A computational study proposes an orbital-coupling model built from two local features: the number of metal atoms around a hydrogen atom, called its coordination number, and the distances between nearby atoms. In the reported confined-defect setting, the rule treats hydrogen-metal interactions as coordination-dependent and hydrogen-hydrogen interactions as following a distance-decay rule.
The most visible test came at the scale of hydrogen release from tungsten. A large simulation reproduced experimental hydrogen retention and a bimodal thermal-desorption profile, with peaks near 550 K and 800 K. In other words, the calculation produced two distinct temperature ranges in which hydrogen left the simulated material. The comparison gives the model a target beyond individual atomic sites, while remaining a simulation-to-experiment comparison.
The model’s first tests
To build and test the descriptor, the authors used density functional theory calculations. The DFT setup used VASP with the PAW and PBE approaches, a 400 eV plane-wave cutoff, and both fully relaxed fixed-volume optimizations and fully fixed calculations. Vacancy-cluster calculations used a 4 by 4 by 4 supercell containing 128 atoms and a 3 by 3 by 3 k-point mesh. The study also constructed 12 symmetric tilt grain-boundary structures.
Solution energy was the study’s common yardstick for comparing hydrogen sites. It was calculated by taking the total energy of a defect containing n hydrogen atoms, subtracting the energy of the same defect without hydrogen, and then subtracting n times a reference hydrogen energy. That reference was defined as half the energy of a hydrogen molecule in vacuum. This lets the calculations compare whether different regions of a defect are energetically favorable for hydrogen within the model.
In tungsten vacancy clusters, stable hydrogen sites were mainly found on inner-surface and subsurface layers. The relationship between the descriptor and solution energy changed sign between regions rather than following one straight trend, producing a broken-line pattern. The finding illustrates the model’s central premise that local coordination changes how distance-based interactions translate into a site’s calculated energy.
A separate test used a tungsten vacancy structure labeled V6. The predicted hydrogen potential-energy surface, a map of calculated energy across possible positions, agreed with DFT in its overall shape, its energy peaks and its continuous changes through central, surface and subsurface regions. The supplied analysis characterizes that agreement as qualitative and reports no numerical surface-error measure.
When distance changes the answer
The model also assigns a distinctive distance rule to hydrogen pairs confined in vacancy clusters. The reported hydrogen-hydrogen interaction followed an inverse distance law: as the separation between the two atoms changed, the interaction followed that distance raised to the 7/2 power. The authors associate that exponent with hydrogen-metal s-d orbital coupling. The calculation therefore gives the distance dependence a specific form in a confined defect.
The same framework was applied to a tungsten vacancy loop. Here, hydrogen solution was favorable near the vacancy-core region but unfavorable at the outer edge. The result shows that the model assigns different solution preferences to different parts of one defect, rather than treating the loop as a single uniform trap.
At tungsten grain boundaries, the analysis was most accurate when it used the two nearest tungsten neighbors. Repulsive hydrogen-metal coupling was more prominent than attractive coupling. The model was also extended to alpha-iron vacancy clusters and molybdenum grain boundaries, where the reported patterns were described as analogous to those in tungsten.
From atomic sites to a release curve
For the larger-scale release calculation, the researchers used a GRACE foundation model to calculate hydrogen-hydrogen binding in vacancy clusters and hydrogen solution at vacancy loops. Large-scale OKMC simulations then coupled those atomistic hydrogen-vacancy energetics with evolving irradiation microstructures to calculate thermal-desorption spectra. This step carried the local defect rules into a calculation of hydrogen release.
In the reported comparison, the double peak was associated with an approximately 0.20 eV difference in hydrogen-hydrogen interaction and a solution-energy difference greater than 0.40 eV per hydrogen between V1 and Vn≥5. The reported comparison linked the two parts of the simulated profile with different vacancy-cluster environments. That link is an interpretation of the model comparison, not a direct experimental isolation of a single mechanism.
What the calculations do not settle
The evidence remains limited to the systems and calculations reported here. The supplied analysis gives no numerical uncertainty estimates for the descriptor fits, the potential-energy-surface comparison or the thermal-desorption prediction, and it provides no independent quantitative validation of the extension to additional metals and defect types. The reported cross-metal results cover alpha-iron vacancy clusters and molybdenum grain boundaries, with analogous patterns to tungsten; they do not establish that the rule will work for every metal, defect shape or hydrogen concentration. The study is therefore a computational demonstration across the listed cases, not universal validation.
Paper data and sources
Original title: Universal orbital-coupling rules for hydrogen-defect interactions in bcc metals
Authors: Qianxi Zhu, Hao-Xuan Huang, Zichen Zhang et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-10-07
DOI: Not available
Original paper · Full text