Skip to content

2.1 Isolated boundary conditions and box convergence

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.

2.1.1 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

Original schematic: Isolated boundary conditions and box convergence. No numerical results are claimed.
Original schematic: Isolated boundary conditions and box convergence. No numerical results are claimed.

2.1.2 Worked investigation

Intuition and prerequisites. A molecule placed in a large box is not automatically isolated. Atomic neighbor periodicity and the electrostatic boundary problem are separate parts of the model. For a polar molecule, periodic replicas can interact even when atoms are far apart. Begin with baseline A, a working nonperiodic Poisson solver, and case 02's force check. This experiment concerns neutral water; charged periodic systems introduce compensating-background and reference issues that require a separate protocol.

Original input delta. Keep SUBSYS/CELL/PERIODIC NONE and DFT/POISSON/PERIODIC NONE together. Compare cubic lengths 12, 16 and 20 angstrom. Recenter every geometry: place O at half the box length and add the same relative H vectors as A. Use POISSON_SOLVER WAVELET only with the supported geometry/build. Keep physical coordinates in angstrom, not SCALED T; otherwise changing the box would stretch the molecule too.

Workflow. 1. Generate the three boxes analytically while holding bond lengths and angle fixed. 2. Verify minimum distances to boundaries and inspect the molecule visually. 3. Run fixed-geometry energy/force calculations with identical basis, cutoff and SCF tolerance. 4. Compare total energy, the three atomic forces and, if printed, the dipole under the same origin convention. 5. Translate the molecule by 0.2 angstrom in x within the largest box and repeat. 6. As an explicitly different model, run XYZ/PERIODIC electrostatics at the same box sizes; do not mix these energies into the isolated convergence sequence.

Output interpretation and checks. A stable isolated result should become insensitive to further box growth within your target tolerance. Translation sensitivity can diagnose representation or boundary artifacts. The periodic sequence may approach the isolated limit slowly because long-range interactions persist after short-range overlap disappears. Compare energy differences with care: a neutral molecule's dipole is origin-independent, whereas a charged system's dipole is not. Do not infer box convergence from the visual absence of atoms crossing a boundary.

Pitfalls and limits. A solver may support only particular periodicity/cell combinations. Obey its documented restrictions rather than matching keyword strings by appearance. Increasing the box at fixed grid resolution increases memory and cost. A diffuse anion is especially demanding and may be artificially confined or electronically unstable with the chosen functional; it is not just a larger neutral-water test. For a finite cluster containing solvent, nonperiodic boundaries also mean a physical surface exists, which changes the question compared with bulk solution.

Exercise. Define separate tolerances for energy and maximum force difference before running. Design a cheaper box that passes both, and report a failed test if neither does. Explain why scaling all coordinates with the cell would confound the exercise.

Diagram. Draw the same water shape in three boxes, with unchanged bond lengths. Beneath, contrast isolated field lines terminating at the far boundary with periodic replicas. EN: molecule fixed, box varied, isolated electrostatics, periodic images. This is a model diagram, not computed field data.

2.1.4 Sources and further reading