Operando Raman characterization of unique electroinduced molecular tautomerization in zero-gap electrolyzers promotes CO₂ reduction

Significance

The development of electrocatalysis research favors the utilization of zero-gap membrane electrode assembly (MEA) electrolyzers for future applications. However, understanding of the “solid–liquid–gas” interface of MEA remains limited due to the lack of operando techniques that can function under extreme current densities (>100 mA cm⁻²). Herein, we developed an operando Raman method, by which we identified the presence of a molecular tautomerization at the 4-mercaptopyridine-modified Cu catalyst surface, which significantly improves the CO₂ electrolysis performance at the MEA interface but not in a flow cell or H cell under the same electrolyte conditions. The triple-phase interface of MEA is shown to offer opportunities to drive electrified chemistry that may be utilized in developing electrocatalysis and other applications.

Abstract

Membrane electrode assembly (MEA) represents an advanced type of electrochemical device currently widely used in various electrocatalysis applications [e.g., electrochemical CO₂ reduction reaction (CO₂RR)], featuring no explicit catholyte flow and a unique “solid–liquid–gas” triple-phase interface. Herein, we identify a peculiar electroinduced thiol to thione tautomerization of 4-mercaptopyridine (4MPy) molecule on Cu catalyst surfaces at this triple-phase interface driven by cathodic polarization. This leads to a significant performance improvement of CO₂RR on Cu with a C₂₊ Faradaic efficiency of over 80% with more than 60% C₂H₄, as well as a 300 mV reduction of cell voltage compared to bare Cu. A home-designed MEA-type operando Raman cell enables mechanistic studies directly under a current density of over 100 mA cm⁻², elucidating the intricate impacts of the 4MPy tautomerization on the local catalytic environments under real reaction conditions. Surprisingly, this tautomerization does not occur in other commonly utilized electrolyzers, e.g., flow cell and H-cell, even with the same catalyst and electrolyte conditions. The direct contact with the electrolyte in the latter cells was found to cause rapid desorption of 4MPy from the catalyst surface before its possible chemical transformation. These results highlight the opportunities of utilizing surface molecular tautomerization to promote CO₂RR performance and using the triple phase of MEA to drive reactions that would otherwise be hard to happen in classical electrochemical devices of similar conditions.

The growing supply of renewable electricity by intermittent energy resources (e.g., solar and wind) and the demand to reduce carbon dioxide emission have stimulated intense research interests in developing electrochemical technology that can convert electricity into solar fuels, e.g., water splitting and electrochemical CO₂ reduction (CO₂RR) (1–3). Classic electrocatalyst research emphasizes the structure–activity relationship of the catalysts (4). However, there are emerging consensuses that the local environment of electrocatalysts (5, 6), including pH (7), water structure (8), and electrolyte ions (9–11), should be viewed as an indispensable component of the catalysis processes, and sometimes, the most important factors determining the device performance (12).

Particularly, surface molecular engineering (13, 14) has been recognized as an effective way to modify the catalyst properties through a judicious choice of the attached molecules. This is analogy to enzymatic catalysis, where the primary and secondary structures around the reaction center can dramatically affect the activity of the active sites (15, 16). For CO₂RR, various molecular modifiers have been reported to improve the performance of the catalysts [e.g., Cu (17–19), Ag (19)], including pyridine (19–21), imidazole-based molecules, etc., favoring the selective conversion of CO₂ into C₂H₄ (18, 19), HCOO⁻ (17, 21, 22), and CO (7, 8, 23, 24). For example, dimerization of N-arylpyridinium on Cu can improve the Faradaic efficiency of C₂H₄ up to 72% with an energy efficiency of 20% (18). Yet despite the proposed mechanisms, the complex microenvironment at the electrochemical interface makes the determination of precise mechanisms rather challenging. This is particularly true for the triple-phase (solid–liquid–gas) junction of the membrane electrode assembly (MEA) (25), an emerging device for electrocatalysis research, where the lack of an explicit liquid layer at the cathodic side renders a unique electrochemical environment. Theoretical calculations, e.g., density functional theory (DFT), have been widely adapted to reveal the catalysis mechanisms and achieved many insights (26–31). However, the divergences between the simplified, idealized model and the intricate convolution of mass transport and electrocatalytic processes at the triple-phase junction still result in uncertainties in the underlying mechanism (25, 32, 33), especially under reaction conditions with current densities over 100 mA cm⁻². Thus, direct operando characterization at the triple-phase junction of MEA is essential to obtain a molecular understanding of chemical processes at this emerging electrified interface.

Beyond CO₂RR, studying the chemistry of surface molecules at the MEA interfaces is fundamentally crucial. The unique electric double layer and electron donating and withdrawing properties of the electrode can alter the properties of surface molecules to differ from those in homogenous solution (34–36) and enable new surface chemistry. Previous studies have shown that electrochemical polarization can create an equivalent Hammett effect (34) on different moieties of the molecules and control its selectivity for chemical reactions. Also, the pK_a (37–39) [e.g., glycine (39)] and the proton transfer behavior of molecules (40) at the electrode interface can differ from those in bulk solution due to the change of solvation environment (39) and electrochemical polarization, which was shown to improve reaction rates by orders of magnitude (40). Considering the above unique zero-gap nature of the triple-phase junction at the MEA devices process, i.e., no explicit liquid layer, it is thus intriguing whether the zero-gap nature of the MEA interface would facilitate new surface chemistry and drive reactions that would be otherwise hard at classical solid–liquid electrified interface.

Herein, we report a 4-mercaptopyridine (4MPy) molecule-modified Cu catalyst (4MPy-Cu) in MEA devices. This molecule was recently reported to enhance the CO₂ or CO conversion in flow cell configuration and in the presence of Nafion (17, 41), which itself can form complex molecular microenvironments. Herein, different from previous studies, we utilized 4MPy modification onto Cu catalysts without any Nafion and thus examined the direct modification mechanisms of the molecule. To shed more on the molecular origins of 4MPy in performance improvements, other molecules with a similar structure as 4MPy were carefully examined. Most significantly, we developed an operando Raman spectroscopy directly conducted onto the MEA device, allowing mechanistic studies of 4MPy within MEA devices at a current density over 100 mA cm⁻² which was not possible previously. To distinguish the importance of electrolyzers in driving surface reactions, we further conduct in-situ spectroscopic studies of 4MPy-modified catalysts in other commonly utilized electrolyzers (like flow or H-cell) with distinct reaction mechanisms revealed. Overall, the present study has revealed not only the mechanisms by which 4MPy modification improves the CO₂RR performance under operando conditions but explored the differences of electrochemical interface in various electrolyzers for driving new surface chemistries.

Results and Discussion

The impact of 4MPy on the CO₂RR performance of Cu catalysts was first evaluated at different current densities with a zero-gap MEA device (Scheme 1), consisting of a gas diffusion electrode (GDE) of Cu (cathode), an anion exchange membrane, a Ni anode, and 1 M KOH as the anolyte. Herein, the Cu GDE was prepared by sputtering Cu onto polytetrafluoroethylene (PTFE) membranes following previous procedures (SI Appendix, Fig. S1) (42). For the 4MPy-modified Cu GDE (termed as 4MPy-Cu), a 4MPy-containing ink was sprayed onto the above Cu GDE (SI Appendix, Figs. S2 and S3). The successful attachment of 4MPy was evidenced by the X-ray photoelectron spectroscopy (XPS), where the S 2p peaks of 4MPy-Cu shifts from 161.1 eV to 161.9 eV, consistent with the known formation of Cu-S bond (43). Meanwhile, the N 1s spectrum does not experience apparent changes after the attachment (SI Appendix, Fig. S4). Therefore, the binding of 4MPy onto Cu is mainly through the Cu-S bond (41, 44). The 4MPy modification on Cu leads to immediate improvements in CO₂RR efficiency (Fig. 1 A and B and SI Appendix, Table S1), achieving a C₂₊ selectivity of 82.5 ± 4.3% with an ethylene selectivity of 60.5 ± 4.1% at 200 mA cm⁻², much larger than the bare Cu catalysts (53.7 ± 2.9% and 40.1 ± 3.4%, respectively); similar improvements can also be observed at higher current densities (Fig. 1B). For practical applications, the cell voltage is a crucial figure of merit that affects the overall energy efficiency of the device (44, 45). Interestingly, the 4MPy modification significantly reduces cell voltages. At 200 mA cm⁻², 4MPy-Cu requires a cell voltage of only 2.6 V, 300 mV smaller than that of bare Cu (2.9 V), representing a relative low voltage in literature at similar current densities (SI Appendix, Table S2 and Fig. S5); similar decreases of cell voltages were also observed at other current densities (Fig. 1B). A home-built operando voltage analysis (SI Appendix, Fig. S6) and differential electrochemical mass spectrometry (DEMS, SI Appendix, Fig. S7) were then conducted on these MEA devices, which revealed that the voltage drops mainly occurred at the cathode side. Benefiting from both the improved C₂ selectivity and reduced cell voltage, the overall energy efficiency (EE) of 4MPy-Cu is much higher than that of Cu (20.8 ± 1.1%), reaching a maximum EE for C₂₊ products at 200 mA cm⁻² up to 34.4 ± 1.9% (Fig. 1C and SI Appendix, Table S2). Besides, the stability of Cu is much lower than that of 4MPy-Cu, accompanied by a rapid decline in the FE_C2+ (SI Appendix, Fig. S8). Thus, 4MPy provides a facile approach that not only improves the performance of Cu-PTFE but also stabilizes the operation of the devices.

Scheme 1.

Comparison of surface modification strategies utilized in previous studies and this work to improve the electrochemical CO2RR performance. (A) Schematic of the zero-gap membrane electrode assemblies (MEA) used in the present study, consisting of Cu-based gas diffusion electrodes (Cu-GDE), an alkaline exchange membrane (AEM), and the Ni foam as the anode. The grey dotted box is the zoom-in of the triple-phase junction of the MEA interface. Notably, there exists no explicit liquid flow in the cathode of MEA; both the water permeated from the anolyte through AEM and precipitated from the humidified air contribute to the formation of the ionic conductive liquid phase, creating a unique solid-liquid-gas phase with water only partially filling the electrode surface. (B) Previous work report molecular modification either by direct deposition of molecular modifier or by surface dimerization of molecules, involving only in situ mechanisms, with no electrolyzer dependence reported. The present study reports the electro-induced molecular tautomerization at the Cu surface by operando studies under reaction conditions (>100 mA cm⁻²), where the cathodic polarization of the electrode introduced the 4MPy transformation from a thiol form into a thione form, which occurred only inside MEA but not in flow cell or H-cell under the same electrolyte.

Fig. 1.

Molecular origins of 4MPy modification to improve CO₂RR of Cu toward C₂₊ and C₂H₄ products. (A) Schematic of 4MPy-modified Cu catalysts (4MPy-Cu) with the 4MPy directly coated onto the reference Cu catalyst. (B) Faradaic efficiency and cell voltage of 4MPy-Cu and Cu catalysts at various current densities during CO₂RR. (C) Comparison of key CO₂RR performance merits between 4MPy-Cu and Cu at a current density of 200 mA cm⁻². (D) Faradaic efficiency of Cu modified by other small molecules with similar structure as 4MPy measured at 200 mA cm⁻² during CO₂RR, including 2MPy, pyridine (Py), and thiophenol molecules. (E) Comparison of CORR Faradaic efficiency and cell voltage of Cu and 4MPy-Cu catalysts at various current densities.

To better understand the molecular origins for above performance enhancements, a “molecular surgery” approach was adapted, in which the Cu catalysts modified with several molecules that are structurally like 4MPy (Fig. 1D). Particularly, mercaptopyridine with -meta and -ortho positions were studied, termed as 2MPy and 3MPy (SI Appendix, Figs. S9 and S10), respectively. Impressively, despite the structural similarity, none of these molecules can lead to performance improvements similar to those of 4MPy. For example, the pyridine modification (removing -SH unit of 4MPy) leads to a much lower performance than those of 4MPy and a significant increase of HER (Fig. 1D); surface modification with thiophenol molecules and 2MPy/3MPy molecules shows much less performance improvement that of 4MPy. Therefore, the presence of N and S sites in the para-position as in 4MPy seems essential for maximizing the performance improvements, as further discussed below.

Noticeably, CO₂RR involves two key steps, including the conversion of CO₂ to CO and the subsequent reduction of CO into multicarbons. As the above results only demonstrate the overall improvement of CO₂RR by 4MPy modification, it is intriguing which step 4MPy affects most. Thus, we conduct further CORR studies on 4MPy-Cu and Cu to scrutinize the impact of 4MPy on the CO conversion process. Impressively, despite the large performance improvement of 4MPy for CO₂RR, both 4MPy-Cu and Cu demonstrate similar CORR performance (Fig. 1E). Thus, 4MPy modification likely affects mostly the conversion of CO₂ to CO during CO₂RR while affecting relatively less the intrinsic activities of CO conversion into multicarbon products, evidenced by their similar CORR performance. Noticeably, the molecular modification mechanisms for improved CO₂RR are often attributed to the reduced energy barrier of CO–CO coupling or CO hydrogenation based on DFT calculations. The above results show further control measurements of CORR is beneficial for identifying key steps that molecular modification functions on. As demonstrated below, 4MPy modifications would improve the CO₂RR by enhancing the local CO supply and promoting CO₂ conversion to CO, without necessarily changing the intrinsic C₂ coupling barrier.

Below, we resolve the detailed mechanisms by which 4MPy affects the conversion of CO₂ to CO utilizing operando Raman spectroscopy. Noteworthy, the ordinary vibrational spectroscopic studies of electrocatalysts in MEA would take the catalysts out of the MEA devices and study them in separate H-type Raman or ATR-IR cells due to the difficulties in conducting operando measurements over 100 mA cm⁻² in MEA devices. However, this approach could be problematic as the H-type spectroscopic cells cannot fully reflect the characteristics of the triple-phase interfaces in MEA and may resolve mechanisms different from those occurring in the actual MEA devices (46, 47). Therefore, direct mechanistic studies in an MEA-type spectroscopic cell are desired to reflect the accurate catalytic mechanisms under reaction conditions; as illustrated below, this is particularly vital for studying the molecular modification due to the facile desorption of molecules from the electrode surface. Therefore, we home-designed an MEA-type operando Raman cell (Fig. 2A and SI Appendix, Fig. S11) by opening an optical window in the anode plate of the MEA device, with the Raman signal measured directly from the anode side through the membrane and focused onto the catalyst. Performance measurements show that this operando Raman cell has identical CO₂RR performance as standard MEA devices (SI Appendix, Fig. S12), thus ensuring realistic mechanism studies of the electrode surface, though not completely excluding the inevitable local variance due to the presence of a hole in the anode. Also, to better avoid the blistering of the electrode and AEM interface by the bubbles during large current measurements, we developed a hot-pressing procedure that compressed the catalysts directly onto the membrane to enable a compact interface; this strategy was also verified not to affect the performance of the MEA cells (SI Appendix, Fig. S13) and significantly reduce the impact of bubbles and enable operando Raman measurements at a current density up to ~100 mA cm⁻², representing a true operando analysis. Fig. 2B shows the evolution of the Raman spectra of Cu during CO₂RR at an incremental current from 0 to 70 mA cm⁻², i.e., directly at the current corresponding to the large current devices. As the current increases, a notable Raman peak at 1,070 cm⁻¹ and a peak at 2,046 cm⁻¹ can be identified, appearing at approximately 1 and 10 mA cm⁻², respectively. The former peak is known to arise from CO₃²⁻ (48), generated through the reaction of CO₂ with the local OH⁻ (2OH⁻ + CO₂ → CO₃²⁻ + H₂O) of alkaline KOH, being a critical loss pathway of CO₂ and destabilizing the electrolyzer by salt precipitation. The gradual CO₃²⁻ formation at higher current densities indicates an increase of local alkalinity near the Cu surface, consistent with the expected local generation of OH⁻ from HER and CO₂RR. Meanwhile, the latter peak is known to be the key reaction intermediates-adsorbed surface CO, CO_ads (49–51), with an onset current of 10 mA cm⁻². Interestingly, for 4MPy-Cu, besides the above two peaks, additional multiple peaks exist in the range 1,200 to 1,800 cm⁻¹ (Fig. 2C and SI Appendix, Fig. S14). These peaks originate from both the surface 4MPy molecules and the membranes (SI Appendix, Fig. S15 and detailed peak assignments in Table 1). Yet distinct from the abovementioned Cu catalysts, the increase of current in 4MPy-Cu does not lead to the growth of CO₃²⁻ peak at 1,070 cm⁻¹ (Fig. 2C). Meanwhile, the CO_atop onset appears approximately at 2 mA cm⁻², much earlier than the 10 mA cm⁻² of the bare Cu. Thus, the introduction of 4MPy on the Cu surface reduces the generation of CO₃²⁻ from the CO₂ reaction with OH⁻ and reduces the onset currents of CO_atop, consistent with improved performance of CO₂RR for 4MPy-Cu. More importantly, the gradual increase of cathodic current on 4MPy-Cu leads to the rapid growth of several peaks (~1,345, 1,402, 1,530, and 1,640 cm⁻¹)—most notable at 1,640 cm⁻¹, as well as a simultaneous decrease of the peak at ~1,573 cm⁻¹; the spectra stabilized afterward at approximately 20 mA cm⁻². These remarkable operando spectra indicate a surprising chemical transformation of 4MPy at the Cu surface under cathodic polarization; that is, the function form of 4MPy modification differs from its initial thiol form. A similar transition can also be observed when the measurements were conducted under the Ar or CO atmosphere (SI Appendix, Fig. S16); thus, the transformation of 4MPy is driven by the cathodic potential and independent of the presence of CO₂, CO, or Ar. Interestingly, as shown below, when conducting similar measurements on those of 3MPy-/2MPy-Cu, no spectral changes can be observed. Thus, the operando structural transformations of molecules only happen on the 4MPy-modified Cu surface, consistent with its greatest improvement of CO₂RR performance.

Fig. 2.

Electroinduced tautomerization of 4MPy and its impact on the local environment of CO₂RR illustrated by operando Raman directly conducted within an MEA electrolyzer. (A) Schematic of the operando Raman cell design, where an optical window was opened in the anode flow plate of the MEA device, with the Raman signals measured directly from the anode side through the membrane and focused onto the catalyst. (B and C) Operando Raman spectra of Cu (B) and 4MPy-Cu (C) in 1 M KOH under CO₂ flow at different current densities. Three regions are plotted including, the 900 to 1,700 cm⁻¹ corresponding to the bicarbonate and 4MPy regions, 1,900 to 2,100 cm⁻¹ corresponding to the CO_ads region, and the 2,800 to 3,600 cm⁻¹ corresponding to the water O–H stretching region. The dashed line in b represents the reference Raman spectrum from the AEM (Sustainion, X37-50-grade T) and the white line plots a representative spectrum of Cu during CO₂RR. The white lines in (C) plot the representative spectra of 4MPy-Cu at low and high current densities, respectively, with the most notable change region marked by the yellow arrow. (D) Comparison of the measured spectra (from Fig. 2C) with the theoretical calculated Raman spectra of the pristine 4MPy (thiol form), protonated 4MPy, and the tautomerized 4MPy (thione form). (E) Ex situ Raman spectra of the pristine 4MPy-Cu film exposed in an alternating pH = 3.4 and 13.6 environment. (F) Ex-situ Raman spectra of the after-reaction 4MPy-Cu film exposed in pH = 13.6 solution for a prolonged period.

Table 1.

Lists of Raman peak frequencies and their assignment of 4MPy-Cu in the as-prepared state and after conversion during operating conditions in the 600 to 1,800 cm⁻¹ region

As-prepared 4MPy-Cu*	4MPy-Cu after conversion†	Peak assignments‡
	688	C–H out-of-plane asys bend
	778	C–H out-of-plane asys bend
1,017	1,008	Ring breathing
1,088	1,091	Trigonal ring breathing with C = S
1,213	1,212	C–H in-plane sys bend
1,289	1,247	C–H in-plane asys bend
	1,338	Ring def with C = C asym stretch
	1,396	Ring def with C = C asym stretch
	1,499	Ring def with C = C sym stretch
	1,532	Ring def with C = C asym stretch
1,582	1,578	Ring stretch with N
1,605	1,645	Ring stretch with N

^*Data obtained from the 4MPy-Cu dry film before the current-controlled SERS experiment.

^†Data obtained after finishing the current scan from 0 to 160 mA cm⁻².

^‡Data and assignment obtained from ref. 52.

Interestingly, 4MPy was previously known to have complex protonation and resonance changes due to local environments and pH (53–55). The above dramatic performance improvement of CO₂RR by 4MPy and the remarkable operando spectroscopic transition intrigue whether protonated and tautomerized 4MPy could account for the observed performance improvements. To evaluate the origins for the above Raman transition, we conducted DFT calculation on both the protonated and tautomerized 4MPy to evaluate the Raman spectral changes by these two processes. The calculations show that both the protonated and tautomerized structures would cause a notable peak shift from 1,592 to 1,627 cm⁻¹, consistent with the observed transformation of 4MPy spectra (1,573 to 1,640 cm⁻¹). However, limited by the accuracy of the DFT calculations and complex electrochemical environments, the calculated features are insufficient to conclude the abovementioned transition to protonation and/or a tautomerization process (Fig. 2D and SI Appendix, Fig. S17). To better distinguish these two possibilities, we thus designed a pH titration measurement. It is hypothesized that if the spectral transformation originated from a direct protonation process, it should be reversible with the pH change. This is verified by the pH swing experiments directly conducted on a pristine 4MPy-Cu film, where a reversible peak change between 1,573 and 1,640 cm⁻¹ can be observed when the pH was switched between pH = 3.4 and 13.6 using HCl and KOH solutions (Fig. 2E). However, if the above transformation is caused by the electrodriven tautomerization, it may not be reversible with a pH change. Thus, after the reaction in MEA, we took the 4MPy-Cu film out, rinsed it with DI water, and then exposed it directly to a KOH solution with pH = 13.6. Noticeably, the Raman spectra of the after-reaction film remain unchanged even overnight, while the protonated sample in the pH swing measurements immediately changes after exposure to pH = 13.6 (Fig. 2F). Therefore, the operando observed spectral changes originate from the electrodriven tautomerization processes rather than direct protonation. Besides, the reduction-induced dimerization of 4MPy were also excluded by ¹H NMR analysis, ultrahigh-performance liquid chromatography analysis, and the Raman DFT calculations (SI Appendix, Figs. S17–S19 and Note 5).

Notably, the absence of protonation of 4MPy under high alkalinity conditions without applying electrochemical potentials agrees with the low pK_a of surface 4MPy (4.6 ± 0.5) (56). Yet, recent studies have demonstrated that an electrified interface can behave as an effective molecular moiety to push electrons into the adsorbed molecules upon cathodic polarization, processing an effective Hammett parameter (34). Therefore, with the increase of the cathodic potentials, more electrons would be pushed into the S atom through the Cu-S bond and subsequently trigger a resonance transformation into a thione form. As the resulting negatively charged N site is known to process a strong nucleophilicity, it would allow the subsequent protonation and formation of N-H, finishing the electrodriven tautomerization transformation processes. Further XPS characterization of the after-reaction film indeed detected a profound N-H at 400.4 eV of the after-reaction film (SI Appendix, Fig. S20), much larger than those observed in the pristine film (SI Appendix, Fig. S4), agreeing with the proposed tautomerization mechanisms. We highlight that despite the known feasible resonance change of 4MPy (52–54), the above results characterized the presence of tautomerization in an alkaline environment within MEA at a current density over 100 mA cm⁻². Further operando studies on both 2MPy- and 3MPy-modified Cu show no changes of Raman spectra during the MEA measurements (SI Appendix, Fig. S21 and Note 6) distinct from those 4MPy (Fig. 2C). This result agrees with the absence of tautomerization of 2MPy/3MPy on the electrode surface and its much weaker performance than 4MPy.

Besides the above changes of 4MPy under reaction conditions as well as changes of CO₃²⁻ and CO_ads, we also observed a much weaker water O-H stretching (2,800 to 3,600 cm⁻¹) in the Raman spectra of 4MPy-Cu as compared to bare Cu; similar reduction of water access can be also observed in D₂O (2,400 to 2,800 cm⁻¹). These results suggest a reduced water accessibility at the 4MPy surface (SI Appendix, Fig. S22). The contact angle measurements also show that the after-reaction 4MPy-Cu films have a much more hydrophobic surface than those of pristine film with a water contact angle changing from 54.0 ± 2.5° to 106.0 ± 1.5°, respectively (SI Appendix, Fig. S23). Therefore, collectively, these above operando Raman measurements reveal several key characters of 4MPy-Cu under conditions as compared to bare Cu catalysts a) the much lower generation of CO₃²⁻ b) the earlier formation of CO_ads c) an electrodriven molecular tautomerization d) a reduced surface water accessibility. With these insights, the mechanisms by which 4MPy modifications promote the conversion of CO₂ to CO can be rationalized as follows: Under operational conditions, the cathodic potentials of 4MPy-Cu drive the surface tautomerization of the 4MPy molecule. As a result, the electrode becomes more hydrophobic with reduced water accessibility and lower local alkalinity at the surface, which reduces the CO₂ loss arising from the formation of CO₃²⁻ and effectively enhances the local concentration of CO₂ supply. This cathodically shifts the onset potentials for CO_ads, enriches the surface concentration of CO_ads, and then facilitates the CO–CO coupling, accounting for the observed higher C₂₊ selectivity and reduced HER. Besides the above mechanisms, it may also be wondered whether the tautomerized 4MPy could directly catalyze CO₂ conversion to CO. Herein, the operando Raman spectra of 4MPy under CO₂ and Ar demonstrate similar features, thus providing no evidence supporting the formation of any adducts between 4MPy and CO₂. This is also consistent with previous vigorous studies concluding that the direct CO₂ catalyzing by pyridine-based molecules seems rather less likely (57, 58). Yet, given the slight performance improvements of 2MPy-/3MPy-Cu than bare Cu, there still exists some other minor changes of local environments caused by mercaptopyridine modifications that improve the performance of CO₂RR beyond tautomerization, which is not further discussed.

So far, the above results have demonstrated the existence of tautomerization at the MEA electrode surface and its improvement of CO₂RR performance through tuning the local catalytic environment. Yet notably, the ongoing developments of electrocatalysis techniques also favor flow cells as another type of large-current electrolyzer (Fig. 3). Compared with MEA, the flow cells expose the cathode to direct contact with the electrolyte (i.e., 1 M KOH), thus placing the catalysts into a different local environment (58). Also, previous studies utilizing 4MPy were reported in a flow cell configuration (41). Hence, it is intriguing whether the observed tautomerization processes would also happen inside flow cells. To address this, we thus further compare the performance of 4MPy-Cu and Cu during CO₂RR and CORR in a flow cell. Consistent with the above results in MEA, 4MPy modification does not affect the CORR process with both 4MPy-Cu and Cu demonstrating similar performance (SI Appendix, Fig. S24). For CO₂RR, 4MPy-Cu still demonstrates an initial higher FE than Cu, like the MEA devices, but to a much lesser extent. These results are different from previous studies (17, 41), reporting promotion effects of 4MPy for the formation of formate and acetate in the case of CORR and CO₂RR, respectively. The exact reasons for these divergences remain unclear but may be attributed to convoluted differences of GDEs, Cu catalysts, and usage of Nafion between these studies. Nevertheless, it is found herein that the improved CO₂RR performance of 4MPy-Cu herein diminishes rapidly in the flow cell, approaching the performance of Cu after ~ 1 h electrolysis (Fig. 3 A and B). Meanwhile, the after-reaction XPS characterization revealed that 4MPy is no longer present on the Cu surface, in stark contrast to the stable attachment of 4MPy in MEA (Fig. 3 C and D). Thus, the deteriorated performance can be attributed to the rapid desorption of 4MPy-Cu in the flow cell which was also observed previously but not discussed in detail (17, 41). This desorption further agrees with the operando Raman measurements of 4MPy-Cu conducted in a flow cell (SI Appendix, Fig. S25), where the Raman peaks of 4MPy decreased rapidly upon successive linear sweep voltammetry (Fig. 4E). NMR measurements on the after-reaction solution further confirmed the presence of desorbed 4MPy in the electrolyte (SI Appendix, Fig. S26). The cyclic voltammogram (CV) showed that the detachment occurs at the potentials of approximately −0.45 V vs. RHE (SI Appendix, Fig. S27 A and B) in both the H-cell and flow cell. In comparison, no desorption peaks can be observed in the CV of 4MPy in the MEA in the similar potential ranges (SI Appendix, Fig. S27C). Most strikingly, the operando Raman spectra of 4MPy in the flow cell demonstrate an absence of above tautomerization processes in MEA, with the spectra remaining unchanged even at current densities above 100 mA cm⁻². Interestingly, we further conduct similar Raman measurements of 4MPy in an H-cell type Raman cell where the 4MPy-Cu is also in direct contact with 1 M KOH; both CV measurements and the Raman results show similar results as those flow cell (SI Appendix, Fig. S28), namely, revealing the similar absence of tautomerization in traditional H-cells as well. Therefore, these results lead to a surprising but compelling conclusion that tautomerization can only occur inside the MEA cell but not in the flow cells nor in H-cells; The slight initial higher performance of 4MPy-Cu than Cu should then be attributed to other minor contributions of pristine 4MPy similar to 2MPy/3MPy in regulating the microenvironments for CO₂RR to CO (e.g., hydrophobicity changes) rather than tautomerization.

Fig. 3.

Schematic mechanisms of 4MPy tautomerization to facilitate the C₂₊ generation during CO₂RR at the triple phase of MEA. Under operational conditions, the cathodic potentials drive a rapid tautomerization of the 4MPy molecule and subsequent protonation (process 1). Concurrently, the electrode becomes more hydrophobic with reduced water accessibility (process 2). Both the reduced HER (process 3) and the pH buffering effect provided by the tautomerized 4MPy would result in lower local alkalinity at the surface. This reduces the CO₂ loss arising from the formation of CO₃²⁻ (process 4), enhances the local concentration of CO₂ supply, and thus cathodically shifts the onset potentials for CO_ads. As a result, the suppressed water accessibility and enhanced surface CO_ads facilitate the CO–CO coupling (process 5), accounting for the observed higher C₂₊ selectivity and reduced HER for 4MPy-Cu.

Fig. 4.

Differences in the functional mechanisms of 4MPy between MEA and flow cell devices during CO₂RR. (A and B) Stability of Faradaic efficiencies of C₂H₄ and H₂ during CO₂RR in MEA (A) and flow cell (B) for 4MPy-Cu and Cu. (C and D) XPS spectra for N 1s (C) and S 2p (D) of 4MPy-Cu before and after CO₂RR in MEA and flow cell. The detailed peak fittings and discussion are included in SI Appendix, Table S3. (E) Raman spectra of 4MPy-Cu in 1 M KOH during CO₂RR with three consecutive scans of the current densities from 0 to 120 mA cm⁻²; tautomerization is negligible in the flow cell despite the same catalysts utilized. (F and G) Proposed mechanisms for the different surface chemistry of 4MPy in MEA (F) as compared to the flow cell (G). The absence of explicit catholyte flow in MEA causes a weaker surface alkalinity; this, together with the limited water access due to surface hydrophobicity, creates a local environment enabling the stable attachment of 4MPy and its subsequent tautomerization. In comparison, direct contact with strong alkaline catholyte in the flow cell leads to rapid desorption of 4MPy, thus eliminating the possibility of tautomerization.

We note there has been a long-time suspicion that the triple-phase junction at the MEA interface may represent a unique electrochemical environment for surface reaction to occur. However, to the best of the author’s knowledge, no direct evidence has clearly illustrated this difference. The above observation thus represents an example of its class and confirms the unique properties of MEA devices in driving surface tautomerization processes. Due to the lack of experimental tools, direct operando visualization of the microscopic triple-phase junction within the MEA device remains a great challenge for the community (59, 60). Yet in comparison with flow cells and H-cells, the MEA lacks explicit catholyte in the device, with the water phase in the cathode contributed from the humidified gas flow and permeated water from the anion exchange membrane side, thus forming an electrode surface only partially covered by water for ionic conduction. As such, the water layer would process lower alkalinity at the interface of MEA due to the dilution of the permeated electrolyte by pure water from the humidified CO₂. It is hypothesized that both the limited water content and its lower alkalinity may account for the much more stable attachment of 4MPy and its subsequent tautomerization in MEA upon cathodic polarization. By contrast, the complete immersion of the molecular modified electrode into the high alkaline solution in flow cells represents a severely harsh environment with strong alkalinity and leads to rapid detachment of the molecules before its possible tautomerization both in flow and H-cells. Further control measurements were also performed on 4MPy-Cu in a flow cell with 1 M KHCO₃ as the electrolyte to mimic the lower pH environments existing at the MEA interface. As shown in SI Appendix, Fig. S29, this lower pH indeed partially reinitiates the tautomerization processes in the flow cell, though weaker than those of MEA; with prolonged time, the 4MPy also slowly detached from the surface. Thus, it is reasonable to conclude that the low alkalinity together with the unique triple phase of MEA devices contributes to the tautomerization of 4MPy in MEA and drives the unique interfacial chemistry.

Notably, with recent emerging interests in utilizing non-Faradic processes at the electrified interface to drive noble surface chemical reactions, the above identification of tautomerization only at the triple-phase junction of the MEA device instead of flow cell at the same electrolyte conditions demonstrates further opportunities in utilizing microenvironments in designing surface reactions at electrified interfaces besides the impact of the electrode potential. The extraordinary stability provided by the triple-phase junction of MEA is also inspiring to the applications of many homogeneous molecular catalysis as many these molecules desorb when heterogenized onto support electrodes before reaching its catalytic potential. Further exploration of these catalysts in MEA devices may provide a strategy to stabilize them and expand their usage in a much larger range of potentials.

To conclude, this study demonstrates a simple molecular modification strategy by 4MPy molecules which significantly improve the C₂₊ and C₂H₄ selectivity of Cu during CO₂RR in MEA at industrial current densities. Together with the reduced cell voltage, the energy efficiency was boosted from 20.8 ± 1.1% to 34.4 ± 1.9% (200 mA cm⁻²). Both CORR and “molecular surgery” experiments show that 4MPy improves predominantly the conversion of CO₂ to CO, and the para position of N and S atom is important for this improvement. Further mechanistic studies by operando Raman directly at large current densities (over 100 mA cm⁻²) of MEA revealed that the key to success originates from the limited access to surface water and lower local alkalinity provided by the surface modification. An electron-induced tautomerization of 4MPy was identified to occur under reaction conditions, accounting for the above changes of local environments and representing an operando observation of thiol and thione transformation at the electrified MEA interface. Notably, this tautomerization can only occur in the MEA cell but not the flow cell at the same electrolyte conditions. Therefore, further utilization of molecules with electron-induced tautomerization properties, e.g., thiol–thione or imine–enamine tautomerization is of great promise to optimize the local environment for CO₂RR on Cu for improved selectivity. The unique electrochemical interface of MEA may also be further explored to develop new electrified chemistry and applied to another electrochemical research beyond CO₂RR.

Materials and Methods

Materials and Chemicals.

All chemicals, including4-mercaptopyridine (ACROS organics, 96%), 2-mercaptopyridine (ACROS organics, 99%), pyridine (ACROS organics, ≥99%), thiophenol–Na (Sigma-Aldrich, ≥99%), and KOH (Sigma Aldrich, 99.999%) were used as received. A polytetrafluoroethylene membrane (Sartorius™, 0.45 μm pore size) was received from Beijing Zhongxingweiye Instrument. 3-mercaptopyridine was synthesized according to previous literature with detailed synthesis procedure and characterization shown in SI Appendix, Note 4.

Synthesis of Sputter Cu.

The bare Cu-based GDE was prepared by sputtering 200 nm Cu (Cu target, sputtering rate ~0.6 Å s⁻¹) directly onto a piece of PTFE membrane using a magnetron sputtering system.

Synthesis of 4MPy-Cu.

For the preparation of 4MPy-Cu, a catalyst ink consisting of 1 mL isopropanol and 1.1 mg 4MPy was sprayed onto the Cu/PTFE membrane with a square of 10 cm². The mass loading is approximately 1 mM cm⁻² after drying under vacuum. The same procedure and mass loading were further used for the preparation of 2MPy-Cu, Py-Cu, and thiophenol–Cu electrodes.

Operando Raman Measurements.

Operando Raman spectroscopy measurements were carried out using a Horiba Xplora™ Plus Raman microscope with a homemade MEA cell (Fig. 2A and SI Appendix, Fig. S10). A 638 nm laser was used as the excitation source and the power is typically attenuated to 7.3 mW for the measurements. A long-working distance objective lens (10.6 mm, Olympus, ×50, NA = 0.5) was used for focusing and collecting the incident and scattered laser light.

Electrochemical CO₂/CORR Performance Test and Product Analysis.

Electrochemical CO₂RR/CORR measurements were performed in both flow cell and MEA setups. The MEA setup is a complete 1 cm² CO₂ or CO electrolyzer which includes a titanium anode flow field, 306 L Stainless Steel cathode flow field and associated nuts, bolts, and insulating kit. The catalyst-deposited PTFE electrode (GDE) was attached to the cathode by copper tape and electrically connected. The Cu side of the GDE faced toward the membrane. The copper tape was protected by Kapton tapes to avoid electrical contact with the membranes or electrolytes. A Sustainion^® membrane (Dioxide Materials) was activated in 1 M aqueous KOH solution for 24 h, washed with water, and then used as the anion-exchange membrane (Sustainion, X37-50-grade T). Ni foam (Shanghai Tankii Alloy Material Co., Ltd.) is used as the anode. An Autolab PGSTAT302N was used as the electrochemical workstation to apply voltage to the cathode and anode. No iR compensation was applied. 1 M KOH electrolyte was used as the anolyte and was circulated using a gas–liquid mixed flow pump (GaossUnion EC200-02) with a robust soft PVC tubing (McMaster-Carr, 5231k134). The electrolyte flow rate was kept at 50 mL min⁻¹. The flow rate of the CO₂ or CO gas flowing into the gas chamber was kept at 50 sccm by a digital gas flow controller (Sevenstar D08-1FB). Notably, CO₂ or CO flowed through a humidifier (1/3 full of Milli-Q water, room temperature) prior to the MEA.

In the flow cell setup, the catalyst-deposited GDE was used as the working electrode (cathode), while Ni foam acted as the counter electrode (anode). An anion exchange membrane (Fumasep FAA-PK-130) was used as the separation membrane. The flow cell consisted of GDE, anion exchange membrane, and nickel anode, and 1 M KOH was used as the electrolyte to circulate through the anode and cathode zones by means of a peristaltic pump (SHENCHEN LabV1-III). CO₂ or CO was vented from the back of the GDE using a mass flow controller with a flow rate of 30 mL min⁻¹, and an Hg/HgO electrode was inserted as a reference electrode in the cathode electrolyte bath.

The gaseous products of both CO₂/CORR were analyzed by gas chromatography (Fuli GC9790Plus) equipped with a thermal conductivity detector (TCD) and a flame ionization detector (FID). High-purity Argon (99.99%) was used as the gas carrier. Liquid products were quantified by ¹H NMR spectroscopy (600 MHz Agilent DD2 NMR Spectrometer) using dimethyl sulfoxide (DMSO) as the internal standard.

Computational Details.

The DFT calculations were performed with the Gaussian 16 software package (61). All the ground state geometries were optimized using the PBE0 functional (62) with the 6-311+G(d,p) basis set (63–65), incorporating empirical dispersion corrections with BJ-damping(GD3BJ) (66, 67). Geometry optimizations are calculated in both the gas phase and aqueous phase, with the latter simulations employing the implicit solvent model Solvation Model Based on Density (SMD) (68). Vibrational frequency analysis was performed at the same computational level for the Raman activities for each normal mode. The Raman spectroscopy under 638 nm incident light was simulated using the Multiwfn software (69), with vibrational frequencies corrected using a scale factor of 0.9594 (70) to account for anharmonic effects. VEDA 4 software (71) was used in the identification of Raman active modes.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.