Unlocking the electronic and adsorption origin of doping-engineered WS₂ monolayers: a DFT investigation with machine learning validation

Defect-engineered tungsten disulfide (WS₂) monolayers provide a tunable platform for clarifying how local dopant chemistry that converts weak gas adsorption into measurable electronic perturbation. Here, N-doped and N-O co-doped WS₂ monolayers are investigated to decouple adsorption strength from electronic-response descriptors toward CO, CO₂, and NH₃ sensing. Machine learning force field (MLFF)-assisted DFT calculations show that MLFF reproduces the DFT energy-force relationships and phonon trends for pristine WS₂ and N-WS₂, whereas the larger deviations for N-O-WS₂ indicate that co-doped configurations should be interpreted together with direct DFT descriptors. Complementary 10 ps ab initio Molecular Dynamics (AIMD) simulations further show bounded energy fluctuations without obvious thermal drift, supporting the finite-temperature stability of the optimized doped structures within the present simulation window. Pristine WS₂ exhibits negligible gas-induced perturbation, including weak adsorption, long gas-surface distances, limited charge redistribution, and minor work-function changes. N and N-O substitutions distort the local W-S coordination environment and introduce dopant-derived states near the Fermi level, increasing surface polarizability. NH₃ exhibits the strongest response in the doped WS₂ due to lone-pair-induced polarization that significantly modulates the surface dipole. Although N-O-WS₂ binds NH₃ slightly more strongly, N-WS₂-NH₃ delivers the largest relative work-function response of 6.04%, indicating that adsorption strength and electronic sensitivity are distinct descriptors. Energy-resolved conductivity-sensitivity analysis identifies NH₃ as the most effective electronic perturbant for the doped WS₂ systems, although the dominant response maxima are offset from the Fermi level and therefore represent spectral modulation rather than directly accessible low-bias sensing performance. Picosecond recovery times indicate rapid reversible desorption but limited residence time. Overall, MLFF-assisted DFT identifies doped WS₂ as a reversible weak-physisorption platform with preferential NH₃ electronic response rather than a strongly adsorbing gas-sensor material.