Size Effects of Highly Dispersed Bismuth Nanoparticles on Electrocatalytic Reduction of Carbon Dioxide to Formic Acid

Electrocatalytic reduction of carbon dioxide into value-added chemical fuels is a promising way to achieve carbon neutrality. Bismuth-based materials have been considered as favorable electrocatalysts for converting carbon dioxide to formic acid. Moreover, size-dependent catalysis offers significant advantages in catalyzed heterogeneous chemical processes. However, the size effects of bismuth nanoparticles on formic acid production have not been fully explored. Here, we prepared Bi nanoparticles uniformly supported on porous TiO2 substrate electrocatalytic materials by in situ segregation of the Bi element from Bi4Ti3O12. The Bi-TiO2 electrocatalyst with Bi nanoparticles of 2.83 nm displays a Faradaic efficiency of greater than 90% over a wide potential range of 400 mV. Theoretical calculations have also demonstrated subtle electronic structural evolutions induced by the size variations of Bi nanoparticles, where the 2.83 nm Bi nanoparticles display the most active p-band and d-band centers to guarantee high electroactivity toward CO2RR.

electron microscopy (SEM) (Phenom Pro Desktop, USA) was used to obtain SEM images. The transmission electron microscope (TEM) and high-resolution TEM (HRTEM) were used on a JEM-2000EX transmission electron microscope with an acceleration voltage of 200 kV. X-ray photoelectron spectroscopy (XPS) was obtained with a Thermo ESCALAB-250 instrument (USA). 1 H NMR spectra were recorded on Bruker AVANCE III HD 500 NMR spectrometer. Raman spectra were acquired on a micro-Raman spectrometer using a 532 nm laser as excitation source. The electron paramagnetic resonance (EPR) spectra were recorded using EPR spectrometer (Bruker, Germany).

CO 2 RR measurement
In electrochemical experiments, the electrochemical workstation (CHI 760E) and a cation exchange membrane (Nafion 117, Dupont) were used. In the H-type cell, 35 mL electrolyte (0.1 M KHCO 3 ), graphite rod, and Ag/AgCl electrode (saturated KCl solution) were used as electrolyte, counter electrode, and reference electrode, respectively. For working electrode, 200 μL of catalyst ink (5 mg mL −1 , containing 0.5 wt% Nafion solution) was loaded onto a 1 × 1 cm 2 carbon paper. The electrolyte was degassed by bubbling high-pure CO 2 before the electrochemical tests at least 0.5 h under magnetically stirring (350 rpm) and kept CO 2 -saturated during the whole test process at 20 mL min −1 by using a mass flow controller. For flow type cell testing, 1 mg cm −2 catalyst was loaded gas diffusion electrode as the cathode and Pt sheet was using as the anode for water oxidation. 1 M KOH was using as electrolyte and circulated around the anode and cathode at a flow rate of 10 ml min −1 during the CO 2 RR. Linear scan voltammetry (LSV) measurements were conducted at −0.2 V to −1. 4 V vs. RHE in H-type cell containing 0.1 M KHCO 3 or at 0 V to −1.4 V vs. RHE in a gas-diffusion flow cell containing 1 M KOH with a scan rate of 10 mV s −1 . Applied potentials were converted to the reversible hydrogen electrode (RHE) using equation E (vs. RHE) = E (vs. Ag/AgCl) + 0.197 V + 0.059 V × pH. Electrochemical impedance spectra (EIS) were tested in the range of 10 5 ~ 10 −2 Hz with amplitude of 10 mV. The gas products were analyzed by online gas chromatography (GC) systems every 14 min equipped with a thermal conductivity detector and flame ionization detectors, and high purity argon gas (99.999%) was used as the carrier gas for the GC. The liquid products were analyzed using NMR spectroscopy by adding 630 μL electrolyte into the 70 μL D 2 O (deuterated water) containing dimethyl sulfoxide solution (DMSO, as an internal standard) with water peak suppression method.
Faradaic efficiency: The Faradaic efficiency (FE) can be calculated as the following where, w is the number of electrons transfer involved in the reaction (for example, twoelectron transfer for formate, CO, and H 2 products), n is the number of moles of product formed over time, F is the Faraday constant, and Q is the total amount of charges through the CO 2 RR process in this time.

In situ ATR-FTIR
In situ ATR-FTIR experiments were performed using a HgCdTe (MCT) detector equipped with liquid nitrogen cooling and purification with dry air. A PIKE electrochemical cell was mounted on a VeeMax III ATR accessory with 60° Au-coated silicon prism loaded catalysts. All ATR-FTIR measurements had a spectral resolution of 4 cm −1 , an optical velocity of 1.8988, and a gain of 1. Three-electrode electrochemical system with the silicon prisms as work electrode, Ag/AgCl as reference electrode, and Pt wires counter electrodes was used.

DFT calculations
Density functional theory (DFT) calculations are carried out in this work based on the CASTEP packages to investigate the size effects of Bi nanoparticles on TiO 2 . 1 For all the exchange-correlation interactions, the generalized gradient approximation (GGA) and Perdew-Burke-Ernzerhof (PBE) functionals are chosen to realize the sufficient description. 2-4 Meanwhile, a 380 eV cutoff energy has been applied for the calculations, which is generated based on the ultrafine quality of cutoff energy and ultrasoft pseudopotentials. In addition, we adopt the Broyden-Fletcher-Goldfarb-Shannon (BFGS) algorithm with coarse k-point settings for all the energy minimization. 5 The convergence tests have proved that these settings are sufficient to reveal the changes in both energy and electronic structures. Considering the computational loadings and efficiency, we have cleaved the Bi nanoparticles from the (104) surface with 10, 14, and 26 atoms of Bi atoms in Bi-TiO 2 -800, Bi-TiO 2 -700, and Bi-TiO 2 -600, respectively.
The nanoparticle sizes are 0.59 nm, 0.80 nm, and 1.19 nm, respectively, which are considered as small, medium, and large sizes to investigate the size effect. TiO 2 is cleaved from (100) surfaces of anatase TiO 2 with four atomic layers with 144 atoms.
Moreover, the 20 Å vacuum space is introduced in the z-axis direction to supply sufficient space on the catalyst surface for all the relaxations. Moreover, we have applied strict convergence requirements for all the calculations in this work as follows.