Ce Promotion of In₂O₃ for Electrochemical Reduction of CO₂ to Formate

Abstract

In₂O₃ is a promising electrocatalyst for CO₂ electroreduction (CO₂ER) to formate. In₂O₃ nanoparticles doped with Pd, Ni, Co, Zr, and Ce promoters using flame-spray pyrolysis were characterized and evaluated in a gas diffusion electrode for the CO₂ER. Doping results in slight shifts of the In binding energy as probed by XPS, which correlates with a change of the Faradaic efficiency to formate (FE_formate) in the order Ce-doped In₂O₃ > Zr-doped In₂O₃ > In₂O₃ > Pd-doped In₂O₃ > Ni-doped In₂O₃ > Co-doped In₂O₃. However, the differences in CO₂ER performance are caused mainly by the different extent of In₂O₃ reduction. Co-doped In₂O₃ is prone to complete reduction to a stable Co–In alloy with a low FE_formate due to a high hydrogen evolution activity. The stabilizing effect of Ce on In₂O₃ is further demonstrated by an X-ray absorption spectroscopy study of a set of Ce-doped In₂O₃ samples (10, 50, 90 at%), highlighting that reduction of In₂O₃ is suppressed with increasing Ce content. Optimum performance in terms of FE_formate is obtained at a Ce content of 10 at%, which is attributed to the stabilization of In₂O₃ under negative bias up to −2 V. At higher Ce content, less active CeO₂ is formed. The highest FE_formate of 86% observed for In₂O₃ doped with 10 at% Ce, at a current density of 150 mA/cm², compares favorably with a FE_formate of 78% for In₂O₃.

Introduction

CO₂ is considered a circular carbon source for the chemical industry and transport sector. Converting CO₂ to chemicals and fuels using renewable energy sources could replace fossil resources in these sectors, thereby closing the carbon cycle. Electrochemical CO₂ reduction offers a direct pathway to utilize renewable electricity to convert CO₂ into base chemicals, which can be further converted to other chemicals and fuels.¹ Typical products of the CO₂ electroreduction reaction are CO, ethylene, ethanol, and formic acid.² Among these, formic acid and formate have the potential to serve as feedstock for chemicals and as a liquid hydrogen carrier relevant to energy storage in chemical bonds.^3,4 Electrochemical CO₂ reduction to formate can achieve high Faradaic efficiency at industrially relevant current densities, making it one of the most promising electrochemical processes for commercialization.^1,5 Typical electrocatalysts for CO₂ reduction comprise the metals Sn, In, and Bi.^2,5,6 Such catalysts reach Faradaic efficiencies close to 100% at current densities >100 mA/cm².⁵

The reaction mechanism of CO₂ electroreduction (CO₂ER) to formic acid involves an oxygen-bound formate intermediate on the catalytic surface.⁷ The carbonyl pathway that leads to undesired CO byproduct competes with this formate pathway. A high oxophilicity of a metal catalyst, therefore, stabilizes the formate intermediate, which is thought to increase the selectivity of the formate pathway over the carbonyl one.⁸ On the other hand, metal oxides are widely believed to be the most active for CO₂ER to formate. Still, their stability against reduction to the metal under the reducing CO₂ER conditions is heavily debated.⁹⁻¹³ Several theoretical studies showed that defects in metal oxides, such as oxygen vacancies, can act as CO₂ adsorption sites.^14,15 Alternatively, it is proposed that surface hydroxyls can react with CO₂ to form a carbonate intermediate before^15,16 or instead of the formate intermediate.¹⁷ A major benefit of metal oxides over metallic catalysts is the higher kinetic barrier of the former for the main competing hydrogen evolution reaction (HER).^18,19 Experimental studies to understand the role of oxide surfaces indicate that oxygen vacancies and hydroxyls are involved in reducing CO₂ to formate. Nevertheless, these studies are primarily conducted at low current densities in H-cell configurations employed in conventional electrochemical studies, which are, for instance, more conducive to coupling with surface-sensitive characterization techniques such as infrared spectroscopy.^{16,17,20−22} Industrial applications require high current densities, implying operation at more negative potentials. Convincing experimental evidence for the importance of oxides as the main catalyst during the CO₂ER at industrially relevant conditions is lacking. Broekmann and co-workers utilized operando Raman, X-ray absorption spectroscopy (XAS), and X-ray diffraction (XRD) to study the impact of experimental conditions at which Sn- and Bi-oxide phases remain stable utilizing a gas diffusion electrode (GDE) configuration.^11,13,23 They observed a strong correlation between the presence of Sn-oxide and Bi-oxy-carbonate phases and high Faradaic efficiency (FE) toward formate. Such a study has not been undertaken yet for In-based CO₂ER catalysts. Wang et al. showed by operando XAS that a mixed In–Sn-oxide shell was maintained on an In–Sn alloy electrocatalyst during CO₂ER at a potential up to −1.2 V vs RHE in an H-cell configuration.²⁴

Here, we report on In₂O₃-based catalysts for CO₂ electroreduction to formate. In₂O₃ nanoparticles doped with Pd, Ni, Co, Ce, and Zr were obtained using flame synthesis. It has been demonstrated that flame spray pyrolysis (FSP) is suitable to incorporate other elements in In₂O₃ nanoparticles.²⁵ Doping of the In₂O₃ with these transition metals can modify the catalytic properties of the In₂O₃ surface and its stability against metal reduction, thereby impacting the CO₂ER to formate in terms of FE and stability. Pd and Co were chosen because of their activity toward reduction of CO₂ to formate at low potential.^26,27 Ce and Zr were selected, as their oxides form strong interactions with In₂O₃,²⁸ which can stabilize their dispersion and impact oxygen vacancy formation as reported to be beneficial for thermal CO₂ hydrogenation.^25,29 During our investigations, we found that Zr and especially Ce doping positively impacted the FE to formate. This led us to investigate in more detail mixed oxides of Ce and In prepared by FSP by varying the Ce content from 0% to 90%. The catalytic performance of the metal oxides is evaluated using gas diffusion electrodes at industrially relevant current densities in the range of 100–200 mA/cm². The samples are investigated by electron microscopy, XRD, XAS, and X-ray photoelectron spectroscopy (XPS) to elucidate the promoters’ influence on the catalytic activity and stability of In₂O₃ as a catalyst for the CO₂ER. This work is also available as part of a PhD thesis.³⁰

Methods

Catalyst Preparation

In₂O₃ Nanoparticles

Nanoparticles (NPs) of In₂O₃ were prepared by flame spray pyrolysis (FSP) using a Tethis NPS10 setup. A 0.15 M In(NO)₃ solution was prepared by dissolving an appropriate amount In(NO)₃·5H₂O (99%, Alfa Aesar) in an equivolumetric mixture of ethanol (99.9%, Biosolve) and 2-ethylhexanoic acid (99%, Sigma-Aldrich). This synthesis was modified by introducing dopants during the FSP preparation, namely 5 wt % Co(NO₃)₂·6H₂O (98%, Sigma-Aldrich), 5 wt % Ni(NO₃)₂·6H₂O (98.5%, Sigma-Aldrich), 5 wt % Pd(OCOCH₃)₂ (98%, Sigma-Aldrich), 5 wt % Zr(acac)₂ (99.9%, Sigma-Aldrich), and 10–90 at% Ce(acac)₂ (acac = acetylacetonate, 99.9%, Sigma-Aldrich). For comparison with XPS analysis, the weight-based concentration of the dopant is converted to the atomic concentration with respect to total metal content as listed in Table 1 and Table 2. The solution was then injected into the nozzle of the FSP setup using an injection rate of 5 mL/min. The flame was fed with a flow of 3.0 L/min O₂ and 1.5 L/min CH₄, to which a dispersion flow of 5.0 L/min O₂ was added. The solid particles were collected from a quartz filter placed downstream of the flame region. The (doped) In₂O₃ solids were sieved over a 250 μm steel sieve. The Ce–In-oxide catalysts are denoted by their atomic Ce content, i.e., Ce(50)-In₂O₃ having a Ce content of roughly 50 at% with respect to total of In and Ce content.

Table 1. Physicochemical Properties of (Doped) In₂O₃ Nanoparticles Synthesized by FSP.

	Loading dopant			Particle size		XPS analysis
Catalyst	Nominal (wt%)a	Nominal (at%)a	Surface (at%)b	dTEM (nm)	dXRD (nm)	In2O3 (at%)	In(OH)3/defects (at%)
In2O3	–	–	–	6.2 ± 1.4	9.2	87	13
Zr–In2O3	4.9	7.5	5.7	6.3 ± 1.5	8.3	85	15
Ce–In2O3	9.6	9.7	11	6.0 ± 1.6	8.2	94	6
Pd–In2O3	5.3	6.8	7.7	5.8 ± 1.7	9.0	92	8
Ni–In2O3	5.0	11.0	16	7.3 ± 1.5	9.1	82	18
Co–In2O3	5.0	11.1	11	6.7 ± 1.6	9.3	64	36

^aICP elemental analysis.

^bXPS surface analysis.

Table 2. Physicochemical Properties of Mixed Ce–In-Oxide Nanoparticles Synthesized by FSP.

	Ce loading		Particle size			XPS analysis
Catalyst	Nominal (at%)1	Surface (at%)2	dTEM (nm)	dXRD In2O3 (nm)	dXRD CeO2 (nm)	In2O3 (at%)	In(OH)3/defect (at%)	Ce4+/Ce3+ ratio
In2O3	–	–	6.8 ± 2.0	8.6	–	80	20	–
Ce(10)-In2O3	11.0	8.0	6.2 ± 1.5	7.8	–	92	8	3.2
Ce(50)-In2O3	51.5	38	6.6 ± 1.7	–	4.6	89	11	4.2
Ce(90)-In2O3	92.6	83	6.0 ± 2.0	–	7.1	55	45	5.3

Gas Diffusion Electrode (GDE) Preparation

A suspension of the catalyst was prepared by weighing appropriate amounts of (doped) In₂O₃, Vulcan XC-72R activated carbon (AC) as support and PTFE nanoparticles (Sigma-Aldrich, 1 μm) and dispersing these solids by sonication in 3 mL isopropanol (99.9%, Sigma-Aldrich) and 1 mL ultrapure water. Nafion ionomer solution (D-520, dispersion, 5% w/w in water and isopropanol, Alfa Aesar) was added to reach a mixture containing 20 wt % Nafion with respect to solid particles. The solution was sonicated for ∼30 min. The ink was sprayed on a commercial gas diffusion layer (GDL, ELAT LT1400W, FuelCellStore) with a spray gun. The total loading of the catalyst on the GDE, as determined by weighing the GDL before and after catalyst deposition, amounted to 0.30 ± 0.06 mg/cm² for the (doped) In₂O₃ nanoparticles and 0.47 ± 0.07 mg/cm² for the mixed Ce–In catalysts. For operando XAS analysis, a GDE was prepared without activated carbon and PTFE particles employing a higher catalyst loading of 1.04 ± 0.12 mg/cm² on a commercial GDL (Sigracet 22BB, Ion-Power).

Characterization

Inductively Coupled Plasma Optical Emission Spectrometry (ICP-OES)

The elemental composition of the FSP-synthesized nanoparticles was determined using an AMETEK ICP optical emission spectrometer (Spectroblue) with axial plasma viewing, equipped with a free-running 27.12 MHz generater (1400 W). Around 25 mg of the sample was dissolved in 5 mL concentrated H₂SO₄ at 220 °C. After cooling, the solution was diluted to 50 mL with demineralized water. Two solutions were prepared to perform the measurements in duplicate. A calibration line with concentrations between 0 to 20 mg/L In, 0 to 10 mg/L Ce, 0 to 5 mg/L Zr, and 0 to 3 mg/L for Co and Ni was used. A second calibration line with concentrations between 0 and 5 mg/L Pd was used due to the overlap of the emission lines of Pd and In.

X-ray Photoelectron Spectroscopy (XPS)

XPS was performed with a K-Alpha XPS apparatus (Thermo Scientific) using an Al anode (Al Kα monochromatic irradiation, 1486.6 eV) operating at 72 W. A spot size of 400 μm was used. Survey and core level spectra were recorded at pass energies of 200 and 50 eV, respectively. The background pressure inside the analysis chamber was kept below 8 × 10^–8 mbar. During measurements, a maximum pressure of 3 × 10^–7 mbar was observed due to using low energy Ar⁺ ions for charge neutralization. Catalyst powders were pelletized before XPS analysis to prevent the charging of the insulating In₂O₃ particles. All spectra were energy corrected by the U‴ component at 916.7 eV of Ce.³¹ After energy correction of the Ce-containing catalysts, the resulting C 1s peak position of adventitious carbon was used to correct the energy scale for the samples that did not contain Ce. The core level spectra were deconvoluted using a Shirley background and using GL(30) line shapes for most contributions. The metallic In peak was fitted using an LA(1.1,1.7,3) line shape obtained from an In reference sample, which was etched in the vacuum chamber of the XPS apparatus (Figure S8a). The Ce 3d core-line spectra were fitted according to the literature.³¹⁻³³

X-ray Absorption Spectroscopy (XAS)

Operando XAS analysis was performed at the ROCK beamline at the SOLEIL synchrotron (Paris, France). Data was collected in fluorescence mode at the In K-edge (27940 eV). An electrochemical cell developed in-house by SOLEIL was used. The GDE prepared as described above was used as the working electrode. A gas flow chamber with a Kapton window was placed at the back of the working electrode, while the catholyte chamber was in front of the working electrode. The cell was positioned at a 39° angle with respect to the incoming X-ray beam, and the fluorescence signal was measured from the back of the electrode by an avalanche-photodiode fluorescence detector positioned perpendicular to the beam. A Nafion proton exchange membrane (N324, Ion Power) was used to separate the catholyte and anolyte chambers. Pt foil and Ag/AgCl (6 mm, redoxme, measured at 0.23 V vs SHE at pH 1) electrodes were used as counter and reference electrodes, respectively. An IviumStat potentiostat was used to set the potential and record the electrochemical response of the working electrode. The anolyte (0.5 M H₂SO₄) was pumped using a peristaltic pump at a flow of 12 mL/min. Catholyte (0.5 M KHCO₃) was introduced in the catholyte chamber. A mass spectrometer (MKS Cirrus LM99 Analyzer) was used to detect CO₂, H₂ and CO in the gas exit line from the gas chamber at the back of the GDE. Catholyte was not pumped around to avoid leakage through the GDE into the gas exit toward the mass spectrometer. The catholyte was refreshed in each experiment. CO₂ was passed at the back of the GDE at a flow rate of 5 mL/min (flow-by mode). A spectrum was recorded every 0.5 s. As 10 spectra were averaged to increase the signal-to-noise ratio, the temporal resolution of the measurements was 5 s. Energy calibration was done using an In foil. The X-ray absorption near-edge structure (XANES) data were background-subtracted, normalized, and fitted using linear combination analysis with XAS viewer as implemented in the Larch software package. The extended X-ray adsorption fine structure (EXAFS) analysis on k³-weighted In K-edge data was also performed with XAS viewer. Scattering paths were calculated with FEFF8 using crystal structures of In₂O₃, In and In(OH)₃. The amplitude reduction factor S₀² (0.77) used for fitting the Ce–In-oxide EXAFS spectra was determined by fitting the two In–In scattering paths of the first shell of reference In foil, fixing the coordination numbers to, respectively, 4 and 8 (Table S2).

X-ray Diffraction (XRD)

XRD was performed using a Bruker D2 Phaser diffractometer using Cu Kα radiation (1.5406 Å). The XRD diffractograms of catalyst powders were recorded between 2θ = 10° and 90° at a scan rate of 1 s/step and a step size of 0.02°. XRD patterns on a GDE were obtained at a scan rate of 2 s/step and a step size of 0.01°. The crystallite size was calculated using the Scherrer equation as implemented in the DIFFRAC.EVA software package.

Wide Angle X-ray Scattering (WAXS)

Ex situ WAXS measurements were performed at beamline ID31 of the ESRF synchrotron (Grenoble, France). The incident photon energy was 75 keV (l = 0.0165 nm) using a Pilatus CdTe 2 M detector set up in the Debye–Scherrer geometry. Kapton tubes were filled with powder samples and sealed with bee wax before analysis. The WAXS peaks were fitted with a Voigt function to determine the full width at half-maximum (fwhm). The average crystal size was calculated using the Scherrer equation. Rietveld refinement was performed using GSAS-II software. In₂O₃ and CeO₂ crystal structures were used to fit the diffraction patterns of the as-prepared catalysts. The patterns were fitted between 2θ = 1.5 and 9. The background was fitted using a chebyschev-1 function with 3 coefficients. Instrument parameters were determined by fitting a reference CeO₂ sample and fixed for all samples. Sample displacement parameters were determined by fitting the pure In₂O₃ sample and fixed for all Ce-doped In₂O₃ samples. The fitting was performed by subsequent fitting of the phase fractions, micro strain and lattice parameter in the given order. After several iterations (>3), the size and lattice parameters were fitted simultaneously. The fit of the Ce(50)-In₂O₃ sample was optimized by subsequent fitting of the atom fractions, atom displacement and thermal displacement.

Transmission Electron Microscopy (TEM)

Transmission electron images were obtained using an FEI Tecnai (type Sphera) instrument operating at an acceleration voltage of 200 kV and an FEI cryoTITAN instrument operating at an acceleration voltage of 300 kV. Catalyst particles were dispersed in ethanol via ultrasonication and deposited on a holey Cu support grid. The particle size and lattice constants of the catalysts were obtained using the software ImageJ.

Scanning Transmission Electron Microscopy Energy-Dispersive X-ray Spectroscopy (STEM-EDX)

The distribution of In and Ce in the In₂O₃–Ce samples was studied using STEM-EDX. Measurements were carried out on a FEI cubed Cs-corrected Titan operating at 300 kV. In₂O₃–Ce samples were dispersed in ethanol via ultrasonication and deposited on a Cu support grid with a holey carbon film. Elemental analysis was done with an Oxford Instruments EDX detector X-Max^N 100TLE.

Electrochemical Experiments

Electrocatalytic CO₂ Reduction

In flow-through mode, CO₂ electroreduction was performed in a flow cell with an electrode with a geometric surface area of 1 cm2 (Micro Flow Cell, ElectroCell). A Nafion proton exchange membrane (N324, Ion Power) separated the catholyte and anolyte chambers. The Nafion membrane was pretreated in a 10 vol % H₂O₂ solution (diluted from 33 vol % solution, VWR Chemicals) at 80 °C for 1 h and subsequently in a 3 M H₂SO₄ (≥95%, Merck) solution (80 °C, 1 h). An Autolab 302N potentiostat was used for electrochemical analysis for cyclic voltammetry and chrono-potentiometry measurements. A three-electrode system was used with a Pt mesh as a counter electrode and a leakless Ag/AgCl electrode (1.6 mm, EDAQ, measured at 0.37 V vs SHE at pH 1) as the reference. The as-prepared GDEs were used as working electrodes. Potentials are specified against the Ag/AgCl reference electrode. As catholyte, 0.5 M KHCO₃ solution was prepared by dissolving high purity KHCO₃ (analysis grade, Merck) in ultrapure water (18.2 MΩ·cm). As anolyte, 0.5 M H₂SO₄ solution was used by diluting H₂SO₄ (>95%, Sigma-Aldrich) in ultrapure water. The catholyte solution was degassed by N₂ and saturated with CO₂ before use. The backside of the GDE was supplied with a gaseous flow of 10 mL/min CO₂. The catholyte and anolyte liquid flows were set at 50 mL/min. Cyclic voltammetry and galvanostatic measurements were performed for 1 h at a current density of 200 mA/cm² during and after which the catholyte was analyzed for products. Faradaic efficiencies toward formate were calculated from the ratio of charge that is needed to produce the obtained formate over the total amount of charge that was consumed at the cathode:

with [FA] being the concentration of formate in the cathode compartment, V the catholyte volume, n the number of consumed electrons (two) per formate molecule, F Faraday’s constant, J the current density, and t time. The experiments were repeated at least twice with a fresh GDE with the same loading to estimate the experimental error. The current densities are reported based on the exposed geometrical surface area of the GDE. This study uses the Faradaic efficiency and current density as major indicators for the catalyst performance and is used to analyze and optimize the cathodic catalyst. As pointed out by, e.g., the group of Seger and co-workers,^34,35 for industrial applications, the system also needs to be optimized for CO₂ conversion efficiency, which is considered outside the scope of this work.

High-Performance Liquid Chromatography (HPLC)

Dissolved reduction products were analyzed by HPLC. A Shimadzu HPLC containing a Polar C18 column (Luna Omega 3 μm, 150 × 3.0 mm) was used to determine the formic acid concentration. A potassium phosphate buffer (pH 2.0) was used as eluent, which was prepared by dissolving 1.4150 g KH₂PO₄ (≥99%, Sigma-Aldrich) and 1.6854 g H₃PO₄ solution (85%, Merck) in 1.0 L ultrapure water. The column oven temperature was set at 30 °C and the UV–vis detector cell temperature at 40 °C. Samples were taken from the catholyte by diluting 100 μL of electrolyte to 1.0 mL with 900 μL 0.1 M H₃PO₄ solution (1:10) to convert bicarbonate to CO₂. An injection volume of 10 μL and a total eluent flow of 500 μL/min was used. The retention time of the formic acid peak was about 2 min. Reference solutions from 263.434 mM down to 5.268 mM formic acid (≥98%, Sigma-Aldrich) were used to obtain a calibration curve for HPLC analysis. Two stock formic acid solutions (263.434 mM and 105.60 mM) were prepared by diluting 1.0 mL formic acid to 100 and 250 mL in 0.5 M KHCO₃ in ultrapure water and were further diluted for reference solutions. The calibration solutions were also prepared by 1:10 dilution in 0.1 M H₃PO₄ (85%, Merck) and analyzed in triplicate. Peak integration was performed using the ICIS algorithm included in the Shimadzu software package.

¹H Nuclear Magnetic Resonance Spectroscopy (¹H NMR)

Dissolved reaction products were analyzed by ¹H NMR spectroscopy. NMR spectra were recorded using a 400 MHz Bruker spectrometer. An aliquot of 450 μL was taken directly from the catholyte, followed by the addition of 50 μL of D₂O (99.9%, Sigma-Aldrich) containing 10 mM DMSO (>99.9%, Biosolve) and 50 mM phenol (≥99%, Sigma-Aldrich) as internal standards. Standard solutions for HPLC calibration were used to verify the DMSO and phenol concentrations. The NMR signal due to protons of water was suppressed by a solvent presaturation sequence. The signal-to-noise ratio was improved by averaging 48 scans with a 16 s delay time. The resulting spectra were calibrated with reference to the DMSO peak position, set at 2.600 ppm in accordance with Kuhl et al.³⁶ Then, the peak belonging to the carbon bonded proton in formic acid was observed at 8.33 ppm.

Gas Chromatography

Gaseous products were analyzed online using a TRACE 1300 gas chromatograph (Thermo Fisher). The permanent gases (H₂, CO, CO₂) and the hydrocarbons (methane, ethylene) were analyzed on separate channels. A Hayesep Q precolumn and a Shin-Carbon ST column with TCD were used for permanent gas analysis. An Al₂O₃/KCl column with FID was used to separate the light hydrocarbons.

Results and Discussion

Catalyst Characterization

In₂O₃ nanoparticles and In₂O₃ nanoparticles doped with Zr, Ce, Pd, Ni, and Co were obtained by FSP. Their physicochemical properties are summarized in Table 1.

The Zr, Pd, Ni and Co contents were targeted at 5 wt % with respect to the total weight of the particles, while the content of the Ce dopant was aimed at 10 wt %. The atomic and weight loadings of the promoters are given in Table 1. The average size of the In₂O₃ nanoparticles determined by TEM was 6.2 ± 1.4 nm (Figure S1). The size of the doped In₂O₃ nanoparticles was close to that of pure In₂O₃ with the average particle size falling in the range of 5.8–7.3 nm with a nearly dopant-independent size distribution. This shows that doping did not strongly affect the particle size (Figure S1). Both In₂O₃ and doped In₂O₃ particles are made up of cubic In₂O₃ as follows from the XRD patterns in Figure S2. The dopants are highly dispersed, as seen from the absence of other diffraction lines. This suggests that the promoters are doped into the In₂O₃ lattice, albeit they can also be present as very small oxide particles.³⁷ The shift of the In₂O₃ diffraction lines for the Ce–In₂O₃ sample toward lower angles indicates lattice expansion by incorporating Ce ions in the In₂O₃ crystal structure. The radii of Ce⁴⁺ and Ce³⁺ ions of respectively 101 and 115 pm are larger than the radius of 81 pm of In³⁺. The cations of the other dopants in their most likely oxidation states are too similar to that of In³⁺ to judge their incorporation in the In₂O₃ lattice from XRD.^25,38 The average crystal sizes of the (doped) In₂O₃ nanoparticles calculated by the Scherrer equation are also listed in Table 1. The crystallite sizes of the In₂O₃ particles doped with Pd, Ni and Co are very similar to that of In₂O₃ (∼9 nm), while those for the Zr- and Ce-doped In₂O₃ particles are slightly smaller (∼8 nm).

XPS was used to analyze the surface composition of the samples. The surface concentration of the dopants is determined from the core level XPS spectra of In and the dopants. The results are listed in Table 1. The surfaces of Pd- and Ni-doped In₂O₃ are slightly enriched in dopant compared to their nominal contents. While the Zr-doped In₂O₃ nanoparticles contain slightly less Zr at the surface than the nominal content, the bulk and surface dopant contents for Ce- and Co-doped In₂O₃ are nearly the same. Overall, the minor differences observed point to a high dispersion of the dopant with some indications of dopant enrichment in the surface for Ni and Pd. For a similarly prepared Ni-doped In₂O₃ catalyst used in CO₂ hydrogenation, it was found that Ni is highly dispersed as Ni cations in the bulk and at the surface of In₂O₃.³⁹ The group of Pérez-Ramirez also used FSP to prepare a range of metal-doped In₂O₃ nanoparticles and found that most metal dopants were evenly distributed in the bulk and surface of In₂O₃.²⁵Figure 1 shows the In 3d XPS spectra of the In₂O₃ and doped In₂O₃ samples. The spectra consist of two main features representing the In 3d_5/2 and In 3d_3/2 states. Two contributions are included to fit these In spectra. The main peak at an In 3d_5/2 binding energy of 444.4 eV is attributed to In₂O₃,⁴⁰ The shoulder, with a higher binding energy of 445.4 eV, is due to In-hydroxide or defects in In₂O₃ (Figure S3).⁴⁰ Doping with Ce and Zr results in a small shift (0.2 eV) of the binding energy of the In 3d_5/2 peaks. The main In 3d_5/2 binding energy shifts from 444.4 to 444.6 eV, as was also observed for Zr doping by Pinheiro Araújo and co-workers.^41,42 This shift indicates a higher effective charge on the In ions, which can be caused by transfer of electrons from In to Ce and Zr.⁴³ The contribution of In(OH)₃/In₂O₃-defects does not change upon Zr doping and decreases for Ce. An opposite shift in the binding energy is observed when In₂O₃ is doped with Ni and Co, resulting in a shift of the main In 3d_5/2 peak to respectively 444.2 and 443.8 eV. The Co-doped In₂O₃ sample contains a higher contribution of In(OH)₃/In₂O₃-defects of 36% as compared to the contribution of 13% for In₂O₃. This, together with the relatively large binding energy shift of In and the homogeneous distribution of Co in the In₂O₃ particles, indicates the well-mixed nature of the Co–In-oxides as observed previously for FSP-prepared Co-doped In₂O₃.²⁵ There are no significant changes in the In 3d spectra upon introducing Pd in In₂O₃. The same trends in binding energies are observed when comparing the O 1s spectra (Figure S4). The O 1s binding energy of lattice oxygen in In₂O₃ is observed at 529.9 eV and shifts to 530.1 eV upon doping with Ce and Zr, while a slight shift to lower energies of 529.7 and 529.5 eV is observed for the Ni- and Co-doped In₂O₃, respectively.

Figure 1.

XPS spectra of the In core-line region of the FSP-synthesized doped In₂O₃ catalysts. All spectra were energy corrected by the U‴ component at 916.7 eV of cerium. The resulting C 1s binding energy (measured at 285.1 eV) was used to energy correct spectra of samples that contain no cerium.

The XPS spectra of the dopants were deconvoluted, as shown in Figure S5. The Ce content in the In₂O₃ particles is 9.7 at%, close to the nominal Ce content (Table 1). Deconvolution of the Ce 3d spectrum of the Ce-doped In₂O₃ using contributions of Ce³⁺ and Ce⁴⁺ reveals a substantial amount of 24% Ce³⁺ (Figure S5a).^37,44 The existence of Ce in the +3 oxidation state suggests the presence of oxygen defects.⁴⁵ Ce³⁺ might be included in the lattice of In₂O₃ as supported by the shift of the XRD diffraction lines in Figure S2.^46,47 Ce⁴⁺ could also substitute In³⁺ in the lattice of In₂O₃,⁴⁶ but the presence of a significant amount of Ce⁴⁺ can most likely be attributed to the presence of small CeO₂ particles, which can also generate Ce³⁺.⁴⁵ The Co 2p_3/2 region was fitted using a model from the literature (Figure S5d).^48,49 The satellite feature at 786 eV indicates the predominant presence of CoO and Co(OH)₂ and, in line with this, the Co 2p_3/2 region could be fitted by contributions due to Co²⁺ at 782 and 780 eV. A minor feature at a lower binding energy of 779.3 eV can be attributed to fully oxidized Co₃O₄.⁴⁸ These results are in line with previous reports where Co forms small CoO_x clusters and dopants in FSP-synthesized In₂O₃.²⁵ The Ni 2p_3/2 core-line region spectrum contains a main peak at 855.5 eV and a broad satellite feature at 861 eV (Figure S5c). A small feature at 854 eV indicates the minor presence of NiO.³⁹ The main peak at 855.5 eV has been attributed to Ni³⁺ before,⁵⁰ but this assignment is not unequivocal. Based on reference spectra, it could also be assigned to Ni(OH)₂,^50,51 but a more likely explanation is that it represents small NiO clusters in strong interaction with In₂O₃.³⁹ The Pd 3d core-line region spectrum reveals a large variety of Pd species. Besides the presence of PdO and a feature assigned to oxygen-deficient PdO, some metallic Pd is also present (Figure S5e).⁵²⁻⁵⁴ Overall, XPS indicates that the dopants cause minor (Ce–In₂O₃) to significant (Co–In₂O₃) shifts in the In 3d binding energy, indicative of their introduction in the In₂O₃ lattice. Nevertheless, it is also clear that part of the dopants are present as separate oxide phases. Regarding doping, Zr and Ce cause a slight increase of the In 3d binding energy for In₂O₃ of 0.2 eV, reflecting a slightly higher positive charge of In³⁺.⁴¹ Doping with Ni and Co results in an opposite shift toward lower binding energies.

CO₂ Electroreduction

The GDEs containing the In₂O₃-based electrocatalysts, carbon and PTFE were evaluated for their catalytic performance in the CO₂ER in a flow cell. The measurements were carried out at (geometrical) current densities of 100, 150, and 200 mA/cm² to compare the catalyst at industrially relevant conditions. In a previous study, the electrode configuration employed here was optimized to avoid mass transport limitations and ensure sufficient catalytic surface area.⁵⁵ We earlier showed that, above a catalyst loading of 0.17 mg/cm², a current density of 200 mA/cm² can be obtained without affecting the Faradaic efficiency toward formate (FE_formate). Here, using an electrode with a catalyst loading of 0.5 mg/cm², we could reach a current density of up to 300 mA/cm² while maintaining a constant FE_formate (Figure S7). This indicates that CO₂ transport is not limiting the FE_formate at the employed current densities. The main products observed here were formate, CO, H₂, and hydrocarbons. The FE of hydrocarbons, mainly CH₄, was always less than 1%. Figure 2 shows the FE toward these products as a function of the current density. Figure S6 shows the potentials required to maintain these current densities. At current densities of 100 and 150 mA/cm², the FE_formate reaches ∼86% for In₂O₃ and the Zr- and Ce-doped In₂O₃. For Pd- and Ni-doped In₂O₃, the FE_formate is slightly lower, namely at 79% and 74%, respectively, at these current densities. The FE_formate for Co-doped In₂O₃ is very low at 37% due to the much higher FE to H₂. Increasing the current density to 200 mA/cm² results in a minor but significant difference in FE_formate between Zr- and Ce-doped In₂O₃ (85 and 86%, respectively) compared to In₂O₃ (81%). The presence of Pd shows, at all current densities, the highest FE to CO, which is most likely due to the formation of metallic Pd or Pd-hydrides, which are active in CO₂ reduction to CO instead of formate.^56,57 The data show a trend of increasing FE toward H₂ and CO when a lower potential is applied to maintain a given current density (Figure S6). This can be attributed to a lower overpotential of the competing HER and CO₂ to CO reduction reaction for the samples that contain Pd, Ni, and Co.

Figure 2.

Faradaic efficiency as a function of current density on doped In₂O₃ catalysts after 1 h CO₂ER in 0.5 M KHCO₃. Catalysts nanoparticles are deposited on a GDE (0.30 ± 0.06 mg/cm²). Experiments were done in triplicate (Note: the larger deviation from 100% in the total Faradaic balance for the Co-doped In₂O₃ is caused by an underestimation of the hydrogen concentration in the exit gas stream due to the high quantities of hydrogen exceeding the linear region of hydrogen detection in the GC setup).

Surface Characterization of Ce- and Co-Doped In₂O₃

Based on the galvanostatic measurements at a current density of 200 mA/cm², the FE_formate decreases in the order Ce-doped In₂O₃ > Zr-doped In₂O₃ > In₂O₃ > Pd-doped In₂O₃ > Ni-doped In₂O₃ > Co-doped In₂O₃. This order trends with the minor shift in the binding energy of the main In³⁺ 3d_5/2 state, suggesting that the impact of dopants on the electronic state of In cations at the surface correlates with the CO₂ER to formate. Care must be taken, however, as some of the dopants can be active for electrochemical CO₂ reduction, especially for the HER. Moreover, some of the compositions might reduce to the metallic form during the CO₂ER. To study this aspect further, we characterized three samples in more detail by cyclic voltammetry (CV).

Figure 3 compares the CVs of In₂O₃ doped with Ce and Co with the one of In₂O₃. The CV of In₂O₃ contains a cathodic peak at −1.0 V vs Ag/AgCl, which corresponds to the reduction of In³⁺ to metallic In, and an anodic peak at −0.9 V vs Ag/AgCl due to reoxidation of metallic In to In³⁺.²¹ As such, these potentials represent the reversible character of the In³⁺/In⁰ redox couple in this sample. The oxidation peak in Co–In₂O₃ is shifted to a significantly higher anodic potential of −0.4 V vs Ag/AgCl, while the reduction is hardly affected. Based on the observation of the identical location of the reduction peak, and the absence of other peaks, we speculate that the CV still represents the In³⁺/In⁰ redox couple and that doping Co in In₂O₃ makes the In³⁺/In⁰ redox couple significantly less reversible. This is attributed to Co doping of the metallic In phase obtained upon reduction. This results in a higher barrier for oxidation of Co-doped In compared to monometallic In. We explain this by the formation of a Co–In alloy for which the In oxidation potential shifts closer to the oxidation potential of Co metal to Co²⁺.⁵⁸ On the other hand, XPS analysis before and after CO₂ER reveals a slight increase in Co-concentration from 12 to 15 at% (Figure S8d). This can indicate that Co migrates to the surface of the particle, which can explain the shift of the oxidation peak closer to that of pure Co. As such, the formation of a monolayer of Co cannot be fully excluded. Doping with Ce does not affect the oxidation and reduction potentials of In₂O₃.

Figure 3.

Cyclic voltammetry in 0.5 M KCHO₃ saturated with CO₂, on GDEs with In₂O₃ (0.35 mg/cm², blue line), Ce–In₂O₃ (0.46 mg/cm², orange line), and Co–In₂O₃ (0.27 mg/cm², red line). Scan rate: 100 mV/s.

XPS before and after CO₂ER was used to analyze possible changes in the surface composition. The electrocatalysts used for CO₂ER were transferred through air to the XPS apparatus, which might lead to oxidation of the surface. Figure 4 shows the In 3d spectra for the fresh and used In₂O₃, Ce–In₂O₃, and Co–In₂O₃ GDEs, which were used to determine the surface composition of the samples (Table S1). The surface concentration of Ce decreased from 10.1 to 5.6 at% after 1 h CO₂ER, indicating that there was some migration of In to the surface of the nanoparticles during the electrochemical reaction or dissolution of Ce from the surface. On the contrary, the surface concentration of Co increased from 12 to 15 at% (Figure S8c, d). The enrichment in Co of the surface might indicate the formation of a Co-rich alloy close to the surface during CO₂ER.

Figure 4.

XPS analysis of the In 3d core-line region before and after CO₂ER using a GDE with pure In₂O₃ (a, blue), Ce-doped In₂O₃ (b, orange), and Co-doped In₂O₃ (c, red) GDEs. After electrochemical CO₂ reduction, the electrode is removed from the electrochemical flow cell and transferred through air to the XPS apparatus.

The spectra were deconvoluted to determine possible changes in the oxidation state (Figure 4). The In XPS spectra of the used samples contain an additional metallic contribution at an In 3d_5/2 binding energy of 443.8 eV (Figure S8a). The In⁰ contribution is slightly lower (7.2 at%) for used Ce–In₂O₃ than for used In₂O₃ (10.6 at%) as shown in Table S1. The In 3d spectrum of used Co–In₂O₃ sample differs the most from the fresh sample among the set. The contribution of the In³⁺ due to In₂O₃ has significantly decreased, while the main feature is at 444.1 eV. Most likely, this reflects the reduction of a significant fraction of In to metallic In. The higher binding energy compared to metallic In can be due to the doping with the slightly more electronegative Co in the lattice. XRD analysis of the GDE after CO₂ER hints at the formation of some metallic In (Figure S8b). The diffractogram does not provide enough resolution to conclude on the presence of In₂O₃ or a CoIn₃ alloy, which is due to overlap of peaks from abundantly present PTFE.⁵⁹

XPS indicates that the oxidation state of the In₂O₃ phase is influenced by the dopants during CO₂ER. Ce might inhibit the reduction of the In₂O₃ phase as observed from the lower metal In content after CO₂ER, while In₂O₃ reduction is promoted by Co. Metallic Co could be an active catalyst for HER.^60,61 The relatively large surface contribution of Co on the surface, possibly forming a Co–In alloy, most likely contributes to the increased FE toward H₂ observed in Figure 2. However, the increased reduction of In₂O₃ to metallic In caused by Co could also negatively impact the adsorption of the oxygen-bound formate intermediate, decreasing the FE_formate.¹⁰

Ce–In Oxide Catalysts

To further investigate the effect of Ce dopant on In₂O₃ and its activity in CO₂ER toward formate, a range of new catalysts were synthesized by FSP with Ce contents of 0, 10, 50, and 90 at%. The main physicochemical properties of the Ce–In₂O₃ catalysts are given in Table 2. The catalysts are denoted by their atomic Ce content, i.e., Ce(50)-In₂O₃ having a Ce content of roughly 50 at%.

TEM was used to determine the shape and size of the Ce–In₂O₃ nanoparticles, while STEM-EDX revealed the nanoscale distribution of Ce with respect to In. The TEM images and particle size distributions given in Figure 5a and 5b show that the Ce–In₂O₃ samples are predominantly made up of particles with a size of 6–7 nm as also observed for the other doped In₂O₃ catalysts. This shows that the particle size for these samples is hardly affected by the composition, with Ce-rich compositions giving nearly the same average particle size. The shape of the In₂O₃ particles is mostly spherical, although some cubes are also observed. An increase in the Ce content leads to the formation of octahedral particles, which is the dominant shape of CeO₂ prepared by FSP.⁶² The TEM images in Figure 5 reveal the crystallinity of the particles. The lattice spacings as determined from the TEM images (Figure 5c and Figure S9) show a small increase in the In₂O₃ (222) lattice spacing from 0.29 nm for the In₂O₃ sample to 0.30 nm for the Ce(10)-In₂O₃ one. From the images of Ce(50)-In₂O₃, we can derive the presence of particles with lattice spacings of 0.29 and 0.30, which can be related to (222) planes of (Ce-doped) In₂O₃ as well as 0.31 nm, which corresponds to the (111) plane of CeO₂. For Ce(90)-In₂O₃, the lattice spacing of 0.31 nm is most frequently observed, indicating the predominance of CeO₂ in this sample. HAADF imaging in combination with EDX gives further insight into the dispersion of In and Ce in the particles (Figure 5d and 5e). Ce is well dispersed over the In₂O₃ particles in Ce(10)-In₂O₃. For the Ce(50)-In₂O₃ sample, different phases can be observed with smaller particles predominantly containing In surrounding larger particles rich in Ce. The Ce(90)-In₂O₃ appears to consist of CeO₂ particles, in which In is present at high dispersion next to some regions with more clustered forms of In-oxide.

Figure 5.

(a) TEM images of the FSP synthesized In₂O₃ and Ce–In₂O₃ catalysts with corresponding size distributions (b). (c) HRTEM images and FFT showing the high (mono-) crystallinity of the nanoparticles and lattice spacing. (d–e) STEM-EDX imaging of the particles showing the high dispersion of In and Ce atoms in the mixed catalysts.

Synchrotron-based XRD (sXRD) was employed to study the phases present in the Ce–In₂O₃ samples in more detail (Figure 6). Qualitatively, the patterns suggest that Ce(10)-In₂O₃ is mainly composed of In₂O₃ in which Ce was doped, while Ce(50)-In₂O₃ and Ce(90)-In₂O₃ are predominantly made up of CeO₂ in which In was doped. The (222) and (111) diffraction lines of respectively In₂O₃ and CeO₂ were fitted by a Voigt function to estimate the crystallite size using the Scherrer equation. The results are listed in Table 2. The crystallite size of In₂O₃ of 8.6 nm determined by sXRD is in good agreement with the value of 9.2 nm determined by lab-based XRD (Table 1). The same holds for the crystallite size of Ce(10)-In₂O₃ determined by sXRD (8.2 nm) and XRD (7.8 nm). Based on Rietveld refinement of the sXRD patterns, there are no signs of CeO₂ admixtures in the Ce(10)-In₂O₃ sample, indicating that possible CeO₂ admixtures should either be very small or absent (Figure S12). The (222) diffraction line of pure In₂O₃ is located at 30.59 degrees, corresponding to a lattice spacing of 2.92 Å. A small shift to lower diffraction angles for Ce(10)-In₂O₃ indicates an increase in the lattice parameter. The In₂O₃ lattice parameter was determined by Rietveld refinement at 10.19 Å, corresponding to an increase of 0.05 Å compared to pure In₂O₃, which confirms the incorporation of Ce in the In₂O₃ lattice (Figure 6b and Figure S12). The sXRD pattern of Ce(90)-In₂O₃ contains a dominant (111) diffraction line representative of the CeO₂ fluorite phase. The small shift to higher diffraction angle indicates inclusion of In in the CeO₂ lattice. Scherrer analysis shows that the average size of the CeO₂ particles is 7.1 nm. No In₂O₃ phase could be fitted using Rietveld refinement, indicating that possible In₂O₃ admixtures should either be very small or absent. The CeO₂ lattice parameter was determined at 5.40 Å (Figure S12). The minor decrease (0.01 Å) compared to pure CeO₂ confirms the inclusion of some In in the CeO₂ lattice. The Ce(50)-In₂O₃ appears to consist of a mixture of CeO₂, which contains more doped In than Ce(90)-In₂O₃, as indicated by the relatively large shift of the CeO₂ diffraction lines, and Ce-doped In₂O₃. The presence of Ce-doped In₂O₃ can be deduced from the shoulder at the (220) peak of CeO₂. Further Rietveld refinement confirms the presence of ∼50 wt % of both phases. The lattice parameters of the In₂O₃ and CeO₂ phases are respectively 10.43 and 5.37 Å. The changes in lattice parameter confirm the inclusion of Ce in the In₂O₃ phase and In in the CeO₂ phase (Figure S12). Scherrer analysis reveals that the CeO₂ particles in this sample are much smaller at 4.6 nm than in the other samples. Comparing these findings to the STEM-EDX images in Figure 5e, it can be said that the Ce(50)-In₂O₃ consists of small 5 nm sized CeO₂ particles doped with In, which support small In₂O₃ particles doped with Ce.

Figure 6.

(a) sXRD patterns of Ce–In₂O₃ catalysts synthesized by FSP. In₂O₃ and CeO₂ phases are characterized according to PDF cards 00-044-1087 and 01-071-4199 respectively. Besides In₂O₃, Ce(10)-In₂O₃, Ce(50)-In₂O₃, and Ce(90)-In₂O₃, a reference sample of CeO₂ (top) is analyzed. (b) Zoom in on the In₂O₃ (222) and CeO₂ (111) diffraction lines showing the shifts indicative of lattice expansion and contraction by Ce- and In-dopants, respectively.

The surface composition of the Ce–In₂O₃ catalysts was analyzed by XPS and listed in Table 2. The surface Ce contents are 8.0, 38, and 83 at% for respectively Ce(10)-In₂O₃, Ce(50)-In₂O₃, and Ce(90)-In₂O₃. The slightly lower Ce contents compared to the nominal values of 11, 51.5, and 92.6 at% hint at surface enrichment of In as also found for the Ce–In₂O₃ sample in the initial set. Qualitatively, this is also in line with the STEM-EDX images for Ce(50)-In₂O₃, showing small In₂O₃ surrounding larger CeO₂ particles (Figure 5e). The metal distribution in the other samples seems to be more homogeneous (Figure S10 and Figure S13).

The XPS spectra of the In 3d region given in Figure 7a were fitted in the same way as described above. This results in two contributions, a main feature due to In₂O₃ at an In 3d_5/2 binding energy of 444.5 eV and a small one at a higher binding energy due to In(OH)₃/In₂O₃-defects. Figure 7a shows the small shift in the binding energy of 0.2 eV of the main In 3d_5/2 peak in Ce(10)-In₂O₃, as was also observed for Ce-doped In₂O₃ in Figure 1. The binding energy and the shift compared to In₂O₃ do not change significantly for the Ce(50)-In₂O₃ and Ce(90)-In₂O₃ samples. The contribution of In(OH)₃/In₂O₃-defects increases, however, with increasing Ce content. Especially, the Ce(90)-In₂O₃ contains a large contribution of 45% of such states (Table 2), which can be attributed to the high dispersion of In-oxide in this sample in line with the high In dispersion observed by STEM-EDX (Figure 5e).

Figure 7.

(a) XPS analysis of the In 3d core-line region of pelletized catalyst particles of In₂O₃, Ce(10)-In₂O₃, Ce(50)-In₂O₃, and Ce(90)-In₂O₃. (b) XPS analysis of the Ce 3d core-line regions of the Ce–In catalysts. The spectra were deconvoluted into features belonging to Ce³⁺ and Ce⁴⁺ oxidation states in the same way as reported by Muravev et al.³³ Ce³⁺ and Ce⁴⁺ related features are shown in blue and green, respectively. All spectra were energy corrected by the U‴ component at 916.7 eV of Ce.³¹ The resulting C 1s binding energy (measured at 285.3 eV) was used to energy correct the spectrum of pure In₂O₃.

Figure 7b shows the Ce 3d XPS spectra of these samples including the Ce³⁺ and Ce⁴⁺ contributions. The Ce³⁺ content is lowest for Ce(90)-In₂O₃ (16 at%), These Ce³⁺ ions reflect oxygen vacancies in the CeO₂ lattice.⁴⁵ The Ce³⁺ content increases to 19 at% in Ce(50)-In₂O₃ and 24 at% in Ce(10)-In₂O₃, the latter value being similar to the one for Ce-doped In₂O₃ in Figure S5a. The higher Ce³⁺ content at a lower nominal Ce content is most likely due to the smaller size of the CeO₂ particles. At the lowest Ce content (highest In content), it can also be that the 3+ oxidation state is preferred for Ce, substituting In. We note that Ce₂O₃ occurs in the same cubic crystal structure as In₂O₃ with the Ia3 space group with a Ce–O bond distance of 2.4 Å (In–O bond distance is 2.2 Å), but no Ce₂O₃ phase was found from Rietveld refinement.^45,63,64

X-ray absorption near edge spectroscopy (XANES) spectra at the In K-edge of the In₂O₃ and Ce–In₂O₃ samples are shown in Figure 8a. All XANES spectra are very similar to those of In₂O₃, reflecting the predominant 3+ oxidation state of In. The XANES spectra of reference In metal is plotted in Figure S15, showing the lower edge energy of metallic In. In₂O₃ forms a cubic crystal structure of the Ia3 space group and contains two different In³⁺ sites, forming (distorted) InO₆ octahedra with In–O bonds around 2.2 Å.⁶⁵ The extended X-ray absorption fine structures (EXAFS) in Figure 8b show an intense first In–O coordination shell. The second shell corresponds to an In–O–In coordination shell with two different bond distances between the In atoms in the cubic In₂O₃ crystal structure.⁶⁵⁻⁶⁸ This second shell decreases with increasing Ce content, which can indicate a decrease of the size of the In₂O₃ particles or the replacement of In in the second shell by Ce. A weak second shell remains for the Ce(90)-In₂O₃ sample. Fits of these EXAFS are given in Figure S16. The fit parameters in Table S2 reveal a decrease in the second shell ln–O-In coordination number from 11.8 ± 1.7 for In₂O₃ to 4.1 ± 3.0 in Ce(90)-In₂O₃ (Table S2). This result is qualitatively in agreement with the findings from STEM-EDX, XPS and sXRD that Ce(90)-In₂O₃ contains small In-oxide particles covering CeO₂.

Figure 8.

(a) In K-edge XANES of for In₂O₃, Ce(10)-In₂O₃, Ce(50)-In₂O₃, and Ce(90)-In₂O₃ before CO₂ER. (b) The k³ weighted FT-EXAFS spectra (In K-edge) of In₂O₃ and Ce–In₂O₃ catalysts before CO₂ER.

The catalytic performance of the Ce–In₂O₃ samples was evaluated in the same way as discussed above. The FE_formate as a function of the current density are given in Figure 9. In line with the result in Figure 2, the Ce(10)-In₂O₃ sample exhibits an increased FE_formate compared to In₂O₃ at current densities of 150 and 200 mA/cm². The maximum FE_formate for this sample is 86 ± 3% at 150 mA/cm², which compares favorably to 78 ± 3% for In₂O₃. The higher Ce content in Ce(50)-In₂O₃ results in a maximum FE_formate of 83 ± 2% at a current density of 150 mA/cm². At the highest Ce content in Ce(90)-In₂O₃, the FE_formate decreases slightly to 75% and 74% at respective current densities of 150 and 200 mA/cm², which is below the performance of In₂O₃. It should be noted that these values are obtained with a relatively low In content in the GDE (∼0.05 mg In₂O₃/cm²). This is below the minimum necessary catalyst loading observed in our previous study with this electrode configuration.⁵⁵ To determine whether CeO₂ plays a synergistic role with respect to In₂O₃, we determined the performance of In₂O₃ at a loading of the GDE of 0.06 mg/cm², which is nearly the same as the In₂O₃ loading used for the Ce(90)-In₂O₃ (∼0.05 mg/cm²). The results in Figure 9 for current densities above 100 mA/cm² clearly show that CeO₂ has a positive influence on In₂O₃ by the substantially higher FE_formate for the Ce(90)-In₂O₃. The lower FE_formate for the GDE with an In₂O₃ loading 0.06 mg/cm² stems from the much larger contribution of the HER. The catalytic activity of CeO₂ as shown in Figure S11 demonstrates that CeO₂ is a poor electrocatalyst for CO₂ reduction, reaching a FE_formate of 17% at 100 mA/cm². The FE_formate decreases to 8% at 200 mA/cm². HER is the main reaction here competing with formate production in addition to the formation of a small amount of CO. On the Pd-, Ni-, and Co-doped In₂O₃ samples, the FE_formate and average applied potential decreased compared to on In₂O₃ (Figure S6), most likely due to a lower kinetic barrier for hydrogen evolution. The applied potential on Ce(10)-In₂O₃ and Ce(50)-In₂O₃ are respectively lower and similar compared to on In₂O₃ (Figure S14). Nevertheless, the higher FE_formate points to a lower kinetic barrier for CO₂ reduction to formate. On Ce(90)-In₂O₃ and CeO₂, the applied potential increases, while the FE_formate decreases compared to on In₂O₃. This is most likely, because pure CeO₂ is a poor electrocatalyst for both hydrogen evolution and CO₂ reduction and cannot maintain a high current density without sufficient In. To understand the changes in intrinsic catalytic activity of the Ce-doped or Ce-supported In₂O₃, such as the onset potential for CO₂ER, potentiostatic measurements on a rotating disk electrode could give better control over the applied potential and more insight in the overpotential of the catalysts toward CO₂ electroreduction.

Figure 9.

Faradaic efficiency as a function of current density on In₂O₃ and Ce–In₂O₃ catalysts after 1 h CO₂ER in 0.5 M KHCO₃. Catalyst nanoparticles are deposited on a GDE (0.47 ± 0.07 mg/cm²). Experiments were repeated in duplicate to obtain the error bars.

The increase in FE brought about by the Ce in Ce(10)-In₂O₃ could be caused by a change in the electronic structure of In₂O₃. The higher binding energy of the valence electrons on the In atom might stabilize the In₂O₃ phase. To understand the contribution of Ce to the electronic structure of In₂O₃, the catalysts were further analyzed using operando XAS.

Structural Evolution of Ce–In Catalysts during CO₂ER

Using operando XANES at the In K-edge, the chemical state of In was characterized during cyclic voltammetry and chronoamperometry on a modified GDE during CO₂ reduction. The temporal resolution of the XANES measurements (0.5 s/scan) was used to follow the spectral changes at the In K-edge during cyclic voltammetry. The potential on the electrode was cycled between −0.4 V and −2.0 V vs Ag/AgCl for 3 CVs before CO₂ reduction was followed at constant potentials of −1.5 and −2.0 V vs Ag/AgCl. Figure S17 shows the corresponding CVs for the In₂O₃ and Ce–In₂O₃ samples. During the first cathodic scan for In₂O₃ (Figure S17a), the current density increases from the onset potential of −1.1 V vs Ag/AgCl onward until the most negative applied potential of −2.0 V vs Ag/AgCl. Linear combination fitting (LCF) reveals the changes in the In oxidation state during CV cycling (Figure 10a). At −1.1 V vs Ag/AgCl, the increase in current coincides with rapid reduction of In₂O₃ to metallic In, obtaining almost 80 at% metallic In after completion of the cathodic cycle. XANES spectra at −1.0 V, −1.5 V, −2.0 V vs Ag/AgCl and EXAFS spectra during the first cathodic cycle show that In₂O₃ reduction continues during the cathodic cycle until the potential and current return to the nonfaradaic region in the CV (Figure S18). During the anodic cycle, a minor reoxidation feature of metallic In starting at −0.9 V vs Ag/AgCl can be observed, which results in the presence of ca. 35 at% In₂O₃. During the second cycle, the maximum cathodic current density is lower than in the first cycle, and two cathodic peaks are observed. The first cathodic peak starting at −0.9 V vs Ag/AgCl coincides with the reduction of reoxidized In₂O₃, yielding again a metallic In contribution of 80 at%. The second reduction feature starts around −1.4 V vs Ag/AgCl but does not lead to a significant change in the In speciation according to XANES. Most likely, this feature represents the reduction of CO₂. In the third cycle, only minor reoxidation occurs. Reduction during the third cathodic scan results in a nearly constant In speciation of 90% metallic In. Subsequent CO₂ reduction at constant potentials of −1.5 and −2.0 V vs Ag/AgCl does not further change the In speciation (Figure 10c). Contrary to observations in our previous work using in situ Raman spectroscopy,⁵⁵ a small amount of In₂O₃ remains present up to −2.0 V vs Ag/AgCl. The presence of residual In₂O₃ may have a role in the CO₂ reduction reaction, as it was argued to be a more active phase for CO₂ reduction to formate.¹⁷ A remaining peak of the In–O shell in the EXAFS confirms the presence of In–O and/or In–OH (Figure S19b). Fitting of the first In–O shell shows that the In–O coordination number decreased to 0.8 ± 0.5 and the In–O bond distance decreased from 2.2 to 2.0 Å. All this suggests a predominantly metallic In phase with some residual In–O bonds.

Figure 10.

Potential resolved evolution of In speciation determined by linear combination analysis of XAS spectra for In₂O₃ (a) and Ce(10)-In₂O₃ (b) during cyclic voltammetry in 0.5 M KHCO₃ at a scan rate of 10 mV/s. The potential is displayed on the X-axis in chronological order, sweeping back and forth between −0.4 V and −2 V vs Ag/AgCl. The green scattered plot gives the recorded current density, reaching a maximum of −2.0 V vs Ag/AgCl at every cycle. (c, d) Time-resolved evolution of In speciation determined by linear combination analysis of XAS spectra for In₂O₃ (c) and Ce(10)-In₂O₃ (d) during chrono-amperometry at −1.5 V and −2.0 V vs Ag/AgCl after 3 CVs.

The reduction of In₂O₃ is slower in the presence of Ce as can be judged from the CV-XANES experiment for Ce(10)-In₂O₃. After the first CV, the metallic In content is only 30 at% (Figure 10b). During the subsequent 2 cycles, more In is reduced, attaining a metallic In content of 50 at%, significantly less than the nearly full reduction (90 at%) obtained for In₂O₃. During chronoamperometry at −1.5 V and −2.0 V vs Ag/AgCl, the reduction of In₂O₃ to metallic In continues, reaching 60% metallic In at −2.0 vs Ag/AgCl after 20 min (Figure 10d).

At a higher Ce content of 50 at%, the reduction of In₂O₃ to metallic In is strongly suppressed (Figure 11a). The large cathodic current observed for the Ce(50)-In₂O₃ sample during the first CV (Figure S17c) could be due to the reduction of CO₂, CeO₂,^100,101 or any contamination in the experiment. No further reduction was observed during chronoamperometry at −1.5 V vs Ag/AgCl. Nevertheless, at a constant potential of −2.0 V vs Ag/AgCl, a slight reduction of In₂O₃ to metallic In was seen (Figure 11c). Suppression of In₂O₃ reduction by Ce is also evident for the Ce(90)-In₂O₃ sample, despite the lower signal-to-noise ratio of the XANES spectra (Figure 11b and Figure S20). The low quality of the spectra due to the low In content hampered meaningful LCF. Reoxidation of In was followed at open circuit potential (OCP) after the CO₂ER in all experiments (Figure S21). A slight reoxidation of In to In₂O₃ was observed for all catalysts. The In₂O₃ fraction was found to increase 2% for In₂O₃ and 8% for Ce(10)-In₂O₃ after 5 min.

Figure 11.

Potential and time-resolved evolution of In speciation determined by linear combination analysis of XAS spectra for Ce(50)-In₂O₃ (a) and Ce(90)-In₂O₃ (b) during cyclic voltammetry in 0.5 M KHCO₃ at a scan rate of 10 mV/s. (c) Ce(50)-In₂O₃ catalyst during chrono-amperometry at −1.5 V and −2.0 V vs Ag/AgCl after 3 CVs.

The suppression of In₂O₃ reduction correlates well with the observed decreased efficiency toward H₂ and increased efficiency toward CO₂ electroreduction. Potentiostatic analysis is needed to provide more information for the change in onset potential toward H₂O reduction and CO₂ reduction. Besides a high current density, a stable catalyst is also required for industrial applicability. This study is limited to measurements of 1 h to identify the effect of In₂O₃ promotion. If Ce leaches out from the catalyst, as might be suggested by XPS analysis, stabilization of the Ce–In mixed phase could be explored by adding a structural promotor, as was demonstrated, for example, for In-sulfides by the group of Gao.⁶⁹

Conclusions

In₂O₃ nanoparticles were doped with Pd, Co, Ni, Zr, and Ce using flame spray pyrolysis. XRD and XPS show a high dispersion of the dopants in In₂O₃ in addition to some segregated dopant oxide phases on In₂O₃. Galvanostatic measurements at a current density of 200 mA/cm² show a decreasing FE_formate in the order Ce-doped In₂O₃ > Zr-doped In₂O₃ > In₂O₃ > Pd-doped In₂O₃ > Ni-doped In₂O₃ > Co-doped In₂O₃. XPS analysis shows that an increasing FE_formate goes along with increasing binding energy of the In atoms, suggesting that the dopants can slightly modify the surface properties of the In₂O₃. Overall, the highest FE_formate is obtained with Ce-doped In₂O₃ (86%), which compared favorably with the FE_formate of 81% for In₂O₃. The lowest FE_formate of 37% is observed at the other extreme for Co-doped In₂O₃. Further characterization indicates stabilization of a metallic Co–In alloy in the latter catalyst, which can explain the significant contribution of the HER. Overall, only Ce and Zr doping of In₂O₃ improve FE_formate. The Ce content of Ce-doped In₂O₃ nanoparticles was then varied at 10, 50, and 90 at% by FSP. The resulting materials are typically In₂O₃ with Ce dopants and CeO₂ clusters at low Ce content, while CeO₂ with In dopants and In₂O₃ clusters make up the catalyst at Ce contents of 50 and 90 at%. Compared to In₂O₃, FE_formate increases for samples that contain 10 and 50 at% Ce and decreases slightly at 90 at% Ce. The highest FE_formate of 86 ± 3% is obtained for Ce(10)-In₂O₃ at a current density of 150 mA/cm², significantly higher than the value of 78 ± 3% for In₂O₃. In situ XAS analysis shows that almost all In₂O₃ reduces to metallic In during cyclic voltammetry and CO₂ER at −2.0 V vs Ag/AgCl. At a Ce content of 10 at%, the degree of In reduction is much lower at 60%, while a further increase to 50 and 90 at% almost completely suppresses In₂O₃ reduction. The improved FE_formate at a low dopant level is attributed to the stabilization of In₂O₃, which presumably has a lower activity in the competing HER reaction than metallic In. A too-high Ce content decreases the FE_formate due to the formation of CeO₂, which is a poor catalyst for the CO₂ER to formate.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acscatal.4c02619.

• Additional tables and figures, containing experimental data, material characterization, and images referred to in this article (DOCX)

Author Contributions

Tim Wissink: Conceptualization, Investigation, Validation, Visualization, Writing. Floriane A. Rollier: XAS and WAXS investigation. Valerii Muravev: XAS investigation, Review and editing. Jason M.J.J. Heinrichs and Rim C.J. v.d. Poll: TEM imaging. Dimitra Anastasiadou and Nikolai A. Kosinov: XAS investigation. Jiadong Zhu: Synthesis In₂O₃ particles. Marta C. Figueiredo and Emiel J.M. Hensen: Conceptualization, Writing, Review and editing.

The authors declare no competing financial interest.