Halo8: a dataset of chemical reaction pathways incorporating halogen chemistry

# Halo8 Database Halo8 is a ASE-based database containing total 20,116,288 structures, each with properties computed by DFT. You can directly access the common properties below with the following code: CoordinatesEnergyForce ```pythonfrom ase.db import connect db_path = 'your/path/to/db' with connect(db_path) as db: for row in db.select(): atoms = row.toatoms() coords = atoms.get_positions() # Coordinates energy = atoms.get_potential_energy() # Energy forces = atoms.get_forces() # Force``` All of the extra properties below are stored in row.data and can be accessed by their keys: dand_idHOMO_idxHOMO_levelLUMO_idxLUMO_levelMulliken_chargesLowdin_chargesDipole_momentNuclear_repulsion_energyElectronic_energyOne_electron_energyTwo_electron_energyExchange_energyCorrelation_energyDispersion_correction ```python from ase.db import connect db_path = 'your/path/to/db' with connect(db_path) as db: for row in db.select(): your_property = row.data['keys_for_property'] ``` The dand_id field follows the format: [Source]_[Formula]_rxn[XXXXX]_[Y] Source: T1x (reactions from the Transition1x dataset) or Halogen (newly added halogen-containing reactions) Formula: chemical formula of the system rxnXXXXX: reaction index number Y: structure index within a given reaction Each reaction pathway consists of 10 images in total: 1 reactant (Y=0), 8 intermediate NEB images, and 1 product (Y=9). The final block of 8 intermediates plus the fixed endpoints form the converged MEP. For example, if the Y indices for a reaction are: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 .... 178 179 180 181 182 183 184 185 The converged MEP of this example is: 0 → 178 → 179 → 180 → 181 → 182 → 183 → 184 → 185 → 9 (10 images total) The initial interpolation path (Y=1-8) and subsequent NEB optimization cycles (Y=10-17, 18-25, ...) are not part of the final pathway. The highest-energy image along the final MEP corresponds to the transition state. Note that all geometries are optimized at the GFN2-xTB level and should not be interpreted as DFT-level stationary points.