Host−Guest Chemistry Meets Electrocatalysis: Cucurbit[6]uril on a Au Surface as a Hybrid System in CO2 Reduction

The rational control of forming and stabilizing reaction intermediates to guide specific reaction pathways remains to be a major challenge in electrocatalysis. In this work, we report a surface active-site engineering approach for modulating electrocatalytic CO2 reduction using the macrocycle cucurbit[6]uril (CB[6]). A pristine gold surface functionalized with CB[6] nanocavities was studied as a hybrid organic−inorganic model system that utilizes host−guest chemistry to influence the heterogeneous electrocatalytic reaction. The combination of surface-enhanced infrared absorption (SEIRA) spectroscopy and electrocatalytic experiments in conjunction with theoretical calculations supports capture and reduction of CO2 inside the hydrophobic cavity of CB[6] on the gold surface in aqueous KHCO3 at negative potentials. SEIRA spectroscopic experiments show that the decoration of gold with the supramolecular host CB[6] leads to an increased local CO2 concentration close to the metal interface. Electrocatalytic CO2 reduction on a CB[6]-coated gold electrode indicates differences in the specific interactions between CO2 reduction intermediates within and outside the CB[6] molecular cavity, illustrated by a decrease in current density from CO generation, but almost invariant H2 production compared to unfunctionalized gold. The presented methodology and mechanistic insight can guide future design of molecularly engineered catalytic environments through interfacial host−guest chemistry.


■ INTRODUCTION
The electrocatalytic reduction of CO 2 using renewable energy sources offers an attractive route to produce storable carbonneutral fuels. 1,2 For this technology to become commercially viable, catalysts that promote the challenging multi-electron/proton transfer reaction with minimal energy losses and high rate and selectivity need to be developed. Currently, CO 2 reduction suffers from significant overpotentials, sluggish kinetics, and low selectivity control as well as the low solubility of CO 2 in aqueous electrolytes. 3−5 Gold (Au) is generally considered the most active electrocatalyst for CO generation with a catalytic onset potential ca. 300 mV more negative than the standard potential for CO 2 to CO reduction on flat electrodes. 6−8 The catalytic mechanism on Au is believed to involve a surface-bound COOH (*COOH) intermediate, which is subsequently further reduced by 1 e − /1 H + transfer to yield CO (*CO) that is finally liberated from the surface. Theoretical predictions indicate that stronger binding of the *COOH intermediate relative to *CO may be needed for enhanced CO evolution activity. 9,10 However, the binding energies of *COOH and *CO on electrocatalysts generally correlate; 11 therefore, stabilization of *COOH would simultaneously stabilize *CO and thus disfavor its required desorption. This scaling relationship impedes a straightforward approach to independently fine-tune surface properties for higher activities, hampering the development of low overpotential single-site metal catalysts. 12 To this end, different approaches such as alloying, 13,14 nanostructuring, 15 and molecular surface modification 16−20 have recently gained attention. Nevertheless, molecular engineering of an electrocatalytic surface site remains to be challenging. Nature, on the other hand, has developed very effective catalytic systems that surpass the limits of scaling relationships. Enzymes such as carbon monoxide dehydrogenase and formate dehydrogenase are able to reversibly interconvert CO 2 at the thermodynamic potential with quantitative product selectivity at their precious-metal free active sites. 21−23 While details about their reaction mechanisms are still under debate, a hydrophobic active site cavity and the stabilization of intermediates by a highly specific second coordination sphere are believed to be crucial for their high activity as well as selectivity. 9,24,25 In our efforts to create analogous synthetic reaction environments, we envision supramolecular host−guest chemistry as a powerful tool to modulate electrocatalytic reactions by selectively favoring certain species due to highly specific binding affinities. Cucurbit[n]urils (CB[n] where n = 5−8) are a class of promising host molecules for this purpose. These barrel-shaped organic macrocycles are water-soluble and have found broad applications in materials chemistry, drug delivery systems, sensing, and catalysis ( Figure 1a). 26,27 CB[n]s are able to shift the reduction potential of electrochemical processes by selectively stabilizing certain species. 28,29 For example, halfwave potentials of ferrocene and methyl viologen can be positively shifted upon complexation with CB [7] due to different stabilization of the oxidized versus reduced species within the cavity.
The cavities of CB[n] are highly hydrophobic 30 and bind a plethora of small organic guests (for n ≥ 6) as well as gas molecules (for n ≤ 6). 26 In aqueous solutions, a large number of gases (e.g., CH 4 , acetylene, and cyclopentane) form inclusion complexes within CB [6]. 31 Their negatively polarized carbonyl-lined portals exhibit high affinity toward metal surfaces allowing for facile supramolecular functionalization of electrodes. 32 In contrast, other hosts such as pillararenes, cyclodextrins, and calixarenes typically rely on additional steps for electrode immobilization, which often afford increased distances between the host cavity and the surface. 33−35 Their ability to directly affect surface-bound species is thereby diminished.
It has been shown spectroscopically and by gravimetric measurements that CB [6] can be used as a highly porous solidstate adsorbent for gaseous CO 2 with high selectivity over CO, rendering this molecule to be attractive for combining host− guest chemistry and electrocatalytic CO 2 reduction studies. 36,37 To the best of our knowledge, neither has CO 2 complexation within CB [6] in solution been demonstrated nor has CB [6] been used in electrocatalytic CO 2 reduction.
In this work, we study CB [6] as a prototypical synthetic organic cavity system for facile manipulation of the heterogeneous CO 2 reduction on Au, a well-studied model electrocatalyst ( Figure 1b). First, the host−guest interaction of CO 2 with CB [6] is studied in homogeneous aqueous solution both experimentally and computationally. Subsequently, we employ surface-enhanced infrared absorption (SEIRA) spectroscopy combined with electrochemistry to investigate the behavior of CO 2 within CB [6] on the Au surface during electrocatalysis.

■ RESULTS AND DISCUSSION
Interaction of CO 2 with CB [6] in Solution. The inclusion of CO 2 in CB [6] was monitored in aqueous KHCO 3 (0.1 M) using Fourier-transform infrared (FTIR) spectroscopy in transmission mode ( Figure 2a). The asymmetric stretching mode of 12 CO 2 was found at 2343 cm −1 , which is in line with previously reported values for 12 CO 2 in solution. 38 In the presence of 8.44 mM CB [6], an additional shoulder was detected at 2333 cm −1 . This band was exclusively observed in a 12 CO 2 -purged electrolyte solution ( Figure S1) and therefore assigned to an interaction between CO 2 and CB [6] (CB [6]·CO 2 ). The corresponding 13 CO 2 asymmetric stretching vibration 38 identified at 2278 cm −1 displayed an identical shift by 10 cm −1 to 2268 cm −1 , confirming this band to originate from CO 2 .
The observed shift matches the previously reported values for the asymmetric stretching mode of CO 2 confined in solid CB [6] at 2336 cm −1 . 36,37 Addition of spermine, a well-known CB [6] guest with a high binding affinity, 26 led to disappearance of the band at 2333 cm −1 , which is in line with a competitive displacement of CO 2 from the cavity. Moreover, the IR signature of the CB [6] framework remained unchanged in the presence of N 2 and 12 CO 2 / 13 CO 2 , indicating that the CB[6]·CO 2 (binding) interaction is weak and nondirectional ( Figure 2b). In contrast, spermine addition gave rise to a strong additional band at 1727 cm −1 . This band can be rationalized by a shift of the carbonyl stretching mode from 1743 cm −1 by 16 cm −1 as a result of electrostatic interactions with the positively charged amine functional group of spermine. 39 Analogously, the unaltered CB [6] carbonyl band in the presence of CO 2 , when compared to the CB [6]·spermine host−guest complex, suggests that CO 2 is not merely associated with the carbonyl portal region of CB [6]. In conclusion, the IR data strongly suggests that CO 2 is bound inside the CB [6] cavity in solution with an IR marker band located at 2333 cm −1 . We note that the binding of CO to CB [6] could not be probed due to its low solubility in the reaction solution.
The assignment of the CO 2 interaction with CB [6] is further corroborated by density functional theory (DFT) and molecular dynamics (MD) simulations (Table S1 and Figures  S2−S5). Both methods show a stronger binding of CO 2 over CO in CB [6], which is in line with previously reported DFT results. 40 A free-energy profile based on umbrella sampling of MD simulations of CO 2 and CO is shown in Figure 3a. The results suggest that CO 2 and CO need to overcome a low energy barrier when entering the cavity and reach an energy

ACS Catalysis
Research Article minimum at the cavity center. CO 2 was found to bind with two possible conformations in the pocket (second local minima found at z = ±3.1 Å). A snapshot of the final state after 22 ns of a MD simulation in KHCO 3 (0.1 M) is shown in Figure 3b. The CO molecule was initially placed in the cavity center and was found to be eventually replaced by two CO 2 molecules at the given concentration. The free-energy profiles were used to calculate the binding constants K a,CO2 = 105 M −1 and K a,CO = 5 M −1 for CB [6]·CO 2 and CB [6]·CO, respectively (binding free-energies of 6.4 and 4.1 kcal mol −1 , respectively, see the Supporting Information for computational details).
To quantify the binding of CO 2 and CB [6] experimentally, fluorescent dye displacement measurements were performed ( Figure 4a and Figure S6). This technique utilizes the fluorescence of the host−guest complex of CB [6] with an indicator dye, in this case trans-4[4-(dimethylamino)styryl]-1methylpyridinium (DSMI, inset of Figure 4a). DSMI is weakly fluorescent on its own but strongly fluorescent in the complexed state. 41 The competitive binding of CO 2 and DSMI to the CB [6] cavity (Figure 4b) was used to determine a binding constant (K a,CO2 ) of 250 ± 100 M −1 in 0.1 M KHCO 3 for the CB[6]·CO 2 host−guest complex. The experimental finding was qualitatively confirmed with a second indicator dye ( Figure S6). This sizable affinity is in good agreement with the computational data. Unfortunately, CO afforded only a small change in fluorescence owing to its low solubility, which prevented an accurate quantification of its binding constant. However, well-established trends of gas affinities and corresponding volume/packing coefficients 31 along with the performed DFT and MD calculations suggest a weaker binding of CO than that of CO 2 to CB [6]. These findings indicate that there is no product inhibition (i.e., CO occupying the CB[6] pocket due to a higher binding affinity than that of CO 2 ), which is a general concern in (supramolecular) catalysis. 42 Au Electrode Modification with CB [6]. The binding of CB [6] to Au surfaces has previously been reported 32,43 but rarely quantitatively. The effect of the solvent and the nature and concentration of the electrolyte were suspected to have a large effect on the surface adsorption equilibrium. Solubility of CB [6] in water is typically very low (<100 μM), 44 but it can be readily increased by more than 2 orders of magnitude through addition of cationic species to the solution. 45 Besides the increased solubility, the interaction of cations with CB [6] influences the kinetics of host−guest complex formation. 46 As an additional complexity, cations themselves have shown to effect CO 2 reduction on heterogeneous electrocatalysts. 47−50 Considering these multiple intricate effects of cations in the system, we employed an aqueous KHCO 3 (0.1 M) electrolyte solution, which is also commonly used in CO 2 reduction catalysis. 51,52 This concentration fulfills the necessity for supporting electrolytes during electrocatalytic CO 2 reduction and overcomes the limited solubility of CB [6] while Figure 2. FTIR spectra of CB [6] solutions recorded in the transmission mode. (a) Absorbance spectra of asymmetric stretching modes of 12 CO 2 in aqueous KH 12 CO 3 (0.1 M) and 13 CO 2 in NaH 13 CO 3 (0.1 M) solution. The reference spectra for dashed spectra were the N 2 purged electrolytes. Spectra of CB [6] from which the CB-free CO 2 reference spectrum was subtracted display a shoulder at 2333 and 2268 cm −1 . The addition of spermine (strong CB [6] cavity binder) leads to disappearance of the 2333 cm −1 band. The broad negative band at ca. 2360 cm −1 results from small atmospheric CO 2 variations between measurements. (b) IR spectra of the region below 1800 cm −1 representing the vibrational modes for CB [6] in 12 CO 2purged 0.1 M KH 12 CO 3 and 13 CO 2 -purged 0.1 M NaH 13 CO 3 , which exhibit no band shifts in comparison to the N 2 -purged 0.1 M KH 12 CO 3 solution. Addition of spermine leads to an additional band at 1727 cm −1 assigned to carbonyl functional groups of CB [6], presumably occurring due to the interaction of the spermine nitrogen and CB [6] carbonyl ring. The spectrum of the CB[6] 13 CO 2 -purged sample was multiplied by a factor of 4 for better comparison (see experimental details in the Supporting Information). Snapshot of MD simulation with one "deep cavity"-bound CO 2 , one CO 2 loosely associated with carbonyl ring, and one unbound CO.

ACS Catalysis
Research Article minimizing the effects of cations on the host−guest complexation as well as on CO 2 reduction to a minimum.
The adsorption of CB [6] on Au was first quantified by quartz-crystal microbalance (QCM) measurements at opencircuit potential in N 2 -purged KHCO 3 as shown in Figure 5a. The coverage was estimated using the Sauerbrey model, 53 assuming a flat gold surface and a van der Waals radius of "rigid" CB[6] of 14.4 Å. 45 A theoretical monolayer coverage of ca. 80 ± 20% for 1.69 mM CB [6] was determined with a Langmuir isotherm. Considering the highest density of hexagonally packed circles within a plane (ca. 90%), the data obtained from QCM is close to that for an optimally covered surface.
QCM was not only applied to study surface adsorption processes by mass changes but also further utilized to obtain insights into the viscoelasticity of the adsorbed layer by studying its dissipation behavior. 54 This can be particularly useful to gather information on changes of layer packing or, importantly, multilayer formation. 55,56 Measurements indicated that, for concentrations of >0.8 mM, the dissipation starts to increase, suggesting the formation of multilayers of CB [6] ( Figure S7). Multilayer buildup of CB [6] on surfaces has been observed previously for CB [6] in solutions containing Na + . 57 Furthermore, formation of chain-like coordination polymers between CB [6] and K + has been reported in the solid state. 58,59 The interaction of CB [6] with the Au surface was further probed by SEIRA spectroscopy, which selectively monitors the Au−electrolyte interface (see below). 60,61 The vibrational band pattern of CB [6] revealed significantly different relative intensities between SEIRA and transmission measurements due to surface selection rules, indicating a preferential binding geometry of CB [6] on the surface compared to the isotropic case in solution ( Figure S8). A notable shift of the carbonyl stretching mode from 1743 cm −1 in solution to 1737 cm −1 on the Au surface further suggests an interaction between (one of) the carbonyl-lined portals of CB [6] with the gold surface, which is in line with previous reports. 32,43,62 Figure 5b shows the normalized intensity of the intense 1483 cm −1 marker band

ACS Catalysis
Research Article of CB [6] as a function of the CB [6] concentration in solution.
No notable increase was observed for concentrations higher than ca. 1.3 mM ( Figure S9), matching the concentration at which multilayer formation was observed in QCM measurements ( Figure S7). As the SEIRA intensity I decreases significantly with distance d from the electrode surface (I ≈ d −6 ), potential multilayers of CB [6] cannot be monitored. 63 SEIRA as well as X-ray photoelectron spectroscopy (XPS) indicated a stable surface-bound CB [6] layer after overnight soaking in a CB[6]-free 0.1 M KHCO 3 aqueous electrolyte solution ( Figures S10 and S11).
SEIRA Measurements of CB[6]·CO 2 under Polarization. SEIRA spectroscopy has recently gained increased attention in the field of CO 2 reduction catalysis due to its high sensitivity for surface processes. 64−67 Typically, the strong IR activity of CO and HCO 3 − /CO 3 2− is exploited to track specific bands to monitor product stability and binding, the bicarbonate/CO 2 equilibrium, and the role of cations. In this study, we specifically focus on the role of CO 2 and its fate upon inclusion into CB [6] during electrocatalytic turnover (see Figure 6).
Upon purging the electrolyte solution containing 1.69 mM CB [6] with CO 2 , the band at 2333 cm −1 assigned to CB[6]·CO 2 appeared, indicating that CO 2 can efficiently penetrate the CB [6] cavity when the host molecule is bound to the Au surface. Compared to the isotropic case in the solution, the CB[6]·CO 2 band showed an increased intensity in the SEIRA spectra (Figure 6a) relative to the unaltered CO 2(aq) band at 2343 cm −1 , suggesting an increase in the total CO 2 concentration (i.e., free CO 2(aq) and CB [6]·CO 2 ) at the Au− electrolyte interface. We speculate that the binding of CB [6] to the Au surface together with its ability to host CO 2 acts as a driving force for CO 2 accumulation at the electrode surface.
SEIRA spectroscopic experiments in a three-electrode configuration were performed to investigate the behavior of the supramolecular complex on the surface at reductive potentials. To the best of our knowledge, the nature of the interaction as well as the stability of CB [6] on Au under electrochemical polarization have not yet been assessed in the literature. SEIRA spectra recorded at various potentials revealed that CB [6] starts to reversibly desorb from the electrode at −0.7 V versus the reversible hydrogen electrode (RHE; Figure 6). This is further confirmed by ex situ XPS measurements ( Figure S12). Note that the anchoring of the carbonyl-lined portal of CB [6] is significantly different to chemisorption through commonly employed thiol groups on Au that suffer from low electrochemical stability under reductive conditions. 68 Though several groups have recently reported that multidentate thiol-anchoring provides improved stability of cocatalysts for CO 2 reduction, reductive desorption effects were not discussed in detail. 18,19 As shown in Figure 6b,c, the intensity of the CB[6]·CO 2 band started to decline at potentials below ca. −0.4 V versus RHE, matching the catalytic onset for CO 2 reduction (see below). This observation indicates the reduction of CO 2 inside the Au-immobilized CB [6] cavity. Mere desorption of CB [6] hosting CO 2 at very reducing conditions, which would also cause a decay of the 2333 cm −1 band, can be excluded because the observed SEIRA signals of CB [6] remain constant at potentials that already afford a significant decrease of the CB[6]·CO 2 band (Figure 6c). Therefore, the results suggest that the effects of CB [6] host−guest chemistry can be exploited for CO 2 reduction at least at the desirable low overpotential region between −0.4 and −0.7 V versus RHE. Furthermore, the data imply that a part of the remaining surface-bound CB [6] has an empty cavity at potentials more negative than −0.8 V versus RHE.
Plotting the 2343 cm −1 band observed on unmodified Au surfaces and in the presence of CB [6] yielded matching trends suggesting that CO 2 reduction of "free" CO 2 proceeds in both cases in a similar environment ( Figure S13). A potential shift of the 2333 cm −1 intensity curve by approximately 40 mV at potential values affording turnover was noted, which may be related to an altered CO 2 reduction reaction within the cavity. CB[6]-bound CO was not detected as expected from the low binding affinity and solubility of CO. Moreover, CB [6] marker bands in the CO 2 -purged electrolyte were found to match those in Ar-purged solutions ( Figure S14) also at very negative potentials. This indicates that the structure and conformation of CB [6] is maintained throughout the large potential window and under CO 2 binding, suggesting that there is no change in CO 2 binding affinity particularly at negative potentials.
Electrocatalysis. Stepped constant-potential electrolysis experiments were performed to monitor the influence of CB [6] on the activity and selectivity of CO 2 reduction on Au. The cell design with online gas sampling was based on recent

ACS Catalysis
Research Article reports 69,70 and is described in detail in the Supporting Information (Figures S15−S17). Additional characterization including diffusion layer thickness and electrochemical surface area determination following a recent report for more standardized data acquisition 71 is also reported in the Supporting Information (Figures S18−S20). The system showed good stability for 12 h ( Figure S21). Figure 7 compares partial current densities at various potentials of Au electrodes in solutions of CB [6] of varying concentrations to cover the CB [6] monolayer and multilayer regimes. The results of unmodified (blank) Au agree with recently published findings. 7 Self-assembly of CB [6] prior to catalysis without CB [6] in the electrolyte solution during electrocatalysis showed marginal differences compared to unfunctionalized Au electrodes ( Figure S22), which is presumably due to a less favorable adsorption equilibrium of CB [6] at negative applied potentials over the timespan of the electrocatalysis experiments.
In contrast, the presence of CB [6] in the solution was found to selectively alter the CO partial current density (j co ), while the hydrogen evolution current (j H2 ) remained almost unchanged (Figure 7). In the potential window from −0.4 to −0.8 V versus RHE, j co curves were observed to shift to more negative potentials with increasing CB [6] concentrations, corresponding to a suppression of j co with CB [6] in the solution (Figure 7a; further details of electrocatalysis data in Figures S23−S27). At 0.39 mM CB [6], an approximately 100 mV more negative potential was required to achieve a comparable j co than in the absence of CB [6]. No signals of CB [6] degradation were observed by nuclear magnetic resonance (NMR) spectroscopy ( Figure S28), and 13 CO 2 electrocatalysis experiments yielded exclusively 13 CO ( Figure  S29). Upon decreasing the potential to be more negative than −0.8 V versus RHE, a convergence of the j CO in the absence and presence of CB [6] with different concentrations was observed (Figure 7a). This observation can be rationalized by the potential-induced desorption of CB [6] from the surface starting to occur at potentials more negative than −0.7 V versus RHE (Figure 6c).
Mechanistic Interpretation. The experimental data demonstrate that the electrocatalytic reduction of CO 2 on Au has been effectively altered by CB [6] encapsulation. This effect can be explained on a molecular basis through the macrocycle's impact particularly on the rate-determining step (RDS) of the catalytic process. In this respect, an initial electron transfer to CO 2 to form a Au−COO − species and a proton-coupled electron transfer to form directly a Au− COOH species are often discussed as possible candidates. 4,64,72−74 Both cases would yield a Tafel slope of 118 mV dec −1 based on the initial electron transfer as the ratelimiting step, which matches our Tafel slopes of approximately 120 mV dec −1 for unmodified and CB [6]-coated Au surfaces ( Figure S30). Thus, the unaffected Tafel slopes in the presence of CB [6] as well as the fact that only CO is generated allow us to draw the conclusion that the macrocycle-mediated catalytic CO 2 reduction likely proceeds via the same elementary steps as on bare Au. The altered catalytic performance however suggests that the new reaction environment induced by CB [6] has significantly altered the kinetics of relevant reaction steps. This can be rationalized by an altered energy landscape for the reaction through specific interactions on CO 2 and *COO − within CB [6] compared to the blank Au interface. This is provided by the following qualitative arguments: (1) The hydrophobic nature of the cavity lowers the availability of water molecules to stabilize surfacebound intermediates through hydrogen bond formation. Previous molecular dynamics simulations suggest approximately 3−4 water molecules in the cavity of solubilized CB [6] without a guest in the aqueous environment. 75 In line with this report, our MD simulations show that an average of less than one (0.62) water molecule is present within the space defined by the CB [6] carbonyl oxygen rims when CO 2 or CO is adsorbed in the CB [6] cavity. (2) Cations such as K + have previously been shown to have a large effect on the CO 2 reduction reaction by stabilizing intermediates exhibiting a permanent dipole moment. 50 However, MD simulations show that the inclusion of K + into CB [6] is energetically unfavorable ( Figure S3), and DFT-optimized structures of CB [6] with K + on a Au surface suggest that the distance between K + and the encapsulated surface-bound intermediate COOH is too large for a stabilizing interaction ( Figure 8).

ACS Catalysis
Research Article solution ( Figure 3 and Table S1). This indicates that CO 2 is stabilized within CB [6] and thus exhibits a reduced capability for subsequent reactions without additional stabilization of the reaction intermediate.
The altered environment within CB [6] will eventually affect the thermodynamics and kinetics of formation of encapsulated surface-bound species. To gauge such effects, DFT calculations were performed particularly for the Au−COOH species formation (details in the Supporting Information, Figures  S31 and S32). Calculations for Au−COO − were not possible due the high level of complexity regarding the uncompensated charge. For CO 2 , the results show a horizontally aligned molecule in the vicinity of the bare Au surface (Figure S32), whereas a more perpendicular orientation was found within CB [6], suggesting that inclusion affords a different stabilization of the substrate near the electrode surface.
Next, the reaction free-energy difference (ΔΔ R G) of the reaction from Au−CO 2 to Au−COOH within CB [6] and at the blank Au surface was calculated. An increase of ∼8 kcal mol −1 (0.36 eV) was determined for the reaction proceeding in CB [6] compared to blank Au (Table S2). Thus, the calculation provides evidence for a notably increased energy required for the reaction upon encapsulation, which is proposed to result from the absence of solvent/cation-mediated stabilization of the product, that is, Au−COOH as discussed qualitatively above as well as from additional stabilization of the substrate CO 2 within CB [6]. This effect is expected to be even more pronounced when considering Au−COO − , which should be energetically higher due to the absent charge compensation by the proton as well as the lack of other electrostatic interactions in the cavity as discussed above.
As a result, a higher-energy RDS is expected, leading to an overall decrease in j CO (at the same applied potential) and therefore providing a molecular basis for explaining the increased overpotential required to reduce CO 2 inside the CB [6] cavity. In agreement, at a 0.39 mM CB [6] concentration corresponding to a quasi-monolayer of CB [6] on Au, an indicative shift in j co of the CB[6]-modified electrode by ∼100 mV as aforementioned was observed. In this context, increasing the CB [6] concentration to 1.59 mM has little effect as the surface coverage is not significantly increased as demonstrated by QCM and SEIRA measurements ( Figure   5). However, at 6.53 mM CB [6] where multilayer formation occurs, further, stronger suppression of the CO current density is apparent. Thus, for high CB [6] concentrations, j CO is likely governed by additional factors, such as a hindered CO 2 uptake into CB [6] and CO release due to the adsorbed multilayers corresponding to an effective blocking of the electrode surface.
Host−Guest Design Criteria for Electrocatalysis. The rational manipulation of the heterogeneous interfacial catalytic process by creating well-defined reaction environments opens up promising avenues to selectively alter the catalytic transformation to desired outcomes in the future. Based on our findings, we propose three design criteria to rationally achieve an improved molecular cocatalyst system for CO 2 reduction to CO: (1) The induced reaction environment should be modulated to selectively stabilize the surface-bound intermediate COO − /COOH to help break scaling relationships. 9 Recent publications discuss the ability of amines 76 and amides 18,19,77 in a ligand construct as well as imidazolium-based ionic liquids 78,79 to stabilize CO 2 reduction intermediates via hydrogen bonding and/or electrostatic interactions on bare electrodes. In this respect, we envision that secondary guests within the CB cavity might be used to accommodate specific functional groups to selectively tune the active site environment. The larger CB [8] is able to incorporate two molecules within its cavity, which offers a large degree of tunability of the electrostatic and chemical environment within the cavity. 26 (2) The inclusion of the substrate CO 2 should facilitate consecutive reactions, for example, by enabling specific interactions of CO 2 with the cavity analogous to enzyme active sites. We observed a 10 cm −1 shift toward lower wavenumbers for CO 2 inside the CB [6] cavity. This may suggest a decreased bond order of CO 2 . Nevertheless, this effect might not be dominant in the presented system due to the lack of stabilization of a key intermediate (COOH/COO − ). The probing of interactions of CO 2 with a secondary guest in the CB cavity via vibrational spectroscopy might enable systematic studies of host·substrate binding prior to catalytic turnover. (3) The kinetics of CO 2 uptake and CO release needs to be sufficiently fast to accomplish highly catalytic activities. A simple model calculation (see the Supporting Information) suggests that ingress and egress kinetics of CO 2 and CO, respectively, should be higher than 20 s −1 to obtain >5 mA cm −2 for j CO . A high binding constant for CO 2 and low binding constant for CO should be maintained. The binding constant is defined as the ratio of ingress and egress kinetics, 46 and the egress kinetics for CO 2 and ingress kinetics for CO should therefore be slow. While enzymes have perfected this independent tuning of complexation kinetics and thermodynamics, it remains difficult to realize this in synthetic systems. The use of cations as "lids" for the CB portals has been suggested to influence host−guest complex formation kinetics. 45 This might enable a pathway through a variation of cations in the electrolyte solution to specifically alter the kinetics of substrate inclusion or product release. Moreover, addition of competitively binding non-redox-active guests could

ACS Catalysis
Research Article allow for further rational tuning of CO 2 ingress/egress kinetics by subtly changing the host−guest equilibria as well as electronics and sterics.
The manifold of new pathways to pursue shows that the CB[n]−Au hybrid system provides us with proposed measures and a modular platform to selectively tune different reaction steps of the heterogeneous CO 2 reduction at Au surfaces and to mechanistically explore electrocatalysis at the inorganic− organic interfaces. This allows for detailed investigations of the reaction mechanism by rational experimental design toward a complete understanding of the process. Such knowledge is a major prerequisite for the development of next generation electrocatalysts.

■ CONCLUSIONS
We present the concept of surface-adsorbed supramolecular host−guest chemistry in heterogeneous electrocatalysis. The combined spectroscopic, analytical, and electrocatalytic results suggest that CO 2 is captured within the cavity of the supramolecular host-molecule CB [6] and electrocatalytically reduced inside the cavity when adsorbed on the gold surface. An increased concentration of CO 2 is observed at the electrode−electrolyte interface due to the binding of CB[6]·CO 2 .
The confinement of CO 2 reduction into CB [6] nanocavities enables molecular tunablity of the local chemical environment on the electrocatalyst surface. The presented data indicates that the CB [6] complexation strongly and selectively affects the electrocatalytic reduction of CO 2 to CO, while the hydrogen evolution activity remains essentially unchanged. The resultant hybrid organic−inorganic composite can thus be regarded as a model system for synthetic cavity-based heterogeneous electrocatalysts. Notably, a plethora of different research fields spanning from metal−organic frameworks 80,81 and covalent−organic frameworks 82 to molecular catalysis 83−86 and synthetic biology 87 are also heavily focused on tuning the local environment and coordination spheres around a catalytic active site. The development of such model systems and analytical methodology as presented here may help us to gain insight in this complex and important field of research. In the context of CO 2 reduction electrocatalysis, this may offer new routes to overcome scaling relationships and rationally tune the selectivity and activity of electrodes also beyond Au. Thus, this concept bridges the gap between enzymatic, molecular, and heterogeneous catalysis and may open up new avenues for utilizing fine-tuned supramolecular units as synthetic catalytic pockets.

* S Supporting Information
The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acscatal.9b04221.