Essential role of lattice oxygen in methanol electrochemical refinery toward formate

Abstract

Developing technologies based on the concept of methanol electrochemical refinery (e-refinery) is promising for carbon-neutral chemical manufacturing. However, a lack of mechanism understanding and material properties that control the methanol e-refinery catalytic performances hinders the discovery of efficient catalysts. Here, using ¹⁸O isotope–labeled catalysts, we find that the oxygen atoms in formate generated during the methanol e-refinery reaction can originate from the catalysts’ lattice oxygen and the O-2p-band center levels can serve as an effective descriptor to predict the catalytic performance of the catalysts, namely, the formate production rates and Faradaic efficiencies. Moreover, the identified descriptor is consolidated by additional catalysts and theoretical mechanisms from density functional theory. This work provides direct experimental evidence of lattice oxygen participation and offers an efficient design principle for the methanol e-refinery reaction to formate, which may open up new research directions in understanding and designing electrified conversions of small molecules.

Lattice oxygen of catalysts plays an essential role in promoting the methanol electrochemical refinery towards formate.

INTRODUCTION

The industry sector is estimated to contribute around 30% of the world’s annual greenhouse gas (GHG) emissions, with the chemical sector significantly contributing to these figures (1). A key challenge to achieving 2050 climate targets is transitioning from existing fossil fuel–based processes to more sustainable and carbon-neutral concepts and processes. Electrifying the chemical industry will have an enormous global impact (2). Therefore, implementing innovative electrochemical refinery (e-refinery) concepts (3) is one of the most promising transitional paths to address carbon-neutral manufacturing in the time frame needed to address 2050 climate objectives. Moreover, this concept offers better control of the driving force of the electrocatalytic reactions for selectivity tuning and thus opens up a substantial opportunity for optimizing partial reduction or oxidation conversions (4, 5).

In this regard, we focus on the methanol e-refinery reaction toward formate. On the one hand, methanol is an excellent energy vector, with high energy density, a simple structure, and high stability under alkaline conditions. Moreover, methanol can be mass-produced via CO₂ reduction and biomass conversion (6). Thus, developing carbon-neutral chemical reactions to upgrade CO₂-/biomass-derived methanol into high-valued chemicals can significantly contribute to GHG mitigation. On the other hand, formate is a key chemical due to its multiple applications across various industries, such as oil and gas, textile, printing, aviation, and others (7, 8). The industrial manufacturing of formate involves an energy and cost-intensive process based on a high-temperature/high-pressure synthesis using a mixture of CO and methanol followed by the subsequent hydrolysis of the resulting methyl formate (9). As a result, the market price of formate is around twice that of methanol (10, 11). In addition, methanol e-refinery could be a more efficient anode reaction to replace the sluggish oxygen evolution reaction (OER) in water electrolysis, which consumes nearly 90% of the electricity driving these reactions (12) but electro-generates a product of negligible economic value (O₂). Therefore, methanol e-refinery toward formate is considered an attractive technology, which has high energy efficiency while concurrently producing chemicals with even higher economic value than methanol (13).

To rationally design methanol e-refinery catalysts, an in-depth mechanistic understanding of the reaction is necessary. In addition, the conventional methanol oxidation reaction targets at the complete oxidation of methanol to CO₂/carbonate, which involves four deprotonation processes and the oxidation of the resulting CO intermediate (14). For this methanol-to-CO₂/carbonate process, quite a few factors have been investigated, including the coordination environment of the reaction center (15), the composition (16), and the defects of the catalysts (17). Recently, the possible influence of lattice oxygen is proposed in a study of density functional theory (DFT) by Chandrasekharan Meenu et al. (18). In contrast, the reaction mechanism of methanol e-refinery toward formate is not well established to date. Trial-and-error method is still the primary approach to searching for methanol e-refinery catalysts. Correspondingly, only a few materials have been reported for the methanol e-refinery reaction toward formate, i.e., Co(OH)₂ (19), Ni_0.33Co_0.67(OH)₂ (20), LaCo_0.5Fe_0.5O₃ (21), Ni_0.75Fe_0.25Se₂ (22), Ni₃C (23), NiFe LDH@NiMo (24), NiO/NF (25), Co_xP@NiCo-LDH (26), Ni₂P-CoP (27), Ni₃S₂-CNF (28), Cu₂Se/Co₃Se₄ (29), Ni(OH)₂ (30), CuO (13), and NiS (31). In particular, to the best of our knowledge, rare attention has been paid to distinguishing surface from bulk catalysis to analyze the degree to which the lattice oxygen participation could promote the methanol e-refinery activity. Hence, it is of great importance to figure out the lattice oxygen–modulated mechanism during the selective methanol conversion, in which a fundamental understanding of the mechanistic pathway based on physicochemical properties is crucial.

This study presents direct experimental evidence for the participation of lattice oxygen during the methanol e-refinery reaction. Gas chromatography–mass spectrometry (GCMS) measurements of ¹⁸O-labeled catalysts (32) reveal that, for highly covalent catalysts such as SrCoO_3−σ and NiO_xH_y, the lattice oxygen participates in the reaction; however, for less covalent LaCoO₃ and FeO_xH_y, the lattice oxygen is not involved. Such distinct behaviors are assigned to the disparate O-2p-band center levels between them, which is identified as an effective descriptor for the methanol e-refinery performance in this work. Notably, NiO_xH_y, with the highest O-2p-band center level among all the investigated catalysts, exhibits a near 100% formate Faradaic efficiency over a wide range of current densities in the membrane electrode assembly (MEA) tests. In light of those findings, we present potential methanol e-refinery mechanisms that encompass the participation of lattice oxygen to explain the correlation between the O-2p-band center levels and formate production rates or Faradaic efficiencies of the catalysts. These findings will initiate the investigation of electronic structure parameters that can serve as activity descriptors to develop more e-refinery processes to convert small organic molecules. Notably, e-refinery chemistry following this mechanism is expected to bypass the limitations inherent in adsorption-based mechanisms where scaling relations constrain performance, thereby potentially offering better control and understanding of the e-refinery processes.

RESULTS

Evidence of lattice oxygen involvement during methanol e-refinery

Inspired by the outstanding OER performance with the activation of lattice oxygen, a series of catalysts with distinct O-2p-band center levels (LaCoO₃, SrCoO_3−σ, FeO_xH_y, and NiO_xH_y) (15, 32, 33) are synthesized and compared, which have thus exhibited disparate reaction routes in OER (32, 34). K-edge x-ray absorption near edge structure (XANES) and extended x-ray absorption fine structure (EXAFS) analyses are applied to the as-synthesized catalysts to investigate the local atomic structure of the central metal ions (Co for LaCoO₃ and SrCoO_3−σ, Fe for FeO_xH_y, and Ni for NiO_xH_y). According to the XANES spectra, the relative K-edge positions of LaCoO₃, SrCoO_3−σ, FeO_xH_y, and NiO_xH_y match with the LiCoO₂, LiCoO₂, FeSO₄, and NiO references, which indicates that the nominal Co, Fe, and Ni valence states are +3, +3, and +2, respectively (Fig. 1, A to C). The Co-O/Co-Co, Fe-O/Fe-Fe, and Ni-O/Ni-Ni bonds of LaCoO₃/SrCoO_3−σ (35, 36), FeO_xH_y (37), and NiO_xH_y (38) are verified by the EXAFS profiles (Fig. 1, D to F), respectively. SrCoO_3−σ has a slightly smaller coordination number than LaCoO₃, according to the slightly weaker peak intensity at 1.4 Å (Fig. 1D). This divergency indicates the lower lattice oxygen stability and the higher lattice oxygen reactivity on SrCoO_3−σ. Figure 1G presents the x-ray diffraction (XRD) patterns of LaCoO₃, SrCoO_3−σ, FeO_xH_y, and NiO_xH_y. No impurity peaks are observed. Then, the density of states (DOS) of LaCoO₃, SrCoO_3−σ, FeO_xH_y, and NiO_xH_y are calculated to confirm the claim in the literature that a notable O-2p–projected DOS discrepancy exists between each pair of the catalysts (15, 32, 33), as exhibited in fig. S1 and fig. S2 (A and E).

Fig. 1. Characterizations of LaCoO₃, SrCoO_3−σ, FeO_xH_y, and NiO_xH_y and the direct evidence of lattice oxygen participation in the methanol e-refinery reaction for SrCoO_3−σ and NiO_xH_y.

(A to C) X-ray absorption near edge structure (XANES), (D to F) extended x-ray absorption fine structure (EXAFS), and (G) x-ray diffraction (XRD) patterns of LaCoO₃, SrCoO_3−σ, FeO_xH_y, and NiO_xH_y. (H) HC¹⁶O¹⁸OH/HC¹⁶O¹⁶OH ratios of ¹⁸O-labeled catalysts (LaCo¹⁸O₃, SrCo¹⁸O_3−σ, Fe¹⁸O_xH_y, and Ni¹⁸O_xH_y) in contrast with unlabeled comparisons (LaCoO₃, SrCoO_3−σ, FeO_xH_y, and NiO_xH_y). (I) Schematic band diagrams of LaCoO₃, SrCoO_3−σ, FeO_xH_y, and NiO_xH_y. As presented, the shift of the O-2p bands toward the O₂/H₂O redox couples (4OH⁻ → O₂ + 2H₂O + 4e⁻ E^θ = 1.23 V versus RHE in alkaline conditions) makes the lattice oxygen of the catalysts a more preferential nucleophilic attacking species than OH⁻ during the methanol e-refinery reaction (32). Here, a.u. and E_F represent arbitrary units and Fermi level energy, respectively.

To check whether lattice oxygen is involved during the methanol e-refinery reaction, the ¹⁸O ratios in the methanol e-refinery product (formate) are detected for ¹⁸O-labeled LaCoO₃, SrCoO_3−σ, FeO_xH_y, and NiO_xH_y and unlabeled comparisons. In this way, the possible influence of natural ¹⁸O abundance (~0.2%) (32) can be ruled out. The catalyst particles are labeled with ¹⁸O according to the previously reported approach (32), i.e., by 10-min chronoamperometry (CA) in H₂¹⁸O (95%)–labeled 1 M KOH solution at 1.6 V versus reversible hydrogen electrode (RHE). The corresponding CA curve for NiO_xH_y is presented in fig. S3 as an example. Then, 3-hour CA is performed on those electrodes in an unlabeled 1 M KOH + 1 M CH₃OH solution after rinsing them with ¹⁶O water to remove H₂¹⁸O. Last, the electrolyte is neutralized by an HCl solution followed by a distillation process to obtain a formic acid solution, which is ready for the GCMS test.

In the GCMS results, the mass-to-charge ratio (m/z) signal of 46 represents HC¹⁶O¹⁶OH, of 48 represents HC¹⁶O¹⁸OH, and of 50 represents HC¹⁸O¹⁸OH. The signal of m/z = 48 and m/z = 50 are normalized by the signal of m/z = 46. At least three independent GCMS tests are conducted for each HCOOH solution. The results for m/z = 48 (HC¹⁶O¹⁸OH) and m/z = 50 (HC¹⁸O¹⁸OH) are displayed in Fig. 1H and fig. S4, respectively. There is no normalized m/z = 48 (Fig. 1H) or m/z = 50 (fig. S4) detected on LaCo¹⁸O₃ and Fe¹⁸O_xH_y compared to that on LaCoO₃ and FeO_xH_y, which indicates that no lattice oxygen is involved during the reaction. For SrCo¹⁸O_3−σ and Ni¹⁸O_xH_y, the normalized signal of m/z = 48 is above the unlabeled control experiments (Fig. 1H), which represents the successful ¹⁸O labeling and the participation of lattice oxygen during the methanol e-refinery reaction (no detectable signals are shown for m/z = 50, as shown in fig. S4). The existence of HC¹⁶O¹⁸OH (with one oxygen from the oxide lattice; m/z = 48 in Fig. 1H) requires the activation of lattice oxygen sites and the formation of oxygen vacancies during the methanol e-refinery process. The participation of lattice oxygen has been found in OER (32, 39, 40), and the physical origin of this phenomenon is probably the same for both methanol e-refinery and OER. As shown in Fig. 1I, the O-2p-band centers of SrCoO_3−σ and NiO_xH_y are closer to the Fermi levels than that of LaCoO₃ and FeO_xH_y, and, when the catalysts’ O-2p states are sufficiently close to the Fermi levels (approaching the redox potential of the O₂/H₂O couple), the participation of lattice oxygen will become thermodynamically favorable (32). Thus, the elevation of the O-2p band may also be the physical origin of the participation of lattice oxygen during the methanol e-refinery reaction.

The divergence in the HC¹⁶O¹⁸OH ratios of the two pairs of catalysts (LaCoO₃, SrCoO_3−σ, FeO_xH_y, and NiO_xH_y) confirms that the OER mechanism discrepancy resulting from different O-2p-band center levels (32, 34) can be extended to the methanol e-refinery reaction. Therefore, different methanol e-refinery activities are anticipated between them where the O-2p-band center levels might play a role.

Methanol e-refinery performance descriptor identification

On the basis of the cyclic voltammetry (CV) results in Fig. 2A, catalysts with higher O-2p-band center levels (SrCoO_3−σ and NiO_xH_y) tend to exhibit higher current densities than lower-level ones (LaCoO₃ and FeO_xH_y) at a fixed potential in 1 M KOH + 1 M CH₃OH. After CV tests, 3-hour CA and ion chromatography (IC) tests are conducted to investigate the selective methanol e-refinery capabilities toward formate for the four catalysts. Figure 2B presents the IC spectra for NiO_xH_y as an example. The product analysis results are summarized in Fig. 2 (C and D): The formate production rates and Faradaic efficiencies of SrCoO_3−σ and NiO_xH_y (catalysts with higher O-2p-band center levels) are higher than that of LaCoO₃ and FeO_xH_y, i.e., a proportional relationship appears between the O-2p-band center levels and the methanol e-refinery performances. Given the observed dependency relationship, we conclude that the O-2p-band center level is a promising methanol e-refinery performance descriptor.

Fig. 2. Electrochemical performances of LaCoO₃, SrCoO_3−σ, FeO_xH_y, and NiO_xH_y.

(A) CV curves in 1 M KOH and 1 M KOH + 1 M CH₃OH at 10 mV s⁻¹. (B) Representative IC measurements for NiO_xH_y. (C and D) Production rates and Faradaic efficiencies for all detectable products. Nuclear magnetic resonance (NMR) is applied for the detection of formaldehyde (fig. S5 shows a representative NMR spectrum); ion chromatography (IC) is used to quantify formate and carbonate; gas chromatography (GC) is used to quantify carbon monoxide and oxygen (fig. S6 shows the representative GC spectra).

To further consolidate the relationship between the O-2p-band center levels and the methanol e-refinery capabilities, another three catalysts (Fe₂NiO_xH_y, FeNiO_xH_y, and FeNi₂O_xH_y) are synthesized with the same method as FeO_xH_y and NiO_xH_y. Considering that the performance difference between FeO_xH_y and NiO_xH_y is more obvious than that of LaCoO₃ and SrCoO_3−σ, we have expanded the Fe_mNi_nO_xH_y system to strengthen this relationship. The seven as-synthesized catalysts (LaCoO₃, SrCoO_3−σ, FeO_xH_y, Fe₂NiO_xH_y, FeNiO_xH_y, FeNi₂O_xH_y, and NiO_xH_y) are all analyzed by CA, and the resulting electrolytes are sent for IC tests to measure the HCOO⁻ concentrations. Figure 3 (A and B) summarizes the variations of formate production rates and Faradaic efficiencies versus the catalysts’ O-2p-band centers. The proportional relationship between formate production rates/Faradaic efficiencies and the O-2p-band center levels in Fig. 3 (A and B) confirms the role of the O-2p-band center level as a methanol e-refinery performance descriptor. In what follows, possible methanol e-refinery mechanisms are studied to interpret this correlation.

Fig. 3. Methanol e-refinery mechanisms toward formate.

(A) Formate production rate and (B) Faradaic efficiency variations versus catalysts’ O-2p-band center levels. (C) Different reaction diagrams between NiO_xH_y and FeO_xH_y. (D and F) Possible reaction mechanisms for SAM and LOM. (E and G) Computed free energy changes for the two schemes on LaCoO₃, SrCoO_3−σ, FeO_xH_y, and NiO_xH_y.

Methanol e-refinery mechanisms

With the help of DFT calculations, a set of promising methanol e-refinery reaction schemes is proposed. Through the participation of lattice oxygen, the schemes can illustrate the essential dependency relationship between O-2p-band center levels and the methanol e-refinery performances of the catalysts.

Discriminated by the involvement of lattice oxygen (the isotope-labeled experiments in Fig. 1H), SrCoO_3−σ and NiO_xH_y with higher O-2p-band center levels exhibit a different methanol e-refinery mechanism from LaCoO₃ and FeO_xH_y, as illustrated in Fig. 3C. As a result, we have established two reaction routes for the four catalysts, namely, surface adsorption mechanism (SAM) and lattice oxygen participation mechanism (LOM), as presented in Fig. 3 (D and F). For SAM, the first three steps are deprotonation processes, forming a *CHO intermediate, and they are followed by two hydroxide-attacking steps to produce HC¹⁶O¹⁶O⁻ (21, 23, 41–44). For LOM, the first three steps are also deprotonation processes, but, this time, with the shift of attacking species from OH⁻ to lattice oxygen, *¹⁶O¹⁸OCH is formed. After HC¹⁶O¹⁸O⁻ is released, an oxygen vacancy will be created. Subsequently, a proton-electron transfer step and a deprotonation step will take place to regenerate the metal-oxygen surface (32). Figure 3 (E and G) summarizes the energy variations of the two reaction routes for the four catalysts, and figs. S7 to S10 present the corresponding DFT models. Two distinct features are noteworthy: (i) LOM exhibits a lower rate-determining step (RDS) energy on SrCoO_3−σ and NiO_xH_y and a higher RDS energy on LaCoO₃ and FeO_xH_y; (ii) the RDS energy of the LOM route on SrCoO_3−σ and NiO_xH_y is lower than that of the SAM route on LaCoO₃ and FeO_xH_y.

The first feature indicates the domination of LOM on SrCoO_3−σ and NiO_xH_y and SAM on LaCoO₃ and FeO_xH_y. This observation interprets the isotope-labeled experiments well: O-2p-band elevation from LaCoO₃ and FeO_xH_y to SrCoO_3−σ and NiO_xH_y changes the methanol e-refinery mechanism. As illustrated above, the mechanism shift in methanol e-refinery is attributable to the similar physical origins of OER (45). To be specific, increasing the O-2p-band center level from LaCoO₃ and FeO_xH_y to SrCoO_3−σ and NiO_xH_y is associated with the shift of O-2p band of the catalyst toward the O₂/H₂O redox potential (Fig. 1I). Correspondingly, the catalyst surface becomes more negatively charged when equilibrated with the electrolyte and preferentially acts as a Brønsted base, making it a more suitable attacking species than OH⁻ during the intermediate oxidation process. As a result, the O-2p-band increment from LaCoO₃ and FeO_xH_y to SrCoO_3−σ and NiO_xH_y leads to the replacement of the nucleophilic attacking species (from OH⁻ to lattice oxygen) (45), as the highlighted steps show in Fig. 3 (D and F). Note that the *CHO oxidation step where the attacking species shifts is the RDS for almost all the four catalysts (except for the LOM on SrCoO_3−σ), after which mechanism deviation begins to appear. The second feature is assigned to the promotion effect of LOM for the methanol e-refinery reaction. In other words, O-2p-band elevation from LaCoO₃ and FeO_xH_y to SrCoO_3−σ and NiO_xH_y changes the mechanism from SAM to LOM and boosts the formate production rate during the reaction (Fig. 3A).

From Fig. 3B, the formate Faradaic efficiencies of the seven studied catalysts have a similar near-linear relationship with the catalysts’ O-2p-band center levels as formate production rates do. To thoroughly illustrate this phenomenon, it is essential to explain the selective promotion effect of O-2p-band increment toward formate production. Thus, the LOM reaction routes for the other two detectable by-products, namely, oxygen and carbonate, are also calculated with DFT (NiO_xH_y is used as an example). The suggested schemes, reaction energy variations, and DFT models are presented in figs. S11 to S13, respectively. The RDS of HCOO⁻ is lower than that of the other two by-products. This result coordinates with the trend in Fig. 3B, especially when the O-2p-band center reaches a relatively high level (the right part of the figure) where LOM governs the reaction. However, when the O-2p-band center level is not that high (the left part of the figure), SAM can also play a role; as a result, slight variations occur in this area.

It is likely that metal oxides/hydroxides with the rate-determining step occurring at the *OCH further oxidation step will adhere to this principle. On the basis of the DFT results, all the four catalysts demonstrate a rate-determining step at the *OCH → *OCHOH stage with the SAM reaction route. Substituting the nucleophilic attacking species (from OH⁻ to lattice oxygen) creates an opportunity to overcome the bottleneck of the SAM route by bypassing the formation of *OCHOH, which could potentially enhance the performance of the methanol e-refinery reaction. The proposed methanol e-refinery mechanisms in Fig. 3 (D and F) and fig. S11 not only consolidate the role of the O-2p-band center level as an efficient methanol e-refinery performance descriptor but also provide promising theoretical references for the catalysts that selectively oxidize methanol to formate.

MEA tests

On the basis of the promising features of NiO_xH_y, this anode is integrated in a 2 cm–by–2 cm (reaction area) MEA cell arrangement (21). With the hydrogen evolution reaction (HER) coupled at the cathode side, the catalyst’s electrochemical performance is evaluated. Figure 4A (inset) exhibits a diagram of the MEA cell. As shown in Fig. 4A, an anion exchange membrane separates the cathode and anode chambers. NiO_xH_y (2 mg cm⁻²) and commercial Pt/C (1 mg cm⁻²) are used as the anode and cathode catalysts, respectively. Polarization curves are obtained by cycling 1 M KOH + 1 M CH₃OH and 1 M KOH at 60°C with a flow rate of 10 ml min⁻¹ in the anode (Fig. 4A). Voltages are recorded for the two anolytes at current densities of 10, 25, 50, and 75 mA cm⁻² (geometric area) after 1-min stabilization. Because the methanol e-refinery reaction happens at a lower potential than OER on NiO_xH_y, the voltage of 1 M KOH + 1 M CH₃OH is lower than that of 1 M KOH at a fixed current density. After that, 1-hour chronopotentiometry is conducted at current densities of 10, 25, 50, and 75 mA cm⁻² (geometric area), and the electrolytes are analyzed by IC to quantify the concentration of HCOO⁻. Figure 4B presents the production rate and Faradaic efficiency of HCOO⁻. The HCOO⁻ current density, defined as formate Faradaic efficiency multiplied by the total current density, is used to compare NiO_xH_y with other catalysts reported in the literature (Fig. 4C) (19–22, 30, 46). The MEA performance of NiO_xH_y outperforms other transition metal catalysts in the literature, according to Fig. 4C.

Fig. 4. MEA test on NiO_xH_y.

(A) Polarization curves with 50 ml of cycling 1 M KOH and 1 M KOH + 1 M CH₃OH. (B) Production rate and Faradaic efficiency of HCOO⁻. (C) Comparison results of HCOO⁻ current densities between NiO_xH_y and the literature.

To fully illustrate the advantages of the methanol e-refinery MEA design, the cell’s economic profit and global warming effect are calculated. When conducting the economic assessment, the initial plant investment (for the electrolysis and distillation processes), the following operation and maintenance expense, the 39% government tax, and some other practical factors are taken into consideration (11, 47, 48). The prices of H₂, CH₃OH, HCOOH, and electricity are set as 1.26 U. S. dollars (USD) kg⁻¹ (49), 0.40 USD kg⁻¹ (50), 0.74 USD kg⁻¹ (51), and 0.012 USD kW⁻¹ hour⁻¹ (52), respectively. Because of the initial enormous investment for the plant, the cumulative net profit (defined as total earnings minus total costs) for the first several years will always be negative. According to Fig. 5A, with the continuous refreshment and maintenance of the catalysts and the equipment (those costs have been included during calculation), the plant will obtain positive cumulative net profits after year 4 (inclusive) at a current density of 75 mA cm⁻² (geometric area). The detailed calculation methods for the cumulative net profit of the methanol e-refinery MEA cell at a current density of 50 mA cm⁻² (geometric area) are presented in the Supplementary Materials. Moreover, the environmental impacts (calibrated with equivalent carbon dioxide emission) of the two MEA cells, namely, the methanol e-refinery + HER and OER + HER cells, are evaluated by life cycle assessment (LCA). During LCA, the whole life cycle of the reaction is taken into consideration (from the extraction of resource material to the waste treatment), which is also referred to as from cradle to grave. The LCA process is composed of two main steps. The first one is the collection of the life cycle inventory data. In this step, the information on the resource materials, the energy input, and the waste emission in the life cycle is collected. The second step is the assessment of the data, after which a figure on the environmental impacts of the activity is created (53). As shown in Fig. 5B, the methanol e-refinery cell tends to release less carbon dioxide compared to the OER + HER cell.

Fig. 5. Economic and global warming assessments of cogeneration of formate and hydrogen on NiO_xH_y.

(A) Cumulative net profits of methanol e-refinery + HER. (B) Global warming effect for the two electrolysis systems, i.e., methanol e-refinery + HER versus OER + HER.

DISCUSSION

With the direct experimental evidence, we have proven the participation of lattice oxygen in the methanol e-refinery reaction to formate and discovered an efficient methanol e-refinery performance descriptor, i.e., the O-2p-band center level of the catalyst. NiO_xH_y, with the highest O-2p-band center level among the seven studied catalysts, presents a near 100% formate Faradaic efficiency over a wide range of current densities in the MEA tests. Moreover, through the participation of lattice oxygen, a reaction pathway is provided for the methanol e-refinery reaction, which could interpret the pivotal role of O-2p-band center level and could potentially extend to other small molecules e-refinery reactions. The e-refinery processes following this mechanism are promising to circumvent the inherent limitations of adsorption-based mechanisms where scaling relations confine performance, thus potentially providing a better understanding and control of electrified conversions of other small molecules.

MATERIALS AND METHODS

Synthesis and bulk characterization

LaCoO₃ and SrCoO_3−σ are synthesized by a conventional sol-gel method (54). FeO_xH_y, Fe₂NiO_xH_y, FeNiO_xH_y, FeNi₂O_xH_y, and NiO_xH_y are synthesized by simultaneously dropping 0.1 M Na₂CO₃ and 0.1 M metal nitrates with the pH of the mixture kept at around 9 (55, 56). A Bruker D8 Advance XRD machine equipped with Cu-Kα radiation (λ = 1.5418 Å) is used to characterize the crystal structure of the catalysts. The morphological information is examined with field emission scanning electron microscopy (SEM; FEI Nano430). ASAP TriStar II 3020 is used to obtain the Brunauer-Emmett-Teller–specific surface areas of the catalysts. The corresponding adsorption and desorption curves are shown in figs. S14 (A and D) and S15. X-ray Absorption Fine structure for catalysis (XAFCA) beamline of Singapore Synchrotron Light Source is used to collect the XANES and EXAFS data. High-resolution transmission electron microscopy (HRTEM) is conducted by a JEOL 2100F transition electron microscope with 200-kV accelerating voltage. To conduct x-ray photoelectron spectroscopy (XPS) measurements, a PHI-5400 machine equipped with a position-sensitive detector and an Al Kα beam source (250 W) is used. The angle resolution is set to 45°, the binding energy resolution is set to 0.8 eV, and the detection limit is set to 80-K counts per second (CPS). The Ar ion sputtering speed is 0.28 nm/s over an area of 300 μm by 300 μm, using a voltage of 12 kV and a current of 4.2 mA. The measurement chamber’s base pressure is maintained at 3.0 × 10⁻⁷ Pa. The C 1s spectra’s adventitious carbon peak (284.8 eV) is used to calibrate the obtained results, and the O 1s XPS spectra are deconvoluted using the Lorentz-Gaussian fitting method. More details are provided in the Supplementary Materials.

Electrochemical measurements

An ink-cast method is used to prepare the working electrodes (32, 57). The catalyst is loaded on a 2.0 cm–by–1.0 cm graphite paper substrate, with a loading area of 0.6 cm by 1.0 cm (double-side loading). The loading amount is 0.114 mg_cat cm⁻². The ink is prepared through dissolving catalyst, acetylene black carbon, and Nafion in a mixture of isopropyl alcohol and water. A three-electrode system is applied to perform the electrochemical measurements, with Pt and Hg/HgO as the counter electrode and the reference electrode, respectively. The potentials manifested in the results are converted to the RHE scale and corrected by Ohmic losses with the following equations, where SHE represents standard hydrogen electrode

$$E(\mathrm{vs}.\mathrm{RHE})/V=E(\mathrm{vs}.\mathrm{Hg}/\mathrm{HgO})/V+{E^{\mathrm{\theta }}}_{\mathrm{Hg}/\mathrm{HgO}}(\mathrm{vs}.\mathrm{SHE})/V+0.059\times \mathrm{pH}/V$$ (1)

$${E^{\mathrm{\theta }}}_{\mathrm{Hg}/\mathrm{HgO}}(\mathrm{vs}.\mathrm{SHE})/V=0.098/V\mathrm{vs}.\mathrm{SHE}\mathrm{at}25{}^{\circ }\mathrm{C}$$ (2)

$$E_{\mathrm{real}}/V=E(\mathrm{vs}.\mathrm{RHE})/V-iR$$ (3)

in which i is the current and R is the electrolyte resistance measured by electrochemical impedance spectroscopy. CA measurements are performed at 1.7 V versus RHE for 3 hours in 1 M CH₃OH + 1 M KOH. More details are provided in the Supplementary Materials.

Product analysis

¹H–nuclear magnetic resonance (¹H-NMR) spectrum is recorded using a Bruker AV 400-MHz NMR spectrometer for the quantification of formaldehyde. Sample solution (500 μl) and D₂O (56 μl) are mixed for the test. Figure S5 is a representative NMR curve. GC is conducted using an Agilent 7890A GC machine for the detection of CO and O₂. The measurements are conducted at an oven temperature of 50°C and an argon gas flow rate of 9 ml/min. GCMS is measured on a GC machine of Agilent 7890B and a mass-selective detector of 5977B. The samples are neutralized by concentrated hydrochloric acid solution and distilled at around 220°C to obtain HCOOH solutions before they are sent for GCMS tests. Preconfigured Dionex ICS-1000 integrated system is used to perform the IC tests with the conductivity detection method. The eluent (0.038 M KOH) flow rate is 1.2 ml/min, and the column temperature is 35°C. More details are provided in the Supplementary Materials.

DFT calculations

The spin-polarized DFT calculations are performed with the Vienna Ab initio Simulation Package (58) using the projector augmented wave method (59, 60). The Perdew-Burke-Ernzerhof functional with the generalized gradient approximation is used to describe the exchange-correlation interactions (61). A cutoff energy of 500 eV is applied for the plane-wave expansion of the electronic wave function. The adsorption and desorption of methanol and formate molecules are simulated on the 1 × 1, 1 × 1, 2 × 2, or 1 × 1 supercell surfaces of LaCoO₃, SrCoO_3−σ, FeO_xH_y, or NiO_xH_y by placing a single molecule on each surface. The Brillouin zones are sampled with a Monkhorst-Pack grid of 5 × 7 × 1, 3 × 7 × 1, 5 × 3 × 1, and 1 × 7 × 1 for the LaCoO₃ (110), SrCoO_3−σ (101), NiO_xH_y (011), and FeO_xH_y (104) surface calculations, respectively. The surface indexes are determined by the corresponding HRTEM images in figs. S14 (C and F) and S16. The energy and force convergence criteria are set to be 10⁻⁵ eV and 0.01 eV Å⁻¹, respectively. To avoid vertical interactions between slabs, a vacuum space of 20 Å is used in the z direction for the slab models. Grimme’s DFT-D2 method (62) is applied to account for the van der Waals interaction. The solvent effects are not taken into account as they are determined to have minimal impact on the free energies. As evidence, we have computed the free energies for the LOM on NiO_xH_y while considering solvent effects. Notably, these values closely match those obtained when solvent effects are not taken into account (table S2). Consequently, we can assert the reliability of our calculations, despite the omission of solvent effects. More computational details are provided in the Supplementary Materials.

Supplementary Materials

This PDF file includes:

Supplementary Text

Figs. S1 to S20

Tables S1 to S5

Supplementary Notes

References