Introduction

The ability of elemental copper to catalyze the electrochemical conversion of CO2 into economically valuable products such as hydrocarbons and oxygenates is unmatched by other single-metal catalysts that generate only carbon monoxide, in either the pure (CO) or hydrous (HCOOH) form [1,2,3,4]. The electrochemical reduction of CO2 into C1–C2 products is notably endergonic [5,6,7], and the associated current densities for the Cu-catalyzed transformations register in the lower end of the mA/cm2 scale [3]. The search for an electrocatalyst simultaneously more active and better product-selective than Cu poses a steep challenge: A set of “heuristic rules,”Footnote 1 extracted solely from experimental results, for the production of hydrocarbons and oxygenates [8, 9], must be transcended especially when the target is the preferential formation of “energy-rich” molecules more chemically reduced than CO.

Reactivity-structure correlations traditionally underpin the rational design and optimization of catalysts. Single-crystal electrodes represent model surfaces whose well-defined structure and composition can be tracked, at the atomic level, before and after the catalytic reaction. For instance, it has been shown that CO2 reduction (CO2R) at Cu(111) electrode surfaces gives, at certain potentials, methane as the favored product on Cu(111); at Cu(100) electrodes, ethylene is the dominant product [10, 11]. The yields are invariably low because product exclusivity remains elusive. Operando electrochemical scanning tunneling microscopy (ECSTM) has revealed that the surfaces of Cu(111) and Cu(100) correspond to the nominal crystallographic orientation prior to and at the onset of CO2R [12]. The case is, however, drastically different for Cu(110) and polycrystalline Cu, Cu(pc); the physical attributes of the latter are more proximate to those of “real-world” catalysts. In the case of Cu(pc), under CO2R potentials in 0.1 M KOH (pH 13) solution, a still-undetermined number of topmost layers undergoes surface reconstruction to form a well-ordered (100) facet [13]. The dynamic nature of the Cu(pc) surface in KOH under operando conditions has provided the impetus of the present investigation, viz., to ascertain whether the same surface reorganization also transpires in KHCO3, a widely employed supporting electrolyte in the CO2R corpus. It is of fundamental importance to determine the generality or specificity of the phenomenon of the surface reconstruction of Cu under CO2R potentials.

Experimental

The experimental procedures have been described in detail elsewhere [14] and will only be summarized here. Surface visualization studies by electrochemical scanning tunneling microscopy (EC-STM) were conducted using a Nanoscope E microscope (Digital Instruments, Veeco, Santa Barbara, CA) that had been refurbished (Advanced Surface Microscopy, Inc., Indianapolis, IN) and equipped with a built-in three-electrode potentiostat. A Pt counter electrode and a miniature leakless Ag/AgCl (3.4 M KCl) reference electrode (Innovative Instruments, Inc., Tampa, FL) were fitted inside an STM electrochemical cell custom-crafted from Kel-F (Emco Industrial Plastics, Inc., Cedar Grove, NJ). A 99.99%-pure polycrystalline Cu disk, 1.0 mm in thickness and 10 mm in diameter (GoodFellow, Coraopolis, PA), served as the working electrode. Potentials are referenced to the standard hydrogen electrode (ESHE), for direct correlation to thermodynamic free-energy changes. The interconversion between the ESHE and ERHE scales is given by the equation: ESHE = ERHE − 0.059 pH.

The STM tips were prepared by an electrochemical etch of a 0.25 mm diameter tungsten wire (Sigma-Aldrich, St. Louis, MO) in 0.6 M KOH at 15 VAC. All images were acquired after 1-h polarization at − 0.9 V; a high-resolution scanner was employed in constant-current mode without post-scan operations.

A two-step cleaning sequence of the Cu working electrode, prior to use, was employed: a mirror-finish metallographic polish with a suspension of polycrystalline diamond (Buehler, Lake Bluff, IL) at a grain size of 0.05 μm, followed by a 10-s electropolishing in 85% H3PO4 (Sigma-Aldrich) at 2.0 V with a Pt counter electrode. The electrode was then ultrasonicated in, and later rinsed with, deaerated Nanopure water. The polished sample was not thermally annealed.

Analytical-grade KHCO3 and KOH (Sigma-Aldrich) were used to prepare 0.1 M of the electrolyte solutions. The 0.1 M KHCO3 solution was deaerated for at least 1 h in ultrahigh purity N2 (Airgas, Radnor, PA); 0.1 M KOH was treated identically. Voltammetric scans were carried out at a scan rate of 50 mV/s; the reported voltammograms represented results from the 10th potential cycle.

Results and Discussion

Previous work on operando electrochemical scanning tunneling microscopy (ECSTM) depicted a dramatic reconstruction of the polycrystalline Cu surface, in alkaline solution, into the (100) facet after a constant-potential hold at − 0.9 V [13]. The same potentiostatic protocol is revisited in N2-purged 0.1 M KHCO3 at a pH of 8. The tableau of ECSTM images in Fig. 1 captures the temporal evolution of the polycrystalline electrode surface. After 2 h, multiple domains collectively formed wide regions with right-angular corners such as those marked by broken lines. On the third hour, further surface-atom rearrangement gave rise to a square adlattice indicative of a Cu(100) surface. The results were reminiscent of the Cu(pc) → Cu(pc)-[Cu(100)] transformation in 0.1 M KOH; the acquisition of atomically resolved images smaller than 10 nm × 10 nm was, however, onerous.

Fig. 1
figure 1

Electrochemical scanning tunneling microscopy images of polycrystalline Cu, Cu(pc), acquired during a constant-potential hold at − 0.90 V in N2-purged 0.1 M KHCO3. Dashed lines are provided as visual guide

Full size image

Cyclic voltammetry of the reconstructed surface, Cu(pc)-[Cu(100)], in 0.1 M KHCO3 bubbled with N2 was initiated at − 0.9 V to avoid disruption of the long-term potential hold. Figure 2 superimposes the voltammograms of the pre- and post-reconstructed Cu(pc) electrodes. The steady-state current-potential profile of the Cu electrode, prior to reconstruction, was consistent with the well-documented reduction-oxidation signatures of Cu in alkaline conditions [15]. Two distinct voltammetric features render support to the visual narrative chronicled by ECSTM: (1) Among the low-index facets of Cu, the (100) surface has been reported to manifest the lowest overpotential for the hydrogen evolution reaction (HER) [16]. Hence, Cu(pc)-[Cu(100)] in Fig. 2 is not unexpected to manifest a positive shift in the onset potential of HER. Such nuanced difference in electrochemical behavior corroborates the geometric evidence from ECSTM, which clearly points to the monocrystallinity of the reconstructed surface. (2) The area of the peak at ca. − 0.20 V, ascribed to the cathodic reduction of Cu(I) oxide to zerovalent Cu [14], was found to decrease after the reconstruction. The result implies a decrease in the electrochemically active surface area of the Cu electrode; that is, the topographic homogeneity of the polycrystalline Cu surface improved after reconstruction.

Fig. 2
figure 2

Cyclic voltammogram of polycrystalline Cu, Cu(pc), in N2-purged 0.1 M KHCO3, before and after a constant-potential hold at − 0.90 V for at least 3 h. Scan rate = 50 mV/s

Full size image

The Cu(pc)-[Cu(100)] sample originally prepared in N2-purged 0.1 M KHCO3 was further examined in N2-purged 0.1 M KOH. Although the applied potential was nominally kept at − 0.90 V during the replacement of the electrolyte solution, perturbations in the local electric field of the Cu electrode were inevitable as the bulk pH changed from 8 to 13 and the ionic environment transitioned from HCO3−-rich to predominantly OH−. Thus, the initial wide-scale ECSTM image in Fig. 3 revealed irregularly shaped domains with symmetry properties indecipherable from those of a pure (100) surface. Indications of surface-atom reorganization were noticeable after 30 min: narrow regions emerged with straight edges parallel to each other. After an hour, atomically resolved images of the Cu(100) square adlattice were unambiguously discernible from an area of 2.5 nm × 2.5 nm at numerous locations on the surface.

Fig. 3
figure 3

Electrochemical scanning tunneling microscopy images of Cu(pc)-[Cu(100)] acquired at a constant potential of − 0.90 V in 0.1 M KOH. The reconstructed surface was initially prepared in 0.1 M KHCO3. Dashed lines are provided as visual guide

Full size image

The facile structural reversion in 0.1 M KOH was consistent with the resemblance of the cyclic voltammograms recorded before and after the 1-h polarization at − 0.90 V (Fig. 4). Only a slight decrease (ca. 16%) in the peak area after the potential hold was evident for the Cu2O-to-Cu cathodic reaction at ca. − 0.40 V. The restoration of the (100) facet was also marked by a marginally small shift in the HER onset potential.

Fig. 4
figure 4

Cyclic voltammogram of Cu(pc)-[Cu(100)] in 0.1 M KOH, before and after a constant-potential hold at − 0.90 V for at least 1 h. The reconstructed surface was initially prepared in 0.1 M KHCO3. Scan rate = 50 mV/s. Experimental details are described in the text

Full size image

Juxtaposition of the voltammograms for the two Cu(pc)-[Cu(100)] samples, one prepared in N2-purged 0.1 M KHCO3 (Fig. 2) and the other regenerated in deaerated 0.1 M KOH (Fig. 4), reveals a pH-dependent shift of the onset potential of HER. For instance, to achieve a current density of − 0.1 mA cm−2, an applied potential of ca. − 0.8 and − 1.0 V was needed, respectively, at pH 8 and pH 13. The negative shift of the onset potential of HER in alkaline media qualitatively agreed with the Nernstian pH correction of − 0.059 V (13–8) = − 0.30 V, although the observed shift was much smaller at − 0.20 V. Such deviation may emanate, in part, from the fact that the bulk pH often differs substantially from the more crucial pH at the electrical double-layer region [17]. Furthermore, the variable baseline asymmetry, typical of many voltammetric scans in alkaline medium, precludes a precise comparison of onset potentials under these conditions.

It may be noted that, at − 0.9 V, the potential at which the ECSTM images were acquired, the extent of HER, albeit still incipient, was more pronounced in 0.1 M KHCO3 than in 0.1 M KOH. In 0.1 M KHCO3, the applied potential was competitively channeled between HER and surface reconstruction, and the resultant rate of the latter process was consequently more sluggish than that in alkaline solution. The reconstructed surfaces in both electrolyte systems, however, still converged to the same structure, i.e., Cu(pc)-[Cu(100)]. At potentials not far removed from the onset of the electrochemical reduction of CO2, the polycrystalline-to-monocrystalline transformation of the Cu electrode surface remained unaffected by the change in supporting electrolyte from the bicarbonate system at pH 8 to the strongly alkaline medium at pH 13. These results demonstrated that the surface reconstruction of Cu under the present experimental conditions was unaffected by the presence of supporting electrolyte ions in solution, viz., OH−, HCO3−, CO32−, H3O+, and K+. The accumulated evidence indicates that the surface reconstruction of polycrystalline Cu is a general phenomenon driven primarily by the negative potential applied on the electrode.

Summary

The polycrystalline surface of copper, Cu(pc), in N2-purged 0.1 M KHCO3 underwent surface reconstruction at a constant potential of − 0.90 V (SHE), near the onset of the electrochemical reduction of CO2. Atomic-level visualization from electrochemical scanning tunneling microscopy (ECSTM) depicted a kinetically hindered formation of the (100) facet, a surface-atom rearrangement similar to that previously reported in CO2-free 0.1 M KOH. Nuances in the voltammetric features rendered additional support to the polycrystalline-to-monocrystalline surface transformation. The reconstructed Cu surface generated at pH 8 in 0.1 M KHCO3 can be subsequently examined electrochemically and structurally at pH 13 in 0.1 M KOH, with minimal disruption, as long as the electrolyte exchange is executed potentiostatically at −0.90 V. At near-neutral-to-alkaline conditions, ions such as OH−, HCO3−, CO32−, K+, and H3O+ exert essentially no influence on the reorganization process. The Cu(pc) → Cu(pc)-[Cu(100)] surface reconstruction under CO2 reduction potentials is a general phenomenon driven primarily by the rather negative applied potential.