Size Dependent CO2 Reduction Activity of Ag Nanoparticle Electrocatalysts


 Coinage metals (Au, Cu and Ag) are state-of-the-art electrocatalysts for the CO2 reduction reaction (CO2RR). Size-dependent CO2RR activity of Au and Cu has been studied, and increased H2 evolution reaction (HER) activity is expected for small catalyst particles with high population of undercoordinated corner sites. A similar consensus is still lacking for Ag catalysts because the ligands and stabilizers typically used to control particle synthesis can block specific active sites and mask inherent structure-property trends. This knowledge gap is problematic because increased performance and catalyst utilization are still needed to improve economic viability. We combined density functional theory, microkinetic modeling, and experiment to demonstrate a strong size-dependence for pristine Ag particles in the sub-10 nm range. Small diameter particles with a high population of Ag edge sites were predicted to favor HER, whereas CO2RR selectivity increased towards that of bulk Ag for larger diameter particles as the population of Ag(100) surface sites grew. Experimental results validated these predictions and we identified an optimal particle diameter of 8-10 nm that balanced selectivity and activity. Particles below this diameter suffered from poor selectivity, while larger particles demonstrated bulk-like activity and reduced catalyst utilization. These results demonstrate the size-dependent CO2RR activity of pristine Ag catalysts and will help guide future development efforts.


Abstract.
Coinage metals (Au, Cu and Ag) are state-of-the-art electrocatalysts for the CO2 reduction reaction (CO2RR). Size-dependent CO2RR activity of Au and Cu has been studied, and increased H2 evolution reaction (HER) activity is expected for small catalyst particles with high population of undercoordinated corner sites. A similar consensus is still lacking for Ag catalysts because the ligands and stabilizers typically used to control particle synthesis can block specific active sites and mask inherent structure-property trends. This knowledge gap is problematic because increased performance and catalyst utilization are still needed to improve economic viability. We combined density functional theory, microkinetic modeling, and experiment to demonstrate a strong sizedependence for pristine Ag particles in the sub-10 nm range. Small diameter particles with a high population of Ag edge sites were predicted to favor HER, whereas CO2RR selectivity increased towards that of bulk Ag for larger diameter particles as the population of Ag(100) surface sites grew. Experimental results validated these predictions and we identified an optimal particle diameter of 8-10 nm that balanced selectivity and activity. Particles below this diameter suffered from poor selectivity, while larger particles demonstrated bulk-like activity and reduced catalyst utilization. These results demonstrate the size-dependent CO2RR activity of pristine Ag catalysts and will help guide future development efforts.

Introduction.
The electrochemical CO2 reduction reaction (CO2RR) is a leading candidate for producing sustainable chemicals and fuels from waste carbon emissions. 1,2 Carbon monoxide (CO) is an appealing product because it is a versatile chemical building block, it only requires two electrons per molecule, and it can be formed with high-current density and Faradaic efficiency (FE). 3 The H2 evolution reaction (HER) can also occur at CO2RR-relevant electrochemical potentials, and competition between the two reactions can reduce product selectivity and energy efficiency.
Optimizing nanoparticle size is one approach to maximize catalyst usage and reduce total metal loading. Size-dependent CO2RR vs. HER is well understood at Au and Cu nanoparticles thanks to agreement between experimental and theoretical investigations, [15][16][17][18] and it is generally agreed that a higher density of undercoordinated corner sites on smaller diameter nanoparticles can increase H2 production and reduce selectivity toward CO. However, there is still a lack of fundamental consensus on how size-dependent CO selectivity and activity evolve for pristine Ag nanoparticles because the ligands, capping agents, and/or stabilizers often used to control nanoparticle size, shape, and crystallographic orientation can block, passivate, or modify specific surface sites. [19][20][21][22][23][24] This poses a challenge for optimizing Ag electrocatalysts because understanding the inherent CO2RR vs. HER selectivity is needed to develop higher activity catalysts and reduce metal usage.
Ag nanocatalysts generally produce higher geometric current densities than bulk Ag particles and polycrystalline Ag foils, 11,[21][22][23][24][25][26][27][28][29] but the use of various synthetic techniques has made it difficult to identify consistent structure-property relationships. For example, the increased performance of oxide-derived [26][27][28] and AgCl-derived 29 Ag nanocatalysts has been attributed to increased surface area, enhanced intermediate stabilization, unique surface defects, preferential crystallographic orientations, and/or the presence of residual chloride ions. Size-dependent CO2RR performance of cysteamine-capped Ag nanoparticles was attributed to sulfur atoms at the catalyst-ligand interface modifying intermediate bonding. 22,23 The performance of citrate-stabilized Ag nanoplates 21 and nanoprisms 24 was attributed to preferential expression of specific crystallographic faces and formation of stacked "super structures," but the identification of different active sites in nanoplates vs. nanoprisms and the role of citrate capping agents leave unresolved questions. Finally, previous calculations have also predicted that Ag particles as small as 2 nm may still retain high CO2RR selectivity. 30 The above noted examples indicate that size, surface structure, and crystallographic orientation can all impact the CO2RR activity of Ag nanocatalysts, but a clear understanding of the inherent, size-dependent activity of pristine Ag catalysts is still lacking in the literature.
Here, we used a combination of density functional theory (DFT), kinetic rate theory calculations, and microkinetic modeling (MKM) to predict the size-dependent CO2RR performance of pristine Ag nanocatalysts. DFT and MKM calculations determined that CO2RR and HER activity were dominated by electrochemically-accessible Ag(100) sites and Ag edge sites, respectively. These results were then combined with Ag nanoparticle Wulff constructions 21,22,31,32 to predict size-dependent CO2RR and HER reactivity trends, showing increased CO2RR selectivity with nanoparticle diameter as the population of Ag(100) sites increased and the population of Ag edge sites decreased. Interestingly, it was the Ag edge sites that impacted HER, whereas undercoordinated corner sites are believed to most strongly impact HER from Au and Cu nanoparticles. [15][16][17][18] Finally, we directly validated our computational results by preparing and evaluating the CO2RR performance of a series of pristine Ag nanocatalysts with well-controlled diameters.
Experimental data revealed increased CO2RR selectivity and activity with increasing particle diameter, and we identified Ag particles in the 8-10 nm range as the optimum balance of selectivity and catalyst usage on a total atom basis. Our results showed that pristine Ag catalysts demonstrate similar size-dependent CO2RR selectivity trends as other coinage metals, albeit with unique CO2RR vs. HER active sites, and fill a current gap in the literature that is needed to maximize the performance of Ag-based CO2RR catalysts. This information provides new insight into a state-ofthe-art CO2RR catalyst and will help guide future materials design and optimization efforts.

Results and Discussion
We used Wulff constructions to analyze the size-dependent surface structure of Ag nanoparticles with diameters between 0.8-9.8 nm (Figure 1a). 21,22,31,32 The resulting nanoparticle structures were based on the DFT-calculated surface energies of the two low-index Ag(111) and Ag(100) facets, and the relative fraction of (100), (111), edge, and corner sites were extracted for each particle diameter. Figure 1b presents the total atomic population vs. nanoparticle diameter, showing the number of electrochemically-inaccessible bulk (interior) atoms rapidly increased with nanoparticle diameter. Figure 1c summarizes the relative population of electrochemicallyaccessible surface sites, excluding the interior bulk atoms. This presentation clearly shows the fraction of accessible corner and edge sites quickly decreased with particle diameter, while the fraction of accessible Ag(111) and Ag(100) sites increased with particle diameter.
DFT, rate theory calculations, and MKM were used to predict site-specific CO2RR and HER rates at an experimentally-relevant potential of -1.0 V vs. the standard hydrogen electrode (SHE).
Similar methodology was previously used to predict CO2RR kinetic barriers at Cu surfaces, 33,34 and specific computational details are provided in Supplementary Information. We considered CO and H2 formation through the following elementary steps, where * represents a vacant active site or bound intermediate: We employed a method to determine the potential-dependent reaction barrier (∆G # ) of an electrochemical reduction via Bronsted-Evans-Polanyi extrapolation of the transition state barrier of an analogous non-electrochemical hydrogenation step. 35 The elementary step with the largest reaction barrier (∆G max # ) was considered to be the rate limiting step that established the forward 36 The resulting kf values were then incorporated into our MKM to predict product formation rates (r = molecules/atomAg/s) based on local concentration of reactants at the catalyst surface. This approach provides additional kinetic insight compared with exclusively using the computational hydrogen electrode (CHE) methodology to identify thermodynamically-challenging, potential-limiting reaction steps based Gibbs free energy changes. 37,38 In particular, rate determining and potential limiting steps may differ once transition state barriers are considered.
Relative CO2RR and HER rates at an applied electrochemical potential of -1V vs. SHE were calculated for each site (Figure S1), and Ag(100) was predicted to be the most CO2RR active site. Table S1 show Ag(100) demonstrated the smallest ∆G max (CO2RR) # and largest kf (CO2RR) among the sites considered, and we identified the proton-coupled electron transfer to CO2 that results in O-H bond formation to form *COOH (step i) as the rate limiting CO2RR step. The corresponding transition state involved a water assisted H-shuttling model as shown in Figure 2b.

Figure 2a and
The adsorbed *CO2 was significantly bent relative to its linear gaseous counterpart, with a ~130 o O-C-O bond angle and a Bader charge of -0.89e that resembled an adsorbed *CO2 δ− species. 36 This adsorbed *CO2 δspecies was hydrogen bonded to a nearby H3O δ+ -like species (a Bader charge of +0.77e summed over H3O). Thus, a structure resembling a CO2 δ− -H3O δ+ pair is seen in the transition state with net Bader charge of -0.12e that indicated partial reduction.
*COOH formation was also the rate limiting elementary step for CO2RR at Ag edge sites and the transition state involved a similar water assisted H-shuttling model ( Figure S2). Elementary step ii (*CO formation) was the rate limiting process for both Ag(111) and Ag corner sites ( Figure S2). In this transition state configuration, the OH fragment has moved away from the *COOH moiety, with a hydrogen bond to a H of *H2O. Furthermore, the surface H has moved towards the H2O suggesting the shuttling of H is in progress. The larger ∆G max (CO2RR) # values associated with these transition states produced smaller kf (CO2RR) values and resulted in lower predicted CO2RR rates compared with Ag(100) or Ag edge sites.
On the other hand, Ag edge sites were predicted to be the most HER active (Figure S1) with the lowest ∆G max (HER) # and largest kf (HER) (Figure 2a and at all sites when only considering Gibbs free energy changes ( Figure S4). 30 The discrepancy between thermodynamically-determined potential limiting steps and kinetically-determined rate limiting steps highlights the importance of including transition state barriers to predict the relative CO2RR and HER activity of catalyst sites.
We then predicted size-dependent CO2RR and HER rates using the site-specific reaction rates and the population of each site as a function of nanoparticle diameter (Figure 2d). Deconvoluting the size-dependent rate contributions revealed that CO2RR and HER activity was dominated by Ag(100) and Ag edge sites, respectively (Figure 2d and S5). Electrochemically-accessible Ag(100) sites were responsible for 88-99% of the total CO2RR rate for particle diameters above 2 nm. Rates initially increased with particle diameter as the population of Ag(100) sites became larger, but values decreased slightly above a diameter of 4 nm as the number of electrochemicallyinaccessible bulk (interior) atoms became larger. This points to decreased catalyst utilization based on total metal content, but surface-normalized CO2RR rates steadily increased with particle diameter when only considering electrochemically-accessible surface atoms ( Figure S6). On the other hand, the Ag edge sites shown in Figure 2c contributed ~95-99% of the total HER activity at all particle sizes, and the relative HER rate decreased with particle diameter as the population of accessible Ag edge sites became smaller. These results predict that improved CO2RR performance should be observed with increasing Ag nanoparticle diameter, but decreased catalyst utilization may occur above a critical catalyst size.
To validate these computational predictions, we created a series of pristine Ag nanoparticles with well-controlled diameters between 5-10 nm and evaluated their size-dependent electrocatalytic activity (details in Supplementary Information). Ag nanoparticles were prepared via e-beam evaporation onto a highly oriented pyrolytic graphite (HOPG) support in an ultra-high vacuum (UHV) chamber, providing a model system mimicking conventional carbon-supported Ag electrocatalysts. SEM images in Figure 3a show the average diameter grew from ~5 nm at a loading of 0.7 nmolAg/cm 2 to ~10 nm at a loading of 4.4 nmolAg/cm 2 . Particle diameters correlated well with Ag loading ( Figure S7) and XPS confirmed all particles formed as metallic Ag ( Figure   S8). Our experimental approach eliminated the need for capping-agents, surfactants, or other surface-bound ligands typically used for size-or shape-controlled nanoparticle synthesis. [19][20][21][22] This allowed accurate comparison between theoretical and experimental results since organic ligands or other capping agents may block specific surface sites and/or modify the intrinsic catalytic properties. 22 Samples were transferred from UHV and their electrocatalytic properties were evaluated in CO2 saturated 0.1M KHCO3 using an H-cell with a custom-built electrode holder ( Figure S9).
Contributions from catalyst-free regions of the HOPG surface were subtracted from the raw data ( Figure S10), and CO and H2 FEs were calculated from the detected products and HOPGsubtracted electrolysis charge (details in Supplementary Information). Potential-dependent CO2RR experiments identified a maximum FECO at -1.2V versus the reversible hydrogen electrode (RHE) for both Ag nanoparticles and a bulk, polycrystalline Ag foil (Figure S10c), which is consistent with previous reports using single-crystal and polycrystalline Ag electrodes. 21,22,26,39 This cathodic potential was selected to evaluate the size-dependent electrocatalytic activity because it achieved the highest FECO and produced sufficient product for accurate CO and H2 quantification at all loadings, and Figure 3b presents the FECO and FEH2 at -1.2V vs. RHE versus Ag particle diameter. The total FE for CO and H2 production at -1.2V was close to 100% for samples with Ag loadings ≥ 1 nmolAg/cm 2 (Table S2), which validated our protocol for subtracting contributions from the HOPG substrate. The FECO increased with Ag diameter from ~20% FECO for ~5 nm diameter particles (<1.0 nmolAg/cm 2 ) to >90% FECO for particle diameters above 8 nm (>2.2 nmolAg/cm 2 ). Accordingly, FEH2 decreased from ~60% to < 5% FEH2 with increased particle diameter. In comparison, a bulk Ag foil produced ~90% FECO at -1.2V (Figure 3b), which indicates Ag particles above ~8 nm diameter demonstrated "bulk-like" product selectivity. A similar trend size-dependent FECO trend was also observed at -1.0V and -1.1V vs. RHE ( Figure   S11), but the small current densities produced in this low loading regime made H2 quantification difficult below -1.2V vs. RHE. Finally, the convergence with bulk TOF values above ~8 nm diameter does not necessarily conflict with previous reports that observed increased geometric current density from nanostructured Ag catalysts. 11,[21][22][23][24][25][26][27][28][29] Catalyst loadings in these previous studies were typically several orders of magnitude higher than this study, and reported enhancements in current density from Ag nanocatalysts largely stemmed from a higher fraction of electrochemically-active surface area compared with bulk catalysts. In our low-loading regime (<4.5 nmol/cm 2 ) the Ag nanoparticles formed at low coverages and contained a smaller number Ag atoms than the bulk Ag foil (27.4 nmol/cm 2 based on XPS). This produced low geometric current densities (0.6-1.2 mA/cm 2 ), but it allowed us to accurately measure intrinsic activity differences and identify sizedependent trends. Our results show size-related effects are important in the sub-10 nm range and we identified an optimum particle size of 8-10 nm that balanced CO2RR selectivity and catalyst utilization. Particles below this diameter suffered from reduced CO2RR selectivity, whereas particle diameters above ~10 nm will experience low catalyst utilization on a total atom basis.
In conclusion, we have coupled DFT, rate theory calculations, microkinetic modeling, and experimental evaluations to identify the size-dependent CO2RR activity of pristine Ag nanoparticles prepared without ligands, capping agents, or other stabilizers. Ag demonstrated a decrease in CO2RR activity at smaller sizes that is similar to other Au and Cu coinage metal catalysts. However, the specific sites responsible for CO2RR and HER were unique. We found that CO2RR and HER activity was dominated by Ag(100) and Ag edge sites respectively, whereas corner sites did not contribute significantly to HER. Small Ag particles favored HER due to a higher population of Ag edge sites. CO2RR selectivity increased with particle diameter as the relative population of Ag(100) sites grew, and the population of Ag edge sites decreased. Our experimental results also allowed us to identify 8-10 nm as the optimum Ag particle diameter to maximize both CO2RR selectivity and catalyst usage. This information provides new insight into the inherent CO2RR activity of Ag nanocatalysts and should be useful for future catalyst design.

Methods
Computational methods. Ag nanoparticle models were represented via the Wulff construction to quantify the density of active surface sites for a given cluster diameter. 31 software. 41 Periodic Ag(111) and Ag(100) slab models were used to represent the active sites on the low-index plane of the metal catalyst. Edge sites were represented using the internal step edge of a Ag(211) surface, as has been done previously. 30  Scanning electron microscopy (SEM) characterizations were conducted on a FEI Quanta 600F microscope operated at 20 kV equipped with an energy-dispersive X-ray (EDX) detector. SEM images were analyzed with the ImageJ software (version 1.52a) to estimate the average Ag particle diameters, and the minimum particle size in the ImageJ analysis was set to 1 nm 2 to avoid measuring pixelated noise.
Electrochemistry measurements. Electrochemical CO2 reduction experiments were carried out in a gas-tight H-Cell. A custom-built, polytetrafluoroethylene (PTFE) electrode holder sealed the base of the catholyte chamber and allowed measurement of the HOPG-supported samples ( Figure   S9). The catholyte and anolyte chambers were filled with 0.