Hydroxide promotes carbon dioxide electroreduction to ethanol on copper via tuning of adsorbed hydrogen

Producing liquid fuels such as ethanol from CO2, H2O, and renewable electricity offers a route to store sustainable energy. The search for efficient electrocatalysts for the CO2 reduction reaction relies on tuning the adsorption strength of carbonaceous intermediates. Here, we report a complementary approach in which we utilize hydroxide and oxide doping of a catalyst surface to tune the adsorbed hydrogen on Cu. Density functional theory studies indicate that this doping accelerates water dissociation and changes the hydrogen adsorption energy on Cu. We synthesize and investigate a suite of metal-hydroxide-interface-doped-Cu catalysts, and find that the most efficient, Ce(OH)x-doped-Cu, exhibits an ethanol Faradaic efficiency of 43% and a partial current density of 128 mA cm−2. Mechanistic studies, wherein we combine investigation of hydrogen evolution performance with the results of operando Raman spectroscopy, show that adsorbed hydrogen hydrogenates surface *HCCOH, a key intermediate whose fate determines branching to ethanol versus ethylene.

T he electrochemical generation of C 2+ fuels and chemicals from CO 2 and H 2 O enables the storage of intermittent renewable energy [1][2][3][4][5][6] . Substantial progress has been made in producing gaseous ethylene from the CO 2 reduction reaction (CO 2 RR), and the Faradaic efficiency (FE) now exceeds 70% at an overpotential of 0.55 V (ref. 7 ).
By contrast, the electrochemical conversion of CO 2 to liquid ethanola promising renewable fuel with high energy density and compatibility with existing storage and transportation infrastructurehas seen more limited progress thus far. Indeed, today's best CO 2 electrocatalysts fail to provide majority ethanol production, instead preferring ethylene [8][9][10][11] .
The optimization of intermediate binding energetics provides a framework in which to evaluate and design for desired electrocatalytic performance 12,13 . For CO 2 RR specifically, the binding energy of CO (ΔE CO ) is an important descriptor that has enabled the prediction of a number of promising CO 2 RR candidates 14 . Experimental studies have similarly followed the correlation between adsorbed CO (CO ad ) as a function of alloying/doping with elements, including Zn (ref. 15 ), Ag (refs [16][17][18] ), Au (ref. 19 ), S (ref. 20 ), B (ref. 21 ), and N (ref. 22 ), as well as with engineering of facets 23 and morphology [24][25][26] .
In spite of these impressive efforts, the FE of CO 2 -to-ethanol remains below 25% if one focuses on studies that achieve commercially-relevant current densities (>100 mA cm −2 ). Even if one includes results down to 6 mA cm −2 , it has reached only 29% FE.
The scaling relationships among the carbonaceous intermediates 27 along the multi-step reduction pathway to ethanol mean that it is difficultif only a single site, and thus one degree of freedom, is relied upon to engineer catalyst adsorption energie s-to optimize ΔE CO simultaneously with the initial CO 2 adsorption; as well as to optimize the site for ensuing carbon-carbon coupling; and to optimize also the subsequent hydrogenation step. The formation of ethanol with high FE will rely on accessing experimental degrees of freedom that engineer these steps.
We reasoned that, since hydrogen (H ad ) co-exists with carbonbased intermediates during CO 2 RR, controlling its presence could potentially offer a new handle to help break the scaling relations. There exist hints at this possibility in the prior literature, such as in the proposed ethylene pathway in which hot water hydrogenates adsorbed *HCCOH, the penultimate reaction intermediate for both ethanol and ethylene 28 . Experimentally, the direct involvement of water in producing ethanol from CO 2 RR was clarified in a recent isotopic study 29 .
We, therefore, pursued means to activate near-surface water molecules with the goal of boosting the production of ethanol.
Our thinking was that cleaving the Cu-C bond of adsorbed *HCCOH could thereby be promoted, favoring thereby the electroproduction of ethanol.
We begin with an investigation of how surface H ad affects the selectivity of CO 2 RR on Cu. To this end, we designed a catalytic system that allows us to construct both hydroxide-and oxidedoped Cu having tunable surface H ad coverage. DFT studies reveal that this new catalyst facilitates water dissociation and favors H ad formation. We then synthesize Pourbaix-stable hydroxide-doped and oxide-doped Cu catalysts and investigate them both ex situ and in situ. In all cases, we achieve a notable increase in the ratio of ethanol to ethylene production, documenting fully a doubling on the doped-Cu catalysts compared to Cu.
The best of these, Ce(OH) x -doped-Cu, reaches a FE of 43% for ethanol at an operating current density of 300 mA cm −2 .
Mechanistic studies indicate that surface H ad favors the ethanol pathway over ethylene.

Results
DFT investigations of the effect of H ad . Since OH − plays a beneficial role in promoting carbon-carbon coupling 7,30 , CO 2 RR electrocatalysis is carried out today in neutral or alkaline aqueous environment. In this environment, H 2 O molecules serve as the proton source for CO 2 RR.
Cleavage of the H-OH bond is needed to form H ad on catalytic surfacethe Volmer step in the hydrogen evolution reaction, HER. This accounts for the slower HER rate in alkaline and neutral media relative to that in acidic media 31 . Previous studies have demonstrated that the introduction of hydroxides or oxides increases H ad coverage by accelerating the water dissociation step 32,33 . The approach tunes H ad without the need to modify the bulk pH.
We thus reasoned that doping Cu with a stable hydroxide or oxide could enhance the surface H ad ; yet allow us to maintain the alkaline environment that favors carbon-carbon coupling. We carried out DFT calculations on Ce oxide-and Mn oxide-doped-Cu(111) (see the Methods section for details). Ce and Mn oxides were chosen due to their Pourbaix-stability under the reducing potentials used in CO 2 RR 34 . We examined the water dissociation energy and the H adsorption energy (E H ) on both bare and doped-Cu(111). Figure 1 and Supplementary Fig. 1 show that water dissociation is more favorable on the oxide/hydroxidedoped-Cu surface (by 0.48 eV and 0.39 eV for Ce and Mn oxide, respectively) in comparison with that on the pure Cu. The adsorption of hydrogen is also stabilized on the oxide-doped-Cu surface. E H is more favorable on Mn oxide/Cu than that of Ce oxide/Cu, which suggests that doping metal oxides on Cu provides the mean to control the extent of local H ad . These findings suggested that building such hybrid catalysts could enable us to investigate and exploit H ad in CO 2 RR.
Catalyst synthesis and characterization. To synthesize the hybrid catalysts, we began with Cu-sputtered-polytetrafluoroethylene (Cu/ PTFE) as the substrate 7 , and we deposited either hydroxides or oxides via electrochemical or sputtering methods (Methods). The electrochemical deposition of metallic hydroxide was carried out in a neutral electrolyte containing the corresponding metallic nitrate as the precursor 35 ( Supplementary Fig. 2). A cathodic current was first applied to the Cu/PTFE electrode to generate OH − from nitrate reduction. The metallic hydroxide was then deposited onto the Cu surface via chemical reaction between the metallic cation and locally-generated OH − . During the electrochemical deposition, the color of the Cu surface turned to brown.
To understand the nature of the doped hydroxides, we carried out characterization of the cerium hydroxide-doped-Cu/PTFE sample (denoted Ce(OH) x /Cu/PTFE) using scanning electron microscopy (SEM), transmission electron microscopy (TEM) and scanning transmission electron microscopy (STEM), as well as Xray diffraction (XRD) and X-ray photoelectron spectroscopy (XPS). Figure 2a shows the typical 3-dimensional networked structure after Ce(OH) x (Fig. 2a) electrochemically deposited into Cu/PTFE fibers. We propose that this structure facilitates CO 2 gas penetration to the triple-phase reaction region. STEM elemental mapping shows a homogeneous distribution of Cu and Ce throughout a single fiber (Fig. 2b). High-magnification SEM images of Ce(OH) x /Cu/PTFE further reveal substantially uniformly-decorated nano-islands on the surface ( Fig. 2c) with an average size of 18 nm and a typical range of 6-30 nm ( Supplementary Fig. 3). High-resolution TEM reveals the interface between Ce(OH) x and Cu/PTFE (Fig. 2d). The corresponding Fast Fourier Transform (FFT, the inset of Fig. 2d) pattern matches that of Cu 2 O (111), indicating that the Cu/PTFEbased sample was partially oxidized. The oxidation of Cu is also seen in the XRD patterns of Cu/PTFE ( Supplementary Fig. 4a), in which the diffraction peaks corresponding to both Cu and Cu 2 O are observed.
The absence of observable lattice spacings in HRTEM, and a corresponding lack of crystalline peaks in XRD, suggest that the electrochemically-deposited Ce(OH) x exists in an amorphous structure, in agreement with the previous reports 35 . Highresolution XPS spectra for the Cu 2p region further show the co-existence of both metallic and oxidized states ( Supplementary  Fig. 4b). The Ce 3d spectra shows the co-existence of Ce 4+ and Ce 3+ , indicating that the deposition of cerium species (Supplementary Fig. 4c, d) was indeed achieved. The O 1s spectra confirm that the cerium species exist as hydroxide (Supplementary Fig. 4e).
Since the chemical states of metals are dependent on the applied potential 36 , we carried out operando X-ray adsorption spectroscopy (XAS) to monitor the oxidation states of Cu and Ce during CO 2 RR electrocatalysis by looking into the Cu K-edge and Ce L 3 -edge, respectively. We found thatin agreement with the XPS results -Cu species were slightly oxidized before the reaction (Fig. 2e, Supplementary Fig. 5). However, once a negative potential had been applied during CO 2 RR, only peaks corresponding to metallic Cu were observed ( Supplementary Fig. 5). No change of Cu local structure (i.e., oxidation state, coordination number, and bond distance) was observed throughout CO 2 RR process 37 (Supplementary Fig. 6 and Supplementary Table 1). Fig. 2f shows that, once a potential of −0.57 V vs. RHE was applied, the chemical state of Ce underwent an initial reduction. The ratio of Ce 3+ /Ce 4+ slightly increased from −0.57 V to −0.64 V vs. RHE, after which it remained unchanged upon furtherincreased reducing potentials. We conclude that an interface is  Fig. 7 and Supplementary Tables 2-5).
The FE for ethanol reached 43% when the Ce(OH) x /Cu/PTFE catalyst was employedwell above the value of 29% for the Cu control. The Ce(OH) x /Cu/PTFE catalyst also achieved an impressive partial current density, 128 mA cm −2 (Fig. 3a), for ethanol, compared to 87 mA cm −2 for the copper control. By comparing the product distributions at the optimal potentials, we found that Ce(OH) x doping had increased the FE toward H 2 by 5% compared to the Cu/PTFE baseline (Fig. 3b). This is consistent with DFT results that indicate that surface H ad is enhanced via accelerated water dissociation and optimized hydrogen adsorption. The ethanol:ethylene ratio increased from 0.65 (Cu/PTFE) to 1.26 (Ce(OH) x /Cu/PTFE), which is the highest among electrocatalysts that achieve a current density of >6 mA cm −2 (Supplementary Table 2). The energy efficiency as a function of partial current density for ethanol (Fig. 3c) reveals that the Ce(OH) x /Cu/PTFE achieved an energy efficiency of 25%. This is, by a factor of 1.6, the highest reported for systems operating above 10 mA cm −2 (Supplementary Table 6).
We also evaluated operating stability of the Ce(OH) x /Cu/PTFE catalyst. It provided stable operation over an initial 6 h at current density 250 mA cm −2 (Fig. 3d). TEM and SEM images of Ce (OH) x /Cu/PTFE electrode after reaction showed the preservation of the hydroxide/Cu interface, as well as of the well-dispersed Ce (OH) x nano-islands on the sputtered Cu surface ( Supplementary  Fig. 8).
We used labeled 13 CO 2 and confirmed that the ethanol was produced from CO 2 . This test indicates that ethanol contamination is not a source of artefactual ethanol (Supplementary Fig. 9a). The small changes (within 3%) in electrochemically active surface area (ECSA) of the Cu/PTFE before and after the deposition of Ce (OH) x also excluded the influence of surface area differences on electrocatalytic performance ( Supplementary Fig. 9b-d). Due to its electrical insulation, the PTFE substrate is not expected to affect ECSA measurements.
To investigate whether the materials design strategy herein offers a general way to tune CO 2 RR selectivity, we further tested other stable hydroxide-and oxide-doped Cu catalysts including Ga(OH) 3 , Mn(OH) 3 , Zr(OH) 4 and TiO 2 (ref. 34 ). A suite of microscopy and spectroscopy analysis confirmed they are structurally analogous with Ce(OH) x /Cu/PTFE (Supplementary Figs. 10-13). Similar electrocatalytic behaviors were observed in the doped samples: compared to Cu/PTFE controls, H 2 and CH 4 production increased, and C 2 H 4 decreased (Fig. 3e). A positive correlation was observed between the FE ratio of ethanol/ethylene and the FE of H 2 (FE H2 ), with TiO 2 /Cu/ PTFE exhibiting the highest ethanol/ethylene ratio of 1.51. This agrees with the hypothesis that enhanced H ad promotes ethanol over ethylene. The HER activities of the samples evaluated using the same flow cell system in Ar atmosphere showed the same trend as the FE H2 during CO 2 RR; i.e., TiO 2 /Cu/PTFE > Zr(OH) 4 /Cu/ PTFE > Mn(OH) 3 /Cu/PTFE > Ga(OH) 3 /Cu/PTFE > Ce(OH) x /Cu/ PTFE, further confirming the enhanced H ad on Cu due to an accelerated Volmer step (Supplementary Fig. 14).
Mechanistic studies. We then sought further mechanistic insight into the selectivity of CO 2 RR. Goddard and co-workers 28 (Fig. 4a, d). The ethylene pathway was proposed to be related to the removal of OH in *HCCOH (Fig. 4e) form *CCH (Fig. 4f); from which *CCH is then further hydrogenated, generating ethylene. In contradistinction, *HCCOH is hydrogenated into *HCCHOH (Fig. 4a-c) in the ethanol pathway.
As seen in Fig. 4d-f, surface water molecules are involved in the removal of OH: the hydroxyl group in *HCCOH is surrounded by five other water molecules with hydrogen bonds. In the transition state, the O-C bond between the hydroxyl group and *CCH dissociates with the help of surface water. In the final state, OH is stabilized by water and *CCH is formed. Thus, surface water plays an important role in the ethylene pathway. In the ethanol pathway, the H ad attacks the *HCCOH (Fig. 4b), forming *HCCHOH, the key intermediate towards ethanol. H ad is only involved in the branching reaction towards ethanol. When we enhance H ad coverage, ethanol selectivity is enhanced (Fig. 4g).
To probe experimentally whether hydroxide modification also impacts the adsorption of carbonaceous intermediates on Cu, we carried out in situ Raman measurements and compared bare Cu/ PTFE with Ce(OH) x /Cu/PTFE across the potential region −0.24 to −0.73 V under CO 2 RR (Supplementary Fig. 15). Due to their short life time, we are unable to provide direct experimental evidence for the *HCCOH intermediates; however, we found negligible influence of Ce(OH) x on adsorbed CO (CO ad )the Raman shift of frustrated rotation, and stretching, associated with Cu-CO, remained in the same position after Ce(OH) x -modification of the Cu surface. Given the scaling relationship between CO ad and other carbonaceous intermediates 14 , we deduced that the electrocatalytic differences between bare Cu/PTFE and Ce (OH) x /Cu/PTFE were unlikely to have originated from changes in the adsorption of carbonaceous species.
It is worth noting that the hydroxide deposition on Cu also promotes the CH 4 production from CO 2 RR. Buonsanti and coworkers 39 recently reported the colloidal synthesis of a class of Cu/CeO 2-x heterodimers that showed a CO 2 -to-CH 4 FE of 54% in KHCO 3 solution, exceeding the physically-mixed and individual controls. With the aid of DFT studies, they assigned the enhanced CH 4 production to the interface comprised of Cu, Ce, and Ovacancy sites that enabled breaking of the CHO*/CO* scaling relation. This mechanism investigated herein may contain analogies with how the hydroxide/Cu interface promotes CH 4 production through the C1 pathway.

Discussion
In summary, we reported an approach to higher-efficiency CO 2to-ethanol conversion levering tuning of the adsorption of hydrogen on Cu. The cerium hydroxide-doped copper catalyst provided a 43% FE at a total current density of 300 mA cm −2 . Mechanistic studies indicated that H ad on Cu favors the ethanol over the ethylene pathway by attacking the Cu-C bond of the *HCCOH intermediate. The findings suggest further avenues to engineer hybrid catalysts that contribute multiple degrees of freedom to the design of multi-step CO 2 reduction reactions.

Methods
Electrode preparation. Cu/PTFE electrodes were prepared by sputtering a Cu layer of 300 nm in thickness onto a PTFE membrane (average pore size of 450 nm) using a Cu target (99.99%) at a rate of 1 Å s −1 .
Using the Cu/PTFE as the substrate, we electrochemically deposited various hydroxides in a three-electrode electrochemical cell. A potentiostat (Metrohm-Autolab, PGSTAT204) was used for the electrodeposition. The Cu/PTFE, a platinum foil and an Ag/AgCl electrode (saturated with KCl) were used as the working, counter and reference electrodes, respectively. The electrodeposition solution comprised 0.1 M KCl as the supporting electrolyte, and 0.025 M corresponding nitrate salts (cerium nitrate, 99.99%, Sigma-Aldrich; gallium nitrate, 99.9%, Sigma-Aldrich; zirconium oxynitrate, 99.99%, Sigma-Aldrich; manganese nitrate, 99.99%, Sigma-Aldrich) as the precursor. A current density of −0.5 mA cm −2 was held for a defined length of time (10,20,30,40, and 50 min) to achieve varied surface coverage of hydroxides. Following the completion of the deposition, the working electrode was rinsed with DI water for at least three times and subsequently dried in N 2 atmosphere. Due to the instability of titanium nitrate, we deposited titanium oxides onto Cu/PTFE via a sequential sputtering of Cu followed by a layer of TiO 2 (5 nm).
Materials characterization. The morphology of the electrodes was characterized using scanning electron microscopy (SEM, Hitachi S-5200) with a 5-kV beam voltage. Transmission electron microscopy (TEM) and elemental mapping images were collected using a Hitachi HF-3300, at an acceleration voltage of 300 kV, equipped with a Bruker energy dispersive X-ray spectroscopy (EDX) detector. The acquisition time in the EDX studies was 3 min. Powder X-ray diffraction (XRD) patterns were recorded using a Bruker D8 using Cu-Kα radiation (λ = 0.15406 nm). X-ray photoelectron spectroscopy (XPS) was conducted on a PHI 5700 ESCA System using Al Kα X-ray radiation (1486.6 eV) for excitation. Operando X-ray absorption spectroscopy (XAS) investigations were carried out at the 9BM beamline of the Advanced Photon Source (APS) located in the Argonne National Laboratory (Lemont, IL). Detailed information regarding operando XAS tests in flow cells is available in a previous report 37 . Cu 2 O, Cu foil, cerium oxide, and cerium oxalate hydrate were used as the reference samples. In situ Raman measurements were performed on a Renishaw inVia Raman Microscope in a modified flow cell and a water immersion objective (×63 ) with a 785 nm laser, using a 5 s integration and averaging 20 scans per region. In the above systems, platinum wire and an Ag/AgCl electrode were used as the counter and reference electrode, respectively.
Electrochemical measurements. Electrochemical studies were carried out using an electrochemical flow cell consisting of a gas chamber, a cathodic chamber, and an anodic chamber. The PTFE-based working electrode was fixed between the gas and cathodic chambers, with the catalysts layer side facing the cathodic chamber (geometric active surface area of 1 cm 2 ). An anion exchange membrane (Fumasep FAA-3-PK-130) was used to separate the anodic and cathodic chambers. All electrochemical tests were conducted on an Autolab PGSTAT204, with an Ag/AgCl electrode and Ni foam being the reference and counter electrodes, respectively. Potentials were converted to the reversible hydrogen electrode scale after iR  correction. Electrochemical impedance spectroscopy (EIS) in the frequency range of 10 5 -10 −1 Hz and an amplitude of 10 mV was used to determine the R value. For performance studies, 1 M KOH was used as the electrolyte, and it was circulated through the cathodic and anodic chambers using peristaltic pumps at a rate of 10 mL min −1 . The flow rate of CO 2 gas through the gas chamber was controlled to be 50 sccm using a digital gas flow controller. Gas chromatography (PerkinElmer Clarus 600) with a flame ionization detector (FID) and a thermal conductivity detector (TCD) was used to analyze the gas products, collected from the end of the gas chamber. 1 H NMR spectroscopy (600 MHz, Agilent DD2 NMR Spectrometer) with water suppression was used to analyze the liquid products, using D 2 O and DMSO as the lock solvent and internal reference, respectively. The hydrogen evolution reaction (HER) activities of various electrodes were evaluated in the same flow cell system, with the flow gas changed from CO 2 to Ar.
DFT calculations. All DFT calculations were carried out using the Vienna ab initio simulation program (VASP; https://vasp.at/) [40][41][42][43] . The projected augmented wave approach 44,45 was used to describe the electron-ion interactions with cutoff energy at 450 eV. The generalized gradient approximation with the Perdew, Burke and Ernzerhof exchange correlation functional was used 46 . For all the cerium related calculations, due to the strong correlations of the partially filled Ce 4f states, we employed the Hubbard parameter, U, to illustrate the on-site coulombic interaction 47 . A U-J value of 4.5 eV was chosen for Ce according to a previous study 48 . In order to illustrate the long-range dispersion interactions between the adsorbates and catalysts, we employed the D3 correction method by Grimme et al. 49 . Brillouin zone integration was accomplished using a 3 × 3 × 1 Monkhorst-Pack k-point mesh. Four layers of Cu(111) surface was optimized, with the top 2 layers relaxed and bottom 2 layers fixed. Two molecular units of CeO 2 were introduced on to the Cu surface and optimized. To keep the model consistent, we replaced Ce atoms with Mn atoms and re-optimized the structure for the Mn oxide calculations. The water dissociation energy was calculated using E dissociation ¼ E H 2 O Ã À E HÃþOHÃ , and the hydrogen adsorption energy was calculated using E H ad ¼ E H Ã þ E slab À 0:5E H 2 , where * designates a surface adsorbed specie.

Data availability
The data that support the findings of this study are available from the corresponding author on reasonable request.