Identification of active catalysts for the acceptorless dehydrogenation of alcohols to carbonyls

The study addresses the challenge of efficiently converting alcohols to carbonyl compounds without external oxidants, producing H2 as a co-product. Traditional aerobic oxidation with O2 or H2O2 is effective but risks over-oxidation and presents challenges for green processing. Acceptorless dehydrogenation is less thermodynamically favored but avoids over-oxidation and enables hydrogen generation from biomass-derived alcohols. Homogeneous catalysts (notably Ru, Ir, Os complexes) are effective but suffer from recycling and additive issues, limiting industrial viability. Heterogeneous catalysts based on transition metals (Ru, Pt, Pd, Re, Ag, Au, Fe, Mn, Cu, Co, Ni) have shown promise. However, a combined, experimentally validated, descriptor-based DFT micro-kinetic framework for rapid in silico screening of metals for alcohol dehydrogenation has been lacking. The paper aims to build and validate an activity–descriptor relationship using DFT and micro-kinetics, benchmarked against 2-octanol dehydrogenation, to guide discovery of active, potentially non-precious catalysts.

Prior work has demonstrated numerous homogeneous acceptorless alcohol dehydrogenation catalysts based on Ru, Ir, and Os complexes, with extensive reviews covering mechanisms and applications in hydrogen storage and synthesis. Heterogeneous catalysis has identified several transition metals and alloys (Ru, Pt, Pd, Re, Ag, Au, Fe, Mn, Cu, Co, Ni) as active for acceptorless dehydrogenation, with studies noting support and particle size effects. Computationally, DFT investigations have detailed adsorption and dehydrogenation pathways of methanol and ethanol on Pt, Pd, Ir, Rh, Ni, and Cu surfaces, and BEP correlations have been used to predict activity trends for methanol, ethanol, and polyalcohols. Descriptor-based kinetic modeling (e.g., CatMAP) has previously been applied to methanol dehydrogenation on stepped surfaces, but a combined experiment–theory, fast screening framework tailored to aliphatic alcohols leading to carbonyls and H2 had not been experimentally validated before this work.

Computational: Plane-wave DFT (VASP) with PAW pseudopotentials and GGA-PBE exchange-correlation was used. An energy cutoff of 400 eV and Methfessel–Paxton smearing with σ = 0.2 eV ensured energy convergence (<1 meV/atom). Structures were optimized to forces <0.02 eV/Å and energy change <1e-6 eV, with a 12 Å vacuum to decouple slabs. Dispersion interactions were included via the density-dependent DDsC correction. Close-packed surfaces modeled: (111)-p(3×3)-4L for Co, Rh, Ir, Ni, Pd, Pt; (0001)-p(3×3)-4L for Ru, Os, Re. Spin polarization was used for Ni and Co. Brillouin zone sampling employed a (5×5) k-point mesh. Transition states were located by NEB and verified by a single imaginary frequency in the stretching analysis. Mechanistic DFT explored methanol dehydrogenation to formaldehyde and H2 via two pathways (alkoxy and hydroxyalkyl) across nine metals, with detailed energetics and structures. Scaling and BEP relations for intermediates and transition states were established against two descriptors: adsorption energies of atomic C* and O* (referenced to CH3OH, H2O, and H2). Mean absolute errors of the linear relations were <0.2 eV. Micro-kinetics: Descriptor-based kinetic modeling used CatMAP to compute turnover frequencies (TOFs) as functions of C* and O* adsorption energies, at 453 K. For methanol, conditions corresponded to 0.1% conversion at 101 kPa CH3OH (PCH3OH ≈ 99.9 kPa; PCH2O·PH2 ≈ 0.1 kPa). To compare with experiments on 2-octanol, methanol-derived DFT data were corrected using gas-phase equilibrium constants (Kp, 453 K): CH3OH→CH2O+H2: 8.40×10^-5 bar; 2-octanol→2-octanone+H2: 7.26×10^-1 bar. Alcohol adsorption and ketone desorption equilibrium constants were scaled by factor (7.26×10^-1 / 8.40×10^-5)^0.5 to generate a 2-octanol TOF volcano map at 453 K and 101 kPa. Coverage effects were assessed; changes in adsorption/activation energies were generally ≤0.1 eV for the dominant hydroxyalkyl path (except CH2O* on Pt at 0.35 eV). Experimental: Alumina-supported Ni, Co, Ru, Pd, Pt and silica-supported Pd, Pt catalysts were synthesized (details in Supplementary Methods). Catalyst properties (reducibility, dispersion, particle sizes) were characterized (Supplementary Tables/Figures). Fixed-bed continuous reactor tests were run at 453 K and 101 kPa with P_OL = 14–20 kPa, PH2 = 44 kPa, WHSV_OL = 3.3–32 h^-1, at conversions <50% (far from equilibrium). Pre-reduction: Pd/Al2O3 at 453 K; Pt/Al2O3 and Ru/Al2O3 at 473 K; Co/Al2O3 at 723 K; Ni/Al2O3 at 773 K. Additional screening computed C* and O* adsorption energies for 294 dilute alloys (single-atom and dimer-atom alloys in p(3×3) with 0.11–0.22 ML solute coverage), and for carbides/nitrides (e.g., Mo2C, Mo2N) on representative low-index facets, including Wulff-selected surfaces such as Mo2C(101), γ-Mo2N(100), β-Mo2N(001). For β-Mo2N(001), a detailed methanol mechanism and micro-kinetics were computed; the effect of co-fed H2 at 453 K with PCH3OH = 14 kPa was examined. Mo2N was synthesized in situ for validation; structure and phase were confirmed by XRD and HRTEM; average particle size ~43 nm; estimated Mo dispersion ~2.5%.

- Mechanistic DFT on Pt(111) shows the hydroxyalkyl pathway is favored over the alkoxy pathway for CH3OH dehydrogenation, with lower barriers (e.g., 0.53 eV for C–H cleavage to CH2OH vs. 0.78 eV for O–H cleavage to CH3O). Across nine metals, preferred pathways vary with oxophilicity; Ni, Co, Ru stabilize alkoxy intermediates strongly. - Descriptor-based micro-kinetic modeling using C* and O* adsorption energies produced a volcano-shaped TOF map with two high-activity zones corresponding to hydroxyalkyl (weaker O binding) and alkoxy (stronger O binding) regimes. - 2-Octanol dehydrogenation TOF map (corrected from methanol using thermochemical scaling) reproduces the same activity trend as for methanol. - Experimental validation (2-octanol, 453 K, 101 kPa, low conversion): TOF per surface metal atom showed Pt as most active, followed by Pd, Ni, Co, Ru. Representative TOFs: Pt/Al2O3 1.19×10^-1 s^-1; Pt/SiO2 1.09×10^-1 s^-1; Pd/Al2O3 2.38×10^-2 s^-1; Pd/SiO2 2.42×10^-3 s^-1; Ni/Al2O3 2.26×10^-2 s^-1; Co/Al2O3 1.62×10^-2 s^-1; Ru/Al2O3 0.87×10^-2 s^-1. Byproduct 1-octene selectivity: <5% for Co, Pd, Pt; <15% for Ni, Ru. Predicted vs. measured TOFs show good parity overall; Pt and Pd measured TOFs are lower than predicted (by ~100× and ~30×), attributable to coking and co-adsorbed H coverage effects (including a modeled drop of Pt predicted TOF to 3.1×10^-2 s^-1 with H coverage). - Screening of 294 dilute alloys identified 12 candidates near the secondary TOF maximum (around EO−EC ≈ 0). However, synthesis protocols for some predicted single-atom/dimer-atom alloys (e.g., AgRu2) are unclear. - Carbide/nitride screening: close-packed facets of Mo2C and Mo2N lie in a low-activity region due to strong O binding, but Wulff-relevant facets β-Mo2N(001) and γ-Mo2N(100) shift toward the active region, approaching the secondary volcano maximum. - β-Mo2N(001) micro-kinetic modeling predicts higher tolerance to co-fed H2 compared with Pt(111); TOF for both declines with increased PH2, but Pt declines more sharply. - Experimental Mo2N test (200 mg, 2-octanol 1.2 mL/min, PH2 = 5.5 kPa): 7% conversion, 100% selectivity to 2-octanone; TOF per total surface Mo of 2.85×10^-3 s^-1 (~30× lower than predicted), likely due to limited expression of (001) facets and site contamination by strongly bound species. - Measured TOF vs. PH2: low PH2 decreases Pt activity (consistent with coking/site loss at low H2), while Mo2N activity is preserved, indicating an operational advantage.

The combined DFT and micro-kinetic approach establishes a two-descriptor volcano for alcohol dehydrogenation activity, linking atomic C and O adsorption energies to TOF. This directly addresses the goal of creating a rapid, predictive screening tool for heterogeneous catalysts. The experimentally observed activity order for 2-octanol dehydrogenation (Pt > Pd > Ni > Co > Ru) aligns with theoretical predictions, validating the model’s capability to rank catalysts and identify mechanistic regimes (hydroxyalkyl vs. alkoxy). The finding that the highest-activity volcano peaks lie outside the energy range of close-packed transition metals suggests intrinsic limitations of pure metals and typical alloys, motivating exploration of alternative chemistries and surface terminations. The alloy screening identifies viable candidates near a secondary maximum, though practical synthesis challenges remain. Transition metal nitrides and carbides, particularly β-Mo2N(001) and γ-Mo2N(100), emerge as promising non-precious alternatives, with both modeling and initial experiments indicating robustness to co-fed H2 and coking. Discrepancies in absolute TOFs (notably for Pt and Pd) underscore the importance of accounting for coverage effects, site blocking by coke, and realistic surface facets in bridging theory and experiment. Overall, the descriptor framework enables rational navigation of catalyst space, providing actionable guidance for designing active, stable, and economically viable catalysts for acceptorless alcohol dehydrogenation.

This work develops and experimentally validates a descriptor-based DFT micro-kinetic framework to identify active catalysts for acceptorless dehydrogenation of alcohols to carbonyls and H2. A two-descriptor volcano (C* and O* adsorption energies) captures activity trends across transition metals, with two high-activity plateaus corresponding to distinct mechanistic regimes. Experimental 2-octanol dehydrogenation confirms the predicted activity ranking (Pt > Pd > Ni > Co > Ru). Large-scale screening finds only a limited set of dilute alloys near the secondary maximum, while carbide/nitride surfaces—particularly β-Mo2N(001) and γ-Mo2N(100)—offer promising non-precious alternatives. Initial Mo2N experiments show selective ketone formation and relative insensitivity to low H2 partial pressures. The established approach provides a platform for rapid in silico catalyst discovery. Future work should focus on: synthesizing phase-pure β-Mo2N and γ-Mo2N with preferential exposure of (001)/(100) facets; developing routes to predicted single-atom/dimer-atom alloys; incorporating explicit coverage, coking, and dynamic site models; and extending the framework to broader alcohol substrates and realistic catalyst morphologies.

- Coverage effects and adsorbate–adsorbate interactions can alter TOFs by orders of magnitude; although catalyst ranking appears robust, absolute rates may deviate (e.g., CH2O* coverage effect on Pt ~0.35 eV). - Discrepancies between predicted and measured TOFs for Pt and Pd likely arise from coking and co-adsorbed H, which reduce accessible active sites; micro-kinetics assumed constant site availability. - The volcano’s global maximum lies outside the adsorption-energy range of close-packed transition metals, limiting improvements achievable through traditional metal alloying. - Alloy candidates identified may lack practical synthesis protocols (e.g., specific single-atom or dimer-atom configurations). - Modeling often used close-packed surfaces, which may not represent the dominant facets under reaction conditions; facet-dependent behavior (e.g., Mo2N(001)/(100)) is critical. - Experimental Mo2N sample likely had a limited fraction of the most active facets and potential site contamination by strongly bound species, contributing to lower-than-predicted TOFs. - Linear scaling and BEP relations (MAE < 0.2 eV) introduce systematic uncertainties that can propagate in micro-kinetic predictions.