Electrochemical Surface Area Quantification, CO2 Reduction Performance, and Stability Studies of Unsupported Three-Dimensional Au Aerogels versus Carbon-Supported Au Nanoparticles.

The efficient scale-up of CO2-reduction technologies is a pivotal step to facilitate intermittent energy storage and for closing the carbon cycle. However, there is a need to minimize the occurrence of undesirable side reactions like H2 evolution and achieve selective production of value-added CO2-reduction products (CO and HCOO-) at as-high-as-possible current densities. Employing novel electrocatalysts such as unsupported metal aerogels, which possess a highly porous three-dimensional nanostructure, offers a plausible approach to realize this. In this study, we first quantify the electrochemical surface area of an Au aerogel (≈5 nm in web thickness) using the surface oxide-reduction and copper underpotential deposition methods. Subsequently, the aerogel is tested for its CO2-reduction performance in an in-house developed, two-compartment electrochemical cell. For comparison purposes, similar measurements are also performed on polycrystalline Au and a commercial catalyst consisting of Au nanoparticles supported on carbon black (Au/C). The Au aerogel exhibits a faradaic efficiency of ≈97% for CO production at ≈-0.48 VRHE, with a suppression of H2 production compared to Au/C that we ascribe to its larger Au-particle size. Finally, identical-location transmission electron microscopy of both nanomaterials before and after CO2-reduction reveals that, unlike Au/C, the aerogel network retains its nanoarchitecture at the potential of peak CO production.

Introduction

The electrochemical reduction of CO2 is increasingly considered a promising approach to tackle ongoing challenges in energy conversion while helping to reduce the emissions of CO2 resulting from anthropogenic activities.[1−4] In particular, the conversion of this CO2 into carbon-based fuels and value-added chemicals using energy from renewable sources like solar or wind can effectively take a step toward closing the carbon-based energy cycle. In addition, the sustainable production of chemical fuels and feedstock materials is advantageous, as it is compatible with the existing infrastructures for transportation and industrial applications, and it could serve to buffer the intermittent nature of renewable energy production technologies.[5]

While the CO2-reduction reaction (CO2RR) can lead to a variety of products (e.g., CO, CH4, C2H4, HCOO–, and CH3OH), only formate (HCOO–) or carbon monoxide (CO) has been found to be economically viable when compared to established, industrial production pathways.[6] In this regard, Au-based catalysts (including alloys, like AuCu) have been reported to display the highest selectivity for the production of CO.[7,8] The seminal experiments by Hori and co-workers, dating back as early as the 1980s, were the first to show a CO faradaic efficiency (FE) of up to 92% at −0.98 V versus the normal hydrogen electrode (VNHE) and ≈2.2 mA·cmgeo–2 using polycrystalline Au.[9] Based on these and subsequent results on polycrystalline and single-crystal model surfaces,[10−13] in recent years the research focus has shifted toward Au-based nanostructured electrocatalysts whose increased surface areas allow reaching the high CO-specific current densities needed for device implementation.[14−19] In doing so, numerous studies have tackled the effect of such Au-nanoparticles’ size and shape (and corresponding surface faceting) on the CO2RR-selectivity.

In electrocatalysis, such relations are customarily inferred by comparing the surface-specific activity (SSA) of a series of catalysts with different particle sizes and/or shapes and corresponding electrochemical surface areas (ECSAs),[20−23] whereby the SSA values correspond to the samples’ ECSA-normalized currents at a potential at which their performance is (mostly) determined by the reaction kinetics. In the CO2-electrocatalysis field, however, studies often fail to perform an accurate assessment of the ECSAs of the involved nanomaterials and/or to utilize this crucial metric to normalize the CO2RR-current density,[24] additionally hindering the comparison of their results with resembling studies. As an example of this, Mistry and co-workers[14] evaluated the surface area of size-selected Au-nanoparticles approximated to spheres and supported on quasi-planar SiO2/Si(111) wafers based on their height and number in atomic force microscopy measurements but performed the corresponding CO2-reduction experiments on equivalent nanoparticles coated on glassy carbon substrates with a far-from-planar surface (i.e., a roughness factor (RF) of ≈10 cmsurface2·cmgeom–2),[25,26] thus putting into question the reliability of their surface-area-normalized current densities. On the other hand, Mascaretti et al.[15] derived their ECSAs from double-layer capacitance measurements known to yield inaccurate values because of the significant deviations in specific double-layer capacitances among porous catalysts, which can be aggravated by an erroneous choice of potential windows and/or scan rates.[27,28] Alternatively, other studies have determined the ECSA of Au electrocatalysts using metal underpotential deposition (UPD) methods based on the potential-controlled adsorption of a (sub)monolayer of Pb or Cu on the Au surface, in what probably constitutes the most reliable technique to this end.[29−32] However, these measurements were systematically performed in a potentiodynamic manner (i.e., recording cyclic voltammograms (CVs) at a given potential scan rate), whereby the slow kinetics of such UPD processes and inaccuracies in the choice of the cathodic inversion potential can lead to significant imprecisions in the inferred ECSA values.[16,17] Therefore, there is a clear need for defining good practices for ECSA evaluation that should allow an unambiguous comparison of the CO2-reduction performance of the different Au-based electrocatalysts found in the literature.

Beyond these considerations, the use of nanostructured catalysts is often combined with that of carbon supports that improve the nanoparticles’ dispersion and concomitantly increase the catalysts’ ECSA, but it can also shift the product selectivity toward undesirable H2 because of the intrinsic H2-evolution activity of C-surfaces.[33−36] This could be possibly circumvented by employing unsupported catalysts, among which aerogels consisting of a three-dimensional (3D) network of interconnected nanowires feature high porosity and large surface areas[37] required for their implementation in device-applicable electrodes. More specifically, noble metal aerogels present an exciting prospect in the field of electrocatalysis, because they combine the high catalytic activity and excellent charge transfer properties of noble metals[38] with an unsupported nature that can be beneficial to the materials’ stability.[39−41] Despite being studied for more than a decade, only a few reports involving electrochemical applications of noble metal aerogels exist, including studies showing their utilization as catalysts for oxygen evolution, oxygen reduction, or ethanol oxidation,[42−44] with only a few studies having analyzed these materials’ potential for CO2-electroreduction.[45]

With these considerations, this work presents a systematic electrochemical study of an unsupported Au aerogel (AuAG) as a CO2-reduction electrocatalyst–a first, to the best of our knowledge. To put the results obtained with this Au aerogel into perspective, we employ two other Au-based materials as performance benchmarks, namely, polycrystalline gold (AuPC) and a commercial catalyst consisting of Au-nanoparticles supported on carbon black (Au/C). We first conduct detailed electrochemical characterization of these three Au-based electrocatalysts to determine their ECSAs on the basis of the surface oxide-reduction charge and of the copper underpotential deposition (Cu-UPD) method in a rotating disk electrode (RDE) setup. This is complemented by identical-location transmission electron microscopy (IL-TEM) measurements of the same nanomaterials, with which we verify their possible degradation during electrochemical conditioning. Finally, we test the CO2 electroreduction performance of AuAG, Au/C, and AuPC in an in-house developed electrochemical cell implementing online gas chromatography for product quantification,[46] and compare the CO2RR performance of these materials with previous literature data for Au-nanocatalysts. This is complemented by additional (postmortem) IL-TEM measurements of the AuAG and Au/C catalysts that are used to assess the changes in their nanostructure following CO2-electroreduction. In this study, we show that Cu-UPD performed with potential holds provides a much more reliable way of estimating ECSAs as compared to using potential-sweeps for Cu-UPD. Additionally, it is revealed that AuAG shows superior selectivity for CO production and higher structural stability during CO2-reduction as compared to Au/C.

Experimental Section

Synthesis of Au Aerogels

The synthesis route of the 3D Au aerogel is analogous to the literature.[47] First, 0.1 mmol HAuCl4·3H2O (>99.9% trace metal basis, Sigma-Aldrich) was dissolved in 492 mL of water. Next, 3 mL of 0.143 M NaBH4 (>96%, Sigma-Aldrich) was quickly added, and the whole solution was stirred for 30 min. Afterward, the solution is split into two equal parts, and 100 mL of toluene (p.a.) was added to each part and vigorously shaken manually for 30 s. The gel pieces were collected at the phase boundary, and subsequently, an acetone exchange was performed. Finally, the gel was transferred to an autoclave (13200J0AB, SPI Supplies) for a solvent exchange to liquid CO2 and later supercritical drying (37 °C and 90 bar).

Chemicals and Gases for the Electrochemical Measurements

The electrolytes employed in the ECSA and CO2-reduction measurements were prepared from 96% H2SO4 (Suprapur, Merck) and KHCO3 (99.95% trace metals basis, Sigma-Aldrich), respectively. The Cu-UPD experiments were performed using CuSO4·5H2O (99.999% trace metal basis, Sigma-Aldrich) and NaCl (≥ 99.999%, TraceSELECT, Fluka). All electrolytes were prepared in the specified concentrations by diluting these salts in the required volumes of ultrapure water (18.2 MΩ·cm, ELGA Purelab Ultra). High-purity N2 and CO2 (6.0 vs 5.5 grades, respectively—Messer Schweiz AG) were bubbled through the electrolytes to saturate them with the respective gases.

Electrochemical Methods for ECSA Quantification

The electrochemical measurements for estimating the ECSAs in acidic media were performed in the RDE configuration in a custom glass cell (Schmizo AG). A gas bubbler facilitated direct bubbling of N2 into the electrolyte before the experimental run for 1 h to remove the dissolved O2 from the electrolyte. Later, the gas bubbler allowed N2 blanketing during the experimental procedure. A gold-mesh (Advent Research Materials) fixed with a PTFE-stopper and a K2SO4-saturated Hg/HgSO4 electrode (RE-2CP, ALS Co. Ltd.) served as the counter and reference electrodes, respectively. The reference electrode was precalibrated versus the reversible hydrogen electrode (RHE) scale in the same electrolyte saturated with H2 by performing H2 evolution/oxidation measurements on a polycrystalline Pt RDE, and hence the potentials mentioned in the ECSA part of this study are expressed as VRHE. The reference electrode was placed inside an electrolyte-filled glass tube with its lower end tapering into a fluorinated ethylene-propylene tube (FEP, Zeus Industrial Products Inc.) plugged with a porous glass frit (Ametek G0300) in contact with the electrolyte.

Three Au-based catalysts were employed as working electrodes during the ECSA study: a polycrystalline Au disk (Pine Research Instrumentation), a 20 wt % Au on Vulcan XC-72 carbon black commercial catalyst (Premetek Co.), and the Au aerogel for which the synthesis is described above. The 5 mm diameter AuPC and glassy carbon disks (HTW Hochtemperatur-Werkstoffe) used to immobilize the powder samples underwent mechanical polishing with 3, 1, and 0.25 μm diamond suspensions (Electron Microscopy Sciences) on a micropolishing cloth (Bühler) in decreasing size order, as to attain a mirror-like finish. After polishing, the disks were sonicated in isopropyl alcohol (99%, VWR) and ultrapure water twice for 5 min each. Following this step, the disks were mounted on an interchangeable rotating ring-disk electrode (RRDE, Pine Research Instrumentation) with a PTFE shroud and an Au ring that did not serve as an active electrode but ensured a tighter, leak-proof assembly of the disks. The resulting electrode assembly was dipped in a 2 M solution of HClO4 (70%, Suprapur, Merck) for 5 min and rinsed with ultrapure water just prior to its use.

For the Au/C and AuAG electrodes, catalyst inks were prepared by adding one part of isopropyl alcohol (99.9%, HPLC grade, Sigma-Aldrich) and three parts of ultrapure water, in that order, to a preweighed amount of Au/C or Au aerogel powder, along with the volume of Nafion solution (5 wt %, Sigma-Aldrich) needed to attain an ionomer-to-carbon or ionomer-to-aerogel mass ratio of 0.20 for Au/C vs AuAG, respectively. Following ultrasonication, catalyst layers were prepared by depositing 10 μL of the resulting ink on the glassy carbon disk embedded in the RRDE tip, as to yield loadings of 15 vs 50 μgAu·cmgeom–2 on a gold-basis for Au/C vs AuAG, respectively. Once dried under a running N2 flow, these Au/C- or AuAG-coated RRDEs (or the AuPC disk, minus the ink-deposition procedure highlighted above) were mounted on a Pine Research MSR Rotating Station and immersed in the N2-saturated 0.1 M H2SO4 electrolyte in which all ECSA measurements were conducted. The working electrodes were immersed in the electrolyte while holding the potential at 1 VRHE, as to avoid the uncontrolled chemisorption of SO42– and HSO4– ions that compete with OH–-adsorption (see eq below).[48]

Following this immersion, the AuPC was electrochemically conditioned by potential-cycling at a sweep rate of 1000 mV·s–1 between 0 and 1.75 VRHE for 100 cycles, because in ref (49) these conditions were found to be adequate to obtain a reproducible electrode surface area. For AuAG and Au/C, this electrochemical conditioning was performed by cycling the potential between −0.04 and 1.64 VRHE at a sweep rate of 50 mV·s–1 for 20 cycles, during which a stable voltammogram was obtained. In contrast to the conditions used for AuPC (vide supra), these milder electrode conditioning parameters were chosen to avoid drastic structural changes in the nanochain network of AuAG and to avoid corrosion of the carbon support in Au/C. To estimate the ECSA based on the charge associated with the electrochemical reduction of the surface oxide built up at this higher inversion potential (i.e., 1.64 VRHE), a specific charge value of 386 μC·cmAu–2 was used for normalization.[48]

Cu-UPD measurements were performed by adding to a known volume of 0.1 M H2SO4 electrolyte the amounts of 70 mM Cu2SO4·5H2O and 70 mM NaCl solutions needed to attain electrolyte concentrations of 0.1 mM Cu2+ and 0.2 mM Cl–, respectively. This addition was performed while holding the working electrode at a potential of 1 VRHE, as to avoid any unwanted Cu deposition. The purpose of adding Cl– ions was to accelerate the kinetics of the Cu-UPD reaction by facilitating coadsorption of Cu-Cl layers.[30,32] To determine the potential value below which bulk deposition of Cu starts taking place, we recorded a cathodic sweep at a scan rate of 20 mV·s–1 from the holding potential of 1 VRHE until a steep increase in the cathodic current indicative of the deposition of Cu multilayers (i.e., Cu-plating) was observed. Then, additional CVs at a rate of 20 mV·s–1 were recorded with different cathodic inversion potentials (Einv,c) close to (if systematically above) the onset of the aforementioned Cu-plating. Next, to explore the effect of time on the Cu-coverage at a given potential, a potential hold was performed at selected Einv,c values for time durations varying from 15 to 600 s. This was immediately followed by an anodic linear sweep from the point of potential hold to 1 VRHE, as to strip the Cu deposited during the given time. Finally, all the anodic currents thus obtained were integrated to quantify the coulombic charge associated with the stripping process. The baseline for the integration of these currents was inferred from a previous CV recorded at 20 mV·s–1 between 0.1 and 1 VRHE prior to the addition of Cu2+ and Cl– ions. The Cu-stripping charges were converted into ECSA values using a normalization charge of ≈370 μC·cmAu–2, which corresponds to the average of the equivalent charges of Cu-UPD monolayer formation on Au(111), Au(110), and Au (100) mathematically derived using the lattice parameters and packing of the surface atoms.[50]

Electrochemical Methods for IL-TEM

The experimental procedure for IL-TEM was inspired by the work of Schlögl and co-workers.[51] An experimental setup similar to that of the ECSA measurements was used, with the exception that the working electrodes consisted of Au finder-TEM grids (Ted Pella Inc.) on which 5 μL aliquots of Au/C or AuAG inks (1:49 diluted as compared to those used for the RDE measurements described above) had been deposited and dried under ambient conditions. These Au finder-grids were specifically chosen because of their relative stability over commonly employed copper grids and in the potential window used for electrochemical conditioning. TEM images of specific spots were acquired for these Au catalyst-Au grid ensembles, following which the outside boundary of the Au TEM grid was welded to an Au wire (99.99+%, 0.2 mm, Advent Research Materials) used to provide electrical contact for the subsequent electrochemical steps. For IL-TEM measurements investigating the effect of the electrochemical conditioning step, 20 CVs between 0.2 and 1.6 VRHE at 50 mV·s–1 were recorded on the TEM grids in 0.1 M H2SO4. On the other hand, IL-TEM analyses concerning the effects of CO2-reduction conditions mimicked the respective electrochemical procedure performed on AuAG and Au/C during the CO2-reduction studies (vide infra) and consisted of a 60 min hold at −0.5 VRHE in CO2-saturated 0.5 M KHCO3. Following the electrochemical steps, the finder-TEM grid was carefully detached from the Au wire and TEM-inspected at the same specific spots as before. This TEM analysis before and after the electrochemical procedure was performed in a JEOL JEM-ARM200F (200 kV, JEOL Limited).

Electrochemical Methods for CO2-Reduction

The Au electrocatalysts were tested for their CO2-reduction performance in an in-house developed electrochemical cell coupled to an online gas chromatograph (GC, 8610C SRI Instruments). The detailed schematic of this cell and the whole measurement setup and product quantification approaches have been discussed in our previous work.[46] Briefly, this two-compartment cell employed a perfluorinated Nafion XL membrane (Chemours) presoaked in 0.5 M KHCO3 as a separator between the working and counter electrode compartments. KHCO3 electrolyte (0.5 M) was first placed in a presaturation tank in which CO2 was continuously bubbled through the electrolyte to remove dissolved O2 and to ensure that the electrolyte was well saturated with CO2 before injecting 3 mL of it into each cell compartment. An Ag/AgCl reference electrode (LF-1, Harvard Apparatus) stored in 3 M KCl solution was utilized for these experiments, with an ≈1 cm2 area of Pt foil (99.99%, Alfa Aesar) serving as the counter electrode. However, all potentials stated in this study are expressed against the reversible hydrogen electrode (VRHE), whereby the aforementioned reference electrode was calibrated by conducting H2-oxidation/evolution measurements on a polycrystalline Pt RDE in a H2-saturated 0.5 M K2HPO4/KH2PO4 buffer electrolyte of the same pH as the CO2-saturated, 0.5 M KHCO3 solution. After assembling the cell and filling it with a presaturated electrolyte, high-purity CO2 was bubbled through both compartments of the cell at a flow rate of 10 mL·min–1 via glass frits (ROBU, 6 mm diameter, porosity #2) for 15 min before electrochemical operation and then continuously during the electrochemical tests. All electrochemical measurements were performed using a potentiostat (VSP-300, Biologic Science Instruments) controlled by EC-Lab software (Biologic Science Instruments). The GC analyzed the CO2-reduction gaseous products every 5 min during 1 h potential holds with the help of an autosampler function on PeakSimple 4.88. The catholyte was examined for ionic CO2-reduction products (e.g., HCOO–) at the end of every one hour potentiostatic measurement using ion chromatography (882 Compact IC Plus, Metrohm AG).

The polycrystalline Au sample used for the CO2-electroreduction measurements was produced by sputtering an ≈210 nm thick layer of Au on 2.5 × 2.5 cm2 glass plates coated with indium-doped tin oxide (ITO), with an ≈10 nm backing layer of chromium between the Au and ITO layers. Thus, this electrode appears referred to as AuPC/ITO in what follows. Glassy carbon plates (Goodfellow Cambridge Limited) with an area of 2.5 × 2.5 cm2 served as the backing electrodes on which Au/C and AuAG catalyst inks were drop-cast to prepare the corresponding working electrodes for CO2RR measurements. Prior to drop-casting, the glassy carbon plates were polished on a Bühler micropolishing cloth with 0.05 μm aluminum slurry to achieve a shiny finish. Next, the plates were sonicated for 5 min each after being immersed in isopropyl alcohol and ultrapure water, respectively. Once dried, the plates were installed in a special drop-casting setup in which a PTFE gasket exposed a 1 cm2 circular area and a 100 μgAu·cmgeom–2 loading of catalyst ink was deposited.[46] Once dried and before placing this glassy carbon-Au catalyst ensemble inside the cell, the catalyst layer was prewetted by transferring the electrodes into a desiccator, placing a few drops of 0.5 M KHCO3 to cover the drop-casted area, and evacuating the desiccator to 30 mbar for 5 min.

The electrochemical procedure involved performing electrochemical impedance spectroscopy measurements at open circuit in a frequency range of 1 MHz–1 Hz to determine the uncompensated solution resistance of the cell. This value varied between 50 and 60 Ω, of which 85% was compensated by EC-Lab, while the remaining 15% was accounted for during data analysis. Next, one CV was recorded at 50 mV·s–1 between −0.1 and 1.6 VRHE followed by three CVs at 20 mV·s–1 in the same potential window to achieve a stable voltammogram of the Au electrocatalysts. Finally, the potential was swept from 0.1 VRHE to the relevant CO2-reduction potential, where chronoamperometric measurements for 1 h and periodic GC injections were initiated simultaneously. Following this 1 h hold, an anodic sweep at 20 mV·s–1 was executed to 1.6 VRHE to strip off any surface species adsorbed on the catalyst surface during CO2-reduction.[52]

Results and Discussion

ECSA Determination of AuPC, Au/C, and AuAG

The meaningful comparison of a newly developed electrocatalyst with well-studied materials of different morphologies and/or architectures requires the evaluation of the ECSAs of these novel and better-established samples. While there are several electrochemical methods for determining ECSA values, their relevance and precision depend on the material under investigation and the underlying assumptions of each method.[53] Upon recording CVs of noble metals like Au, these exhibit characteristic surface oxide formation and reduction pseudocapacitive currents that have been extensively documented in the literature and can be leveraged to calculate the ECSA of Au-based electrocatalysts.[54] In doing so, it is commonly assumed that the associated charges correspond to the chemisorption of OH on the gold surface, according to the equation:[55]

Concomitantly, the ECSA of a given Au catalyst can be estimated by integrating the area under the peak associated with the reduction of this gold (hydr)oxide layer, as to derive the charge related to the number of surface sites electrochemically available for this oxide ad/desorption process.[54] This charge is then divided by a normalization value associated with the formation/reduction of a (hydr)oxide monolayer on an idealized Au surface, yielding the catalyst’s ECSA. Herein lies the biggest shortcoming of this so-called oxide-reduction method of ECSA quantification, because the precise stoichiometry of the oxide formed at a given potential is unknown (and may imply the partial buildup of a more oxidized phase, like Au2O3), and it is hard to establish if this (hydr)oxide formation process is limited to a (sub)monolayer or may imply the buildup of multiple layers.[27,29]

Alternatively, the ECSA can also be quantified taking advantage of metal UPD processes, implying the potential-driven electrodeposition of (sub)monolayer amounts of metal atoms on the surface of a substrate of interest at potentials positive of the theoretical value of bulk deposition (i.e., plating) of the adsorbing species.[33] For Au-based materials, the similarity in atomic radii of Au and Cu leads to strong metal–substrate interactions that drive this UPD process, and thus Cu-UPD has been widely employed for ECSA determination of Au-based materials.[30−32] To this end, a (sub)monolayer amount of Cu is deposited on the Au surface by either cathodic potential sweep or potential hold down to/at the so-called UPD potential beyond which multilayer deposition of Cu begins to take place. This potential sweep or hold is followed by an anodic scan to strip off the adsorbed Cu adatoms. The currents associated with the deposition and stripping of Cu are subsequently integrated to obtain the deposition/stripping charge, which is in terms divided by a normalization charge for the Cu-UPD process (further discussed below) to yield the corresponding ECSA value. Most importantly, the careful experimental determination of the UPD potential at which the Cu-adsorption process reaches its maximum coverage while still being limited to a (sub)monolayer amount is crucial to ensure this method’s reliability, because potential holds negative of this value result in bulk deposition (plating) of Cu on the Au surface and therefore yield overestimated ECSA values.

Based on this comparison among ECSA-determination approaches, we started our study using the surface oxide-reduction method to quantify the ECSA of AuPC. Its voltammogram in N2-saturated 0.1 M H2SO4, displayed in Figure a, features a characteristic, steep increase in anodic current at ≈1.32 VRHE that leads to a broad anodic current related to the oxidation of gold and, upon inversion of the potential scanning direction, is followed by a sharper reduction peak centered at ≈1.16 VRHE, in accordance with previous CVs recorded under the same conditions.[50,54] When the derived oxide-reduction charge is normalized by the value of 386 μC·cmAu–2 stated by Tremiliosi-Filho and co-workers[48] and additionally divided by the geometric surface area of the electrode (0.196 cmgeo2–see the Experimental Section), a RF of ≈1.7 cmAu–2·cmgeo–2 is estimated for AuPC and appears listed in Table . This RF is in line with the values reported for other AuPC electrodes (cf. 1.6 vs 2.1 cmAu–2·cmgeo–2 in refs (48) vs (56), respectively) and generally agrees with what would be expected for polycrystalline surfaces prepared by hand-polishing.[57,58]

Figure 1

(a) Conditioning CVs recorded for AuPC in N2-saturated 0.1 M H2SO4. (b) Cu-UPD CVs with variable cathodic inversion potentials (Einv,c) recorded for AuPC in N2-saturated 0.1 M H2SO4 + 0.1 mM CuSO4.5H2O + 0.2 mM NaCl. The dashed line represents the baseline for integration of Cu stripping peaks. (c) Cu stripping charges calculated from anodic scans after chronoamperometric holds at the potentials shown in plot (b) for different time durations.

Table 1

Summary of RF (cmAu2 cmgeo–2) and ECSA (m2·gAu–1) Values Derived for AuPC, Au/C, and AuAG Using TEM, Au Oxide Reduction, and Cu-UPD Methods (E-Scan and E-Hold + Stripping)

	TEMa	Au oxide reduction	Cu-UPD, E-scan	Cu-UPD, E-hold + stripping
AuPCb		1.7 cmAu2·cmgeo–2	1.1 cmAu2·cmgeo–2	1.6 cmAu2·cmgeo–2
Au/C	27 m2·gAu–1	21 m2·gAu–1	10 m2·gAu–1	17 m2·gAu–1
AuAG		11 m2·gAu–1	2 m2·gAu–1	9 m2·gAu–1

Only applicable to the Au/C sample.

For AuPC, the tabulated values are RFs corresponding to the ratio between the surface area of Au and the geometric area of the electrode.

After this initial RF-quantification for AuPC based on the oxide-reduction method, we proceed to conduct a more reliable RF-determination using Cu-UPD. For this, CVs in the same 0.1 M H2SO4 electrolyte additionally containing 0.1 mM Cu2+ and 0.2 mM Cl– were first recorded with cathodic inversion potentials (Einv,c) of 0.28, 0.26, or 0.24 VRHE (see Figure b). For potentials ≤0.24 VRHE, a steep increase in cathodic current was observed, indicating the unambiguous occurrence of Cu-plating. Following this, potentiostatic Cu-UPD measurements were conducted at the aforementioned Einv,c values, and the Cu-stripping charges recorded in the subsequent anodic sweeps appear plotted in Figure c as a function of the duration of the potential hold. Evidently, the chronoamperometric measurements performed at 0.24 VRHE resulted in a monotonic increase of the Cu-stripping charge with potential-hold time indicative of a bulk Cu-deposition process. On the contrary, potential-holding at 0.28 and 0.26 VRHE led to Cu-stripping charges plateauing over a time interval of 600 s, indicating the occurrence of a Cu-UPD process and the formation of a Cu-(sub)monolayer on the AuPC surface. Using the normalization charge of ≈370 μC·cmAu–2 and considering the Cu stripping charge of ≈622 μC·cmgeo–2 achieved for the UPD potential of 0.26 VRHE result in a RF of ≈1.6 cmAu2·cmgeo–2, commensurate with the value of ≈1.7 cmAu2·cmgeo–2 derived using the oxide-reduction method, and that also appears tabulated in Table . Notably, when this RF-quantification is performed on the basis of the Cu-stripping charge in the corresponding CV measurement with an Einv,c of 0.26 VRHE (i.e., as opposed to using the stripping charge following a sufficiently long E-holding), the result is an RF of 1.1 cmAu2·cmgeo–2 (cf. Table ).

This ≈31% lower value highlights the importance of performing these Cu-UPD measurements using potentiostatic holds followed by anodic scans, as opposed to the CV measurements applied in previous CO2-reduction studies (vide supra) which, as we will further discuss below, lead to underestimated surface areas and correspondingly overestimated SSAs.

Having assessed the ECSA of AuPC, we performed a similar ECSA-quantification study on the commercial, 20% Au/C catalyst. The TEM images of this material shown in Figure a unveil that its nanoparticles exhibit an average particle size of ≈9 nm. Approximating these Au nanoparticles as spheres, we leveraged a particle size distribution (see Figure S1) derived from 200 samplings in the TEM images and calculated a specific surface area of ≈27 m2·gAu–1 for the Au/C catalyst (see section 1 in the Supporting Information for details) that again appears summarized in Table .[59]

Figure 2

IL-TEM images before (a) and after (b) conducting 20 cyclic voltammetry scans at 50 mV.s–1 on Au/C between 0.2 and 1.6 VRHE in N2-saturated 0.1 M H2SO4.

Upon subsequent electrochemical measurements with a gold loading of 15 μgAu·cmgeo–2 in N2-saturated 0.1 M H2SO4, this Au/C catalyst displayed the characteristic CV displayed in Figure a which, compared to the CV of AuPC in Figure a, features a broad double-layer current caused by the additional presence of carbon that is in line with previous observations.[60] Similar to the case of AuPC, the charge under the Au oxide reduction peak in Figure a is normalized by a specific charge of 386 μC·cmAu–2 and further divided by the geometric surface area of the electrode (0.196 cmgeo2) and the catalyst ink loading (15 μgAu·cmgeom–2) to yield an ECSA value of ≈21 m2·gAu–1 listed again in Table .

Figure 3

(a) Conditioning CVs recorded for Au/C in N2-saturated 0.1 M H2SO4. (b) Cu-UPD CVs with variable cathodic inversion potentials (Einv,c) recorded for Au/C in N2-saturated 0.1 M H2SO4 with 0.1 mM CuSO4.5H2O and 0.2 mM NaCl. The dashed line represents the baseline for integration of Cu stripping peaks. (c) Cu stripping charges calculated from anodic scans after chronoamperometric holds at the potentials shown in plot (b) for different time durations.

Once we obtained a reproducible CV for Au/C in N2-saturated 0.1 M H2SO4, Cu2+ and Cl– ions were added to the electrolyte (see the Experimental Section), and the potential was cycled from 1 VRHE to varying Einv,c potentials of 0.30, 0.28, 0.26, or 0.24 VRHE to achieve varying levels of Cu-deposition on the Au-nanoparticles’ surface. These CVs are displayed in Figure b, and the corresponding chronoamperometric and stripping CV measurements used to determine the accurate UPD potential appear plotted in Figure c. Unlike in the case of AuPC, in which bulk Cu-deposition was only observed at potentials ≤0.24 VRHE (see Figure b), for Au/C this effect is already present at 0.26 VRHE, and Cu-stripping charges independent of the potential hold duration (and thus indicative of the exclusive occurrence of a UPD process) are only observed for Einv,c values of 0.28 or 0.30 VRHE. Most importantly, if one considers the Cu-stripping charge of ≈954 μC·cmgeo–2 obtained following a 600 s hold at 0.28 VRHE, along with the Cu-UPD normalization charge of 370 μC·cmAu–2 discussed above, an ECSA value of ≈17 m2·gAu–1 is estimated (see Table ). Interestingly, this ECSA is ≈40%lower than the specific surface area of ≈27 m2·gAu–1 estimated on the basis of the nanoparticles’ size distribution and mentioned above, a disagreement that stems from fractions of these nanoparticles that are not accessible to the electrolyte and thus do not contribute to the electrochemical process (e.g., due to their partial agglomeration and/or anchoring to the support surface),[61] and is in accordance with what has been reported in previous literature for C-supported metal nanoparticles.[59,62] Furthermore, when this ECSA quantification is performed using the Cu-stripping charge from continuous CVs with the same Einv,c value as the UPD potential deduced above (i.e., 0.28 VRHE), it results in an ECSA value of 10 m2·gAu–1 (cf. Table ). This large deviation of ≈40% with regard to the Cu-UPD measurements performed with potential holds is in line with the behavior observed above for AuPC and again emphasizes the downsides of UPD measurements in a potentiodynamic mode.

After establishing the ECSA of this Au/C benchmark electrocatalyst, we shift our focus to the novel 3D Au aerogel (AuAG). The TEM images displayed in Figure a unveil that the unsupported Au network consists of a nanostructure of smooth, interconnected chains with an average web thickness of ≈5 nm and numerous junctions with a larger average diameter. Notably, these irregular dimensions and abundant interconnectivity are in contrast with the observations for Pt-based aerogels, consisting of a necklace of nanoparticles of relatively resembling diameters and with a lower interconnectivity extent, and prevent a reliable estimation of the geometric surface area of this AuAG sample based on its necklace’s bare average diameter.[63,64]

Figure 4

IL-TEM images before (a) and after (b) conducting 20 cyclic voltammetry scans at 50 mV s–1 on AuAG between 0.2 and 1.6 VRHE in N2-saturated 0.1 M H2SO4.

The ECSA determination of the AuAG was pursued in a similar manner to that for Au/C, in this case by preparing RDEs with an aerogel loading of 50 μgAu·cmgeo–2 and recording CVs in N2-saturated 0.1 M H2SO4. These are featured in Figure a and display a narrow double-layer region (consistent with the AuPC CVs–see Figure a) compared to the Au/C voltammograms in Figure a that can be ascribed to the absence of a carbon support in the aerogels. Leveraging the Au oxide reduction method for ECSA determination of this 50 μgAu·cmgeo–2 Au aerogel catalyst layer, we obtain an ECSA value of ≈11 m2·gAu–1, which appears tabulated in Table .

Figure 5

(a) Conditioning CVs recorded for AuAG in N2-saturated 0.1 M H2SO4. (b) Cu-UPD CVs with variable cathodic inversion potentials (Einv,c) recorded for AuAG in N2-saturated 0.1 M H2SO4 with 0.1 mM CuSO4.5H2O. The dashed line represents the baseline for integration of Cu stripping peaks. (c) Cu stripping charges calculated from anodic scans after chronoamperometric holds at the potentials shown in plot (b) for different time durations.

On similar lines to AuPC and Au/C, we performed Cu-UPD measurements on AuAG by introducing Cu2+ into the experimental setup; notably, these measurements were conducted without adding Cl– , because in a series of separate measurements we observed that this anionic species leads to a continuous decline of the Cu-stripping charge (see Figure S2) that we attribute to a Cl–-induced pitting of the aerogel’s Au nanochains which might result in the collapse of their 3D nanostructure and drastically reduce this material’s ECSA.[65,66] Following this preliminary observation, we conducted potential-sweep and -hold measurements sequentially to determine the UPD potential and associated Cu-stripping charges; these results are presented in Figure b,c, respectively. Considering the Cu-stripping charge of ≈1745 μC·cmgeo–2 obtained after a 600 s hold at a UPD potential of 0.26 VRHE, along with a Cu-UPD normalization charge of 370 μC·cmAu–2, a catalyst loading of 50 μgAu·cmgeo–2, and a geometric area of 0.196 cm2, we obtain the ECSA value of ≈9 m2·gAu–1 listed in Table . By comparison, the CV-based Cu-UPD measurement for the same UPD potential of 0.26 VRHE results in an ECSA value of only ≈2 m2·gAu–1, which is ≈80% lower than what is derived from the equivalent Cu-UPD measurements with potential hold. Moreover, this difference between ECSA values derived from both Cu-UPD methods (i.e., with continuous E-scanning vs holding the potential) is in terms ≈2-fold larger than what we observed in the above Au/C and AuPC measurements (see Table ). A likely reason for this could be the absence of Cl– ions during Cu-UPD experiments with the Au aerogel, as compared to those performed on AuPC and Au/C, because this absence of chloride likely results in slower Cu-deposition kinetics. Most importantly, the majority of the studies discussed above in which Cu-UPD CVs were used to estimate Au-nanomaterials’ ECSAs also overlooked the inclusion of Cl– in their UPD measurements and, based on these observations, likely reported largely underestimated ECSAs.[17,18,67] Notably, we believe that this issue can be likely aggravated if the Cu-UPD CVs are recorded at faster scan rates.[17]

In summary, under the assumption that the Cu-UPD method based on potential holds yields the most reliable ECSA values, Table unveils that the Au oxide reduction method slightly overestimates (by ≈10 to 20%) the ECSAs of all three materials discussed in this work. On the other hand, the Cu-UPD method, which employs potentiodynamic deposition of Cu, consistently underestimates the ECSAs, and alarmingly so when the measurements are conducted in the absence of Cl– ions.

IL-TEM Analysis for Electrochemical Conditioning

The electrochemical conditioning of Au surfaces involves potential-cycling to high anodic potentials (≥1.6 VRHE) that can lead to surface reconstruction, Au dissolution, and/or redeposition[66,68,69] and may alter the surface area estimated using the oxide-reduction and Cu-UPD methods applied above. To determine the extent to which these effects may have impacted the ECSA values reported in Table , we employed IL-TEM to gain microscopic insight regarding the morphology of the nanostructured Au electrocatalysts before and after conditioning (i.e., potential-cycling). These IL-TEM measurements were performed on both Au/C and AuAG by mimicking the electrochemical conditioning steps performed in the RDE configuration (i.e., recording 20 CVs at 50 mV·s–1 between 0.2 and 1.6 VRHE in N2-saturated 0.1 M H2SO4–see Figure S3), and the corresponding IL-TEM images before and after conditioning are presented in Figures a,b and 4a,b, respectively. For Au/C, the red squares marked as ‘A’ in Figure a,b depict how an elongated part of the carbon matrix gets crumpled into a smaller volume after cyclic voltammetry, while the red square marked ‘B’ points a part of carbonaceous material that is absent in the postmortem TEM image. These observations hint toward considerable mobility and corrosion of the carbon matrix at these high anodic potentials, consistent with what has been reported previously.[70] Because the Au NPs are embedded in the carbon matrix, its corrosion results in the nanoparticles’ migration and even detachment, as indicated by the red circles in Figure a (absent in the postconditioning image, cf. Figure b). Interestingly, the blue circles in Figure b point at the appearance of ≈3 nm Au NPs in the vacant region around the carbon matrix that cannot be found in Figure a. We hypothesize that these smaller Au NPs were produced as a result of Au dissolution and redeposition as the Au/C was cycled between oxidative and reductive potentials.[71] While the relocation of NPs should not affect the overall ECSA of the Au/C electrocatalyst, their segmentation and possible coalescence would lead to a change in the ECSA. To estimate the impact of these potential-cycling effects on this variable, we evaluated the catalyst’s postconditioning particle size distribution using the same procedures highlighted above and discussed it in section 1 of the SI (see Figure S1b), from which we derived an average particle size of ≈9.5 nm and a corresponding geometric surface area of ≈24 m2·gAu–1. Comparing this value to the preconditioning geometric surface area of ≈27 m2·gAu–1 and considering the error associated with such statistically averaged calculations, these conditioning-induced changes in the surface area can be considered negligible.

Next, we shift our focus to the IL-TEM images of AuAG in Figure b, whereby a cursory glance reveals that the overall 3D structure looks largely unchanged, pointing toward an enhanced stability of this unsupported electrocatalyst during electrode conditioning. However, similar to what was observed for Au/C, the blue circles in the postconditioning TEM images in Figure b indicate the presence of ≈3 to 4 nm Au NPs around the AuAG that were absent in the TEM images of the as-prepared AuAG in Figure a.

Therefore, these Au NPs were also produced because of the Au dissolution–redeposition cycles caused by potential cycling; based on the above conclusions for Au/C, though, we also assume that their appearance does not have a significant effect on the material’s ECSA. Additionally, the red squares in Figure a,b point out minute changes observable in the aerogel network, whereby chain terminations are notably different or the spacing between the chains has changed. On the other hand, this reshaping of the aerogel network is highly localized and does not entail any significant changes in the necklace size, and thus we believe that the electrochemical conditioning does not lead to significant changes in the material’s ECSA.

CO2-Reduction Performance of AuPC/ITO, Au/C, and AuAG

Following this rigorous quantification of the nanocatalysts’ ECSA and the verification that this key parameter is not significantly affected by their electrochemical conditioning, we tested the CO2-reduction performance of AuPC/ITO, Au/C, and AuAG in CO2-saturated 0.5 M KHCO3. The FEs for CO and H2 production (blue vs red symbols) during CO2-reduction on AuAG and Au/C at various potentials are shown in Figure a,b, respectively, while the corresponding results for AuPC/ITO are displayed in Figure S4. Regarding the latter, the current densities recorded during CO2-reduction on AuPC/ITO at potentials > −0.5 VRHE were too low to generate CO and H2 at levels that can be accurately and reliably detected by the GC, and thus these potentials have been omitted from Figure S4. Qualitatively speaking, the FE and CO partial current trends observed for AuPC/ITO are very similar to those reported by Kuhl et al. for polycrystalline Au. More specifically, both polycrystalline samples show a peak FECO at ≈−0.7 VRHE, but higher CO partial currents are reported by Kuhl et al. at moderate overpotentials than our AuPC/ITO (Figure S5).[72] On the other hand, while the CO partial current trends observed by Hori et al. are qualitatively similar to our findings as well, the peak FECO in Hori’s study occurs at a much lower potential of ≈−0.92 VRHE.[9] We hypothesize that these differences may stem from inconsistencies in the experimental

Figure 6

FE of CO (blue) and H2 (red) produced during CO2-reduction on (a) 100 μg.cm– Au aerogel and (b) 100 μg.cm– Au/C (Au basis) in CO2-saturated 0.5 M KHCO3, respectively.

setups used in those studies versus this work; specifically, Kuhl and co-workers used a two-compartment cell similar to ours but with CO2-saturated 0.1 M KHCO3 as the electrolyte, while Hori et al. also used 0.5 M KHCO3, yet in a three-compartment cell with the cathode at its center (and thus under different convection conditions).[46]

On the other hand, the FEs for AuAG and Au/C exhibit quite similar qualitative trends at moderate to low potentials (i.e., ≤ −0.6 VRHE), with CO production being predominant down to ≈−0.8 VRHE and H2 production taking over as the applied potential becomes more negative (see Figure a,b). Additionally, both Au/C and AuAG produced negligible amounts of HCOO– over the whole potential range (< 1%, and hence not shown), which is in accordance with the literature.[73] Moreover, AuAG proves to be more selective than Au/C for CO production, achieving a peak FE of ≈97% for this product at ≈−0.48 VRHE, as compared to the maxima of ≈88% vs ≈83% at ≈−0.67 vs ≈−0.69 VRHE for Au/C vs AuPC/ITO, respectively. Specifically when it comes to the comparison between nanomaterials (i.e., AuAG vs Au/C), this behavior may stem from a suppression of the competing H2-evolution reaction (HER) at high-to-medium potentials on the unsupported Au aerogel, in terms caused by the absence of an HER-active carbon phase in this material. To validate this hypothesis, we performed additional electrochemical measurements in CO2-saturated 0.5 M KHCO3 on a working electrode made of the (Au-free) Vulcan XC-72 carbon black, with a carbon loading of 400 μgC·cm–2 equivalent to the one in the Au/C measurements. Notably, H2 was the only reduction product detected in these measurements, and the H2 partial currents for this C-electrode as well as AuAG and Au/C are displayed in Figure a. As shown in the figure’s inset, these H2 partial currents at potentials ≥ −0.6 VRHE in which the reaction is expected to proceed mostly under kinetic control are 5–8 times higher for Au/C as compared to the Au-free, C-electrode despite the identical carbon loadings in both measurements. Thus, the high H2 partial currents featured by the Au/C sample do not stem from its C-support, contradicting our initial assumption that the higher FEH2 for Au/C vs AuAG at low overpotentials was a result of the presence of a carbon support in the former. Instead, the low H2 selectivity observed for AuAG over Au/C likely stems from an effect of the Au-nanoparticle shape and size on the HER-kinetics, whereby the smaller particles (i.e., higher ECSA) of the Au/C catalyst vs AuAG promote this undesired side-reaction.[14,34]

Figure 7

(a) Partial current densities of H2, normalized with respect to the geometric area, generated during CO2-reduction of AuAG, Au/C, and Vulcan XC-72 in CO2-saturated 0.5 M KHCO3, with the inset showing the zoomed-in picture of partial currents in the kinetically controlled regime and (b) partial current densities of CO, normalized with respect to the ECSA, generated during CO2-reduction on materials reported in this study and in the literature.

Finally, we assess the possible occurrence of particle size/shape effects on the CO2-to-CO SSA of these nanocatalysts by plotting their potential-dependent, ECSA-normalized CO partial currents. To this end, the CO-specific current densities were normalized with respect to the ECSAs determined by the oxide-reduction method in the same electrochemical cell in which the samples were tested for CO2-electroreduction. The ECSA values thus obtained were ≈15 m2·gAu–1 for Au/C and ≈12 m2·gAu–1 for AuAG. While the value obtained for AuAG is quite similar to the one obtained with the same method in 0.1 M H2SO4 in an RDE setup (≈11 m2·gAu–1 ; refer to Table ), a significant decline in the ECSA is observed for Au/C (as compared to ≈21 m2·gAu–1 in the RDE setup) which may be attributed to the presence of 6.6 times more loading and thus, underutilization of the thicker catalyst layer (consistent with our previous observations for a carbon-supported Pd-catalyst).[46]Figure b reveals that AuAG and Au/C feature similar ECSA-normalized, CO partial currents across the potential range within which their performance appears to be under kinetic control (corresponding to potentials ≥ −0.6 VRHE and CO-currents ≤ −0.25 mA·cmAu–2), while larger deviations are observed at lower potentials at which these partial currents reach a plateau, likely caused by a limited supply of reactant stemming from insufficient convection and/or surface poisoning.[13,49,52]

Figure b also includes a comparison with other Au-based nanomaterials reported elsewhere, chosen considering the similarities among the geometry of the electrochemical cell and experimental conditions employed. Among these, Lu and co-workers’ nanoporous gold and Chen and co-workers’ oxide-derived Au-NPs[16,17] feature CO partial currents that are higher than those of Au/C and AuAG when assessed on a geometric area basis (see Figure S5). However, upon normalizing both currents with respect to the Au-ECSA values reported in the same studies, we find that the resulting CO-SSAs are lower or similar to those featured by AuAG and Au/C, respectively (see Figure b). While the nanostructure width of the pore-like and pillar-like nanostructures reported by Kim et al. is 4–10 times higher (≈20 to 30 and ≈50 nm, respectively) than the nanomaterials used in this study, these materials qualitatively resemble the network-like structure of AuAG when observed under the microscope, and hence, are ideal candidates for comparison to our materials.[18] Moreover, as also seen in Figures b and S5, the ECSA- and geometric-surface-normalized partial CO-currents reported by Kim and co-workers for pore-like Au nanostructures that qualitatively resemble the network-like structure of our AuAG are well in accordance with the CO partial current density values reported herein for AuAG and Au/C within the low-current regime associated with kinetic control. On the other hand, significant deviations can again be observed at higher current densities at which mass transport limitations within the catalyst layer start becoming an important factor. Thus, this analysis indicates that the performance of this highly CO-selective AuAG is in line with the trends reported in the past and ultimately points at an absence of particle size effects on the CO2-to-CO selectivity of such Au nanostructures, at least within the size ranges considered in this work and the studies included in the above comparison.

IL-TEM Analysis for the CO2RR

To finalize this work, and inspired by previous studies reporting significant morphological changes in Au-based catalysts in the course of CO2RR ,[74−76] we performed an IL-TEM analysis of the AuAG and Au/C samples before and after CO2-reduction. The experimental setup is similar to that employed during IL-TEM analysis for electrochemical conditioning and the electrochemical protocol mimics that were used during CO2-reduction measurements at −0.5 VRHE, the potential leading to the highest CO FE in the GC cell (see Figure ).The IL-TEM results for Au/C are shown in Figure , whereby the image in Figure a was acquired before the entire electrochemical protocol for CO2-reduction (conditioning CVs and potential holds), and that in Figure b was acquired afterward. At a nanoscale, no significant changes are observed apart from the folding of a part of the carbon scaffold on itself, indicated by the red box. A more drastic change is shown in Figure c, though, whereby significant coalescence of numerous Au NPs in one spot is observed after 1 h CO2-reduction hold at −0.5 VRHE. Such occurrences were not observed in the sample prior to the CO2-reduction test and, while they are relatively rare, this agglomeration is in accordance with what has been observed in the literature for Au NPs deposited on C supports.[77]

Figure 8

IL-TEM analysis performed on Au/C (a) before and (b) after CO2-reduction at −0.5 VRHE in CO2-saturated 0.5 M KHCO3. The red squares serve as a guide to the eye to point out changes that have taken place. (c) TEM image acquired after CO2-reduction at −0.5 VRHE for 1 h, showing significant coalescence of Au NPs around one spot.

Complementarily, the IL-TEM results for AuAG are shown in Figure . A low-magnification view of the aerogel’s structure is presented in Figure a,b, where no significant changes can be seen apart from some modifications in the way the Au network folds on itself, leading to changes in pore sizes (red squares). The red circles in these images are zoomed in and shown in Figure c,d, which prove that the aerogel preserves the ligament nanostructure during the CO2RR. A minor change is highlighted by the red box, where a part of the aerogel in Figure c cannot be seen any more in the post-CO2RR image in Figure d, because it folded into the bulk of the aerogel. Despite this minor change, this general stability of the Au aerogel nanoarchitecture over the commercially available Au NPs embedded in the carbon support during the CO2RR can be attributed to an inherent metastability of the said aerogel, consistent with previous reports in which Au NPs tend to assemble in a nanochain-like structure very similar to this aerogel upon the CO2RR.[75,76]

Figure 9

IL-TEM analysis performed on AuAG (a), (c) before and (b), (d) after CO2-reduction at −0.5 VRHE in CO2-saturated 0.5 M KHCO3. The red squares serve as a guide to the eye to point out changes that have taken place and the red circles in (a) and (b) are zoomed into (c) and (d), respectively.

Conclusions

In summary, we examined polycrystalline Au, Au on carbon, and unsupported Au aerogels for their ECSAs using the surface oxide-reduction and Cu-UPD methods and found that the potential-hold Cu-UPD measurements provide a much more reliable ECSA quantification as compared to the often-employed potential-sweep Cu-UPD method. IL-TEM analysis performed before and after electrochemical conditioning suggested the occurrence of dissolution and redeposition of Au in the nanostructured Au catalysts, but this is shown to not affect the ECSAs drastically. Despite its lower ECSA, the Au aerogel featured similar ECSA-normalized CO partial current densities and recorded higher FE for CO production as compared to Au/C over the entire potential range, with a peak FECO of ≈97% at ≈−0.48 VRHE. The suppression of H2 evolution during the CO2RR exhibited by the Au aerogel as compared to Au/C is proven to be an effect of the Au-nanoparticle size and shape and not an effect of the absence of a carbon support. Finally, IL-TEM investigation before and after the CO2RR suggests that the Au aerogel is stable during the process and does not undergo significant structural changes. By exploring the CO2-reduction prospects of novel Au aerogel architectures, this work provides a basis for benchmarking future electrochemical investigations involving Au-based (unsupported) nanostructures.