Physics-guided thermodynamic screening of H₂ adsorption on alkaline-earth-modified anatase TiO₂ nanoparticles with...

Understanding how surface modification tunes H₂ adsorption on oxide nanomaterials is important for designing reversible hydrogen-interaction platforms under near-ambient conditions. Here, we examine single-H₂ adsorption on a fixed, idealized ∼1.8 nm anatase-derived TiO₂ nanoparticle reference model, both pristine and surface-modified with alkaline-earth (AE) species (Be, Mg, Ca, Sr, Ba, and Ra), using GATOR, a physics-guided DFTB-based screening framework with LLM-assisted interpretation. Within this framework, the deterministic thermodynamic screening sequence defines the classification, whereas the constrained language model serves only as an interpretive layer rather than as a predictor of adsorption physics. The workflow combines DFTB-derived adsorption energetics, adsorption-mode analysis, finite-temperature thermodynamic descriptors, Langmuir-type coverage estimates, rule-based thermodynamic screening, and post-hoc consistency checks to evaluate local H₂ uptake-release behavior. Electronic adsorption energies are mapped onto finite-temperature descriptors at 298 K, including finite-temperature-corrected adsorption enthalpy surrogates, standard-state free-energy estimates, desorption midpoint temperatures, and pressure-dependent fractional coverages. Across the series, Be-TiO₂ and Ra-TiO₂ occupy excessive-binding/heating-needed regimes, whereas Ca-, Sr-, and Ba-modified TiO₂ remain weak-binding and under-loaded near ambient pressure. Pristine TiO₂ and Mg-modified TiO₂ represent the most balanced strong-binding cases, but both lie above the practical desorption window under the entropy-consistent thermodynamic treatment. Frontier-orbital and conceptual-DFT descriptors change only modestly upon H₂ adsorption, indicating that the observed trends are governed mainly by local adsorption geometry and finite-temperature adsorption thermodynamics rather than by substantial global electronic-structure reorganization. Within the DFTB-based screening framework, pristine TiO₂ and Mg-TiO₂ occupy the most balanced region of the representative adsorption landscape. Higher-level PBE-D3(BJ) calculations support attractive local H₂ interactions for both representative motifs, but they do not reproduce the small DFTB energetic preference for Mg-TiO₂ and place the local Mg-H₂ minimum at a longer separation. Accordingly, the pristine/Mg ordering should be regarded as method-sensitive, and no method-independent energetic preference between these two systems is assigned.