Chemicals and materialsAg powder (99.9%, 50 nm) was purchased from Ningbo Jinlei Nano Materials Co., Ltd. Polyetherimide (PEI) was purchased from Saudi Basic Industries Corporation (SABIC). N-Methyl-2-pyrrolidone (NMP), sulfuric acid (H2SO4), potassium sulfate (K2SO4), potassium bicarbonate (KHCO3), potassium chloride (KCl) and potassium hydroxide (KOH) were purchased from Sinopharm Chemical Reagent Co., Ltd. Nafion 117 proton exchange membranes (PEMs) were purchased from DuPont. All chemicals were used as received without further purification. Electrolyte solutions were prepared using 18.2 MΩ H2O (ultrapure water, from Master-S30UVF water purification system).Catalyst preparationAg hollow fiber (Ag HF) was fabricated by a combined phase-inversion/sintering process (Supplementary Fig. 1)37. Briefly, commercially available polyetherimide (PEI, 24 g) was added to N-Methyl-2-pyrrolidone (NMP, 96 g), followed by ultrasonic treatment for 1 h to obtain a homogeneous and transparent solution. Then Ag powder (80 g) was added to the above solution. The as-obtained mixture was further treated by the planetary ball-milling (using 250 mL zirconia jar and ϕ5 mm zirconia balls) at 300 rpm for 24 h to form a uniform slurry. After cooling to room temperature, the slurry was vacuumed (1 mbar) for 5 h to remove bubbles and then to obtain a casting solution. Next, the casting solution was extruded through a spinning machine and shaped in a water bath via the phase-inversion process. After spinning, the as-formed tubes were kept in a water bath for 24 h to eliminate the solvent completely, followed by stretching and drying in ambient conditions with a humidity of ~28% for 48 h to obtain a green body. The green body was cut into appropriate lengths and then calcinated in an airflow (100 mL/min) at 600 °C (heating rate: 1 °C/min) for 6 h to remove PEI. After being naturally cooled to room temperature, the calcined green body was then reduced in a 5% H2 (argon balance) flow (100 mL/min) at 300 °C (heating rate: 1 °C/min) for 3 h to obtain Ag HF.The Ag HF with an exposed length of 4 cm was stuck into a copper tube using conductive silver adhesive for electrical contact (see Fig. 2g for details), while the end of the Ag HF tube as well as the joint between the Ag HF and copper tube were sealed and covered with gas-tight and nonconductive epoxy. After drying at room temperature for 12 h, a working Ag HF penetration electrode (Ag HPE) was obtained with an exposed geometric area of 0.5 cm2 (S = πDoutL = 3.14 × 400 × 10−4 × 4 ≈ 0.5 cm2, where S is the electrode area, Dout is the outer diameter of hollow fiber, and L is the length of hollow fiber).Ag2CO3-Ag HPE was synthesized from Ag HPE by electrochemical redox activation treatments. Typically, the Ag HPE was subjected to oxidation and reduction treatments on a Biologic VMP3 potentiostat using a three-electrode system in a gas-tight two-compartment electrolysis cell containing a Nafion 117 membrane as the separator, a KCl-saturated Ag/AgCl reference electrode and a platinum mesh (3 cm × 3 cm) counter electrode. The electrolyte solution was CO2-saturated 0.5 M KHCO3, and the CO2 flow rate was kept at 2 mL/min. Prior to the experiments, the electrolysis cell was vacuumized and then purged with CO2 for 30 min. The Ag HPE was electrochemically oxidized at a fixed potential of 2.0 V (vs. Ag/AgCl) for 4 min to obtain Ag2CO3-Ag HPE. Subsequently, the Ag2CO3-Ag HPE was reduced at a fixed potential of −0.50 V (vs. Ag/AgCl) for 10 min to obtain CD-Ag HPE. The CD-Ag HPE possessed the same exposed geometric area of 0.5 cm2 (S=πDoutL = 3.14 × 400 × 10−4 × 4 = 0.5 cm2). For the 10-tube CD-Ag HPE array electrode, the exposure geometric area was 5 cm2 (S = nπDoutL = 10 × 3.14 × 400 × 10−4 × 4 = 5 cm2, where n is the number of hollow fiber tubes). The electrochemical oxidation reaction and reduction reaction obeyed Eqs. (1) and (2), respectively.$$2{Ag}+2{H}_{2}O+{HC}{O}_{3}^{-}\to A{g}_{2}C{O}_{3}+6{e}^{-}+{O}_{2}\uparrow+5{H}^{+}$$
(1)
$$A{g}_{2}C{O}_{3}+5{H}^{+}+6{e}^{-}\to 2{Ag}+{HC}{O}_{3}^{-}+{2H}_{2}\uparrow$$
(2)
In addition, the OD-Ag HPE was also treated in the Ar-saturated 0.5 M KOH with the same electrochemical redox activation treatments as that of CD-Ag HPE. That is, the Ag HPE that underwent 60 s of oxidation and 600 s of reduction in Ar-saturated 0.5 M KOH electrolyte solution, respectively.Material characterizationThe cross-section and surface morphologies of samples were observed via scanning electron microscopy (SEM) with a SUPRRATM 55 microscope using an accelerating voltage of 5.0 kV. Transmission electron microscopy (TEM) investigations were conducted with a JEM-ARM300F microscope operated at 300 kV. X-ray diffraction (XRD) measurements were performed on a Rigaku Ultima 4 X-ray diffractometer using a Cu Kα radiation source (λ = 1.54056 Å) at 40 kV and 40 mA. X-ray photoelectron spectroscopy (XPS) tests were conducted using a Quantum 2000 Scanning ESCA Microprobe instrument with a monochromatic Al Kα source (1486.6 eV). The binding energies in all XPS spectra were calibrated according to the C 1 s peak (284.8 eV).Electrochemical measurementsAll electrochemical measurements were performed using a Biologic VMP3 potentiostat in a two-compartment electrolysis cell at an ambient temperature and pressure. The electrolysis cell comprises two symmetrical compartments made of quartz glass with an inner height of 5.0 cm, an inner length of 5.0 cm and an inner width of 1.5 cm (Fig. 2g, h). The cathodic and anodic compartments were separated by a Nafion 117 membrane, and the electrolysis cell was equipped with a KCl-saturated Ag/AgCl reference electrode in the cathodic compartment and a platinum mesh (3 cm × 3 cm). CO2-saturated H2SO4 containing various concentrations of KCl or different pH was used as catholyte, and 0.5 M K2SO4 + 0.05 M H2SO4 aqueous solution was used as anolyte. Prior to the experiments, the electrolysis cell was vacuumed, and then CO2 was continuously delivered into the cathodic compartment at a constant rate of 30 mL/min for 30 min. During all of the electrochemical measurements, the CO2 flow rate was fixed at 30 mL/min. Note that the input CO2 flow rate was fixed at 30 mL/min during all electrochemical measurements, but the exhaust flow rate was not 30 mL/min at all due to the hydrogen evolution and CO2 consumption. Thus, the outlet flow rate of the electrolysis cell was measured by an independent Alicat® mass flowmeter (Fig. 2i and Supplementary Table 4). The cathodic electrolyte was 3.0 KCl + 0.05 M H2SO4, the anodic electrolyte was 0.5 M K2SO4 + 0.05 M H2SO4, unless otherwise stated.For electrochemical characterizations, the electrochemical impedance spectroscopy (EIS) measurements were performed at −0.2 A/cm2 with a voltage amplitude of 50 mV, and the frequency limits were typically set in the range from 50 Hz to 500 kHz. The effective double-layer capacitance (Cdl) was obtained from the constant phase element (CPE) parameters and the two resistances using the Brug formula37,49:$${C}_{{dl}}={T}^{\frac{1}{P}}\left({\frac{1}{{R}_{s}}}+{\frac{1}{{R}_{{ct}}}}\right)^{\frac{P-1}{P}}$$
(3)
Where Rs is the solution resistance, Rct is the charge transfer resistance, T is CPE constant and P is CPE exponent. For the CO2 electroreduction tests, a Biologic VMP3 potentiostat was used for small current situations (≤ 400 mA), while the Biologic VMP3 potentiostat was connected to a VMP3 booster chassis with a 10 A current option to be used under large current situations (≥ 400 mA). The CO2RR performance of the CD-Ag HPE array electrodes with different tube numbers in the 2-electrode acidic system (pH = 1) was performed using an ANS6050D (50 V, 50 A) direct-current source from ANS Power Co., Ltd. with constant current control applied. In addition, both the anolyte and catholyte were cycled in anodic and cathodic compartments with a fixed flow rate of 50 mL/min by using two identical peristaltic pumps (JIHPUMP BT-50EA 153YX). For the long-term performance test of CO2 electroreduction, the current density was fixed at −2 A/cm2 in CO2-saturated 3 M KCl + 0.05 M H2SO4 catholyte, the exhaust from the cathodic compartment was measured by online GC during the entire 200-hour test.All the current in the main text and supplementary materials were geometrically normalized to the electrode area. All the applied potentials were recorded against the KCl-saturated Ag/AgCl reference electrode and then converted to those versus the reversible hydrogen electrode (RHE) with iR corrections using the following equation:$$E\left({{{\rm{vs}}}}.\, {{{\rm{RHE}}}}\right)=E\left({{{\rm{vs}}}}.\, {{{\rm{Ag}}}}/{{{\rm{AgCl}}}}\right)+0.197V+0.0591V\times {{{\rm{pH}}}}+0.85i{R}_{s}$$
(4)
Where E (vs. Ag/AgCl) is the applied potential, pH indicates the H+ concentrations of the electrolyte solutions (Supplementary Table 7), i is the current density at each applied potential, and Rs is the solution resistance obtained via EIS measurements (Supplementary Table 8). In order to avoid the overcorrected potentials, 85% iR correction was applied as the previous reports17,48. All applied potentials in the main text and Supplementary Information are referred to as RHE, unless otherwise stated. Note that the XY data at different pH values of Figs. 4a, c, e were first converted into XYZ data by the origin® software, to obtain the corresponding contour maps of Figs. 4b, d, f, respectively. Moreover, the XY data at different K+ concentrations of Fig. 5d in manuscript were first converted into XYZ data in origin® software, to obtain the corresponding contour map of Fig. 5f. In addition, the XY data at different current densities of Figs. 6a, d were first converted into XYZ data by the origin® software, to obtain the corresponding contour maps of Fig. 6b, e, respectively. Taking Fig. 4b for example, the X represents current density of Fig. 4a, Y represents pH of Fig. 4a, and Z represents percentage loss of CO2 of Fig. 4a, which were then standard smoothened and transformed back into a virtual matrix. The as-obtained virtual matrix was further presented in the form of a contour map, i.e., Fig. 4b, in consistency with the previous reports28,50.Product quantificationsGas-phase products from the cathodic compartment were directly vented into a gas chromatograph (GC-2014, Shimadzu) equipped with a Shincarbon ST80/100 column and a Porapak-Q80/100 column using a flame ionization detector (FID) and a thermal conductivity detector (TCD) during the electroreduction tests and analyzed online. FID was used for CO quantification (as well as CH4, C2H4 and C2H6), while TCD was used for H2 and CO quantification. In addition, the concentration of unreacted CO2 in the outlet gas was analyzed with an independent Agilent 7890B gas chromatograph (Supplementary Fig. 9). A thermal conductivity detector (TCD) with a carbon molecular sieves column (TDX-1) was used for CO2 quantification. All faradaic efficiencies reported were based on at least three different GC runs, and the error bars of all figures in this work were based on the standard deviations of at least five independent electrochemical tests, unless otherwise stated. High-purity argon (99.999%) was used as the GC carrier gas. In all the CO2 electrolysis tests, only H2 and CO were the gas-phase products, and their faradaic efficiencies were calculated as follows:$${FE}=\frac{{C}_{{product}}\times {10}^{-6}\times {v}_{{outlet}}\times {10}^{-3}\times t\times n\times F}{{V}_{m}\times Q}\times 100\%$$
(5)
where Cproduct is the concentration of the gas-phase products (ppm), νoutlet is the outlet flow rate of the electrolysis cell. Note that the exhaust flow rate of the electrolysis cell was not equal to the input CO2 flow rate at all due to the hydrogen evolution and CO2 consumption. Thus, the outlet flow rate of electrolysis cell was measured by an independent Alicat® mass flowmeter (Fig. 2i). The actual measured outlet flow rate was higher than the inlet flow rate due to HER occurrence (Supplementary Table 3 and Fig. 5f). Thus the actual exhaust outlet flow rates (corrected voutlet) were measured by the Alicat® mass flowmeter (Fig. 2i), which possessed a full scale of 50 sccm with the accuracy ± (0.8% of Reading + 0.2% of Full Scale), as shown in the note d of Supplementary Table 3. That is the experiment errors caused by the Alicat® mass flowmeter could be negligible under the CO2 electroreduction conditions. In addition, the Alicat® mass flowmeter results were calibrated using a mixture (including CO, H2, CO2 and water vapor compositions at room temperature) that approximates the actual outlet gas composition (Supplementary Table 4). t is the reaction time, n is the number of transferred electrons for producing CO or H2, F is the Faraday constant, Vm is the gas mole volume, and Q is the total quantity of the electric charge. The CO formation rate was calculated using the following equation:$${CO\; formation\; rate}=\frac{Q\times {{FE}}_{{CO}}}{F\times n\times t\times S}$$
(6)
Where S is the geometric area of the electrode (cm2).The jCO,limit(gas) is the theoretical limit of CO partial current density with all gas-phase CO2 molecules input into the electrolysis cell were reduced to CO products. The theoretical limits of CO product partial current density, i.e., jCO,lim(gas) were calculated using the following equation37,38:$${j}_{{CO},{lim}({gas})}=\frac{{nF}}{S}\frac{{v}_{{{CO}}_{2}}}{{V}_{m}}$$
(7)
So, the theoretical limits of CO FE, i.e., FECO,lim(gas) were calculated by the following equation:$${{FE}}_{{CO},{lim}({gas})}=\frac{{j}_{{CO},{lim}({gas})}}{{j}_{{total}}}\times 100\%$$
(8)
The CO2 SPCE, the CO2 carbonation and unreacted CO2 were calculated as follows:$${{{\rm{SPCE}}}}=\frac{{{produced\; CO}}}{{{Input}}\, {{{CO}}}_{2}}\times 100\%$$
(9)
$${{CO}}_{2}\, {carbonation}=\frac{{Input}\, {{CO}}_{2}-{Unreacted}\, {{CO}}_{2}-{Converted}\, {{CO}}_{2}\,}{{Input}\, {{CO}}_{2}}\times 100\%$$
(10)
$${Unreacted}\, {{CO}}_{2}={V}_{{outlet}}-{Prodeced\; CO}-{Prodeced}\, {H}_{2}$$
(11)
$${Converted}\, {{CO}}_{2}={Prodeced\; CO}$$
(12)
Where the input CO2 (Vinput) of electrolysis cell was controlled by a Alicat® mass flow controller, and the outlet flow rate (Voutlet) of electrolysis cell was measured by an independent Alicat® mass flowmeter. The actual outlet gas of electrolysis cell contains unreacted CO2, produced CO (from CO2RR) and produced H2 (from the HER). The produced CO and H2 at given current density could be quantified by online GC. Furthermore, to verify our calculation of unreacted CO2, the concentration of unreacted CO2 in the outlet gas was measured by an independent online GC (Fig. 2i).By assuming that the overpotential of oxygen evolution reaction on the anode side is zero, the cathodic energy efficiency for CO was calculated as follows12:$${{EE}}_{{CO}}=\frac{(1.23+(-{E}_{{CO}}))\times {{FE}}_{{CO}}}{1.23+(-E)}$$
(13)
Where ECO is −0.11 V (vs. RHE); 1.23 V is the thermodynamic potential for water oxidation in the anode side.Possible liquid-phase products from the cathodic compartment after CO2 electrolysis for 1 h were analyzed using another off-line GC-2014 (Shimadzu) equipped with a headspace injector and an OVI-G43 capillary column (Supelco, USA). No liquid-phase product was detected by the off-line GC. The post-reaction catholyte solution was also analyzed by a 600 MHz NMR spectrometer (Bruker) for possible liquid-phase products (especially formate and acetate). After an hour of electrolysis, an aliquot of catholyte solution (0.5 mL) was mixed with 0.1 mL (CH3)3Si(CH2)3SO3Na (DSS) (6 mM) and 0.1 mL D2O for use as internal standards. No liquid-phase product was detected by 1H NMR (Supplementary Figs. 6–9).COMSOL multiphysics simulationsA reaction-diffusion model (Supplementary Fig. 10) was used to simulate the local pH and CO2 concentration using COMSOL Multiphysics software in a typical 50 μm diffusion layer22,23,25. All the interactions between species in the electrolyte (CO2, HCO3-, CO32-, SO42-, K+, Cl-, OH-, H+ and H2O) were considered (Supplementary Table 11). Specifically, one end of the one-dimensional simulation area is set as the working electrode surface, and the other side is set as the bulk concentration to describe the bulk electrolyte. We used Henry’s law to calculate the CO2 concentration, assuming that the CO2 fugacity is 1 bar.$${C}_{{CO}2,{aq}}^{0}={K}_{H}^{0}{C}_{{CO}2,{gas}}^{0}$$
(14)
\({{{{\rm{K}}}}}_{{{{\rm{H}}}}}^{0}\) is the Henry’s constant, which can be calculated by using the equation below, where T is the temperature.$${{\mathrm{ln}}}{K}_{H}^{0}=93.457\times \frac{100}{T}-60.2409+23.3585\times {{\mathrm{ln}}}\frac{T}{100}$$
(15)
Due to the high concentration of the ions, the saturated concentration of CO2 in an electrolyte is corrected using the following equations.$$\log \frac{{C}_{{CO}2,{aq}}^{0}}{{C}_{{CO}2,{aq}}}={K}_{s}{C}_{s}$$
(16)
$${K}_{s}\,=\,\sum ({h}_{{ion}}\,+\,{h}_{G})$$
(17)
$${h}_{G}={h}_{G,0}+{h}_{T}(T-298.15)$$
(18)
Cs is the molar concentration and Ks is the Sechenov’s constant.We considered the following homogeneous and heterogenous reactions in our model, which are based on the previously published works22,23,25. The heterogenous reactions take place in the electrolyte as follow:$$2{H}_{2}O+2{e}^{-}\to {H}_{2}+2O{H}^{-}$$
(19)
$${{{\rm{C}}}}{O}_{2}+{H}_{2}O+2{e}^{-}\to {{{\rm{CO}}}}+2O{H}^{-}$$
(20)
In this work, two sets of Butler-Volmer boundary conditions (CO2RR and HER) were set to represent the two main sets of reactions on electrode surface, the kinetic parameters sets for these two reactions were the Tafel slopes of CO (103 mV dec−1) and H2 (56 mV dec−1), respectively, and the FEs of CO and H2 in the model were calculated and simulated based on these parameters.Over the whole domain, the following homogenous reactions occur:$${{{\rm{C}}}}{O}_{2}+{H}_{2}O\, \rightleftharpoons \, {H}^{+}+{HC}{O}_{3}^{-}$$
(21)
$${HC}{O}_{3}^{-}\, \rightleftharpoons \, {H}^{+}+C{O}_{3}^{2-}$$
(22)
$${{{\rm{C}}}}{O}_{2}+O{H}^{-}\, \rightleftharpoons \, {HC}{O}_{3}^{-}$$
(23)
$${HC}{O}_{3}^{-}+O{H}^{-}\, \rightleftharpoons \, C{O}_{3}^{2-}+{H}_{2}O$$
(24)
$${H}_{2}O\rightleftharpoons {H}^{+}+O{H}^{-}$$
(25)
The bulk concentrations and pH values were measured experimentally and implemented in the model. The thickness of the diffusion layer was assumed to be 50 μm. The electrode surface undergoes a reduction reaction, its current characteristics follow the Bulter-Volmer equation and the Nernst equation, its potential characteristics follow the Nernst equation, the concentration of the opposite body phase is set to the corresponding initial concentration, and the ion migration in the simulation area follows the Nernst Planck equation. The solution process is based on the MUMPS (multiple massively parallel spark direct solver) steady-state solver, and the relative tolerance and residual factor are set to 1E−8 and 1, respectively, eight layers of boundary layer subdivision are set on the simulated electrode surface to ensure the accuracy of the simulation results. The pH and CO2 concentration distribution near the electrode surface was calculated by solving the operating current from 0–1000 mA/cm2 at bulk pH of 1, 4, and 7, respectively.The electrode surface reaction follows the BV equation:$${i}_{{loc}}={i}_{0}\left(\exp \left(\frac{{\alpha }_{a}{Fn}\eta }{{RT}}\right)-\exp \left(\frac{-{\alpha }_{c}{Fn}\eta }{{RT}}\right)\right)$$
(26)
Where iloc is the local current density at the electrode/electrolyte interface, i0 is the exchange current density, \({{{{\rm{\alpha }}}}}_{{{{\rm{c}}}}}\) and \({{{{\rm{\alpha }}}}}_{{{{\rm{a}}}}}\) is the cathodic and anodic charge transfer coefficients, η is the activation overpotential.The balance potential follows the Nernst equation:$${E}_{{eq}}=-\frac{\triangle G}{{nF}}$$
(27)
$${E}_{{eq}}={E}_{{eq},{ref}}-\frac{{RT}}{{nF}}{{\mathrm{ln}}}{\prod}_{i}{\left(\frac{{a}_{i}}{{a}_{i,{ref}}}\right)}^{{\nu }_{i}}$$
(28)
Where Eeq is the electrode potential, ΔG is the Gibbs free energy, Eeq,ref is the standard electrode potential, \({{{{\rm{a}}}}}_{{{{\rm{i}}}}}\) is the (electrode reactive ion concentration),\(\,{{{{\rm{a}}}}}_{{{{\rm{i}}}},{{{\rm{ref}}}}}\) is the (standard electrode reactive ion concentration), vi is the reaction stoichiometric number.The transfer of dilute substances follows Fick’s law:$${N}_{i}={J}_{i}=-{D}_{i}\nabla {c}_{i}$$
(29)
$$\frac{\partial {c}_{i}}{\partial t}+\nabla {N}_{i}={R}_{i,{tot}}$$
(30)
Where Ji is the ion flux, Di is the diffusion coefficient (DLi = 1 × 10−7 m2s−1), ci is the concentration of ion, ∇ci is concentration gradient.The transportation of electricity follows the Nernst Planck relationship:$${N}_{i}=-{D}_{i}\nabla {c}_{i}-{z}_{i}{u}_{m,i}F{c}_{i}\nabla {{{\varnothing }}}_{l}={J}_{i}+u{c}_{i}$$
(31)
Where zi is the transfer number (zLi = 1), um,i is the electric mobility coefficient, ϕ is the electrolyte potential.Computational theoretical calculationsThe present first principle DFT calculations are performed by Vienna Ab initio Simulation Package with the projector augmented wave method51. The exchange function is treated using the generalized gradient approximation of Perdew-Burke-Emzerh functional52. For Kohn–Sham wave functions, the cutoff energy of the corresponding plane-wave basis set was set to 450 eV. The K points meshing was obtained from the Monkhorst-Pack scheme53. Grimme’s DFT-D3 methodology was used to describe the dispersion interactions. A vacuum width of of 15 Å along the Z axis was created to ensure negligible interaction. The force convergence criterion was set to 0.02 eV/Å and energy convergence criterion was 10−4 eV. To fully consider the solvation effect, 18 explicit water molecules were optimized, and a local minimum via the hydrogen bond network was formed. The Gibbs free energy change (ΔG) of each step is calculated using the following formula:$$\triangle G\,=\,\triangle E\,+\,\triangle {ZPE}-\,T\triangle S$$
(32)
where ΔE is the electronic energy difference directly obtained from DFT calculations, ΔZPE is the zero point energy difference, T is the room temperature (298.15 K) and ΔS is the entropy change.
Ampere-level CO2 electroreduction with single-pass conversion exceeding 85% in acid over silver penetration electrodes
