Lattice-Nitrogen-Mediated Chemistry Suppresses Hydrogen Evolution for Record Faradaic Efficiency in Ammonia Synthesis

Abstract

Ammonia (NH₃) production using air, water, and electricity offers a transformative route to carbon-free chemical synthesis, addressing global sustainability challenges. However, the hydrogen evolution reaction (HER) in aqueous systems significantly hinders NH₃ selectivity, limiting the Faradaic efficiency (FE) to below ∼15%. Herein, we report an FE of approximately 48% for aqueous NH₃ synthesis, using two-dimensional (2D) nitride catalysts. These catalysts enable lattice nitrogen protonation through the Mars-van Krevelen (MvK) mechanism, effectively suppressing HER. Using operando spectroelectrochemistry, we identified active sites and tracked nitrogen vacancy cycles, providing unprecedented insights into the reaction pathways. Our findings, supported by advanced computational techniques and complementary spectroscopic analyses, highlight the stability and efficiency of the MvK cycle, setting a new benchmark for sustainable NH₃ production.

Introduction

Ammonia (NH₃) is a cornerstone of modern industry, enabling global food security and economic growth. − However, its large-scale production via the Haber–Bosch process is energy-intensive, heavily reliant on fossil fuels, and inaccessible in some regions due to infrastructure limitations. Electrochemical NH₃ synthesis through the nitrogen reduction reaction (NRR) offers a transformative alternative, capable of operating under ambient conditions and achieving carbon neutrality when powered by renewable energy. − This approach also opens the door to decentralized NH₃ production, addressing global inequities in access to this critical resource. −

Despite its promise, the NRR efficiency is hindered by the competing hydrogen evolution reaction (HER) and the high activation energy required to break the NN triple bond. Traditional NRR mechanisms, whether associative or dissociative, struggle with low energy efficiency. − Recently, single-atom catalysts such as Fe_SA–N–C and Rh₁/MnO₂ have demonstrated promising FE of up to 56.55 and 73.3%, respectively. However, atomically dispersing metals in a stable and uniform manner often involves complex synthesis procedures, which can be difficult to scale or reproduce. The Mars-van Krevelen (MvK) mechanism offers an innovative pathway, leveraging lattice nitrogen atoms as intermediates. − Protonation of the lattice N leads to the formation of N vacancies following the NH₃ desorption, which are then replenished via the N₂ feed gas. For instance, thermocatalytic studies on Co₃Mo₃N have demonstrated that lattice nitrogen can be reversibly exchanged during ammonia synthesis, with direct experimental evidence from isotopic labeling confirming the MvK pathway. , Recent theoretical studies have also shown that the MvK pathway is kinetically more favorable than the conventional Langmuir–Hinshelwood type pathway on several metal nitrides, particularly those with stable lattice N vacancy and moderate vacancy formation energies. Microkinetic modeling on rock salt-type VN(100), for example, has revealed that the rate-determining step for MvK is significantly more accessible than N₂ dissociation in associative routes and that lattice hydrogenation proceeds more readily at lattice N sites than metal sites. However, realizing this mechanism requires catalysts with stable nitrogen vacancies and high electronic conductivity. , Transition-metal nitrides, including MNenes, have emerged as strong candidates due to their tunable stoichiometry, electronic properties, and structural stability. − However, existing bulk transition-metal nitrides, like VN and TiN, degrade over the course of NRR because of the instability of N vacancies. ,

Prior theoretical investigations on the MvK mechanism have largely focused on surface functional groups and axial reaction mechanisms, − with limited exploration of its edges, apart from one study on porous MNenes. Similarly, studies on related MXene materials such as Ti₃C₂T _x , and Ti₂CO₂ have not adequately addressed edge phenomena. To fully understand the existence and evolution of nitrogen vacancies at MNene edges for NRR, an integrated computational and experimental approach is urgently needed.

In this work, we investigate the Ti₂NT _x MNene as a catalyst for NRR in aqueous environments, achieving an FE of 47.5 ± 2.8%. Through integrated experimental and computational approaches, including isotopic labeling, operando X-ray absorption spectroscopy (XAS), density functional theory (DFT), and ab initio molecular dynamics (AIMD), we demonstrate that the Ti₂NT _x catalyst operates via a stable and sustained MvK mechanism. Notably, we identify edge sites as the primary loci of nitrogen vacancy formation and regeneration, supported by experimental results and theoretical calculations. These findings represent a significant step toward scalable, sustainable NH₃ synthesis and advance the fundamental understanding of MNene catalysis. A schematic of our proof-of-concept of the MvK mechanism compared with the conventional associative and dissociative pathways is shown in Scheme .

1. Illustration of conventional electrochemical surface N₂ adsorption/activation mechanisms vs. unique MvK pathway exhibited by Ti₂NT _x MNene.

^a The MvK mechanism enables high NH₃ FE in aqueous environments.

Results and Discussion

Synthesis and Characterization of Ti₂NT _x MNene

The Ti₂NT _x MNene catalyst was synthesized through a modified etching process adapted from our prior work (Figure a). Characterization by X-ray diffraction (XRD) and Raman spectroscopy confirmed phase purity and structural integrity, consistent with previous reports (Figure b,c). In the XRD patterns (Figure b), the characteristic (002) peak shifted from 12.9° in the Ti₂AlN precursor to 9.6° in the Ti₂NT _x MNene, corresponding to an increase in interlayer spacing from 0.68 to 0.92 nm. This expansion indicates the removal of the Al layer in Ti₂AlN and the formation of well-exfoliated Ti₂NT _x sheets. Notably, no diffraction peaks associated with residual fluoride salts (LiF, KF, and NaF) were observed, indicating the effectiveness of the acid washing step and the high phase purity of the final material. Raman spectroscopy (Figure c) further supported the Al etching, showing the disappearance of the Ti₂AlN precursor’s characteristic vibrational modes and the emergence of a distinct ∼400 cm^–1 band associated with Ti atomic plane vibrations in Ti₂NT _x . Additionally, a shift and broadening of the A_1g mode was observed, consistent with lattice distortions and modifications induced by surface terminations during the etching process. To elucidate the chemical state and atomic structure, ex situ XAS was performed at the Ti K-edge. X-ray absorption near-edge structure (XANES) and Fourier-transformed extended X-ray absorption fine structure (FT-EXAFS) analyses revealed changes in Ti valence states and bond environments during synthesis (Figure d,e). Using a calibration curve derived from first-derivative spectra of Ti reference materials (Figure S1, Supporting Information), we determined Ti valence states of 2.1 for the Ti₂AlN MAX phase, 3.6 for the multilayer (ML) MNene, and 3.9 for the delaminated (DL) MNene. These changes correspond to the replacement of Ti–Al metallic bonds in the MAX phase with Ti–N and Ti–O bonds in MNene, which is induced by etching and surface functionalization. EXAFS analysis (Figure S2) identified peaks at 1.5 Å (Ti–N/Ti-O), 2.5 Å (Ti–Ti), and 3.7 Å (Ti–N–Ti multiple scattering), consistent with literature values. − These features were preserved across all synthesis stages (Figure e), indicating that the conversion from MAX to MNene retains the necessary local bonding motifs essential for catalytic performance.

1.

(a). Schematic of Ti₂NT _x MNene synthesis via oxygen-assisted molten salt fluoride treatment of the Ti₂AlN MAX phase (detailed procedure in Experimental Section). (b). X-ray diffraction patterns of Ti₂AlN MAX (red) and delaminated Ti₂NT _x MNene (blue) were measured on a zero-diffraction silicon holder. (c). Raman spectra of Ti₂AlN MAX (red) and delaminated Ti₂NT _x MNene (blue) were collected with a 532 nm laser. (d). Ti valence as a function of edge energy from XANES spectra for reference compounds (hollow circles), Ti₂AlN MAX (red), multilayer Ti₂NT _x (green diamond), and delaminated Ti₂NT _x MNene (blue square). (e). Fourier transform of normalized EXAFS spectra for Ti₂AlN MAX (red), multilayer Ti₂NT _x (green), and delaminated Ti₂NT _x MNene (blue), using a k weight of 2 for EXAFS fitting.

NRR Performance of Ti₂NT _x MNene Catalyst

The Ti₂NT _x MNene catalyst demonstrated an excellent NRR performance in neutral electrolytes (LiCl, NaCl, and Na₂SO₄, 0.1 M). Cyclic voltammetry (CV) and linear sweep voltammetry (LSV) measurements identified an NRR onset potential near 0 V (vs RHE) and an HER onset at approximately −0.8 V, providing a favorable electrochemical window for NRR, especially compared with acidic systems (Figures a, S3 and S4, Supporting Information). The FE increases with decreasing cathodic potential in Na₂SO₄ and achieved an FE of 47.5 ± 2.82% for NH₃ production, with comparable performance observed in LiCl and NaCl electrolytes (Figures a,b, S13 and S14). Post-experiment XAS revealed a correlation between the Ti oxidation state and catalytic performance. For example, an increase in Ti oxidation state from ∼3.6 at 0 V to ∼4.0 at −0.2 V could be responsible for the decline in FE from 47.5 to 12.2% (Figures c, S17). Chronoamperometry and CV measurements further supported this observation, showing that the sample operated at lower cathodic potentials retained higher current density, whereas performance degradation is observed at higher cathodic potentials (Figures d, S18). This observation is likely due to the MvK mechanism, as lattice nitrogen is extracted to form NH₃ under NRR conditions, and nitrogen vacancies are created within the Ti₂NT _x lattice. In the absence of complete nitrogen replenishment from the N₂ gas phase at higher cathodic potentials, these vacancies become susceptible to the incorporation of oxygen from the electrolyte, leading to localized surface oxidation. These results underscore the importance of potential-dependent control over vacancy dynamics and the conditions under which the MvK cycle remains reversible.

2.

(a). Faradaic efficiencies for NH₃ synthesis were obtained across various neutral electrolytes. (b). NH₃ Faradaic efficiency and yield rate of Ti₂NT _x MNene in N₂-saturated 0.1 M Na₂SO₄ electrolyte at different potentials after 4 h of chronoamperometry. (c). XANES-derived edge positions of the MNene catalyst after 4 h of chronoamperometry at 0 V vs RHE (red diamond) and −0.2 V vs RHE (blue diamond), indicating Ti valence changes. (d). Cyclic voltammograms before (red) and after (blue) 4 h chronoamperometry at 0 V vs RHE, measured at a scan rate of 50 mV s^–1.

Recent studies have highlighted that ubiquitous nitrogen-containing species, such as NH₄ ⁺ and NO _x , can reach levels up to ∼40 nmol, even when using ultra high purity gas supplies. Additionally, ammonia contamination from ambient air and the prolonged sitting of the electrolyte can introduce further complications, potentially leading to overestimation of NRR performance, a challenge particularly critical for catalysts with inherently low ammonia yield rates. While control experiments, such as isotope labeling with ¹⁵N₂, are typically used to resolve such ambiguities, they become difficult to apply in systems governed by the MvK mechanism, where the catalyst inherently contains lattice ¹⁴N. As a result, one effective strategy to reduce the influence of ammonia contamination is to move away from low yield regimes, that is, to increase the total amount of electrochemically produced ammonia so that background contamination becomes negligible relative to the signal. To address this, we strategically employed a Ti₂NT _x catalyst spray-coated onto a gas diffusion layer (GDL) electrode with a geometric surface area of 4 cm² for the NRR experiments (Figure S15a). This GDL electrode includes a lid on the backside through which N₂ gas was purged at 2 mL min^–1, ensuring direct and continuous access of N₂ to the catalyst surface via microporous channels, effectively overcoming N₂ solubility limitations. Incorporating this setup, the observed total NH₃ produced increased significantly compared with the experiments carried out using a low geometric area (0.192 cm²) glassy carbon electrode. Therefore, the observed NH₃ production is primarily from electrochemical NRR instead of contamination. Overall, the use of the GDL configuration provides greater confidence in the NRR results, as the total ammonia production is well above the threshold of interference from ambient ammonia contamination.

Determining the Mars-van Krevelen Mechanism

The observed high FE of the MNene garners interest in the NH₃ formation mechanism. To study this, the feed gas was changed to analyze the production rate and product identity (Figures , S19–S21). First, Ar rather than N₂ was used to achieve an inert atmosphere within the reaction vessel for 68 h. If N from the catalyst was the N source, then there would be considerable NH₃ production. In Figure S19a, during the first four hours of catalytic conditions, ∼6.1 μg of NH₃ was produced, corresponding to 70% of the total lattice N being converted to NH₃. Over 68 h of catalytic conditions, the amount of produced NH₃ stays below the maximum theoretical in the case of 100% lattice N conversion. The slight NH₃ decrease produced after 50 h is attributed to membrane crossover. Furthermore, the NH₃ yield rate of the catalyst starts at ∼7.8 μg cm^–2 h^–1 before exponentially decaying toward zero with time, due to bulk N being protonated and desorbing. The same trend exists for FE due to applied charge going toward stabilizing the vacancy-filled MNene structure (Figure S19b, Supporting Information). The source of the N for NH₃ production was further corroborated through N₂ isotopic exchange experiments repeated on the same electrode in a semibatch mode (Figure ). It can be observed that upon first introduction of ¹⁵N₂ into the system, a mixture of ¹⁴NH₄ ⁺ and ¹⁵NH₄ ⁺ is observed in the 1H-NMR spectrum. After the first introduction of ¹⁵N₂, the electrode was removed from the system and the cell thoroughly cleaned before fresh electrolyte was filled and the cell reassembled. This process was repeated 3 times, and it was observed that the ratio of ¹⁵NH₄ ⁺ to ¹⁴NH₄ ⁺ increases over these trials indicating that bulk nitrogen atoms are being involved in the reaction sequence and getting replaced with the ¹⁵N atoms. This is further shown when switching back to ¹⁴N₂ as the feed, where the ratio of ¹⁵NH₄ ⁺ to ¹⁴NH₄ ⁺ can be seen to rapidly decrease before only ¹⁴NH₄ ⁺ is observed in the NMR spectrum. This sequence is only possible if the bulk nitrogen atoms in the structure are involved in the reaction, leading to mixtures of isotopes. This further explains the stability of nitrogen vacancies as ¹⁵N is filled in the structure.

3.

1H-NMR results from isotopic exchange experiments for the NH₃ detection. Detected products from each experimental run are indicated to the right of their respective spectrum. Results demonstrate the existence of the MvK mechanism in MNenes.

Investigation of Mars-van Krevelen Mechanism Pathway through DFT Free Energy Profiles

We explore three possible MvK mechanisms via DFT: the associative Heyrovsky MvK, MvK_as, (hydrogenations occur on the N-site), the dissociative Heyrovsky MvK, MvK_dis, (N₂ dissociates on the N-site, then undergoes associative steps on the Ti bridge site), and a new associative-dissociative Heyrovsky MvK, MvK_as‑dis, (dissociation occurs after NH₂ formation, followed by migration to the Ti bridge site), further defined in the methods section in SI. The surfaces under investigation include Ti₂N­(OH)­O, Ti₂N­(OH)_2, Ti₂NO₂, and Ti₂NO₂ with edge −OH terminations. These mechanisms are studied under both neutral and acidic electrolyte conditions to provide insights into their MvK behavior. The newly proposed MvK_as‑dis mechanism is additionally investigated (Figures S22 and S23, Supporting Information). We include all three possible MvK mechanisms in our energy profile calculations (Table S1, Supporting Information). Detailed discussion of different MvK mechanisms on various Ti₂NT _x MNene edge surfaces (Figures S24–S27, Supporting Information) is elaborated in the Supporting Information. MvK_as is found to be the most favorable mechanism across all surfaces, with MvK_as‑dis potentially occurring concurrently. Ti₂N­(OH)O has the highest energy barrier for nitrogen vacancy formation, while Ti₂NO₂ requires the least energy. Edge −OH terminations on Ti₂NO₂ increase the energy barrier for vacancy formation but facilitate the spontaneous protonation of replenished vacancies. Controlling pH can modify edge surface protonation, potentially promoting the MvK mechanism. However, a trade-off exists between nitrogen vacancy formation and hydrogenation of replenished vacancies, influenced by the ratio of −OH edge terminations on the Ti₂NO₂ surface.

Timing and Extent of Lattice Nitrogen Protonation

AIMD simulations suggest that nearby hydronium ions in neutral and acidic solutions play a crucial role in the N replenishment step. Specifically, hydrogenation by nearby hydronium ions impedes complete embedding of N₂ molecules, leading to partial embedding that favors the MvK mechanism (Figures a–f, S28–S31, Supporting Information). By “partially embedded,” we refer to one nitrogen atom from nitrogen gas filling into a single nitrogen vacancy site; and “fully embedded” refers to two nitrogen atoms from nitrogen gas filling into a single nitrogen vacancy site. This occurs primarily due to the steric hindrance from N₂H or N₂H₂ molecules and the reduced electronegativity of the N atom after protonation. This distinctive embedding behavior is closely related to whether the N–N bond breaking can happen spontaneously. When full embedding occurs (i.e., both nitrogen atoms fill the single vacancy) (Figures g–l, S33 and S34, Supporting Information), N–N bond breaking does not occur due to strong Ti–N coordination. Alternatively, in cases of partial embedding, N–N bond breaking is observed (Figures c,d, S28–S31, Supporting Information), facilitated by reduced steric hindrance that allows the formation of N₂H or N₂H₂ at the vacancy sites, weakening the N–N bond. We note that prompt hydrogenation from the hydronium ions prevents all of the N₂ gas molecules from entering the N vacancy on the edge sites. Additionally, multiple Ti–N coordination prevents the N atom from leaving the surface, hindering multiple hydrogenations on a single nitrogen, resulting in mostly embedded NHNH as well as if N₂ is close to nearby vacancies and bridge sites before filling into vacancy. Therefore, if two or more protonation steps yield NNH₂ and NNH₃, the adsorbent is likely to move to nearby bridge sites instead of filling into vacancy sites (Figures S34–S37, Supporting Information). Without the replenishment step, the energy barrier for N₂ bond breaking becomes challenging, and no N–N bond breaking is observed in the bridge NRR pathway during our AIMD simulations.

4.

Selected snapshots from 1 ps AIMD simulations of adsorbed N₂ near a nitrogen vacancy N_vac on the Ti₂N­(OH)₂ bilayer model (θ = 1/4). The vacancy filling process may vary from partial to full under different hydronium concentrations, potentially influencing the feasibility of the MvK mechanism. a–f. N₂ partially embeds in * _vac and follows the MvK_as‑dis pathway on the surface in 1.25 M H₂SO₄. (a). N₂ was positioned near * _vac . (b). N₂ partially filled * _vac . (c). After the first protonation, the N–N bond begins breaking. (d). The N–N bond breaks. (e). Protonation of NH leads to NH₂ formation. (f). NH₂ remains on the Ti_top near the interlayer. g–l N₂ fully embeds in * _vac without undergoing NRR in a 1.25 M Na₂SO₄ solution. (g). N₂ was positioned near * _vac . (h). N₂ begins filling the nitrogen vacancy. (i). Without prompt protonation, N₂ fully occupies the nitrogen vacancy. (j). After the first protonation, the NNH formed. (k). The embedded N atom moves upward, and N₂ becomes horizontally embedded and gets trapped in the vacancy site. (l). NNH remains fully embedded, with the N–N bond intact. Cyan, blue, white, green, gray, red, and yellow represent nitrogen atoms above the first layer, nitrogen atoms below the first layer, hydrogen, hydrogen adsorbed on nitrogen, titanium, oxygen, and sulfur, respectively.

Importance of −OH Termination on Edge Surface and Proton Concentration of H Source

Analyzing the −OH/O ratio on the Ti₂NT _x surface is crucial for understanding the MvK mechanisms. −OH termination influences water splitting equilibrium in neutral electrolyte, hydronium concentration in acidic electrolyte, and the embedding and breaking of N₂. AIMD simulations of nitrogen vacancy replenishment demonstrate the Ti₂N­(OH)₂ surface undergoing an associative-dissociative MvK mechanism shown in Figure b–d. From an overall comparison, both OH_edge_O and O_edge_OH terminations promote N₂ embedding, while O_edge_O terminations discourage it (Figures a–h, S38 and S39, Supporting Information). Interestingly, the accessibility of −OH terminations near the edge surface influences the MvK via hydrogen bonding. Among different surfaces, Ti₂N­(OH)O (OHO_edge_O) exhibits more frequent and stronger interlayer −OH bonds, stabilizing the interlayer, but making OH bonds more difficult to break. Therefore, the −OH interlayer termination near the edge may not serve as a hydrogen source, so hydrogen from water splitting or hydronium is required in neutral and acidic systems, respectively, for N₂ reduction.

5.

Summary figure of 1 ps AIMD calculation for MvK mechanism on different Ti₂NT _x edge facets in 2 × 2 × 9 supercell (a–d). in neutral (1.25 M H₂SO₄) and in (e–h). acidic electrolyte (1.25 M NaSO₄) with nitrogen vacancy ( $\theta =\frac{1}{4}$ ) and adsorbed N₂ initially near the vacancy. Illustrations are selected at different snapshots for better visualization. (a). Instead of filling *_vac, N₂ went to *_bri without undergoing NRR. (b). Proceeded and formed NH₂ and NH. (c). Formed N₂H while maintaining the N–N bond. (d). N₂ located on *_bri and approached *_vac without undergoing NRR. (e). N₂H formed while partially embedding into *_vac. (f). N₂ fully embedded horizontally into *_vac and formed N₂H. (g). N₂H formed while partially embedding into *_vac. (h). N₂ stayed on Ti_top without undergoing NRR. Color code: gray, red, blue, cyan, white, green, yellow, and purple represent Ti, O, N below the first layer, N above the first layer, H, and H bonded with N, S, and Na, respectively.

Presence of Nearby NH₃ Adsorption Effect on MvK Mechanism

We investigate the influence of pH and nearby NH₃ on the MvK mechanism on different Ti₂NT _x MNene edge surfaces (Figures a–l, S28, S30–S33, S35–S37, S40–S43, Supporting Information) and summarized in Figures and S44. Overall, we found that higher hydronium concentrations accelerate N₂ reduction but impede vacancy replenishment. SO₄ ^2– ions stabilize hydronium and facilitate proton transfer. AIMD studies reveal that NH₃ desorption and hopping occur on Ti₂N­(OH)­O surfaces, while neither NH₃ nor H₂O embeds defects in all kinds of Ti₂NT _x surface. NH₃ can act as a proton source for nearby NH or NH₂ hydrogenation (Figures S26, S27, S41, and S45–S48 Supporting Information). The presence of NH₃ on the surface promotes traditional NRR pathways over vacancy filling, while N₂ and N₂H filling is more favorable in the absence of NH₃. The complex interactions among NH₃, N₂, and N₂H in desorption and vacancy filling during NRR are further discussed in the Supporting Information.

Stability of N Vacancies during the MvK Mechanism and Catalyst Regeneration

A shortcoming of catalysts that utilize the MvK mechanism is vacancy instability, leading to material degradation. Therefore, it is necessary to investigate the nitrogen vacancy stability; thus, Raman and operando XAS experiments were performed. To analyze the capability for N₂ replenishment, Raman analysis was conducted on the same electrode before and after 68 h of Ar-saturated NRR to induce N vacancy formation, then after 68 h of additional N₂-saturation to replenish the N vacancies (Figure ). This cyclic behavior is observed via analyzing 19 locations on the electrode. On the pristine electrode (Figures a, S49a), 17 of the 19 spots present a spectrum that aligns with the expected Ti₂NT _x spectrum, while 2 match that of the Ti₂NT _x MNene with an emphasized Ti–O vibration peak. This enhanced Ti–O peak aligns with that from TiO₂ materials, therefore the spectra were labeled as a hybrid Ti₂NT _x /TiO₂. Following 68 h under Ar-saturated NRR conditions, the electrode was removed from the cell, briefly dried, and analyzed (Figures b, S49b). Of the 19 spectra, 6 aligned with pristine Ti₂NT _x material and 13 aligned with the hybrid Ti₂NT _x /TiO₂ material. This demonstrates that lattice N is being extracted, forming N vacancies, which are then temporarily filled with oxygen. Finally, the electrode was placed into a fresh cell with electrolyte, and 68 h of NRR conditions were applied with N₂-saturation to replenish the generated N vacancies (Figures c, S49c). Afterward, 16 of the 19 spectra were converted back to pristine Ti₂NT _x , and 3 resembled the hybrid material.

6.

Raman white light images (inset) and spectra of Ti₂NT _x MNene coated on glassy carbon electrode (a) prior to NRR conditions, (b) after 68 h NRR under Ar-saturation to generate N vacancies, and (c) after 68 h NRR under N₂-saturation after Ar-saturation to replenish previous N vacancies. Color coding on the white light images corresponds to the Raman spectrum immediately to the right.

To further investigate the stability of nitrogen vacancy creation and replenishment, we performed FT-EXAFS analysis of the catalyst under various atmospheres to monitor real-time structural changes (Figure a–c). The corresponding XANES data are shown in Figure S51. Under a continuous N₂ feed (Figure a), the material largely retains its structural integrity instead of bulk oxidation, with only minor shifts in Ti–N/Ti-O bond lengths and cyclic changes in the Ti–N–Ti network, consistent with the formation and replenishment of vacancies. Notably, the final spectra closely match the initial, indicating the preservation of the local coordination environment after 2 h of reaction time. The EXAFS features at short radial distances ∼0.9 Å in Figures e and are regarded as nonphysical, arising from Fourier transform artifacts due to limited k-space resolution and do not correspond to real interatomic distances (Figure S50).

7.

Operando Fourier transform of the normalized EXAFS spectra for Ti₂NT _x after being exposed to (a). N₂ as a reactant gas, (b). Ar as an inert atmosphere gas, and (c). N₂ reintroduction as a regeneration gas. A k-weight of 2 was used for EXAFS fitting.

When the feed gas is switched to Ar (Figure b), vacancy formation persists; however, in the absence of N₂, these vacancies remain unfilled, leading to oxide formation as indicated by post-reaction spectra resembling TiO₂ (Figure S2, Supporting Information). Reintroducing N₂ to the system regenerates the pristine MNene structure as the material undergoes catalytic cycles to produce NH₃ (Figure c). The resemblance of the final spectrum in Figure c to the initial spectrum in Figure a and to the spectrum in Figure e further supports this interpretation. Nevertheless, these results highlight the dynamic interplay between vacancy creation, temporary oxide formation, and structural regeneration, underscoring the material’s resilience and catalytic adaptability.

A continuous N₂-purging long-term experiment demonstrated sustained NH₃ formation over 100 h, while also maintaining decently stable current density (Figure S52a). Moreover, XRD analysis of the Ti₂NT _x -coated GDL before and after the 100 h NRR experiment confirmed the retention of the MNene structure (Figure S52b). Raman mapping of the electrode before and after the 100 h NRR, performed over a 1 × 1 cm² area with 441 data points, also indicated the preservation of Ti–N bonds (Figure S52c, S52d). In addition, N₂/Ar feed cycling was also carried out for 14 h with the cycle lasting 2 h (Figure S53). Postelectrolysis XRD and Raman also suggest the preservation of Ti₂NT _x structure after N₂/Ar cycling, which is consistent with the results shown in Figure . Together, these studies only support the presence of a stable MvK cycle and highlight the reversible interplay between Ti–N and Ti–O bonding. However, to fully understand the structural and morphological stability, including possible surface reconstruction or compositional changes, operando and postelectrolysis characterizations should be considered.

Vacancy Filling Stability during MvK Mechanism

To compare the possible oxide formation at vacancy sites, we conducted adsorption free energy calculations for O₂/N₂/Na in the vacancy sites (Table S2, Supporting Information). Overall, the adsorption energy is smaller on Ti₂NO₂ (O_edge_O) for all adsorption sites, and the adsorption tendency for refilling MvK surfaces is O₂ ≫ N₂ > Na for all studied surfaces. The enhanced adsorption affinity of the O₂ bond is primarily attributed to the breaking of the O₂ bond, with one oxygen atom occupying the nitrogen vacancy site and the other occupying the Ti–Ti bridge site. This phenomenon implies potential deactivation due to oxide formation during the MvK mechanism, but as demonstrated earlier, the catalyst can be regenerated with N₂ to continue the catalytic cycle.

DFT Insights into Pair Distributions during MvK Mechanism Filling– Exploring Layer Changes, Charges, and Varied Adsorptions/Vacancies

To investigate changes in the pairwise radial distribution during the MvK mechanism, we analyzed Ti–O, Ti–N, and Ti–Ti pair distances using a combination of DFT calculations and experimental EXAFS data (Figure ). Median pairwise distances were found to be approximately 1.8–2.2 Å for Ti–O pairs, 2.0–2.3 Å for Ti–N pairs, 2.7–3.3 Å for first-neighbor Ti–Ti pairs, and 3.9–4.5 Å for second-neighbor Ti–Ti pairs (Ti–N–Ti).

8.

Pairwise distance boxplot of Ti–O/Ti-N/Ti–Ti of different coordinates grouped by surface layer index (bottom layer 0, top layer 8), pairwise type, and different Ti₂NT _x surface including (a). surface Ti₂N­(OH)O (OHO_edge_O), (b). surface Ti₂N­(OH)₂ (OH_edge_O), (c). surface Ti₂NO₂ with edge −OH termination (O_edge_OH), and (d). surface Ti₂NO₂ (O_edge_O). For each boxplot, pairwise distance of different adsorption type and charge (includes 0, 1, 4 extra electrons for 1 × 2 × 9 slab model; 0, 2, 8 extra electrons for 2 × 2 × 9 slab model for bilayer model) from DFT optimized geometry are included. Each hollow circle represents an overlap for each pair count on each boxplot. It should be noted that we only include the pair distance analysis from layer 4 to 8 due to the fixation of atoms from layers 0 to 3. Neutral charged optimized geometries for different adsorption types (Figures S54–S57, Supporting Information) explain the pairwise distance change of Ti–N due to the MvK mechanism at the vacancy filling step and possibly deactivation by oxygen vacancy filling for the TiNO₂ (O_edge_O) surface.

For Ti–N pairs, a wide distribution and binodal trends were observed across all of the studied surfaces. These trends correspond to nitrogen positions in either half-embedded N₂ molecules (2.8–3.1 Å from Ti–N pairs; Figure a,b) or fully embedded N₂ molecules (3.1–3.2 Å; Figure c,d). Despite a systematic offset between DFT and EXAFS results, these findings align with the increased peaks and fluctuations observed in Figure a,c, which suggest Ti–N and Ti–O bonds between 1.5 and 2 Å.

Variations in Ti–Ti pair distances were more prominent in the top two layers, where vacancy-induced relaxation influenced neighboring Ti atoms up to the second coordination shell. The observed shifts in the Ti–N–Ti peak (EXAFS) were primarily attributed to outer-edge surface relaxation. A violin plot comparison of adsorption types further corroborated these observations, confirming that radial changes stemmed from nitrogen vacancy filling by N₂ or O₂ rather than surface charging effects (Figures a–d, S54–S57).

An intriguing observation was the broad range of same-layer Ti–O distances, attributed to O₂ filling N vacancies. In Ti₂NO₂, for instance, stabilization occurred via interlayer oxygen atoms moving closer to the edge surface (Figures d, S54d). Convergence tests (Figure S58a–c, Supporting Information) on the Ti₂NO₂_(sur-N) surface revealed that edge stabilization was achieved when oxygen coordinated with three Ti atoms rather than two, a behavior unique to Ti₂NO₂.

For surfaces with −OH terminations, hydrogen bonding increased interlayer interactions, reducing edge oxygen electronegativity, and stabilizing −OH terminations. This stabilization prevented the deactivation of nitrogen vacancy sites, thereby preserving catalytic functionality. These findings highlight the critical role of vacancy interactions and terminations in governing the catalytic stability of Ti-based systems for NH₃ synthesis.

Conclusions

This work demonstrates a novel mechanism for NRR electrocatalysis using an emerging class of 2D MNene materials. In neutral aqueous conditions (0.1 M Na₂SO₄), the MNene catalyst achieved an FE of 47.5 ± 2.8%. Comprehensive analyses, including isotopic labeling and operando XAS, DFT, and AIMD simulations, revealed that the MvK mechanism is both thermodynamically favorable and kinetically efficient, with nitrogen vacancies regenerated by N₂ in the electrolyte. Proposed pathways (MvK_as, MvK_as‑dis, and MvK_dis) highlight the critical role of edge and interlayer terminations in the catalytic activity. These findings establish MNenes as a powerful platform for electrochemical NH₃ synthesis, offering significant insights into NRR mechanisms and advancing the development of sustainable, carbon-free ammonia production technologies.

Experimental Section

Synthesis of Ti₂NT _x MNene

Ti₂AlN MAX phase was synthesized via ambient pressure sintering. Ti and AlN were mixed and ground in a 2:1 molar ratio for 10 min. The mix was then placed in a tube furnace and heated to 1600 °C at a ramp rate of 10 °C min^–1 and held for 1 h under continuous Ar flow. The resulting Ti₂AlN pellet was ground in preparation for etching to MNene.

The Ti₂AlN MAX phase was etched through an oxygen-assisted molten salt fluoride etching technique as previously utilized. In short, the MAX phase was mixed in a 1:1 mass ratio with a 59:29:12 wt % mixture of KF, LiF, and NaF, respectively. This mixture was then placed in a tube furnace and heated to 550 °C at a ramp rate of 10 °C min^–1 under an Ar flow. The furnace was kept at 550 °C for 5 h, where halfway through the Ar flow was cut off and the furnace was sealed to the environment. The sealed furnace was then cooled to 450 °C and maintained there for 50 min before cooling to room temperature. The molten salt-treated MAX phase (Ti₂AlN-MST) was collected for acid washing to remove excess salts and etched species.

The Ti₂AlN-MST was washed in 4 M formic acid for 45 min before being filtered and washed with water until a neutral supernatant was achieved. The material was then dried in a vacuum oven at 50 °C overnight, and the multilayer material (Ti₂NT _x ML) was collected.

The Ti₂NT _x ML was delaminated via an intercalation and sonication strategy. Tetramethylammonium hydroxide was used as the intercalating agent by mixing for 4 h. The suspension was then sonicated for 30 min before centrifuging at 3500 rpm until a neutral supernatant was achieved. The material was then resuspended in a small amount of water and left to rest to allow the remaining MAX phase to precipitate from suspension. The colored supernatant was filtered and dried in a vacuum oven at 50 °C overnight.

Material Characterization

Crystalline structure of the material was analyzed by using X-ray diffraction (XRD, Rigaku Miniflex 6G). Raman spectroscopy was performed with a Renishaw inVia Qontor equipped with a 532 nm excitation wavelength laser, an 1800 lines mm^–1 grating, and a 50× long objective lens. XAS measurements were performed in fluorescence mode at the multipurpose beamline for spectroscopy, 12-BM, at the Advanced Photon Source (APS) located at the Argonne National Laboratory (ANL). A defined beam size of 0.5 × 0.8 mm² using slits and an incident photon flux of ∼10¹¹ photons s^–1 were used. XANES data were collected in the vicinity of the Ti K-edge (4966 eV) at ambient temperature. Ti foil, TiH₂, TiN, and TiO₂ rutile were investigated in fluorescence to obtain the reference spectra. XAS data were processed by using the Demeter software package with the built-in AUTOBK algorithm used to normalize the absorption coefficient.

Electrode Preparation for Electrochemistry

MNene was mixed into a mixture of ethanol and 5 wt % Nafion in ethanol and then sonicated for 30 min. This suspension was then dropcast onto a glassy carbon electrode and dried for 30 min at 50 °C under vacuum.

Electrochemical Measurements, Product Quantification, and Determination of Performance

The methods for electrochemical measurements were detailed previously and summarized here. Additionally, suggested protocols from the literature were employed to ensure that produced NH₃ did not originate from contamination sources. , To avoid potential NOx contaminants, 0.1 M KOH base trap and 0.05 M H₂SO₄ acid trap were used to purify the N₂ gas feed (ultrahigh pure, 99.999%). A photograph of the NRR electrochemical setup is provided in Figure S64. The counter and reference electrodes were graphite and Ag/AgCl, respectively. A Nafion-117 membrane was used to separate the anode and cathode chambers of the H-cell. Measurements were conducted using a Bio-Logic SP-300 potentiostat.

NH₃ concentration was detected via UV–vis spectroscopy using the indophenol blue method, and FE was calculated using the same methods previously discussed. ,

Isotopic Exchange Experiments

Isotopic exchange experiments were performed in a semibatch process to minimize gas usage to the system. ¹⁴N₂ and ¹⁵N₂ gases were pressurized to the back of a gas diffusion layer electrode with MNene material spray-coated on. The electrolyte was collected and analyzed via proton nuclear magnetic resonance (1H-NMR) spectroscopy with a 400 MHz Bruker Avance III system. For the ¹⁵N₂ measurements shown in Figure , the electrode was deliberately removed from the electrochemical cell after each 2 h run and immersed in fresh electrolyte for the subsequent run. The primary objective of this experimental design was to demonstrate that via the MvK mechanism, ¹⁵N₂ gas substitutes lattice ¹⁴N within the catalyst. When the ¹⁵N-substituted electrode is placed into fresh electrolyte, the ammonia formed originates from lattice ¹⁵N, as confirmed by the data in Figure . This provides direct evidence of the lattice substitution.

DFT Optimization of Structures

The research utilized first-principles methods implemented in the Vienna Ab initio Simulation Package (VASP) to carry out all computations. Density functional theory (DFT) was employed within the generalized gradient approximation (GGA), Perdew–Burke–Ernzerhof (PBE) functional for electron correlation, and the projector augmented wave (PAW) pseudopotentials for electron–ion interactions. Structural optimization was sampled in the Brillouin zone with the Monkhorst–Pack method using 2 × 3 × 1 for (system 1 × 2 × 9) and 2 × 2 × 1 for (system 2 × 2 × 9) k-point grid. The cutoff energy of 500 eV is deployed, and the convergence tolerance of force is set to be 0.01 eV A^–1, and 10^–5 eV for energy with a Gaussian smearing width of 0.1 eV. The empirical DFT-D3 correction was utilized to describe the VDW interactions to avoid the potential errors of physical adsorption. Moreover, spin polarization is used to accurately describe the electron distribution of Ti₂NT _x . ,− The free energy correction terms of the gas phase, including zero-point energy correction and vibrational entropy, are implemented using normal-mode analysis as in previous work. ,

Ti₂NT _x DFT Edge MNene Model Setup Feasibility

The details of convergence test for the edge slab model and adsorption energy E _ads and N vacancy formation energies E _vac are shown in Figure S58 and Table S3 (Supporting Information). Overall, we found that the 1 × 2 × 9 slab model with the fixbot4 setup can maintain the change of adsorption energy within 0.02 eV when we further increase system layers. Whereas, when we consider the Ti₂N­(OH)O MNene (OHO_edge_O model), we employed the 2 × 2 × 9 slab model to keep vacancy defect ratio at 50% for each surface, which is further explained in the Supporting Information.

Ab Initio Molecular Dynamics (AIMD)

AIMD simulations were employed to test the nitrogen vacancy filling and ammonia desorption for the MvK mechanism. Dynamics were run with the NVT ensemble at a temperature of 298 K with the energy convergence tolerance set to 10^–4 eV, without a force convergence tolerance. The Nose thermostat with a Nose-mass parameter of 0.5 was used with a time step of 1 fs, totaling 1 ps over each simulation.

The simulation cell was reoptimized using the supercell of our previous work’s unit cell. Four different surface terminations were analyzed, starting from Ti₂NO₂ to Ti₂NO₂ with an edge −OH surface termination, Ti₂N­(OH)₂, and the bilayer model of Ti₂N­(OH)O (Figure S54a, S55a, S56a, S57a, Supporting Information). To prevent interaction between the cleaved surface and the periodic system, a 10 Å vacuum layer was added after cleaving the surface along the edge direction of the MNene. To simulate the electrolyte system, we employed pure water, 2.5 M Na₂SO₄, and 1.25 M Na₂SO₄ to represent a neutral electrolyte system. Additionally, 2.5 M H₂SO₄ and 1.25 M H₂SO₄ solutions were utilized to simulate the acidic electrolyte system. The quantities of solutes were determined based on the water density at 298 K (0.997 g cm^–3). To achieve a 2.5 M solution system, only one salt molecule was required due to the cell’s volume. As for simulating 1.25 M solution systems, a larger supercell with twice the length along the basal plane was created. To shorten diffusion distance onto surface in simulation, the adsorbed molecules N₂ and/or NH₃ were placed near the Ti₂NT _x vacancy site on the surface, and then salt was placed around 5 Å away from the surface. Subsequently, solvent molecules (H₂O) were added using the amorphous packing tool in Materials Studio Software to pack the system.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.5c09104.

• Computational methods; proposed electrochemical reaction mechanism of Mars-van Krevelen mechanism; XANES and EXAFS spectra for reference materials; CV and LSV of MNene in 0.1 M Na₂SO₄; chronoamperometry curves; ammonia quantification; ammonia yield and Faradaic efficiency in various neutral electrolytes; NRR experiments using gas diffusion layer electrodes; postelectrolysis characterizations; ¹H NMR quantification; NRR using ¹⁵N₂ isotope; AIMD snapshots and free energy diagrams, and long-term studies using Ar and N₂ gases (PDF)

#.

D.K.Y. and H.E.L. contributed equally to this work.

The authors declare no competing financial interest.