the detection of H 2 ) and a flame ionization detector (FID, for the detection of CO) were applied in our studies. We found that in the studied system H 2 and CO are the only two detectable products, accounting for 100% AE 5% of the current density that was electrochemically measurable. The electrochemically measured current densities were thus subdivided into partial current densities by taking into account the chromatographically determined concentration ratios, as will be shown later in Fig. 2. During operation, aliquots were analysed in intervals of 20 min during steady state electrolyses.
EM Measurements. EM analysis was conducted with a Zeiss Gemini 450 SEM with an InLens secondary electron (SE) and a backscatter electron detector (BSD). An accelerating voltage of 1.5 kV (probe current of 20 pA) and 5.0 kV (probe current of 120 pA) were applied for SE and BSD imaging, respectively. For highangle annular dark-field scanning transmission electron microscopy (HAADF-STEM) combined with energy-dispersive X-ray spectroscopy (EDX) and TEM imaging, an FEI Titan Themis (equipped with a SuperEDX detector) was used with an acceleration voltage of 300 kV.

Results and discussion
In colloidal nanoparticle synthesis, PVP is a widely applied shape-control agent that promotes the growth of specific crystal faces while hindering others [33,34]. In the synthesis of Ag NCs used in this study, PVP -by strongly binding to the (100) facets of Ag-, facilitated the formation of almost perfect nanocubes of side lengths of about 100 nm, as shown in Fig. 1a. The XPS spectrum (Fig. 1b) of a catalyst prepared without carbon support clearly exhibits a strong Ag3d signal, as well as a small peak that can be assigned to the N1s excitation of the PVP molecules adsorbed on the surface of the nanocubes. As shown in Fig. 1b, the applied UV-ozone treatment resulted in a significantly decreased N1s peak intensity. The peak has not disappeared, however, which hints that some PVP still remained on the surface despite the UV-ozone treatment.
Although the adsorbed PVP could, in principle, inhibit the catalytic activity of the nanocubes [35,36], the UV-ozone treated, Csupported Ag NCs showed good performance when applied for the electroreduction of CO 2 . This is demonstrated by Fig. 2a, showing the current density and the product distribution as a function of the applied electrode potential. The current densities shown in Fig. 2a were averaged for 1-hour electrolyses carried out in CO 2 saturated 0:5 mol dm À3 KHCO 3 solutions: for the electrolyses at different potentials, fresh solutions and newly prepared catalysts were applied.
It is known that on Ag, the primary product of CO 2 reduction is CO [37]. The same is true for the carbon-supported Ag NCs, with the addition that compared to plain silver -e.g., a silver foil [14]-the Ag nanocubes exhibit a broader overpotential range for CO production. That is, only a little amount of H 2 is formed at potentials less negative than À1:1 V vs. RHE, and CO 2 reduction generally prevails over hydrogen evolution in the entirety of the studied potential range À1:3 V < E < À0:7 V ð Þ . This observation is in agreement with other reports on nanoparticulate silver catalysts of CO 2 electroreduction [38].
In order to check the stability of the catalyst, we chose the moderate potential value of À1:0 V vs. RHE for a prolonged operation study. As shown in Fig. 2b, the catalyst preserved both its overall activity and its relative selectivity towards the production of CO (the Faradaic efficiency of CO formation was about 80%) for an electrolysis lasting 20 hours.
Nevertheless, since catalysts can maintain their macroscopic activity even as they undergo partial deactivation or decomposi-tion [39], we carried out IL-SEM investigations of the working electrode surface, which -although the overall activity remained unchanged-indeed revealed some degradation.
In Fig. 3 we compare two scanning electron micrographs of the same spot of a working electrode surface; one recorded before (Fig. 3a) and one after (Fig. 3b) a 20-hours electrolysis treatment at À1:0 V vs. RHE, similar to the one used to obtain the data of Fig. 2b. Fig. 3a shows highly isotropic Ag NCs of a side length of about 100 nanometers, distributed evenly on the supporting carbon matrix. As revealed by Fig. 3b, the nanocubes undergo some slight deformation and shrinkage during electrolysis, and, more prominently, some subnanometer sized particles appear on the surface. EDX mapping (Fig. 3c) confirmed that these small particles consist of silver, and are most probably formed as a debris of nanoparticle degradation due to the mechanical impact of gas evolution [16].
In order to get a clearer view of the degradation process of Ag NCs, the above SEM experiment was repeated with a working elec- Potential dependence of the current density and the product distribution, as determined by means of online gas chromatography in an H-type cell for 1-hour electrolyses. Each electrolysis (data points) were carried out using a freshly prepared catalyst and a fresh solution. Curves were created by interpolation. (b) Time dependence of the catalytic performance, as determined by a single electrolysis experiment lasting 20 hours, with subsequent chromatographic head-space analysis (data points). The curve was created by interpolation. trode prepared without the supporting carbon matrix (see the Experimental section for details).
The as-prepared electrode surface is shown in Fig. 4a, exhibiting cubic shaped Ag nanoparticles distributed on the glassy carbon electrode substrate. Somewhat surprisingly, the SEM image of the same spot, recorded after a 20-hours electrolysis, shows practically no degradation and the appearance of just a little amount of the subnanometer sized particles, as shown in Fig. 4b. What is even more surprising is that if we record an SEM micrograph with the same configuration, just of a different spot of the sample -that was not scanned before electrolysis-, the picture gets quite different. Fig. 4c clearly shows slightly deformed Ag nanocubes, along with a significant amount of Ag debris formed during electrolysis.
The micrographs of Fig. 4 very clearly reveal an important pitfall of IL-SEM analysis; namely, that due to electron beaminduced changes of the catalyst surface during the preelectrolysis scan, the sample may get at least partially deactivated for the catalysed process. Due to its decreased electrocatalytic activity, the pre-scanned area of the sample may show no or little changes during the electrolytic process, while other spots (that were not affected by pre-electrolysis SEM scanning) preserve their activity and, in turn, exhibit significant degradation. In other words, the often advertised nondestructiveness of IL-SEM [8,9] should not be taken as granted -at least, not for all catalyst types.
That the effect shown in Fig. 4 can indeed be explained by preelectrolysis electron beam-sample interactions is further demon-strated by Fig. 5, showing an SEM micrograph of a working electrode surface obtained after electrolysis. Only a part (a rectangular segment) of this sample was scanned by SEM before electrolysis took place, and despite that the sample was exposed Fig. 3. IL-SEM investigation of the degradation of carbon-supported Ag nanocubes, used as catalysts of CO2 electroreduction. The same spot of the working electrode surface is shown just before (a) and right after (b) the electrode was used for a 20-hours electrolysis of a CO2-saturated 0:5 mol dm À3 KHCO 3 solution at an electrode potential of À1:0 V vs. RHE. The formation of subnanometer sized Ag particles during electrolysis is revealed by the HAADF-STEM (gray-scale) and EDX scans (red-scale) in (c), recorded postelectrolysis at a pristine location that has not been subjected to an electron beam before.  to the electron beam only for a short time, a marked difference can be observed between the degradation features of the pre-scanned segment and the rest of the surface area. Most notably, the coverage of the pre-scanned area with the subnanometer sized Ag particles is less pronounced, compared to other sites. This hints that the electron beam exerts an effect not only on the Ag nanoparticles but also on the underlying glassy carbon substrate.
Note that provided we refrain from long-time exposure of the sample to the electron beam, the above-described electron beam irradiation effect is hardly noticeable per se. Yet, as shown by Fig. 5, even the irradiation damages that remained undetected dur-  ing pre-electrolysis EM scanning can prove significant when the sample is used for electrolysis and scanned afterwards.
To demonstrate the irradiation effect in itself, we carried out prolonged SEM scans on one of our catalyst samples. As revealed by Fig. 6, the effect of contamination (as visualized by the growth and even the apparent merging of the nanocubes) is more pronounced when larger magnifications are applied (i.e., when the beam is more focused) or when the sample is scanned for longer times.
At first glance, the growing and subsequently merging nanocubes shown in Fig. 6 may resemble the coalescence of Pt nanoparticles observed by Chorkendorff et al. under in situ TEM conditions [22]. Note, however, that under TEM conditions, the accelerating voltage and the electron dose are both much higher than in SEM. Accordingly, the main feature that Chorkendorff et al. described in their study was a shrinkage (and not a growth) of most nanoparticles, with only a few of these displaying actual coalescence [22]. Shrinkage in this study was shown to be an effect of both the high electron dose and the oxidizing atmosphere. None of these are characteristic of our SEM measurements; thus in our case, it seems more straightforward to presume that the beam has little effect on the nanocubes themselves, and it is rather the under-beam formation of a carbonaceous passive layer what is seen in Fig. 6.
Although the SEM images recorded at an accelerating voltage of 1.5 kV may not allow a clear distinction between the core of the nanoparticles and the contamination layer formed around them (Figs. 6 and 7a), the contamination layer can be visualized by EM scans at higher (20 kV) accelerating voltage (Fig. 7b). That under the formed carbonaceous contamination layer the Ag nanocubes preserve their original shape can be confirmed by the HAADF-STEM and the TEM bright field images shown in Figs. 7c and d, respectively.
It is of worth noting that the contamination layer is most probably formed by the PVP capping agent, remnants of which remain adsorbed on the Ag nanocubes despite the applied UV-ozone treatment, and then get polymerized and pinned to the electrode surface by the electron beam [28]. Based on the electrocatalytic degradation pattern shown by Fig. 5, we can assume that some PVP may also remain on the substrate, forming there a carbona-ceous shell that is however presumed to be not as thick as on the surfaces of the nanocubes, where PVP is primarily adsorbed.
The under-beam formation of the passive layer on the surface of nanoparticulate catalysts seems to block the pre-scanned surface even if entirely different settings, and much harsher electrolysis conditions, compared to what was described before, are applied. This is demonstrated by Fig. 8, where we modified a gas diffusion electrode (GDE) with carbon-supported Ag NCs (this time, without the application of UV-ozone treatment), and performed electrolysis by applying a potential of À1:4 V vs. RHE, thus passing through a total charge amount of 1600 C cm -2 . While the identical location SEM images of Fig. 8a and b show no trace of degradation, particle deformation and the appearance of newly formed, small particles is clearly shown by the SEM micrograph of Fig. 8c, recorded at a random spot after the electrolysis. Although as pointed out in [40], in fact any organic contaminations of a catalyst sample may act as source of material for the formation of passive carbonaceous crust layers, the prominent role of PVP in this process is further supported by our numerous IL-SEM studies on PVP-free catalysts, where no such contamination effects were ever seen [11][12][13][14][15][16][17][18][19][20].

Concluding remarks
No effort has so far been made to demonstrate the effect of capping-agent related under-beam passive layer formation on the catalytic behaviour of nanoparticle type electrocatalysts. This is considered worrying, particularly because of the emerging popularity of IL-SEM-based stability studies where the pre-electrolysis scanning can contaminate (and consequently disable) the catalyst sample in a way that the post-electrolysis scan would deceivingly show no degradation.
Using PVP-functionalised Ag nanocubes as model catalysts of CO 2 reduction, we demonstrated how under-beam contamination (a carbonaceous, passive crust formed over the catalyst particles) might account for artefacts in IL-SEM studies in such a way that the experimenter is provided with false comfort with regard to the stability of the catalyst. This paper was written with the aim to direct attention to this possible pitfall of IL-SEM studies, which may especially emerge when IL-SEM is applied on electrocatalysts prepared by a synthesis route involving capping agents.
Apart from the issues that PVP remnants can cause in the interpretation of IL-SEM experiments, it should also be emphasized that shape-forming surfactants may exert further unwanted effects also on the essential catalytic properties. E.g., in case of the system studied here we have to note that if no action (in our case, UVozone treatment) is taken to remove (at least most of) the adhering PVP remnants, this will negatively affect both the selectivity and the stability of the catalyst. In our case omission of the UV-ozone treatment resulted, for example, in the overall Faradaic efficiency (toward CO production) dropping from $ 80% to $ 65%, and a further dropping to below 50% over 2 hours of electrolysis (under conditions similar to those applying for Fig. 2b). The removal of capping agents may be based on plasma/thermal annealing [41] (note that the UV-ozone treatment we applied here proved to be far from ideal), or it may even rely on mere electrochemical methods. Namely, it was recently shown in two independent studies (by our group [42] and by Pankhurst et al. [43]) that capping agent remnants may effectively be removed by the harsh cathodic potentials applied during CO 2 electrolysis. Needless to say, the latter ''operando activation" method [42] does not work for capping agents baked to the catalyst surface by the electron beam in an IL-SEM scenario.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.