1.3 Gaussian basis and GTH potential are separate decisions
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.
1.3.1 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
1.3.2 Worked investigation
Intuition and prerequisites. In GPW, Gaussian functions describe orbitals while an auxiliary grid represents quantities such as the density. Increasing the grid cutoff cannot manufacture missing orbital flexibility. A pseudopotential determines which electrons are explicit and the effective interaction with the core; it is not merely a numerical speed switch. Use the isolated water model, case 02's force interpretation, and the actual installed data files.
Input experiment. In both H and O KIND sections of A, compare DZVP-MOLOPT-GTH against TZVP-MOLOPT-GTH and TZV2P-MOLOPT-GTH if those exact entries exist in the installed BASIS_MOLOPT. Keep GTH-PBE-q1 and GTH-PBE-q6 unchanged. First inspect the record headers, then select a name; never invent a missing basis by editing its label. Retain the same geometry, charge, grid and functional. Record basis-function count and wall time alongside energy and force. The family labels are descriptors, not a universal guarantee that a larger-looking name is a strict nested superset.
Workflow. 1. Establish a grid/SCF setting sufficiently tight that their changes are smaller than the target basis comparison. 2. Perform all basis variants at one fixed geometry. 3. Compare a chemically meaningful energy difference, such as the stretching energy between A and an otherwise identical structure with both O–H distances expanded by 2%, rather than only the absolute energy. 4. Optimize separately with each basis only after the fixed-geometry comparison; distinguish electronic and geometry effects. 5. Repeat an interaction energy using case 16 to reveal basis borrowing that a monomer calculation cannot detect.
Interpretation and checks. Total energies may decrease as flexibility increases, but practical Gaussian families and numerical grids need not produce a perfectly monotonic ladder. Greater basis size can make the overlap matrix ill-conditioned, particularly in condensed systems with diffuse functions. Look for linear-dependence warnings and unstable SCF, not only run time. Count explicit electrons from q values each time a potential changes. For comparisons across pseudopotentials, core reference offsets can differ, so raw total energies cannot be casually subtracted.
Pitfalls. Replacing PBE by another functional without revisiting the pseudopotential approximation can add a model inconsistency. Some research workflows deliberately use a potential generated with a related functional, but that choice requires justification rather than pretending the labels are interchangeable. Short-range MOLOPT variants can be attractive in condensed phases; whether they preserve the observable must be tested. A larger cutoff does not repair basis-set superposition error by itself.
Exercise. Build a two-axis convergence chart: rows are Gaussian basis choices, columns are grid settings. Report the stretching-energy change and maximum force-component change relative to your most stringent calculation. Decide which cheaper pair meets a declared tolerance and which comparisons remain inconclusive. Include exact data-file hashes.
Diagram. Two independent sliders feed a central density sketch. One changes orbital lobes (“Gaussian orbital basis”); the other changes a background mesh (“auxiliary grid”). A separate core icon identifies “pseudopotential / explicit electrons.” A crossed-out single slider states “cutoff alone ≠ basis convergence.”