Construction of hydrophilic-hydrophobic domains in Bi 2 O 3 /nitrogen-doped carbon electrode to boost CO 2 -to-formate conversion

Bi has drawn attention in catalyzing the electrochemical CO 2 -to-formate conversion due to promising selectivity and low cost, but the process suffers from low activity. Herein, we introduce nitrogen-doped carbon (NC) support with hydrophobicity modification to enhance the activity of a binder-free Bi 2 O 3 electrode. Formate partial current on the NC supported Bi 2 O 3 electrode almost doubles compared to Bi 2 O 3 on unmodified support. Furthermore, the hydrophobicity modification with polytetrafluoroethylene (PTFE) significantly extends the stability of NC supported Bi 2 O 3 by diminishing flooding. It also maintains > 90% formate selectivity at a broad potential range from − 0.87 to − 1.27 V (vs. RHE) and shows a formate partial current density of − 100 mA cm − 2 at − 1.37 V (vs. RHE) in 0.5 M KHCO 3 . The improvement is attributed to the synergetic effects of the hydrophilic active sites and the hydrophobic PTFE modified NC support. The unique structure promotes the reactants transport and thus maximizes the active site utilization at the triple-phase interface. This facile microenvironment regulation can be extended to other applications involving gaseous-aqueous phases.


Introduction
CO 2 electrochemical reduction reaction (CO 2 RR) presents a viable method for closing the carbon loop by employing renewable energy sources to transform CO 2 into useful chemicals or fuels [1,2].The production of formate is a practical procedure that requires only two-electron transfer, and the reaction takes place at a reasonably good selectivity.In addition, formic acid/formate has the best economic value when the cost of energy input is considered [3].
Bi catalysts have attracted considerable interest in electrochemical CO 2 -to-formate conversion owing to excellent selectivity with low environmental impact, abundant availability, and affordable price [4,5].High selectivity is attributed to ideal binding of CO 2 reduction intermediates and hydrogen on Bi which promotes CO 2 -to-formate formation while effectively suppressing the competing hydrogen evolution reaction [3,5].However, Bi-electrocatalysts still suffer from low efficiency in near neutral electrolytes as high selectivity and catalytic current density cannot be attained at low overpotentials [6][7][8][9][10].Although high catalytic activities can be achieved in alkaline electrolytes [11,12], the process remains problematic as CO 2 is mainly consumed in the alkaline electrolyte through a reaction with the hydroxide ions to produce carbonates.This creates issues in the downstream liquid product separation process [13].In this scenario, CO 2 RR on Bi-based electrodes in near-neutral electrolytes can offer a viable approach if the performance of the catalyst is improved.
When the catalyst is directly grown on a support, use of costly polymer binders (like Nafion) with a naturally insulating property is avoided.It improves the electron transport and active site accessibility [26][27][28] compared to alternative approach, which includes mixing of catalyst particles with a binder.Another benefit of the binder-free configuration is that it allows us to study the active materials without such secondary effects as the type [29], content [30], and dispersion solvent [31] of binder, which can influence the performance.However, only a few reported NC supported Bi 2 O 3 electrodes were designed for efficient CO 2 -to-formate conversion, combining the advantages of a NC support and binder-free electrode architecture [24].Nevertheless, the narrow potential range of formate production still hinders its practical application, as in industrial production typically the production rate, i.e. current, is fixed while potential fluctuates.It is imperative to fabricate well-performing binder-free NC supported Bi-based electrodes for CO 2 -to-formate conversion in a wide potential range, and further comprehensively study the active materials.
Binders also greatly affect the wettability of the electrodes, which is another important aspect in CO 2 RR as it influences the adsorption and transport of reactants and products at a solid−liquid−gas triple-phase interface [32,33].A super hydrophilic electrode surface might result in flooding and the loss of gas transport pores, which may be a reason for the narrow formate selective potential range of the above-mentioned NC supported Bi 2 O 3 electrodes [24].A very hydrophobic electrode, on the other hand, does not allow for enough contact between the electrode and electrolyte [33].As a result, efforts have been made to adjust the electrode wettability by modulating the catalyst [34][35][36][37][38][39][40][41] or the support [11,42,43].For instance, organic chemicals modified Cu-based materials [37,38,40] and polytetrafluoroethylene (PTFE)-modified NC nanotube [35] could generate hydrophobic surfaces which can trap CO 2 close to active sites and further enhance the performance, while Bi 2 O 3 on hydrophobic carbon fiber outperforms Bi 2 O 3 on hydrophilic ones [11].However, even though notable attempts have been made, systemic comparisons between these two wettability approaches are still missing in existing studies.Besides, none of them emphasizes widely used NC supports.
To fabricate an efficient NC supported Bi-based electrodes, wettability modulation and systemic comparisons between different wettability approaches, particularly when employing a binder-free electrode structure, are required to be explored.To that end, we fabricated binderfree Bi 2 O 3 on a carbon cloth (Bi 2 O 3 ) and first improved its CO 2 RR efficiency by introducing a NC support (Bi 2 O 3 @NC).This resulted in initial improvements in reducing the overpotential for formate production by more than 200 mV and nearly doubled the current density in 0.5 M KHCO 3 .However, changes in the hydrophilicity of the support led to instability and undesirable enhancement of hydrogen evolution at high overpotentials.Therefore, we set investigating the effects of different hydrophobic modifications.Our results show that it is preferable to alter the surface hydrophobicity by modification of the support rather than the catalyst layer itself.Optimum CO 2 -to-formate condition is achieved with the coexistence of hydrophilic Bi 2 O 3 on a hydrophobic PTFEmodified NC support (Bi 2 O 3 @PTFE-NC).The unique structure mitigates the flooding issue effectively.Bi 2 O 3 @PTFE-NC electrode displays over 90% Faradaic efficiency (FE) to formate in a broad potential region from −0.87 V to −1.27 V (vs.RHE) with a maximum value of 94.9% at −0.97 V (vs.RHE).Our systematic approach for constructing hydrophilic-hydrophobic electrodes can be further applied to improve electrocatalytic reactions involving both gaseous/aqueous phases on the electrode surface., ≥99.0%), Nafion dispersion (5 wt%), 2-Propanol (suitable for HPLC, 99.9%), and potassium bicarbonate (KHCO 3 , 99.7%) were purchased from Sigma-Aldrich.Ethanol AA grade was purchased from Altia Oyj.All chemicals were used without further purification.Water from a Synergy purification system (18.2MΩ cm at 25 • C) was used to prepare all aqueous solutions.Carbon Cloth (CC, ELAT-H) of a thickness of 406 μm was purchased from FuelCellsEtc and cut into 2.5 cm×2.5 cm pieces.Nafion 115 membrane was purchased from Ion Power and treated with 5% H 2 O 2 for 60 min, water for 30 min, 0.5 M H 2 SO 4 for 60 min, and finally, 3 times in water all at 100 • C. CO (99.995%) from Woikoski was used as supplied.Leak-free Ag/AgCl reference electrode from Innovative Instruments was used as supplied.

Electrode preparation
Synthesis of Nitrogen-doped carbon nanosheet array (NC) anchored on CC: Zeolite imidazolate framework-8 (ZIF-8) is used as precursor for NC synthesis by a method adapted from a previous report with minor changes [44].In short, CC was pretreated with 0.5 M KMnO for 30 min, thoroughly washed, and dried.40 mL of 0.05 M Zn(NO 3 ) and 40 mL of 0.4 M 2-methylimidazole were stirred vigorously for 1 min, and after that, a piece of pretreated CC was placed vertically in the mixture and aged for 2 h to grow ZIF-8 coating on CC.Subsequently, the ZIF-8 coated CC was rinsed with ethanol 3 times before being dried at 80 • C in the oven for 1 h.NC was synthesized through pyrolysis of ZIF-8 coated CC in an N 2 atmosphere at 800 • C for 2 h with a heating rate of 1 • C min −1 , followed by soaking in 0.5 M H 2 SO 4 overnight to remove the zinc residue.After rinsing several times with water and ethanol, the NC-coated electrode was dried at 80 • C in the oven for 2 h.

Synthesis of PTFE-modified NC (PTFE-NC) anchored on CC:
The as-prepared NC coated electrode was dipped 3 times in 0.2 wt% PTFE solution and dried at 80 • C in the oven for 2 h.The as-obtained supports were further heated in the N 2 atmosphere at 350 • C for 30 min with a heating rate of 5 • C min −1 to remove the surfactant in the PTFE solution.Other concentrations such as 0.1, 0.25, and 0.3 wt% PTFE solutions were also used in the optimization procedure.
Synthesis of PTFE-modified CC (PTFE-CC): CC was dipped 3 times in 0.2 wt% PTFE solution and dried at 80 • C in the oven for 2 h.The asobtained supports were further heated in the N 2 atmosphere at 350 • C for 30 min with a heating rate of 5 • C min −1 to remove the surfactant in the PTFE solution.
Growth of Bi 2 O 3 on the supports (Bi 2 O 3 , Bi 2 O 3 @PTFE-CC, Bi 2 O 3 @NC, and Bi 2 O 3 @PTFE-NC): BiO 2-x was grown on CC, PTFE-CC, NC, and PTFE-NC through a modified solvothermal method adapted from a previous report, respectively [26,45].In a typical procedure, 1 mmol of Bi(NO 3 ) 3 ⋅5 H 2 O was dissolved in 6 mL of ethylene glycol by sonication followed by the addition of 12 mL of ethanol to form a homogeneous solution.A piece of the support was placed aslant in the autoclave and stood for 1 h before thermal treatment at 160 • C for 5 h.After the solvothermal reaction, the electrode was repeatedly rinsed with ethanol and dried at 80 • C in the oven for 2 h.The as-obtained samples were further calcinated in the air at 300 • C for 1 h with a heating rate of 5 • C min −1 to convert BiO 2-x into β-Bi 2 O 3 [46].
Synthesis of PTFE-modified Bi 2 O 3 @NC: BiO 2-x (before calcination in the air) was dipped 3 times in diluted PTFE solution.The concentration of the PTFE solution was again varied between 0.1, 0.2, 0.25, and 0.3 wt%.The electrode was then dried at 80 • C in the oven for 2 h, and the as-obtained samples were further calcinated in air at 300 • C for 1 h with a heating rate of 5 • C min −1 to remove the surfactant in the PTFE solution [47] and convert BiO 2-x into β-Bi 2 O 3 [46].
Air-brushed Bi 2 O 3 /Nafion: 16 mg Bi 2 O 3 was dispersed in a mixture of 160 μL Nafion dispersion and 240 μL 2-propanol by sonication.The as-prepared ink was airbrushed on CC to acquire the loading with approximately 1.5 mg cm −2 .

X-ray diffraction (XRD):
The crystal structure was characterized with XRD using a Rigaku diffractometer and Cu Kα radiation (λ = 0.15418 nm, 45 kV, 40 mA) in a 2θ range of 10 • to 90 • with a scan step size of 0.02 • .
Raman: Raman spectra were recorded on Renishaw inVia™ QON-TOR Raman spectrometer instrument equipped with a 532 nm laser and a CCD detector.
Transmission electron microscopy (TEM): Images of the samples were obtained using a JEOL JEM-2800FS microscope at an accelerating voltage of 200 kV.
Scanning electron microscopy (SEM): Images of the samples were obtained using a Tescan Mira3 microscope at an accelerating voltage of 15 kV and 5 kV for supports and Bi 2 O 3 on supports, respectively.SEM-EDS images were taken using a JEOL JIB-4700 F at an accelerating voltage of 15 kV.
X-ray photoelectron spectroscopy (XPS): XPS measurements were made using the Kratos Axis Ultra system, equipped with a monochromatic Al-K α X-ray source.All measurements were performed with a 0.3 mm×0.7 mm analysis area and the charge neutralizer.The wide scans were performed with 80 eV pass energy and 1 eV energy step, and the high-resolution scans were performed with 40 eV pass energy and 0.1 eV step size.The energy calibration was made using the adventitious carbon C 1 s component at 284.8 eV.All decompositions were made with CasaXPS using GL ( 30) peaks (product of 30% Lorentzian and 70% Gaussian).
CO 2 adsorption: CO 2 gas adsorption experiments at −0.15 • C (273 K) were carried out using a MicrotracBEL BELSorp MAX II (Microtrac BEL) equipment with integrated cooling bath.Samples were pretreated under vacuum at 300 • C for 5 hours using a Microtrac BEL-Prep Vac II Pre-Treatment Station (Microtrac MRB) prior to measurement.
Contact angle measurement: Contact angle measurements were made using Theta Flex optical tensiometer with 12 μL CO 2 -saturated 0.5 M KHCO 3 .

CO 2 bubble adhesion experiment:
The electrode is placed horizontally with a clamp and immersed in CO 2 buffered 0.5 M KHCO 3 .Images of CO 2 bubble adhesion on the electrode were taken by an opensource High-Speed Camera System [48], and the volume of a gas bubble is approximately 12 μL.

Electrochemical tests
Electrochemical flow cell for CO 2 RR: CO 2 electroreduction reactions were performed at room temperature using CO 2 buffered 0.5 M KHCO 3 (pH=7.3)in a flow cell (see the configuration in Scheme S1 and in our recently reported article [49]) coupled with an IviumStat.XRi potentiostat.A Bi 2 O 3 on the support (Bi 2 O 3 , Bi 2 O 3 @NC, Bi 2 O 3 @-PTFE-NC, and PTFE-modified Bi 2 O 3 ) or a support (CC, NC, and PTFE-NC) was directly used as a working electrode (area 2 cm ×2 cm), an Iridium Mixed Metal Oxide (MMO) plate was used as a counter electrode and a leak-free Ag/AgCl electrode was used as a reference electrode.CO 2 was directly bubbled onto the working electrode at 11 mL min −1 , and the electrolyte flow rate was controlled at 11 mL min −1 .
CO 2 was purged for at least 30 min before 10 cyclic voltammetry (CV) cycles were performed from −0.90 V to −1.90 V (vs.Ag/AgCl) at a scan rate of 50 mV s −1 .After recording the CVs, electrolysis was performed with a controlled-potential method from −1.30 V to −2.00 V (vs.Ag/AgCl).During the electrolysis, the gaseous products were monitored by micro gas chromatography (GC, Agilent 990) with two columns, MoleSieve 5 A and PoraPlot, and micro-machined TCD detectors.Liquid products were collected and quantified when electrolysis was finished using High-Performance Liquid Chromatography (HPLC) coupled to AMinex column HPX-87X from BioRad and RI detector.
Faradaic Efficiency (FE) of gaseous product i (CO or H 2 ) is calculated using the equation below: Impedance was collected from 1000 Hz to 1 Hz.Applied potentials were automatically corrected by instruments with built-in 85% iR compensation and converted to the reversible hydrogen electrode (RHE) scale according to the equation shown below:

Double-layer capacitance (C dl ): C dl was determined by recording
CVs of working electrodes between −0.1 V and −0.2 V (vs.Ag/AgCl, 3 M KCl) at different scan rates varying from 20 to 120 mV s −1 in CO 2 buffered 0.5 M KHCO 3.An iridium wire was used as a counter electrode.

Synthesis and characterizations of electrodes
Different supports were first prepared and Bi 2 O 3 was then anchored as explained in the Methods section and shown in Scheme 1. Firstly, NC was anchored on CC through pyrolysis of Zn-based zeolitic imidazolate framework (ZIF-8) [44].The successful synthesis of NC was confirmed by SEM, TEM, Raman, and XRD (Figures S1 and S2).NC arrays maintain the nanosheet morphology of ZIF-8 after pyrolysis and acid wash (Figure S1b-c), which is consistent with the reported literature [44].Only D-band at 1350 cm −1 and G-band at 1590 cm −1 are visible in the Raman spectrum of CC (Figure S2a), while additional A-bands between the D-and G-band resulting from assorted vibrations caused by the presence of heteroatoms become visible in NC (Figure S2b) [50] indicating a successful synthesis.The insertion of nitrogen and removal of Zn were further proved by EDS in Figure S1f-g (N content >8 at% while Zn almost completely removed, ≤0.03 at%).To increase the hydrophobicity of the NC support, PTFE was added to modify the surface (PTFE-NC).After the modification, the NC arrays were partially covered with melting polymer particles as shown in Fig. 1a.The Raman spectrum of PTFE-NC is similar to the one of NC, proving the surface defects did not change during the PTFE modification (Figure S2c).
Secondly, Bi 2 O 3 was grown on three supports (CC, NC, and PTFE-NC) by a widely used solvothermal method and subsequent calcination in air.As shown in Fig. 1b and Figure S3a-c, Bi 2 O 3 nanosheets are densely and uniformly grown on the supports.In this way, the active material can form intimate contact with the supports, which could boost charge transfer [26,51].These nanosheets overlap and interconnect with each other to form large macropores.The secondary structure of an individual nanosheet was further examined by TEM in Fig. 1c-d revealing that each nanosheet is composed of connected nanoparticles of a size of around 10 nm with lattice fringes of 0.319 nm corresponding to the (201) lattice spacing of β-Bi 2 O 3 [52,53].The porous structure with both macropores and nanopores is advantageous for mass transfer and provides abundant active sites for CO 2 RR.In contrast, Bi 2 O 3 airbrushed on CC with the aid of the Nafion binder (air-brushed Bi 2 O 3 /Nafion) shows inhomogeneous distribution of large catalyst aggregates and Nafion covered surface (Figure S3d).
The crystal structure is fine β-Bi 2 O 3 (PDF#27-0050) [46,54], which is confirmed by XRD shown in Fig. 1e.The peak at 26 • is the (002) plane of graphitized carbon from the supports (Figure S2d).The chemical composition of the samples (Fig. 1f-h, and Figure S4) was characterized by using XPS.The peaks at 159.20 and 164.60 eV in the high-resolution Bi 4 f spectrum (Fig. 1f) of Bi 2 O 3 correspond to Bi 3+ in the Bi 2 O 3 component.In the Bi 2 O 3 @NC sample, the Bi 4 f peaks shift to the slightly higher binding energy of 159.35 and 164.75 eV, which may result from the interaction between Bi 2 O 3 and NC.A similar observation has been made when Bi 2 O 3 was anchored on polydopamine-derived NC [24].The Bi 4 f peaks in the Bi 2 O 3 @PTFE-NC sample show nearly the same positions as those in the Bi 2 O 3 @NC sample.Pyrrolic-N (at 400.5 eV) and pyridinic-N (at 398.4 eV) both dominate the N states with pyridinic-N composing more than 70 at% in both Bi 2 O 3 @NC and Bi 2 O 3 @PTFE-NC, which is shown in the high-resolution N 1 s XPS spectra (Fig. 1g).N species in NC have been reported to facilitate CO 2 capture [55] which is beneficial for electrocatalytic processes as more CO 2 can also be utilized by the neighboring Bi 2 O 3 .Both the pyrrolic-N and pyridinic-N in NC have been proven to effectively promote CO 2 -to-formate in bismuth/NC electrocatalysts by a previous study [22].The F 1 s signal present in Bi 2 O 3 @PTFE-NC originates from the PTFE particles coated on the NC support (Fig. 1h).
All the characterizations show that we successfully synthesized binder-free electrodes composed of both supports and Bi 2 O 3 .The morphology, crystal structure, and chemical composition of Bi 2 O 3 show minor changes after the introduction of NC, which could be further used to rationally evaluate the influence of NC supports on the performance of CO 2 RR.

Performance comparison of Bi 2 O 3 and Bi 2 O 3 @NC
A customized flow cell was used to evaluate the catalytic performance of Bi 2 O 3 and Bi 2 O 3 @NC directly as working electrodes in a CO 2buffered 0.5 M KHCO 3 electrolyte.Bi 2 O 3 @NC exhibits slightly lower onset potential and higher current density in preliminary cyclic voltammetry (CV) measurements compared to Bi 2 O 3 as shown in Figure S5.Electrolysis was further carried out on a subset of applied potentials ranging from −0.67 V to −1.37 V (vs.RHE) based on the CVs to define preferable conditions for CO 2 RR.Through analysis of gaseous and liquid reduction products over Bi 2 O 3 and Bi 2 O 3 @NC, only three products are found, which are H 2 , a trace amount of CO, and mainly formate.The selectivity and current density are displayed in Fig. 2a-b, Figure S6a-e, and Table S1.FE of the target product formate on Bi 2 O increases with increasing overpotential in the range from −0.67 V to −1.17 V (vs.RHE) before gradually declining.On this electrode, the FE of formate achieves over 90.0%selectivity at −0.97 V (vs.RHE) and reaches the highest value of 95.6% at −1.17 V (vs.RHE).Formate production rate on an air-brushed Bi 2 O 3 /Nafion electrode is lower than on Bi 2 O 3 (Figure S6f), which is in line with the reported literature [26].Bi 2 O 3 @NC reaches ≥90% formate FE at lower overpotentials (−0.87 V vs. RHE) when compared to Bi 2 O 3 .The partial current density of formate production of both Bi 2 O 3 and Bi 2 O 3 @NC increase as more negative potentials are applied, with Bi 2 O 3 @NC showing more than 2 times the formate partial current density compared to Bi 2 O 3 at most potentials.Notably, Bi 2 O 3 @NC reaches −23.8 mA cm −2 at −0.97 V (vs.RHE) while more than 200 mV higher overpotential is required for Bi 2 O 3 , which achieves a comparable current density value at −1.17 V (vs.RHE).
The electrochemically active surface area (ECSA) was evaluated using the double-layer capacitance (C dl ) and the results are shown in Fig. 2c and Figure S7.C dl of Bi 2 O 3 @NC is approximately 36% higher than C dl of Bi 2 O 3 (1.18 vs. 0.87 mF cm −2 , respectively).Modification of Bi 2 O 3 with NC also enhances the volume of gas (CO 2 ) assimilated by the material, with an approximately 3 times measured increase in the level of CO 2 adsorbed by Bi 2 O 3 @NC when compared with unmodified Bi 2 O (Figure S8).This lowers the charge transfer resistance from 24.38 Ohms for Bi 2 O 3 to 20.18 Ohms for Bi 2 O 3 @NC (Fig. 2d).
A well-accepted mechanism of the two-electron reduction of CO 2 to HCOO − on Bi-electrocatalysts follows the steps below [4,56,57] in [Eqs.
(1), ( 2), ( 3), (4), and (5)]: The enhancement in CO 2 sorption (Eq.( 1)) and the enlarged ECSA offers an increasing number of active sites for the reaction.The lower charge transfer resistance in Bi 2 O 3 @NC compared to Bi 2 O 3 indicates that CO 2 accepts more readily electrons to form CO * − 2 (Eq.( 2)).Moreover, NC in the vicinity of Bi 2 O 3 is reported to enhance adsorption capacity and lower the energy barrier to form formic acid intermediate (HCOO * , Eq. ( 3)) [25,58].To sum up, the performance boost after the introduction of NC near Bi 2 O 3 could be attributed to multiple reasons as an enlarged ECSA, faster kinetics, enhanced charge transfer, and increase in the adsorption capacity of CO 2 and HCOO * .

Modulations of surface hydrophobicity of the Bi 2 O 3 @NC electrode with tuning both support and catalyst layer with PTFE
However, when comparing the potential window where over 90% FE for formate is reached, a problem is evidently present on Bi 2 O 3 @NC.The FE for formate on Bi 2 O 3 @NC first drops to 86.3% at −1.17 V (vs.RHE) and further deteriorates to 73.3% at −1.37 V (vs.RHE), while Bi 2 O 3 still retains a high value of 87.8% at −1.37 V (vs.RHE).The current density of competing H 2 production on Bi 2 O 3 @NC increases around 6 times at this potential when compared to Bi 2 O 3 (Figure S6e).To understand what  may cause this, we must look at the electrode structure more closely: It is generally recognized that CO 2 RR occurs at the solid−liquid−gas triplephase interface.Gaseous CO 2 transport is facilitated by the hydrophobic domains and the presence of these domains also prevents flooding from clogging the gas diffusion pathways.At the same time, protons and dissolved CO 2 molecules are transported with the electrolyte through the hydrated hydrophilic domains [59].Bi 2 O 3 @NC exhibits a faster reaction rate of CO 2 RR (higher partial current densities), and consequently enters the reactant diffusion-dominating potential region earlier than Bi 2 O 3 .Due to the low solubility of CO 2 , the amount of dissolved CO 2 molecules delivered to the active sites through the electrolyte is insufficient for promoting the reaction in this region, which requires the diffusion of extra gaseous CO 2 towards active sites through hydrophobic domains [32,33].The limiting current densities are reported to be inversely correlated with the length of the gaseous CO 2 diffusion pathway, especially at high overpotentials [60].To decrease the length of those pathways, a better solid-liquid-gas triple-phase interface enhancing gaseous CO 2 transfer to the active site must be constructed.
According to the contact angle measurements presented in Fig. 3a, the surface of CC is hydrophobic while the porous Bi 2 O 3 surface on CC shows a hydrophilic feature.In contrast, the electrolyte droplet is almost instantly absorbed by the surface as NC turns to be extremely hydrophilic (Fig. 3b).Growing hydrophilic Bi 2 O 3 on the hydrophilic NC leads to an electrode (Bi 2 O 3 @NC) with super hydrophilicity (Fig. 3b).The loss of hydrophobic domains in Bi 2 O 3 @NC hinders gaseous CO 2 transport towards active Bi 2 O 3 sites efficiently.Moreover, NC may also be activated and favor HER when in contact with protons [61].When these two supports without Bi 2 O 3 are directly utilized as the working electrode, NC exhibits partial current densities of HER that are around two times higher than those of bulk CC (Figure S9).Increasing proton consumption on NC also restricts the CO 2 -to-formate since this conversion also includes a protonation process (Eq.( 3)).As a result, selectivity to formate on Bi 2 O 3 @NC continues to decrease in the reactants (gaseous CO 2 and protons) diffusion-dominating potential region.There appears to be a trade-off between the improved adsorption of intermediates and the increased hydrophilicity brought by the introduction of NC which is evident also in previous studies [24].
To mitigate this issue, we adjusted the hydrophobicity of the electrode surface with a PTFE polymer.PTFE is a synthetic fluoropolymer, which is quite widely used in the electrochemistry field as a coating and binder because of its highly hydrophobic nature [35,41,[62][63][64][65].To obtain an ideal state, where sufficient hydrophobic regions to boost the transfer of reactants and sufficient hydrophilic regions to retain contact with the electrolyte exist [28], the PTFE content needs to be optimized.
Electrodes were then fabricated to weigh the advantages and disadvantages of the two commonly used strategies, which are modulating either the hydrophobicity of the support surface or the catalyst surface using PTFE of different concentrations.
First, to modify the support surface (NC), PTFE solutions with varying concentrations were applied to treat NC using a dip coating method as described in Methods section.When PTFE modification is applied, HER is successfully inhibited to the same degree as on CC (Figure S9) and the surface becomes hydrophobic (Fig. 3c).After the PTFE modification, Bi 2 O 3 was grown on these PTFE-modified NC supports (see Methods and Scheme 1).It is noteworthy that the hydrophilicity of the Bi 2 O 3 surface has only a slight influence on the hydrophobicity of the entire electrode since Bi 2 O 3 @PTFE-NC is still hydrophobic (Fig. 3c), similar as earlier reported to Bi 2 O 3 on hydrophobic carbon fiber [11].To screen the ideal PTFE content, these samples were employed as the working electrodes for CO 2 RR at −1.37 V (vs.RHE) (Fig. 4a).As the concentration of the PTFE solution used in the synthesis is increased to 0.2 wt%, the performance of CO 2 RR is improved since the side reaction HER is suppressed while the total current density remains high (Fig. 4a).However, both FE of formate and total current density drop when the concentration of the PTFE solution is increased further from 0.2 wt%.This is probably due to that higher amount of insulating PTFE polymer particles cover the NC surfaces and disrupt the charge transfer.
The solid−liquid−gas triple-phase interface can also be enhanced by hydrophobicity modifications of the catalyst surface.The PTFE solutions with the same concentrations were applied to modify the Bi 2 O 3 @NC and these modified working electrodes were subsequently also evaluated at −1.37 V (vs.RHE) (Fig. 4b).We indeed notice a suppression of HER with the increasing PTFE coverage.However, this suppression comes at the expense of the total current density which continuously drops with more PTFE introduced.This is likely due to Bi 2 O 3 surfaces being shielded by this insulating polymer (PTFE) from the catalytic reaction.
Both CO 2 -to-formate and HER processes require proton transfer.With the wettability tuning of the support, there are fewer protons in contact with hydrophobic PTFE-NC to favor HER but enough protons are delivered to hydrophilic active sites for CO 2 RR.Through this comparison, surface engineering of support surfaces by growing Bi 2 O 3 on PTFE-NC is unquestionably a superior method over PTFE modification of the catalyst surface.Hence, for further experiments the concentration of the PTFE solution used to modify the NC supports was fixed to 0.2 wt% because it promotes the CO 2 RR on the working electrode more than all other studied electrode modifications.Hydrophobicity of the electrodes is reported to evade the stability challenges due to flooding of the electrode [66].To evaluate the stability of the differently modified electrodes, chronoamperometry at a fixed potential of −0.97 V (vs.RHE) was performed for several hours (Fig. 5a-b).The total current density and formate FE on Bi 2 O 3 @PTFE-NC does not noticeably decrease during the 12 h of electrolysis, indicating good stability of performance.Postmortem SEM analysis shows that the morphology is nanosheet (Figure S10a-b) while the XRD pattern (Figure S10c) shows a mixture of metallic Bi with some Bi 2 (CO 3 )O 2 .This is in accordance with the reported literature [22,67].The high-resolution XPS Bi spectrum in Figure S10d depicts that on the surface the chemical composition is Bi 3+ , which may result from the re-oxidation of the metallic Bi or surface Bi 2 (CO 3 )O 2 as the sample has been under ambient conditions before the XPS measurements.The contact angle of the used electrode shows a slight decrease, but the electrode is still hydrophobic (Figure S11a), and the high-resolution F 1 s spectrum of XPS demonstrates that there is no chemical composition change of the PTFE during longer term electrolysis (Figure S11b).
In contrast, Bi 2 O 3 @NC only shows good selectivity towards formate production and inhibition of HER during the first 1 h of electrolysis.After that, flooding occurs since HER continues to rise so that FE towards formate declines gradually (Fig. 5b).The flooding worsens after 3 h and the total current density becomes quite unstable and starts to decrease, demonstrating that Bi 2 O 3 @NC is not suitable for long-term use.As shown previously, the electrolyte wets both NC and Bi 2 O 3 @NC immediately (Fig. 3b), which may block the gaseous CO 2 diffusion pathway to the active sites [66].Consequently, it lowers CO 2 concentration at the catalyst/electrolyte interface causing CO 2 RR to deteriorate.The electrolyte saturation can also promote salt precipitation, leading to the eventual permeant blockage of CO 2 diffusion and reduced active area [68].
To verify the stability improving effects of PTFE modified support, CC covered with PTFE was also used as a support for Bi 2 O 3 .Bi 2 O 3 @-PTFE-CC exhibits similar short-term performance of CO 2 -to-formate conversion in terms of FE and partial current density for formate production when compared with Bi 2 O 3 (Figure S13a).However, Bi 2 O 3 @-PTFE-CC shows stable behavior over 12 h of electrolysis while flooding occurs on Bi 2 O 3 already after 2 h (Figure S13b-c) demonstrating that hydrophobicity modulation of supports is a universal method to improve the stability of CO 2 RR.
To gain a more comprehensive electrochemical evaluation, shortterm CO 2 electrolysis on the Bi 2 O 3 @PTFE-NC electrode was performed in a wider range of potentials and the results are compared with Bi 2 O 3 @NC (Figure S12 and Fig. 5c-d).Bi 2 O 3 @PTFE-NC shows similar onset potential and total current density to Bi 2 O 3 @ NC (Figure S12a-b), and the same three products -trace quantities of CO, primary formate, and the side product H 2 -are still detected in this potential range (Figure S12c).Compared to Bi 2 O 3 @NC, the formate FE on Bi 2 O 3 @PTFE- NC displays roughly the same values in the low overpotential region and only after applying more negative potentials the formate FE becomes noticeably higher (Fig. 5c and Table S1).For example, at an applied potential of −0.97 V (vs.RHE), the formate FE value reaches 94.9% on Bi 2 O 3 @PTFE-NC while it only reaches 92.9% for Bi 2 O 3 @NC at the same applied potential.Surprisingly, the FE for formate production is 87.6% at −1.37 V (vs.RHE) on Bi 2 O 3 @PTFE-NC, which is more than 14% higher than the value on Bi 2 O 3 @NC and approximately identical to the selectivity on Bi 2 O 3 (87.8%)at the same applied potential.The PTFE modification broadens the potential range, where FE for formate production is ≥90%, to over 400 mV while the range is only 200 mV for Bi 2 O 3 @NC.When the partial current density towards formate production is plotted (Fig. 5d), Bi 2 O 3 @PTFE-NC exhibits a negligibly lower value at low overpotentials but a sound improvement starting from −1.07 V (vs.RHE) compared to Bi 2 O 3 @NC.Particularly at −1.37 V (vs.RHE), it has a 25% increase (−100.8 vs. −80.7 mA cm −2 for Bi 2 O 3 @-PTFE-NC and Bi 2 O 3 @NC, respectively).A 60-hour electrolysis was performed at −0.87 V (vs.RHE), which is the potential of formate FE first reaches over 90% (Fig. 5c).The current density remains stable, and the formate FE undergoes only neglectable decay, still maintaining 87.6% even after 60 h (Figure S12d).The performance of CO 2 RR, including the required overpotential, the partial current density, and the broad potential range for obtaining over 90% formate FE, is higher than many reported Bi-based working electrodes (see summarized CO 2 RR performance of literature in Table S2).
This improvement in the formate FE at high overpotentials is not likely due to the charge transfer resistance and ECSA, since Bi 2 O 3 @-PTFE-NC exhibits similar charge transfer resistance and lower ECSA compared to Bi 2 O 3 @NC (Figure S14).This improvement should originate from the enhanced gaseous CO 2 diffusion in this reactant diffusiondominating potential region due to the hydrophobicity modulation.The hydrophobicity change can further restrict water diffusion and may generate a more alkaline microenvironment as reported [11,40].In our flow-through cell configuration, CO 2 gas is directly bubbled onto and subsequently through the working electrode.Because of its hydrophobic feature in the air, we postulate that the Bi 2 O 3 @PTFE-NC has an aerophilic property in the water, whereas the hydrophilic Bi 2 O 3 @NC would be aerophobic.Both Bi 2 O 3 @PTFE-NC and Bi 2 O 3 @NC were subjected to a CO 2 bubble adhesion experiment in 0.5 M CO 2 -saturated KHCO 3 solution and the results are presented in Fig. 6, and videos in Appendix B-C.The outcomes demonstrate that the CO 2 bubble can enter through the surface of Bi 2 O 3 @PTFE-NC extremely quickly (within 45 ms) confirming its aerophilic property (Fig. 6a and Video in Appendix B).It implies that the surface hydrophobicity of the electrode is a crucial factor in CO 2 gas transport.On the other hand, the CO 2 bubble is easily stuck on the surface of Bi 2 O 3 @NC (Fig. 6b and Video in Appendix C).Even in the flow-through flow cell, Bi 2 O 3 @NC probably could only utilize a relatively little amount of gaseous CO 2 for the reaction.
Hence, we suggest the coexistence of hydrophilic-hydrophobic domains in this electrode for maximum performance (Scheme 2a) where the hydrophobicity of the whole electrode surface is introduced by the super hydrophobic PTFE-NC support and the active sites are hydrophilic.Such a state can be called a Cassie-Wenzel coexistence state [32,42,69], where the surface is neither super hydrophilic nor superhydrophobic.In this state, not only the dissolved CO 2 is utilized in the reaction, but there can also be a concentrated CO 2 gas layer trapped near the aerophilic (hydrophobic) PTFE-NC support as earlier reported which can be used in the reaction [11,32].In this scenario, the formation of the solid−liquid−gas triple-phase interface is boosted, making the active sites more accessible to both gaseous CO 2 and liquid electrolytes.Consequently, the CO 2 molecules from the gas phase, the ions (as proton source) from the liquid phase, and the electron transfer from the PTFE-NC support can greatly promote the CO 2 RR at the triple-phase interface.Contrarily, the state of Bi 2 O 3 @NC is known as the Wenzel state because the active sites and the NC support are both super hydrophilic [32,42,69], and thus, the electrolyte can completely penetrate the electrode surface, preventing the creation of a triple-phase interface and almost eliminating the trapped CO 2 layer (Scheme 2b).In this case, only dissolved CO 2 molecules can get involved in the reaction which can hinder the reaction rate in the CO 2 diffusion-dominating potential region and leads to a drop in formate selectivity.

Conclusion
In conclusion, we fabricated a binder-free Bi 2 O 3 @NC electrode for CO 2 RR.It exhibits enhanced performance at a low overpotential range compared with Bi 2 O 3 .Unfortunately, Bi 2 O 3 @NC suffers unfavorable side effects such as flooding during longer-term use and rapid drop in formate selectivity at a higher overpotential range.The problem is mitigated by tuning the hydrophobicity of the NC support with PTFE coverage before the Bi 2 O 3 growth.The unique structure favors fast gaseous CO 2 transport without hindering the electrolyte contact with active sites of Bi 2 O 3 .The formation of hydrophilic (the Bi 2 O 3 active sites) and hydrophobic (the PTFE-NC support) domains improved the solid−liquid−gas triple-phase interface enhancing gaseous CO 2 transfer to the active sites for CO 2 RR.In addition, with the increasing surface hydrophobicity, the adverse contact between the electrolyte and support is reduced so that not only the competitive HER at high overpotentials is decreased but also flooding is inhibited.This work offers a facile way to fabricate binder-free electrodes and mitigate the flooding issue.It also Fig. 6.CO 2 bubble adhesion experiment on (a) Bi 2 O 3 @PTFE-NC and (b) Bi 2 O 3 @NC.

Fig. 2 .Fig. 3 .
Fig. 2. Potential dependence of selectivity and partial j of formate over (a) Bi 2 O 3 and (b) Bi 2 O 3 @NC.(c) Capacitive current plotted against the scan rate for Bi 2 O 3 and Bi 2 O 3 @NC for the determination of C dl .(d) Nyquist plots of Bi 2 O 3 and Bi 2 O 3 @NC at −0.67 V (vs.RHE) with equivalent circuit model (R1: Ohmic resistance, R2: charge transfer resistance).All the measurements were conducted in CO 2 -saturated 0.5 M KHCO 3 at room temperature.

Fig. 4 .
Fig. 4. Comparison of selectivity and total j at −1.37 V (vs.RHE) in CO 2 -saturated 0.5 M KHCO 3 at room temperature between surface hydrophobicity tuning of (a) the support by modifying NC before Bi 2 O 3 growth and (b) the catalyst layer by directly modifying Bi 2 O 3 @NC using PTFE of different concentrations.

Fig. 5 .
Fig. 5. Stability measurements at −0.97 V (vs.RHE) of (a) Bi 2 O 3 @PTFE-NC and (b) Bi 2 O 3 @NC.Potential dependence on (c) selectivity and (d) partial j of formate over Bi 2 O 3 @PTFE-NC and Bi 2 O 3 @NC.All the measurements were conducted in CO 2 -saturated 0.5 M KHCO 3 at room temperature. 4 charge transfer per mole of gas product, in our case Z=2; F: Faraday constant, 96485 C mol −1 ; ∅ i : volume fraction of gaseous product quantified by GC; v: molar flow rate of CO 2 , mol s −1 ; I: total current when GC sampling, A.