Unveiling the thermodynamic-kinetic trade-off effect on acid sites in zeolite-catalyzed alcohol dehydration

Abstract

Understanding the distinct roles of Brønsted and Lewis acid sites remains a great challenge in designing zeolite catalysts, as their coexistence often obscures mechanistic understanding. Here, we combine solid-state NMR spectroscopy with density functional theory to elucidate the site-specific pathways of ethanol dehydration to ethylene over ZSM-5 zeolite. Two key intermediates are identified: chemisorbed ethanol on Lewis acid sites (LAS) and surface ethoxy species on Brønsted acid sites (BAS), both formed via -OH activation followed by β-H elimination to yield ethylene. Comparative analysis reveals a thermodynamic–kinetic trade-off between the two sites. LAS facilitates low-temperature -OH activation but exhibits high barriers for β-H elimination, limiting ethylene formation. In contrast, BAS requires higher activation energy for -OH activation but enables more facile β-H elimination, promoting ethylene production. This intrinsic trade-off, governed by the thermodynamics of -OH activation, provides a mechanistic basis for understanding and tuning alcohol dehydration on zeolite acid sites.

Clarifying how different acid sites function in zeolite catalysts is essential yet difficult due to their overlapping roles. This study uncovers distinct reaction pathways and a key trade-off that governs ethylene formation from ethanol.

Introduction

Alcohol molecules, including methanol and higher-chain alcohols, are important platform chemicals that can be produced in abundance from various feedstocks such as coal, natural gas, and biomass^1–5. The catalytic dehydration of these alcohols has attracted considerable interest due to its environmental benefits and potential economic advantages over traditional crude oil processes^6–11. Zeolites play a crucial role as catalysts in this transformation, characterized by their tunable active sites, well-defined molecular-sized channels, and excellent thermal stability (Fig. 1)^12–15. However, their high catalytic activity in alcohol conversion often poses challenges in achieving precise product selectivity and managing coking issues. To develop high-performance catalysts, it is essential to thoroughly understand the mechanisms governing alcohol conversion. However, this understanding is complicated by the inherent complexities of the active sites and the surface species involved in the reaction^16–19.

Fig. 1. Zeolite-catalyzed alcohol conversion.

Schematic illustrating alcohol conversion over zeolites to produce fundamental chemicals. The two-stage mechanisms (-OH activation and β-H elimination) on Brønsted and Lewis acid sites proceed in parallel, with a kinetic trade-off wherein each site preferentially facilitates a different step of the reaction.

Both Brønsted acid sites (BAS) and Lewis acid sites (LAS) serve as critical active centers in the dehydration of alcohols over zeolites^20,21. BAS arise from the isomorphous substitution of aluminum for silicon within the zeolite lattice, enabling alcohol dehydration via either concerted or stepwise mechanistic pathways³. In contrast to the well-defined nature of BAS, LAS exhibit greater structural and functional complexity²². The (hydro)thermal treatment of zeolites induces the disintegration of framework Al (tetra-coordinated) into coordinatively unsaturated tri-coordinated framework Al and extra-framework Al species, both of which act as LAS^23,24. Moreover, the spatial proximity of BAS and LAS within the zeolite channel can give rise to synergistic active sites that integrate the properties of both acid types^25–29. Consequently, ethanol dehydration is governed by the interplay of cooperative and competitive interactions between these distinct acidic sites.

The dehydration of alcohols over zeolite catalysts is a complex process comprising multiple sequential steps, including adsorption, activation, and subsequent reaction pathways, which collectively give rise to a diverse array of surface-bound species. In the context of alcohol dehydration catalyzed by BAS, the process is initiated by the activation of the alcohol’s hydroxyl (-OH) group, facilitating the formation of active intermediates such as alkyloxonium ions and surface alkoxy species^30–32. Specifically, in ethanol dehydration, the activation of the -OH group at BAS leads to the generation of surface ethoxy species (SES), which subsequently undergo β-H elimination to yield ethylene³³. While BAS are recognized as primary active sites for ethanol dehydration on zeolites, LAS can also contribute to the overall catalytic activity. Extra-framework LAS have been demonstrated to enhance the formation of SES on BAS at reduced temperatures, thereby promoting ethylene production³⁴. The electron-deficient character of LAS facilitates strong interactions with electron-rich molecules like water and alcohols^22,35–38. However, the detailed mechanisms underlying ethanol activation and conversion on LAS within zeolite remain insufficiently understood, hindering a comprehensive elucidation of the distinct roles played by different active sites in the ethanol dehydration process.

In this work, we elucidate the roles of distinct active sites in ethanol dehydration to ethylene over ZSM-5 zeolite. Three critical surface intermediates are involved in ethanol dehydration over zeolites: surface ethoxy species on Brønsted acid sites (SES-BAS), chemisorbed ethanol on Lewis acid sites (CSE-LAS), and triethyloxonium ions (TEO). Mechanistic analysis reveals that BAS and LAS catalyze ethanol dehydration through parallel two-stage pathways, with each site exhibiting a thermodynamic-kinetic trade-off that governs ethylene formation (Fig. 1). Specifically, LAS preferentially promotes the initial -OH activation at lower temperatures but imposes kinetic limitations on the subsequent β-H elimination step, whereas BAS favors β-H elimination at elevated temperatures. This mechanistic compensation dictates overall catalytic efficiency. Comparable kinetic behaviors are also observed during isopropanol dehydration on zeolites, supporting the generality of this mechanism.

Results

Effect of acid sites on catalytic activity

ZSM-5 zeolite samples (Si/Al = 11.5, Zeolyst) were calcined in static air at 650 °C, resulting in samples designated as ZSM-5-650 (refer to the methods section for detailed procedures). Structural and acidic property analyses reveal the presence of BAS and LAS at concentrations of 347 µmol/g and 85 µmol/g, respectively (as detailed in Supplementary Figs. S1–S4 and Supplementary Table S1). To investigate the origin of the LAS, ²⁷Al magic-angle spinning nuclear magnetic resonance (MAS NMR) spectroscopy was performed on NH₄⁺-exchanged ZSM-5-650 samples (Supplementary Fig. S5) under hydration condition. On ZSM‑5‑650, the 0 ppm signal assigned to octahedrally coordinated Al originates from water coordination to Lewis acidic Al sites. Upon NH₄⁺ exchange, these octahedral species reversibly transform into tetrahedrally coordinated Al, giving rise to a signal at ~ 55 ppm. This reversible coordination change demonstrates that the observed octahedrally coordinated Al species result from water adsorption on framework-associated Al rather than from extra-framework Al species²². The acidic characteristics of the zeolites were further probed using ³¹P NMR spectroscopy, which involved introducing trimethylphosphine oxide (TMPO) as a probe molecule (Fig. 2a). TMPO, a Lewis base, interacts with both BAS and LAS, and the resulting changes in its ³¹P NMR signal provide information about the nature and type of acidity present on the dehydrated zeolite. The observed signals at 76 ppm and 88 ppm were assigned to TMPOH⁺ ions formed on BAS and synergistic Brønsted/Lewis (B/L) acid sites³⁹, respectively. In addition, signals within the 63-69 ppm chemical shift range correspond to interactions between TMPO and framework-associated Al species possessing zero (FTA), one (FTA-OH), or two (FTA-2OH) hydroxyl groups⁴⁰. To further investigate the interactions between hydrogen and Al species within TMPO-adsorbed ZSM-5-650 zeolites, a two-dimensional (2D) ¹H-²⁷Al dipolar-mediated heteronuclear multiple quantum coherence (D-HMQC) NMR experiment was conducted. (Fig. 2b). This spectroscopic technique leverages dipolar coupling, a through-space interaction between nuclear magnetic moments, to reveal spatial proximities and correlations between ¹H and ²⁷Al nuclei. A prominent signal at (2.5, 51) ppm is attributed to Al-OH groups from tri-coordinated Al sites bound to TMPO, which consequently form tetrahedral Al configurations. Furthermore, a signal at (1.8, 51) ppm reflects the interaction between the methyl group of TMPO and these tetrahedral Al sites. Notably, the absence of correlation signals linking Al-OH/TMPO to higher-coordinated Al species (i.e., penta- or hexa-coordinated) confirms the lack of extra-framework Al in the zeolites. In addition, the absence of correlation signals between BAS and framework Al is attributed to the formation of delocalized TMPOH⁺ species. Therefore, these data demonstrate that the predominant acid sites in the ZSM-5-650 samples are BAS and framework-associated Al LAS. It should be noted that upon zeolite hydration, tri-coordinated Al sites exposed under dehydrated conditions would transform into higher-coordination states (tetra-, penta-, or hexa-coordinated Al), leading to a reduction in the number of LAS⁴⁰.

Fig. 2. Acidity characterization and catalytic performance of the catalysts.

a 1D ³¹P MAS NMR spectra of TMPO-loaded (P/Al = 0.5) dehydrated ZSM-5-650 zeolite, with corresponding TMPO adsorption models on BAS and framework Al LAS, along with ³¹P chemical shifts. b 2D ¹H-²⁷Al D-HMQC MAS NMR spectrum of the same sample. c Conversion of ethanol (black), ethylene yield (red), DEE yield (green) and C₃₊ hydrocarbons yield (blue) during ethanol dehydration over ZSM-5-650, ZSM-5-650-AHFS, and ZSM-5-650-Na zeolites. DEE represents diethyl ether; C₃₊ represents hydrocarbons containing 3 or more carbon atoms.

To further delineate the individual roles of BAS and LAS in ethanol dehydration, two additional ZSM-5 zeolite samples were synthesized: one predominantly containing BAS and another primarily comprising LAS (Supplementary Figs. S1–S5). The BAS-rich sample, designated ZSM-5-650-AHFS, was prepared by selectively removing framework Al LAS from ZSM-5-650 using ammonium hexafluorosilicate (AHFS). Conversely, the LAS-rich sample, denoted ZSM-5-650-Na, was obtained by ion-exchanging BAS in ZSM-5-650 with sodium ions (Na⁺) using sodium nitrate (NaNO₃). Ethanol dehydration activity was then evaluated over these zeolites in the temperature range of 200–275 °C (Fig. 2c). At 200 °C, ethanol dehydration over all tested samples primarily yields diethyl ether (DEE) and ethylene. For the BAS-rich ZSM-5-650-AHFS, ethanol conversion reaches 77.8%, with DEE and ethylene yields of 58.0% and 19.8%, respectively. In contrast, the LAS-rich ZSM-5-650-Na exhibits 47.4% ethanol conversion, yielding 43.9% DEE and 3.5% ethylene. The high DEE yields on both zeolites suggest that DEE formation proceeds with a relatively low activation barrier over both BAS and LAS. It is noteworthy that DEE serves as a reactive intermediate in the ethanol dehydration network over zeolites and can subsequently convert to ethylene (Supplementary Fig. S6). However, the markedly lower ethylene yield over LAS-rich zeolite indicates the limited reactivity toward alkene formation (Supplementary Fig. S7). Notably, ZSM-5-650, which contains both BAS and LAS sites, exhibits a significantly higher ethylene yield of ~ 67.3% within the first 2 h. This can be attributed to the cooperation of BAS and LAS in promoting the β-H elimination step, which is further supported by turnover frequency (TOF) analysis (Supplementary Tables S2 and S3).

As the reaction temperature was incrementally raised from 200 °C to 275 °C, all three zeolite samples—ZSM-5-650, ZSM-5-650-AHFS (BAS-rich), and ZSM-5-650-Na (LAS-rich)—show a marked increase in catalytic activity. However, a notable trend emerges with ZSM-5-650: the selectivity toward ethylene, the primary dehydration product, decreases as the temperature increases. This decline in ethylene selectivity is likely attributable to the initiation of secondary reactions, such as oligomerization of ethylene. To further explore how the BAS-to-LAS ratio influences ethanol dehydration, a series of dealuminated ZSM-5 catalysts were synthesized by calcining the parent ZSM-5 zeolite at temperatures ranging from 500 °C to 700 °C, yielding samples with varying BAS-to-LAS ratios (Supplementary Table S1). Catalytic testing of these samples reveals that ZSM-5-600 and ZSM-5-650, characterized by moderate BAS-to-LAS ratios, outperform their counterparts in terms of catalytic efficiency. These results indicate that an optimal equilibrium between BAS and LAS is essential for maximizing the efficiency of ethanol dehydration to ethylene, as illustrated by the performance trends (Supplementary Fig. S8), highlighting the critical role of acid site distribution in achieving high selectivity and conversion under these reaction conditions.

Characterization of surface species during ethanol dehydration over ZSM-5

Investigating the formation and evolution of surface species on zeolites provides crucial insights into the dynamic mechanisms governing ethanol transformation. To characterize these surface species during ethanol dehydration, ¹³C MAS NMR spectroscopy was employed. Specifically, CH₃¹³CH₂OH was reacted over ZSM-5-650, ZSM-5-650-AHFS, and ZSM-5-650-Na at 200 °C, with the resulting surface species analyzed via ¹³C MAS NMR (Fig. 3a). The spectroscopic data reveals a diverse range of adsorbed species on the zeolite surfaces, as evidenced by distinct ¹³C NMR signals spanning the chemical shift range of 60-90 ppm. Signals at 61 and 63 ppm are assigned to the methylene groups of physically adsorbed ethanol, reflecting its intact presence on the surface^9,41. A signal at 69 ppm is attributed to the methylene groups of DEE, a common intermediate to ethylene or byproduct in ethanol dehydration. Notably, signals at 72 ppm and 85 ppm are exclusively observed on ZSM-5-650 and the BAS-rich ZSM-5-650-AHFS. These signals were identified as surface ethoxy species bonded to BAS (SES-BAS) and triethyloxonium ions (TEO), respectively^30–32,42. Our previous experimental and theoretical studies have shown that the reaction pathway involving the TEO intermediate is more favorable than the conventional ethanol dehydration route on ZSM-5, offering a more plausible mechanism for ethylene formation at lower temperatures^32,41. In this pathway, TEO serves as the intermediate leading to the formation of SES-BAS, which then decomposes to ethylene.

Fig. 3. Surface species identification during zeolite-catalyzed ethanol dehydration.

a ¹³C MAS NMR spectra of trapped species resulting from the reaction of CH₃¹³CH₂OH over ZSM-5-650, ZSM-5-650-AHFS, and ZSM-5-650-Na zeolites at 200 °C for 1 min. b ¹³C-{²⁷Al} S-RESPDOR NMR spectra of the same trapped products, obtained with (S) and without (S₀) ¹³C-{²⁷Al} S-RESPDOR dipolar dephasing (recoupling time: 2 ms), with ΔS = S₀ - S. c ¹³C-{²⁷Al} S-RESPDOR build-up data (red dots) for the signal at 66 ppm, along with simulated curves; the brown, green, and dashed black lines correspond to internuclear ¹³C-²⁷Al distances of 2.63, 3.03, and 2.83 Å, respectively. The inset depicts the theoretically optimized local structure of ethanol chemically bound to FTA-2OH. d 2D ¹³C-{²⁷Al} PT-D-HMQC NMR spectrum of trapped products resulting from the reaction of CH₃¹³CH₂OH over ZSM-5-650 at 200 °C for 1 min (recoupling time: 2 ms). The carbon atoms of surface ethoxy species are highlighted in blue. The lower panels illustrate identified surface species on zeolites during ethanol dehydration: DEE (diethyl ether), SES-BAS (surface ethoxy species over Brønsted acid sites), CSE-LAS (chemisorbed ethanol over Lewis acid sites), and TEO (triethoxonium) ions.

Furthermore, a distinct surface species, characterized by a prominent ¹³C NMR signal at 66 ppm, is predominantly observed on the LAS-rich ZSM-5-650-Na. While this species is also detected on ZSM-5-650, it is only marginally observable on the BAS-rich ZSM-5-650-AHFS, indicating a strong correlation with the presence of LAS. Based on these findings, it is deduced that the formation of this specific surface species is primarily driven by framework LAS on zeolite. Remarkably, this species forms readily even at room temperature (Fig. 4), indicating that LAS can effectively activate ethanol. It is proposed to serve as a key intermediate in ethanol dehydration to ethylene over zeolites.

Fig. 4. Temperature-dependent evolution of surface species during ethanol dehydration.

a ¹³C CP MAS NMR spectra of trapped species obtained from CH₃¹³CH₂OH reaction over ZSM-5-650 zeolite at 25-350 °C for 1 min. b Temperature-dependent evolution of CSE-LAS, SES-BAS, TEO and DEE species, derived from the ¹³C MAS NMR spectra in (a). c Schematic illustration of parallel two-stage pathways for ethanol dehydration to ethylene on BAS and LAS. The “K” and “k” represent the reaction equilibrium constant and the reaction rate constant, respectively.

To gain deeper insights into the bonding characteristics of surface species on zeolites, the interactions between these species and framework active sites were further investigated. To this end, ¹³C-{²⁷Al} symmetry-based rotational-echo saturation-pulse double-resonance (S-RESPDOR) NMR experiments were conducted (Fig. 3b), enabling the assessment of spatial proximity and interactions between ¹³C nuclei in surface species and ²⁷Al nuclei in framework active sites^43,44. In this technique, when ²⁷Al nuclei are in close proximity to ¹³C nuclei, irradiation of the ²⁷Al nuclei induces significant dipolar dephasing effects on the ¹³C signals—quantified as the fractional signal reduction (ΔS/S₀)—leading to a pronounced decrease in signal intensity. Conversely, if the ¹³C and ²⁷Al nuclei are spatially distant, irradiation of ²⁷Al results in minimal changes to the ¹³C signals. As depicted in Fig. 3b, the ¹³C signals corresponding to physically adsorbed ethanol (at 61 and 63 ppm) and DEE (at 69 ppm) exhibit negligible alteration upon ²⁷Al irradiation, suggesting that these species are positioned away from the zeolite’s active sites. Similarly, the ¹³C signal of TEO at 85 ppm shows no discernible change with ²⁷Al irradiation, indicating a lack of close interaction between the cationic center of TEO and the anionic framework sites of the zeolite. In contrast, the ¹³C signal at 72 ppm, attributed to SES-BAS, displays a clear dipolar dephasing effect (ΔS/S₀ = 0.52), reflecting a strong spatial coupling with framework Al. Similarly, the ¹³C signal at 66 ppm, observed exclusively in ZSM-5-650 and LAS-rich ZSM-5-650-Na, exhibits substantial dipolar dephasing (ΔS/S₀ = 0.51), signifying a robust interaction with framework Al. Based on this evidence, the 66 ppm signal is assigned to surface ethoxy-like species linked to framework LAS, herein termed as chemisorbed ethanol on LAS (CSE-LAS). To aid in the interpretation of the 66 ppm ¹³C NMR signal, we performed DFT calculations on two types of surface species—chemisorbed ethanol (–OHCH₂CH₃) and directly bonded ethoxy groups (–O CH₂CH₃)—on three representative framework-associated LAS: FTA, FTA-OH, and FTA-2OH (Supplementary Fig. S9). The predicted ¹³C chemical shifts for the chemisorbed ethanol species (CSE-LAS) are 66 ppm (FTA), 68 ppm (FTA-OH), and 64 ppm (FTA-2OH), closely matching the experimental observation. In contrast, the calculated chemical shifts for the ethoxy species on the same LAS types are lower: 58 ppm (FTA), 61 ppm (FTA-OH), and 59 ppm (FTA-2OH). These results confirm that the 66 ppm NMR signal originates from CSE-LAS. The corresponding structure is illustrated in the lower panels of Fig. 3. In situ FTIR spectroscopy was further employed to monitor the formation of surface intermediates during ethanol dehydration over the zeolite catalysts (Supplementary Fig. S10). A distinct absorption band at 2912 cm⁻¹, attributed to chemisorbed ethanol, is observed exclusively in ZSM-5-650 and the LAS-rich ZSM-5-650-Na samples, but is absent in the BAS-rich ZSM-5-650-AHFS. This selective appearance of the 2912 cm⁻¹ band reinforces the conclusion that CSE-LAS forms specifically on LAS, consistent with our NMR results and DFT predictions. Structurally, in CSE-LAS, the oxygen atom adopts a tri-coordinated configuration, resembling the configuration in SES-BAS; however, unlike SES-BAS, where the oxygen is bonded to a silicon atom, in CSE-LAS, it is coordinated to a hydrogen atom, implying distinct reactivity in the ethanol dehydration process.

To further probe the configuration of CSE-LAS species, we experimentally determined the interatomic distance between the carbon atom of CSE-LAS (methylene group) and the Al atom associated with framework LAS. This was achieved by measuring the dipolar dephasing effect (ΔS/S₀) as a function of recoupling time in ¹³C-{²⁷Al} S-RESPDOR NMR experiments (Fig. 3c). By fitting the resulting S-RESPDOR build-up curve, a distance of 2.83 Å is derived, which closely matches the theoretically calculated distance of 2.80 Å, providing structural insight into the surface-bound CSE-LAS on zeolite. Furthermore, to elucidate the coordination environment of Al species during ethanol dehydration, 2D ¹³C-{²⁷Al} population transfer dipolar-mediated heteronuclear multiple quantum coherence (PT-D-HMQC) NMR experiments were performed (Fig. 3d). The PT-D-HMQC results indicate that SES-BAS (72 ppm) interacts with tetra-coordinated Al at 48 ppm. This Al is part of the framework and is responsible for Brønsted acidity in the zeolite. In contrast, the CSE-LAS (66 ppm) correlates with a distinct tetra-coordinated Al species at 53 ppm. This observation indicates that the tri-coordinated framework Al LAS transitions to a tetra-coordinated state upon ethanol adsorption, forming a coordination bond during dehydration. This process mirrors the adsorption behavior of TMPO on LAS (Fig. 2a), where the oxygen atom’s lone pair electrons coordinate with the electron-deficient Al atom of LAS, exemplifying a classic Lewis acid-base interaction.

Understanding the Reactivity of Surface Species in Ethanol Dehydration over ZSM-5

To elucidate the reactivity of surface species during ethanol dehydration over ZSM-5-650, their temperature-dependent (25 °C ~ 350 °C) behavior was monitored using variable-temperature ¹³C MAS NMR spectroscopy (Fig. 4a). Ethanol predominantly binds to the acid sites of zeolites, activating the -OH group to generate surface species like CSE-LAS, SES-BAS, TEO, and DEE, which then convert into ethylene and potentially secondary products at elevated temperatures. The normalized ¹³C NMR signals, plotted against reaction temperature (Fig. 4b), demonstrate the formation and subsequent transformation of CSE-LAS, SES-BAS, TEO, and adsorbed DEE on the zeolite surface, providing insights into the mechanistic steps of ethanol dehydration. This process unfolds in two primary steps: (1) the initial formation of surface intermediates through the activation of ethanol’s -OH group, and (2) the transformation stage, involving the β-H elimination, which leads to ethylene. Energetically, the activation of the -OH group is more facile than the subsequent β-H elimination, resulting in a temperature-driven accumulation of surface intermediates that peaks as the reaction progresses. Notably, CSE-LAS forms at temperatures as low as below 25 °C (Fig. 4a and Supplementary Fig. S11), in stark contrast to SES-BAS and TEO, which require temperatures exceeding 100 °C to emerge. This low-temperature formation of CSE-LAS underscores the exceptional catalytic activity of framework LAS in facilitating ethanol activation and generating chemisorbed ethanol species, compared to the higher thermal threshold demanded by BAS-driven intermediates (SES-BAS, TEO and DEE). DEE appears at a lower temperature than SES-BAS and TEO, suggesting that it serves as a precursor for the subsequent formation of these intermediates, in agreement with our previous theoretical results^32,41. Further ¹³C MAS NMR analysis of the surface species formed after reaction with DEE over ZSM-5-650 shows that DEE is converted to SES-BAS more readily than ethanol, confirming its higher reactivity and its role as a key intermediate in ethanol dehydration (Supplementary Fig. S12). Following their maximum accumulation (Fig. 4b), these surface species undergo decay at elevated temperatures due to their transformation into ethylene. The rate of this decay serves as an indicator of the intermediates’ reactivity in the β-H elimination step. Analysis of the decay rate reveals a reactivity order of TEO > SES-BAS > CSE-LAS > DEE (Fig. 4b), directly reflecting a corresponding order of intrinsic reactivity for these intermediates in the ethanol dehydration reaction.

These observations indicate a mechanistic trade-off in the two-stage ethanol dehydration process, which is governed by the distinct characteristics of the active sites. Specifically, LAS facilitates the activation of ethanol’s -OH group with remarkable efficiency, initiating the formation of CSE-LAS at room temperature. However, the subsequent transformation of CSE-LAS into ethylene, requiring β-H elimination, demands significantly higher temperatures, resulting in a broad evolution curve that reflects a slower and more thermally intensive decay process. Conversely, BAS requires elevated temperatures to activate ethanol’s -OH group and form SES-BAS, yet the transformation of SES-BAS proceeds more readily once formed, exhibiting a narrower evolution curve indicative of a faster and energetically more favorable β-H elimination step. This stark contrast illustrates two parallel reaction pathways on BAS and LAS, each involving distinct energetics in the two-stage mechanism (Fig. 4c): LAS exhibits superior activity in low-temperature ethanol activation but are less efficient in subsequent transformation steps, whereas BAS, while displaying lower initial activity, promotes a more rapid product formation. This dichotomy underscores the distinct catalytic strengths and inherent limitations of each acid site type within the ethanol dehydration pathway.

To further validate the applicability of the proposed mechanistic framework, we extended our investigation to isopropanol dehydration over zeolitic catalysts. As shown in Supplementary Fig. S13, isopropanol is more effectively converted to propene on ZSM-5-650 and BAS-rich ZSM-5-650-AHFS, whereas significantly lower activity is observed on LAS-rich ZSM-5-650-Na, mirroring the reactivity trends observed for ethanol. To probe the underlying surface chemistry, temperature-dependent solid-state¹³C NMR experiments were conducted over ZSM-5-650 in the range of 25–250 °C (Supplementary Fig. S14). At room temperature, chemisorbed isopropanol species are clearly observed on both LAS (76 ppm, denoted as CSI-LAS) and BAS (68 ppm), and their assignments are further supported by ¹³C–²⁷Al S-RESPDOR NMR analysis (Supplementary Fig. S15). The presence of CSI-LAS at low temperature indicates that –OH activation readily occurs on LAS, forming stable chemisorbed species. Upon increasing the reaction temperature above 100 °C, the CSI-LAS signal diminishes and eventually disappears, suggesting its conversion via β-H elimination to propene. Concurrently, significant amounts of adsorbed hydrocarbon species, including alkenes and aromatics, are detected.

Despite this, the enhanced propene formation observed on BAS-rich catalysts (Supplementary Fig. S13) indicates that CSI-LAS is unlikely to be the dominant pathway for propene production. While isopropoxy species on BAS (typically at ca. 80 ppm) are widely considered as the key intermediates for propene formation, they are not detected at low temperatures and remained undetectable throughout the reaction process. This absence is likely due to their high reactivity and rapid turnover at elevated temperatures once formed. Collectively, these findings demonstrate that isopropanol dehydration exhibits behavior analogous to ethanol dehydration, reinforcing the existence of a trade-off between –OH activation and β-H elimination on both BAS and LAS. These results support the general validity of the proposed mechanism for alcohol dehydration over zeolites.

Reaction mechanism of ethanol dehydration reaction

To gain mechanistic insights into the ethanol dehydration reaction and the reactivity of different active sites, density functional theory (DFT) calculations were performed (Fig. 5 and Supplementary Figs. S16–18). The dehydration process was computationally modeled over the framework LAS, including FTA, and its hydroxylated forms (FTA-OH, FTA-2OH). For comparative analysis, ethanol dehydration over BAS was also computationally evaluated. We focused on the TEO-mediated pathway, which has been identified as the most energetically favorable route compared to both unimolecular and bimolecular dehydration mechanisms^32,41.

Fig. 5. Calculated reaction energy profiles for ethanol dehydration on ZSM-5.

Ethanol dehydration pathways over FTA, FTA-OH, FTA-2OH, and BAS. Upper panels: key reaction intermediates and transition states (TS) for SES-BAS/CSE-LAS decomposition to ethylene. Dashed line: simplified SES-BAS formation (see Supplementary Fig. S17 for details). Energies in kcal/mol.

Ethanol dehydration over both BAS and LAS involves a two-stage process: (1) activation of the ethanol -OH group, and (2) β-H elimination leading to ethylene formation. For framework LAS (Fig. 5), the initial step involves direct chemisorption of ethanol, forming CSE-LAS, accompanied by substantial exothermic energy release: − 53.5 kcal/mol for FTA, − 35.4 kcal/mol for FTA-OH, and − 33.4 kcal/mol for FTA-2OH. This indicates a strong stabilization of CSE-LAS, with adsorption strength decreasing as the number of hydroxyl groups on the LAS increases. However, the subsequent β-H elimination from CSE-LAS is endothermic, revealing a critical energetic barrier. A clear inverse relationship emerges between the stability of CSE-LAS (reflected by the exothermic heat of adsorption) and the facility of ethylene formation: greater adsorption strength correlates with a higher energy barrier for decomposition. Specifically, the activation energy for β-H elimination rises in the sequence FTA-2OH (38.8 kcal/mol) ≈ FTA-OH (40.5 kcal/mol) < FTA (51.5 kcal/mol), illustrating a trade-off wherein stronger initial binding to LAS hinders the kinetics of the transformation step.

Unlike the CSE-LAS pathway, ethanol activation on BAS proceeds via SES-BAS formation, with TEO identified as a key intermediate, as supported by DFT calculations (Fig. 5 and Supplementary Fig. S17). This process entails an activation energy of 25.6 kcal/mol (denoted TS3 in Supplementary Fig. S17) for SES-BAS formation, rendering it energetically more demanding than the formation of CSE-LAS. The SES-BAS formation pathway is endothermic, largely due to the transformation of TEO to SES-BAS, and thus becomes more favorable at elevated temperatures, aligning with temperature-dependent experimental observations (Fig. 4). For the subsequent decomposition of SES-BAS into ethylene, the activation energy is calculated at 30.5 kcal/mol (TS in Fig. 5), a value notably lower than those for all framework LAS systems (FTA: 51.5 kcal/mol, FTA-OH: 40.5 kcal/mol, FTA-2OH: 38.8 kcal/mol). To clarify DEE’s role in ethanol dehydration, DFT calculations were also conducted for its formation and conversion over BAS and LAS. On BAS, ethanol and DEE readily interconvert with similar barriers (24.8 vs. 29.2 kcal/mol), and DEE can form TEO (23.5 kcal/mol), which decomposes to SES-BAS and ultimately ethylene (Supplementary Fig. S17). On LAS (FTA, FTA-OH, FTA-2OH; Supplementary Fig. S19), ethanol–DEE equilibrium is also favorable, with low forward/reverse barriers (26.5–28.2 vs. 16.4–18.2 kcal/mol). However, direct DEE-to-ethylene conversion is kinetically hindered (40.8–67.1 kcal/mol). These results indicate that while DEE can reversibly form from ethanol over both BAS and LAS, its further conversion to ethylene is kinetically feasible only via the BAS-mediated TEO route. On LAS, DEE preferentially reverts to ethanol, which then proceeds through the CSE-LAS pathway toward ethylene formation, as illustrated in Fig. 5.

We further conducted kinetic studies of ethanol dehydration using BAS- and LAS-enriched samples, as shown in Supplementary Fig. S20. To ensure intrinsic kinetic behavior, apparent activation energies were determined from Arrhenius plots obtained under low-conversion conditions (below 15%), thereby minimizing the influence of heat and mass transfer limitations. The experimentally determined activation energies are 31.7 kcal/mol for the BAS-rich ZSM-5-650-AHFS sample and 38.5 kcal/mol for the LAS-rich ZSM-5-650-Na sample. These values are in close agreement with the DFT-predicted energy barriers: 30.5 kcal/mol for the decomposition of SES-BAS and 38.8 kcal/mol for the conversion of CSE-LAS (Fig. 5). This strong correlation between experimental kinetics and theoretical calculations provides compelling evidence that the SES-BAS-mediated pathway offers a more favorable route for ethylene formation, highlighting a kinetic advantage over the CSE-LAS mechanism.

These differences in activation energies reflect how BAS and LAS distinctly influence each reaction step. Consequently, the site-specific trade-off observed in the two-stage ethanol dehydration process stems from the intrinsic acid characteristics of BAS and LAS, which control the thermodynamics and kinetics of intermediate formation and transformation. On LAS, the covalent bond strength between SES and the zeolite framework scales directly with LAS acidity, following the order FTA > FTA-OH ≈ FTA-2OH, as evidenced by their respective adsorption heats (− 53.5, − 35.4, and − 33.4 kcal/mol). This stronger bonding, while stabilizing CSE-LAS, elevates the activation energy for β-H elimination due to the increased stability of the carbenium ion-like transition state (Fig. 5 and Supplementary Fig. S18), impeding ethylene formation. Conversely, SES-BAS formation on BAS requires an endothermic -OH elimination step, which is less favorable at low temperatures but becomes viable at higher temperatures due to the decomposition of TEO. Despite this higher energy requirement for formation, SES-BAS exhibits a pronounced carbenium ion character that enhances the ease of β-H elimination, lowering the energy barrier for decomposition relative to CSE-LAS. This disparity in carbenium ion character elucidates why SES-BAS, though more challenging to form, undergoes kinetically more favorable decomposition. Thus, the mechanistic trade-off observed on both BAS and LAS during two-stage ethanol dehydration stems from the balance between adsorption strength, carbenium ion stabilization, and the activation energy required for β-H elimination at each site. In light of the entropy changes associated with adsorption and desorption during ethanol dehydration, Gibbs free energy calculations were performed for the reaction pathways (Supplementary Figs. S21, S22). At 200 °C, SES-BAS exhibits higher Gibbs free energy than CSE–LAS, while the activation Gibbs free energy for the subsequent β-H elimination remains lower for SES-BAS than for CSE-LAS. These findings confirm that the thermodynamic–kinetic trade-off persists under reaction conditions. Given that water is a byproduct of ethanol dehydration, we further assessed its influence on the reaction mechanism. As shown in Supplementary Figs. S23–S25, the inclusion of a water molecule markedly stabilizes all key adsorbed species, intermediates, and transition states compared with the anhydrous models (Fig. 5 and Supplementary Fig. S17). Despite this enhanced stabilization, the thermodynamic-kinetic compensation persists, with the stability order (CSE-FTA-2OH > CSE-FTA > CSE-FTA-OH > SES-BAS) remaining inversely correlated with the β-H elimination barriers (60.4, 51.4, 44.5, and 26.8 kcal/mol, respectively). These results confirm that the mechanistic insights derived under anhydrous conditions remain robust upon inclusion of water.

Discussion

We have revealed a thermodynamic-kinetic trade-off effect occurring in parallel pathways on BAS and LAS during the two-stage ethanol dehydration process. For LAS, ethanol activation via the –OH group occurs readily at relatively low temperatures, resulting in the formation of chemisorbed ethanol species. However, the subsequent β-H elimination step on LAS requires overcoming a comparatively high activation barrier, which limits the overall ethylene production efficiency on these sites. In contrast, BAS require higher temperatures to activate ethanol and generate surface ethoxy species in the initial step, but the following β-H elimination proceeds with a much lower activation energy, thus facilitating ethylene formation more efficiently.

This difference stems from the distinct thermodynamic and kinetic properties governing the –OH activation step on each site. LAS favors exothermic adsorption and stabilizes intermediates at lower temperatures, while BAS rely on endothermic intermediates such as TEO, which become more reactive at elevated temperatures. These contrasting characteristics result in a mechanistic balance, or trade-off, on each site type that governs the rate and efficiency of ethanol dehydration. Furthermore, the similar kinetic patterns observed during isopropanol dehydration to propene over zeolites reinforce the general applicability of this site-specific trade-off mechanism across different alcohol dehydration reactions catalyzed by zeolites. These findings provide molecular-level insight into the distinct functions of LAS and BAS in alcohol conversion over zeolites.

Methods

Sample preparation and characterization

The preparation of calcined ZSM-5 zeolites was carried out as follows: ZSM-5 zeolites (with a Si/Al ratio of 11.5, obtained from Zeolyst) were arranged in a shallow bed configuration on a quartz crucible and subsequently calcined at 500–700 °C for 6 h in air. The heating temperature gradually elevated at a rate of 1 °C/min raised from room temperature to a target temperature. The calcined samples are denoted as ZSM-5-500, ZSM-5-600, ZSM-5-650 and ZSM-5-700, respectively. The ZSM-5-650 sample was washed with ammonium hexafluorosilicate (AHFS) by the following procedure (denoted as ZSM-5-650-AHFS): Catalysts were suspended in AHFS solution at a concentration of 1 g/10 ml with an AHFS: Al ratio of 2. The resultant mixture was heated to 90 °C under intense stirring. After 6 h stirring, the final mixture was filtered and subjected to three successive washings with deionized water. Subsequently, the samples were dried in air at 100 °C overnight and then calcined at 550 °C for 6 h in air with a rate of 2 °C/min raised from room temperature to a target temperature to remove NH₄⁺. Furthermore, the calcined catalyst ZSM-5-650 was also converted to the sodium form (denoted as ZSM-5-650-Na) and ammonium form (denoted as ZSM-5-650-NH₄) through the following procedure: The catalysts were suspended in 1.0 M solution of sodium nitrate or ammonium nitrate respectively, with a concentration of 1 g/10 ml. The mixture was heated to 90 °C under intense stirring for 2 h. The final mixture was filtered and then subjected to three successive washings with deionized water, then, the samples were dried in air at 100 °C for 2 h. This procedure was iterated three times to obtain zeolite in sodium form and ammonium form.

TMPO adsorption was performed on pre-dehydrated samples. Prior to the adsorption of TMPO, the sample was dehydrated on a vacuum line. The temperature gradually increased at a rate of 2 °C min⁻¹ and the sample was kept at a final temperature of 673 K at a pressure below 10⁻³Pa overnight. Typically, solid TMPO was dissolved in CD₂Cl₂ solution, and a certain amount of TMPO solution was added into the sample in the glass sample tube in a N₂ glovebox, followed by removal of the CD₂Cl₂ solvent by evacuation on a vacuum line at 80 °C for 1 h. To ensure uniform adsorption of TMPO molecules in the channels of the zeolite, the sealed sample was further heated at 180 °C for 2 h. Finally, the sample was transferred to the ZrO₂ MAS rotor in the N₂ glovebox prior to the solid-state NMR experiments.

The structure and crystalline nature of H-ZSM-5 zeolites were investigated by X-ray diffractometer (X’Pert³ Powder) using CuKα radiation with a step of 0.02° at a respective voltage of 40 kV and a current of 40 mA. The scanning range was from 5° to 40°. The acid concentration of zeolites was determined by FT-IR of pyridine adsorption experiments. The measurements were conducted on a Bruker Tensor 27 spectrometer. Prior to the measurements, catalysts were pre-activated at 350 °C under high vacuum (< 10⁻⁵ Pa) for 6 h. After cooling down to 25 °C, a background spectrum was collected. Subsequently, an excessive amount of pyridine was then introduced to the infrared cell and held for 10 min to allow equilibrium. The residual pyridine was removed by vacuum at 200 °C for 2 h. The FT-IR spectra of pyridine-adsorbed samples were measured after cooling down to 25 °C.

All the catalytic reactions were conducted in a fixed bed reactor with an inner diameter of 6 mm⁴⁵. The ZSM-5 powder was compacted into pellets between 40-60 mesh. The pellets (0.1 g) were activated at 400 °C in flowing helium for 1 h prior to the reaction. Then both ethanol and isopropanol were reacted upon the zeolite for 1 min at a preset temperature, in which the total gas flow through the reactor was 100 sccm. Then the reaction was thermally quenched by pulsing liquid nitrogen onto the catalyst bed, and the temperature was cooled down to ambient temperature within a very short period (0.2 s). After the reaction was quenched, the reactor containing the catalyst was sealed by means of valves. The sealed reactor was then transferred to a glove box filled with pure N₂ and the catalyst was packed into an NMR rotor for NMR measurements. For the kinetic measurements, a feed containing 12 mbar ethanol was passed over 25 mg of catalyst at a total gas flow rate of 100 mL/min. The reaction temperature was varied between 160 and 225 °C, with the conversion rate kept below 15%. All the effluent was analyzed by online GC-FID chromatograph (Shimadzu GCMS-QP2010) equipped with a Supelco Supel-QTM PLOT capillary column (30 m × 0.32 mm × 15 μm). The temperature programming started at 40 °C (maintained for 1 min), followed by a rate of 5 °C min⁻¹ to a temperature of 140 °C (maintained for 2 min) and a rate of 10 °C min⁻¹ to a final temperature of 250 °C (maintained for 10 min).

Solid-state NMR experiments

All the 1D ¹H, ¹³C, ²⁷Al and ³¹P solid-state NMR experiments were performed at 11.7 T on a Bruker-Avance III-500 spectrometer, equipped with a 4 mm probe, with resonance frequencies of 500.57, 125.87, 130.44 and 202.65 MHz for ¹H, ¹³C, ²⁷Al and ³¹P, respectively. For the entirety of the experiments, a magic angle spinning rate of 10 kHz was employed. For the ¹H MAS NMR experiments, single-pulse excitation was performed using a ¹H π/2 pulse length of 4.0 μs and a repetition time of 2 s. The ¹H chemical shifts were referenced to adamantane (1.78 ppm). For the ¹H → ¹³C CP/MAS NMR experiments, the Hartmann-Hahn condition was achieved using hexamethylbenzene (HMB), with a contact time of 3.5 ms and a repetition time of 2 s. The ¹³C chemical shifts were referenced to HMB (a second reference to TMS). The ²⁷Al MAS NMR spectra were acquired using a small-flip angle technique with a pulse length of 0.3 μs (π/18 flip-angle (calculated from the Al(NO₃)₃ solution)) and a recycle delay of 1 s. The ²⁷Al chemical shifts were referenced to 1 M Al(NO₃)₃ aqueous solution (0 ppm). The ³¹P MAS NMR spectra were recorded with a π/2 pulse length of 3.7 μs, a recycle delay of 30 s and a ¹H decoupling RF strength of 60 kHz. The ³¹P chemical shifts were referenced to ammonium dihydrogen phosphate (0.81 ppm). The ²⁹Si solid-state NMR experiments were carried out at 9.4 T on a Bruker Avance III-400 spectrometer equipped with a 4 mm probe. The ²⁹Si MAS NMR spectra were acquired with a ¹H decoupling RF strength of 54 kHz, a recycle delay of 100 s and a spinning rate of 10 kHz. The ²⁹Si chemical shifts were referenced to kaolinite (− 91.5 ppm).

The 2D ¹H-²⁷Al D-HMQC experiment was performed at 18.8 T on a Bruker-Avance III-800 spectrometer, for which ¹H and ²⁷Al Larmor frequencies are equal to 800.36 and 208.56 MHz, respectively, using a 3.2 mm probe under 20 kHz MAS speed. The recoupling pulses were applied on the proton channel with the SR4 sequence⁴⁶, with the recoupling time of 3.0 ms. And the pulses applied on the Al channel were central-transition-selective pulses with low RF power lasting 12.8 µs for π/2 pulses. The spectra were acquired by averaging 32 scans for each of the 12.5 μs t1 increments with a recovery delay of 1 s. Presaturation has been applied with a train of 10 π/2 ¹H pulses 10 ms apart to spoil residual longitude magnetizations at the beginning of each scan.

The ¹³C-{²⁷Al} symmetry-based rotational-echo saturation-pulse double-resonance (S-RESPDOR) experiments were also performed at 11.7 T on a Bruker-Avance III-500 spectrometer with a 4 mm probe under 10 kHz MAS speed. ¹H-¹³C cross polarization (CP) with a contact time of 5.0 ms was employed to prepare the initial ¹³C signal. The recycle delay was set to 2 s. SR4 dipolar recoupling was used on the ¹³C channel with ν_nut,13C = 20 kHz. Continuous-wave ¹H decoupling with an amplitude of 71.4 kHz was used during the dipolar recoupling, while a SPINAL-64 (small-phase incremental alternation with 64 steps) ¹H decoupling with an amplitude of 75.8 kHz was used during acquisition. A π pulse length of 8 μs was used on the ¹³C channel. The saturation pulse on the ²⁷Al channel with an amplitude of 53.8 kHz and a length of 100 μs = TR was irradiated at the 54 ppm ²⁷Al signal.

The simulations for measuring the distance of ¹³C and ²⁷Al spins were performed using the SIMPSON program⁴⁷. An S-RESPDOR pulse sequence with an SR4 recoupling strategy was simulated to examine the evolution of RESPDOR curves versus ¹³C-²⁷Al distances. To saturate ²⁷Al spins, a pulse with a power level of 53.8 kHz was employed on the ²⁷Al channel. Other parameters such as magnetic field (11.7 T), spinning speed (10 kHz), and saturation pulse length (100 μs) were kept the same as those in ¹³C-{²⁷Al} S-RESPDOR experiments. The RESPDOR curves for different ¹³C-²⁷Al dipolar interactions (distances) were obtained by varying the duration of the recoupling modules. In addition, a uniform powder sample (320 α, β crystallites selected using the REPULSION method⁴⁸ and 9 γ angles) was considered for the simulation.

The 2D ¹³C-{²⁷Al} PT-D-HMQC experiments were also performed at 18.8 T on a Bruker-Avance III-800 spectrometer, for which ¹H, ¹³C and ²⁷Al Larmor frequencies are equal to 800.36, 201.26 and 208.56 MHz, respectively, using a 3.2 mm probe under 10 kHz MAS speed. For cross polarization from ¹H to ¹³C, spectra were acquired using a ¹H pulse width of 5.6 μs, and a contact time of 3000 μs. Repetitive sideband-selective (SS) WURST-80 adiabatic pulses with a length of 98 µs were applied on Al channel during SR4 recoupling to accelerate coherence transfers between ¹³C and ²⁷Al. The frequency offset of the WURST pulses was alternated between 148 kHz. The spectra were acquired by averaging 3600 scans for each of the 33.3 μs rotor-synchronized t₁ increments with a recovery delay of 1 s.

Computational method

The framework tri-coordinated Al Lewis acid sites, without a hydroxyl group (FTA) (Supplementary Fig. S16a), with one hydroxyl group (FTA-OH) (Supplementary Fig. S16b), and with two hydroxyl groups (FTA-2OH) (Supplementary Fig. 16c), are represented using 72 T models. The Brønsted acid site (Supplementary Fig. S16d) of ZSM-5 zeolites is represented by 88 T models. Both sets of models were derived from crystallographic structural data obtained from the IZA Structure Database (http://www.iza-structure.org/databases). All the zeolite models incorporate the full double 10-MR intersection pores characteristic of ZSM-5 zeolite. Both the BAS and LAS were modeled on symmetrically equivalent T12 crystallographic sites within the MFI framework. While the tri-coordinated framework Al species can, in principle, be formed via direct cleavage of Al–O–Si bonds during dealumination, such processes typically generate adjacent silanol groups. These silanol groups can introduce local structural distortions, thereby compromising the stability and accessibility of the resulting Lewis acid sites. To avoid these limitations and in line with previous reports^49,50, we constructed a well-defined tri-coordinated framework Al LAS through controlled Si defect engineering. Specifically, the FTA site (Al(OSi)₃) was modeled by introducing a single Si vacancy adjacent to a tetra-coordinated Al center. The FTA-OH site (HO–Al(OSi)₂) was generated by removing two neighboring Si atoms, and the FTA-2OH site ((HO)₂–Al(OSi)) was formed by introducing three Si vacancies. This approach enables the simulation of stable, isolated LAS with varying degrees of hydroxylation while preserving a realistic framework environment. The BAS was modeled using the Si₁₂–O₂₄(H)–Al₁₂ unit at the T12 position, located at the channel intersection. This configuration was chosen based on its high accessibility to reactants and its demonstrated structural and catalytic relevance in previous reports^51–53. The terminal Si-H bond length was set at 1.47 Å, with the bond oriented in the same direction as the corresponding Si-O bond.

In this work, the active site atoms represented as ball and stick view and the adsorbed hydrocarbon complex were treated as the high-level layer (Supplementary Fig. S9), while the rest of the frameworks were treated as the low-level layer. To maintain the structural integrity of the modeled ZSM-5 zeolite, partial structure optimizations were conducted by relaxing the atoms within the high-level layer, while the remaining atoms were kept fixed at their crystallographic positions. All the transition state (TS) structures were identified using the QST3 method with the Gaussian program. Then the IRC (Intrinsic Reaction Coordinate) method was used to determine the structures of the corresponding reactant and product. Based on the imaginary vibrational frequencies of the optimized TS, we slightly adjusted the positions of the vibrating atoms along the calculated reaction coordinate, in both directions towards the reactant and the product. Finally, we optimized the resulting structures to achieve their minimum energy configurations.

A combined theoretical approach, specifically ONIOM (ωB97XD/6-31 G(d,p):AM1), was utilized for the geometry optimization of adsorption states and transition states (TS). The ωB97XD method, which accounts for both short- and long-range dispersion effects, has been proven effective and has been extensively applied to investigate the reaction mechanisms occurring within zeolites^54,55. Since the AM1 method is known to underestimate the low-level interaction energies, all the energies reported in this work were calculated at the ωB97XD/6-31 G(d,p) level of theory, using the structures optimized previously.

Harmonic frequency calculations employing partial Hessian vibrational analysis (PHVA), including the high-layer active acid site and organic species, were performed to check whether the stationary points found exhibit the proper number of imaginary frequencies. The intrinsic energy barriers, defined as the zero-point corrected electronic energy difference between the transition state and the reactant complex for each elementary step. The corresponding Gibbs free energies at 473 K were then computed within the rigid-rotor harmonic oscillator approximation, incorporating vibrational and thermal corrections obtained from harmonic frequency analysis.

Author contributions

M.H., W.J.C., and C.W. conducted the catalysts preparation, characterization, and catalytic testing; Y.Y.C. performed the DFT calculations; M.H. and Q.W. carried out the solid-state NMR experiments; M.H., Y.Y.C., C.W., J.X., and F.D wrote the manuscript; and C.W. and J.X. designed and directed the project. All authors discussed the results and commented on the manuscript.

Peer review

Peer review information

Nature Communications thanks Kim Larmier, and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data supporting the findings of this study are available within the article and Supplementary Information files. All data are available from the corresponding author upon request. Source data are provided in this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-70418-y.