Engineered MgO Nanoparticles with Tunable Electronic Signatures for Energy Applications.

We couple density-functional tight-binding (DFTB3) with a physics-guided, agentic multimodal AI workflow to map the size-composition landscape of two ultrasmall Zn-doped MgO nanoparticles with radii of 0.8 and 0.9 nm and Zn contents between 0 and 25 at%. For both sizes, Zn incorporation monotonically stabilizes the rocksalt-derived nanoparticles and narrows the energy gap (0.8 nm: from 5.76 to 4.79 eV; 0.9 nm: from 5.00 to 4.53 eV) via an upward HOMO and a downward LUMO shift. The frontier-level evolution yields lower ionization potential and higher electron affinity; hardness decreases, while electrophilicity and the maximum electron transfer increase, with a stronger response in the smaller particle. A practical operating window at 10-15% Zn emerges that balances enhanced electron-accepting propensity with adequate robustness. The multimodal AI model, fusing DOS, geometry, and image features, provides accurate and interpretable corrections to DFTB3 trends under data scarcity. The combined thermodynamic and electronic signatures rationalize energy-oriented uses, particularly polysulfide anchoring in Li-S host architectures and modified dielectric and transport behavior in oxide-based devices.