5.4 Adsorption energies as a reference bookkeeping problem
These are unexecuted teaching inputs and starting models. Original diagrams are schematics, not calculated results. Validate version-specific syntax, licensed or authorized data, numerical convergence and the scientific model before using this workflow.
5.4.1 Model, units and provenance
Keep basis/potential files and executable versions traceable. Grid controls use Ry; common energy/force outputs use hartree and hartree/bohr. Read the unit in each output heading.
Shared inputs, conventions and evidence
5.4.2 Worked investigation
Intuition. An adsorption energy is meaningful only after stating the surface, adsorbate state, coverage and reference reaction. Surface relaxation, molecular deformation and lateral adsorbate interactions can all change the number. Use a neutral molecular adsorbate on case 17's validated insulating slab, with the actual coordinate file and binding orientation supplied by the learner. This is a design protocol, not an invented adsorption prediction.
Original calculation set. Prepare three folders with matched model chemistry: combined slab+M; clean slab with the same supercell, atom constraints and reference relaxation convention; isolated molecule M in an independently converged nonperiodic cell. Keep basis, functional, dispersion model and numerical accuracy consistent. For frozen-component interaction analysis, instead retain the adsorbed geometries in the separated calculations and consider ghost functions. The key input deltas are removal of selected coordinates/KINDs, correct charge/spin, boundary changes only for the isolated reference, and separate PROJECT names. There is no single keyword that makes these references scientifically equivalent.
Workflow. 1. Write the reference reaction S+M→S–M and define Eads=ES–M−ES−EM, so negative values indicate electronic stabilization under this convention. 2. Generate multiple plausible orientations/sites rather than only one guessed pose. 3. Optimize each with a stated set of frozen substrate atoms. 4. Recalculate all references at final numerical accuracy. 5. Separate deformation contributions by comparing isolated relaxed fragments to fragments frozen in the combined geometry. 6. Increase lateral area, slab thickness and vacuum independently. 7. If comparing to experiment, add only justified ZPE, thermal, entropy and chemical-potential corrections, documenting each term.
Interpretation. Different cell areas imply different coverages; dividing energy by adsorbate count does not remove lateral interactions. Adsorption can dissociate the molecule, in which case the reference reaction must account for the locations of all atoms. A lower optimized energy at one site says little about kinetic accessibility. A surface energy ranking at zero Kelvin is not a catalytic turnover prediction.
Checks and pitfalls. Ensure identical constrained-layer conventions between combined and clean slabs. A geometry builder can silently change atom order and invalidate fixed-atom lists. If the adsorbate transfers charge, finite-size and electrostatic alignment issues deserve explicit analysis. Numerical cancellation is strongest when references share compatible settings, but an isolated molecule and periodic slab still require independently appropriate boundaries. Do not compare a PBE slab with a PBE-D3 combined structure and call the difference adsorption.
Exercise. Construct a complete energy ledger for molecular and dissociative adsorption alternatives with balanced atoms and charges. Identify which additional gas-phase or reservoir reference is needed in each case. Report a site-ordering uncertainty from convergence tests rather than extra decimal digits.
Diagram. A thermodynamic bookkeeping triangle connects relaxed fragments, frozen fragments and combined structure. EN: adsorption energy, deformation, frozen interaction, coverage, common model. Draw no numerical site rankings.
5.4.3 Related calculations
- 5.3 Explicit solvation and the difference between interaction and free energy
- 6.1 NVE energy conservation separates timestep and SCF error