HAC Database

Computational Details of the HAC Database 

The HAC dataset was constructed by performing quantum chemical calculations on seven previously reported hydrogen–ammonia combustion reaction pathways using the ORCA software package (version 5.0.4). All electronic structure calculations employed the B3LYP-D3(BJ) functional with the def2-TZVP basis set. The temperature range was set to 500–3000 K, reflecting typical operating conditions of hydrogen–ammonia internal combustion engines, and ab initio molecular dynamics (AIMD) simulations were carried out with time steps ranging from 0.4 to 1.3 fs.

For each reaction channel, 120,000 configurations were sampled via AIMD, yielding a total of 840,000 unique molecular geometries and corresponding energy values. Geometry optimizations of intermediates and transition states (TSs) were performed at the same level of theory—B3LYP-D3(BJ)/def2-TZVP—augmented with the def2/J auxiliary basis set to enable the RIJCOSX approximation for efficient Coulomb and exchange integrals. The self-consistent field (SCF) convergence criterion was set to tightSCF to ensure high numerical precision. 

Transition states were validated by frequency analysis, confirming the presence of exactly one imaginary vibrational mode. Intrinsic reaction coordinate (IRC) calculations were subsequently performed to trace the minimum energy path connecting reactants, transition states, and products, with configurations along these paths included in the dataset.

AIMD sampling was initiated from each verified transition state and conducted over the same temperature range (500–3000 K) with time steps of 0.4–1.3 fs. Each of the seven reaction channels contributed 120,000 snapshots, resulting in a comprehensive dataset of 840,000 geometry–energy pairs that capture the rich conformational and energetic landscape of hydrogen–ammonia combustion chemistry.

Table Description of HAC Reaction Channels