Keeping sight of copper in single-atom catalysts for electrochemical carbon dioxide reduction

Carbon dioxide can be electrochemically converted into valuable multi-carbon products using Cu-based single-atom catalysts. However, transient cluster formation, which is undetectable using ex-situ techniques, may be responsible for C2+ products. Here we discuss these observations to highlight the need for operando characterisation when defining active sites.

techniques to identify specific performance-enhancing structural features. For Cu-N-C materials, the catalyst structure can be determined ex situ before and after electrolysis to observe any irreversible modifications. For example, X-ray photoelectron spectroscopy (XPS) can give the ratio and predominant nature of the nitrogen environments, high-angle annular darkfield scanning transmission electron microscopy (HAADF-STEM) can detect the presence of isolated sites, and X-ray absorption spectroscopy (XAS) can identify oxidation states with X-ray absorption near edge structure (XANES) and coordination environment with extended X-ray absorption fine structure (EXAFS) analysis. These techniques among others can provide a structural basis for computational modelling to determine possible reaction mechanisms and derive structurefunction relationships. However, dynamic alterations that are only observable under working conditions have raised concerns over the true nature of the active site.

Transient structural alterations
Despite ex situ analysis showing that Cu-N-C single sites are retained after electrolysis, operando spectroscopic techniques have identified that the cathodic applied potentials required for CO 2 R can cause reduction of Cu 2+ sites to Cu 0 , which decreases the affinity for N-coordination and gives rise to Cu agglomeration (Fig. 1a). Such an effect was observed in a Cu-N-C material composed of predominantly Cu-N 4 sites, which exhibited high ethanol selectivity (faradaic efficiency for ethanol (FE EtOH ) = 55% with gas recirculation) 9 . Ex situ characterisation using XPS, HAADF-STEM, and XAS showed no signs of transformation, with the same structural motifs observed pre-and post-electrolysis. However, operando XAS at potentials below −0.6 V vs. RHE showed Cu 2+ reduction to Cu 0 and Cu-Cu coordination signals in XANES and EXAFS spectra respectively, signifying the formation of subnanometer Cu clusters. Upon exposure to air, or through the application of oxidative potentials, the Cu-N 4 signals of the starting material were fully recovered while the Cu-Cu signal disappeared, showing that nanocluster formation within the Cu-N-C material is reversible. In other words, re-oxidation of the Cu clusters causes re-dispersion of the Cu atoms into N-chelated single-sites. The reversibility of the process precludes characterisation with ex situ techniques, therefore operando techniques are required to observe these dynamic changes.
Several subsequent systems employing operando techniques have displayed similar behaviour, which is not restricted to solely Cu-N-C materials. A recent study on oxygen-ligated isolated Cu 2+ sites also gave high ethanol yields (FE EtOH = 91%) 10 . Again, where ex situ characterisation showed no structural changes, operando XAS data were able to link the high selectivity to the reversible emergence of Cu n clusters. The same observations of cluster formation have also extended to isolated Cu sites confined in a covalent triazine framework for C 2+ product generation 11 . The absence of Cu ions in solution after application of an anodic potential further confirms that the Cu clusters can dissociate to re-anchor Cu atoms as single-sites within the framework. These examples suggest that C 2+ products could be an indicator of agglomeration. The mechanistic requirements for forming multi-carbon species provide some insights into why this may be the case.
Extensive studies on bulk and nanoparticulate catalysts have shown that C-C coupling involves the dimerisation of surfacebound *CO intermediates 12 . The large interatomic distances typically observed for Cu-N-C catalysts would not allow for this process to take place at single-sites, implying that either an entirely different mechanism is taking place or that reversible restructuration occurs under working conditions (Fig. 1b). Therefore when C 2+ products are observed, cluster formation is highly probable. Recent examples where multi-carbon products were generated from Cu-SACs have suggested different mechanistic explanations. Studies have proposed that CO-CO coupling can occur at single sites 13,14 , or that Cu sites can be sufficiently close in distance to facilitate dimerisation 15 . However, more in-depth characterisation is required to definitively rule out agglomeration and validate these assumptions.
Although agglomeration likely accounts for multi-carbon products, C 1 products can also be formed when clusters are present. In one study, a Cu-polyphthalocyanine catalyst underwent aggregation to form nanoparticles under operating conditions, but the catalyst only generated CO 16 . Adjacent Cu sites on Pd 10 Te 3 nanowires also generated CO and not C 2+ products 17 . Pre-formed clusters have also produced CH 4 or HCOOH as the sole carbon product from CO 2 R, showing that clusters do not always form C 2+ products 18,19 . In essence, although C 2+ products are a likely symptom of agglomeration, they are not a sole indicator and operando techniques are necessary to confirm the structure of the active site. Consequently, researchers should be careful when assessing the true nature of the active site to avoid the pursuit of incorrect rational modification routes.

Understanding reversible cluster formation
SACs incorporating metals other than Cu can also undergo agglomeration, however, the degree of M 0 formation is typically much lower than that of Cu-SACs and often not reversible 20,21 . Therefore, the possible causes behind cluster formation and the reversibility of this process must be understood through assessment of specific Cu-based examples. The dynamic behaviour arises from the combined effects of the metal coordination environment and metal-support interactions. COMMENT CO 2 R requires electrochemical reduction of Cu sites in Cu-N-C catalysts, which decreases their affinity for the support, facilitating atom migration and cluster formation. Consequently, avoiding agglomeration may simultaneously prevent catalysis. Interestingly, the clusters formed by restructuration are highly selective for C 2+ products and are therefore beneficial to selectivity. Lowcoordination Cu-based species take advantage of this unexpected form of structure-function relationship, however synthetic methods for cluster formation are typically complex 18,19 . In situ electrochemical synthesis offers a straightforward alternative route to cluster formation, in which stabilisation methods through support interactions can be used to control the cluster size to target specific products 22,23 . The role of the support in reversible cluster formation is well-illustrated by comparison with molecular catalysts. Solutionbased molecular complexes are susceptible to potential-dependent decomposition, where reduction of the metal centre can increase ligand lability and promote irreversible heterogeneous material deposition, as demonstrated by deposition of Co from cobaloximes and Cu from Cu-cyclam 24,25 . However, with copper (II) phthalocyanine (CuPc) supported on carbon-nanotubes (CNTs) through π-π stacking, cluster formation was reversible as with the Cu-N-C examples 26 . In the same study, a supported porphyrin also showed reversible agglomeration whereas a Cu-cyclam, which does not anchor well to the CNTs, showed irreversible decomposition. Additionally, the reversibility of the process is not only driven by the high affinity of N sites for Cu(II), but is also dependent on particle size where smaller clusters are more easily reverted to atomic sites 27 .
The added understanding of structural attributes from operando analysis has opened new routes to selective CO 2 R. Up to now, operando XAS has proven the most applicable technique to study restructuration, but the absence of M-M signals does not always signify the absence of clusters due to problems associated with disorder and interference 28 . Therefore, alternative verification methods would be desirable. However, the range of available spectroscopic techniques that can observe such changes inoperando is still limited. In situ infrared and Raman spectroscopy are capable of indirectly distinguishing between clusters and single-sites due to differences in signals arising from surfacebound molecules 29 . However, analysis is complicated by dynamic desorption and adsorption of species under electrochemical conditions. The development of emerging in situ spectroscopic techniques such as ambient pressure XPS (AP-XPS) and liquid cell TEM (LC-TEM) could extract more information about the reversible process and identify active structures to provide computational methods with a correct basis to explore reaction mechanisms 30 . Innovative solutions coupling multiple operando techniques will likely prove vital in revealing the true nature of active sites.

Outlook
We have highlighted that reversible cluster formation occurs in Cu-SACs under CO 2 R conditions. The in situ formation of Cu clusters can give rise to higher C 2+ selectivity than conventional Cu catalysts, opening straightforward synthetic routes to valorise CO 2 R. However, the future of SACs for CO 2 R relies on validation of cluster formation to define structure-activity relationships that can guide rational catalyst development. We, therefore, reiterate that operando techniques should be commonplace when using Cu-SACs for CO 2 R due to the prevalence of reversible restructuration.