Introduction

Biomass is the only renewable resource of organic carbons in nature and their conversion to value-added chemicals and liquid fuels is of vital importance in achieving global carbon neutralisation1,2. Cellulose-derived 5-hydroxymethylfurfural (HMF) is widely recognised as a platform chemical for the synthesis of sustainable liquid fuels and chemicals3,4. Particularly, the selective hydrogenolysis of HMF to 2,5-dimethylfuran (DMF) as biofuels and feedstocks of renewable p-xylene has attracted much interest5,6. A great deal of effort has been devoted to develo** supported metal catalysts for this reaction, and state-of-the-art catalysts are based upon Ru, Pd, Pt, Ni and Cu materials7,8,9,10,11. We have designed cobalt oxide-supported ruthenium (Ru/Co3O4) and cobalt/nickel [(Co)Ni/Co3O4] catalysts that show DMF yields of 93% and 70–76%, respectively, at 130 °C for 24 h7,8. Schüth et al. developed a hollow platinum-cobalt bimetallic nanoparticle (PtCo@HCS) catalyst, which achieved a high yield of DMF (98%) at 180 °C for 2 h10. Esteves et al. investigated various supported copper catalysts and identified a high yield of DMF (93%) over Cu/Fe2O3-Al2O3 at 150 °C for 10 h11. To date, metal and harsh reaction condition (i.e., high temperature and/or long reaction time) are almost indispensable to achieve the high yield of DMF. It is widely accepted that the homolytic dissociation of H2 occurs on these metal catalysts, generating free radicals (H·) to drive the subsequent hydrogenolysis12. Recently, it is reported that Hδ− species obtained via heterolytic dissociation of H2 showed enhanced catalytic performance13,14,15,16,17. Thus, the development of new catalysts that can generate Hδ− species hold great promise to promote the hydrogenolysis of HMF under mild reaction conditions.

Although single-atom catalysts can catalyse the heterolytic cleavage of H213,14,15, there is complicity associated with their preparation and thermodynamic stability. Meanwhile, metal oxides with a high concentration of surface defects are reported as emerging catalysts with high activity for the heterolytic cleavage of H216,17,18,19,20,21,Density functional theory studies

In this work, all spin-polarized DFT calculations were carried out using the Vienna Ab-initio Simulation Package (VASP)53. The projector augmented wave (PAW) method54 and the Perdew−Burke−Ernzerhof (PBE)55 functional under the generalized gradient approximation (GGA)56 were applied throughout the calculations. The kinetic energy cut-off was set to 400 eV, and the force threshold in structure optimization was 0.05 eV/Å. We used a large vacuum gap of 15 Å to eliminate the interactions between neighbouring slabs. By adopting these calculation settings, the optimized lattice constant of CoO (P1) is 4.248 Å, which is in good agreement with the experimental value of 4.267 Å57.

The transition states (TS) of surface reactions were located using a constrained optimization scheme and were verified when (i) all forces on the relaxed atoms vanish and (ii) the total energy is a maximum along the reaction coordination but it is a minimum with respect to the rest of the degrees of freedom58,59,60. The adsorption energy of species X on the surface (Eads(X)) was calculated with

$${{{{{{\rm{E}}}}}}}_{{{{{{\rm{ads}}}}}}}({{{{{\rm{X}}}}}})=-({{{{{{\rm{E}}}}}}}_{{{{{{\rm{X}}}}}}/{{{{{\rm{slab}}}}}}}-{{{{{{\rm{E}}}}}}}_{{{{{{\rm{slab}}}}}}}-{{{{{{\rm{E}}}}}}}_{{{{{{\rm{X}}}}}}})$$
(1)

where EX/slab is the calculated total energy of the adsorption system, while Eslab and EX are calculated energies of the clean surface and the gas-phase molecule X, respectively. Obviously, a positive value of Eads(X) indicates an exothermic adsorption process, and the more positive the Eads(X) is, the more strongly the adsorbate X binds to the surface.

The oxygen vacancy formation energy (EOV) was calculated according to

$${{{{{{\rm{E}}}}}}}_{{{{{{\rm{OV}}}}}}}={{{{{{\rm{E}}}}}}}_{{{{{{\rm{slab}}}}}}-{{{{{\rm{OV}}}}}}}+{1/2{{{{{\rm{E}}}}}}}_{{{{{{\rm{O}}}}}}2}-{{{{{{\rm{E}}}}}}}_{{{{{{\rm{slab}}}}}}}$$
(2)

where Eslab-OV is the total energy of the surface with one oxygen vacancy, and EO2 is the energy of a gas-phase O2 molecule.

For the model construction, we built a p(2 × 3) surface slab containing five atomic layers for the CoO(100) surface (a = 12.74 Å; b = 8.67 Å; c = 23.50 Å; α = β = γ = 90°), and the top four CoO layers were allowed to relax, while the bottom atomic layer was kept fixed to mimic the bulk region. A 2 × 2 × 1 k-point mesh was used in calculations of all these models. Note that the on-site Coulomb interaction correction is necessary for the appropriate description of the Co 3d electrons, and all calculations are performed with U = 5.1 eV and J = 1.0 eV, which are consistent with the values determined by previous studies61,62.

In addition, we tested the effect of the spin state of 3d electrons in Co2+ in the optimization of CoO, and found that the high‐spin antiferromagnetic arrangement was the most stable state, and the calculated magnetic moment of 2.74 µB obtained from the difference in spin-up and spin-down densities is consistent with literature reports63,64,65.