Three copper atoms emerged as the strongest gas-binding arrangement in a computer model of a defected hBN/SiC surface. Across clusters containing one to five Cu atoms, Cu3 had the most favorable calculated adsorption energy for each of six tested gases: carbon monoxide (CO), hydrogen (H2), nitrogen (N2), oxygen (O2), hydrogen sulfide (H2S) and carbon dioxide (CO2). For Cu3, the energies were -2.20 eV for CO, -1.03 eV for H2, -1.44 eV for N2, -1.64 eV for O2, -1.67 eV for H2S and -0.36 eV for CO2.
The work is a modeling study, not a measurement on a physical sample. It examines how the copper-to-vacancy ratio and cluster nuclearity, meaning the number of copper atoms in a cluster, relate to computed stability, electronic structure and gas-surface reactivity at boron vacancies in hBN/SiC. The supplied document is an arXiv version 1 preprint dated 26 August 2026.
Where the copper settled
To build the surface, the authors modeled a bilayer hBN/SiC supercell with 100 boron, 100 nitrogen, 64 silicon and 64 carbon atoms. Removing one boron atom created the vacancy used in the calculations. They used density-functional theory (DFT) in VASP with PAW, with PBE for the general exchange-correlation treatment and the screened hybrid HSE03 functional for selected electronic calculations.
Machine-learning molecular dynamics was used to explore how copper was trapped and how clusters developed. The simulations randomly introduced ten boron-vacancy defects, placed Cu atoms 3 Å above the surface, varied the copper count from 5 to 60 and tested Cu-to-vacancy ratios of 1, 2, 4 and 6. Each composition had ten independent runs lasting 5 ns at 300 K, with a 0.5 fs timestep.
At the end of those trajectories, maximum atomic vibration amplitudes remained below 3 Å. Increasing the Cu-to-vacancy ratio was associated with larger vacancy-anchored clusters and more weakly confined Cu atoms at higher loading.
Three atoms stood out at the vacancy
Formation-energy calculations suggested that three Cu atoms were sufficient to passivate one boron-vacancy site. Beyond that size, the binding energy saturated at approximately -2.6 eV per Cu, while larger clusters increasingly involved metallic Cu-Cu interactions.
Nudged-elastic-band calculations, which estimate an energy barrier along a movement path, put copper migration across B-N bonds at approximately 200 meV. Cu-Cu aggregation was effectively barrierless in the modeled comparison, and the trapped-vacancy case had an aggregation energy about 0.55 eV more favorable than the pristine-surface case.
Larger clusters show distinct calculated spin patterns
For clusters with four or more Cu atoms, the calculations found different spin patterns between even and odd nuclearities. Even-numbered complexes remained open-shell doublets, while odd-numbered complexes favored closed-shell singlets. Comparisons of cluster sizes separated by two Cu atoms showed two new states in the calculated band gap, and Cu4, Cu6 and Cu8 had similar calculated states.
The gas tests
On the reference surfaces, gas binding was much weaker. Pristine hBN/SiC had adsorption energies of approximately -0.06 to -0.19 eV for several gases, while the strongest bare-vacancy interaction was only -0.34 eV for H2S. The Cu3 values reached -2.20 eV for CO and -1.64 eV for O2.
For oxygen, O2 adsorption energies across Cu1 to Cu5 ranged from -0.66 to -1.64 eV, and computed charge transfer ranged from +0.49 to +0.75 electron. The O-O bond length increased from approximately 1.21 Å in isolated O2 to 1.30 to 1.39 Å in the molecule-surface complexes, changes consistent with molecular activation.
Carbon monoxide was strongly chemisorbed through Cu-C bond formation, with Cu-C bonds 1.77 to 1.79 Å long. Its internal C-O bond shifted only from 1.13 Å in molecular CO to 1.15 to 1.16 Å on the cluster, a change described as weak activation.
A result confined to the model
The gas calculations used PBE-D3(BJ) for Cu_n@VB with n from 1 to 5, assessing adsorption energies, Bader charge transfer and the most stable adsorption geometries. The broader electronic-structure calculations used DFT in VASP with PAW, PBE generally and HSE03 for selected calculations.
That makes the headline result a comparison within a defined computational set: modeled hBN/SiC cells, selected copper cluster sizes and the six gases listed above. The preprint reports calculated properties rather than measurements on a physical sample. Its authors say supporting data are available upon request.
Paper data and sources
Original title: Nuclearity of Copper Clusters on hBN/SiC Heterostructure Modulates Molecular Adsorption
Authors: Reza Khakpour, Arsalan Hashemi, Xiaoya Chang et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-26
DOI: Not available
Original paper · Full text