Geometry optimization and boundary conditions

The inputs on this page are educational templates that have not been executed for this course. Use an authorized installation, verify version-specific options and establish your own convergence.

Optimize positions at fixed cell

Copy the full first input to water_opt.inp in a new folder with verified data. Retain tested numerical settings, set PROJECT water_opt and RUN_TYPE GEO_OPT, and append this top-level block after END FORCE_EVAL. Do not duplicate GLOBAL or MOTION:

&MOTION
  &GEO_OPT
    OPTIMIZER BFGS
    MAX_ITER 100
    MAX_FORCE 1.0E-4
    RMS_FORCE 5.0E-5
    MAX_DR 3.0E-3
    RMS_DR 1.5E-3
  &END GEO_OPT
&END MOTION

MAX_FORCE/RMS_FORCE are maximum component and RMS criteria in hartree/bohr; MAX_DR/RMS_DR use bohr. They are trial thresholds, not measured errors. BFGS is suitable for this tiny example; larger systems may need LBFGS. A summed total force vector cannot certify all individual components. Optimizer parameters.

Read all required force/displacement criteria and termination reason. MAX_ITER=100 does not promise success. Inspect geometry/warnings and final complete XYZ frame; output often resembles water_opt-pos-1.xyz but numbering depends on history/settings. Interrupted writes may not end with a complete three-atom frame. Official geometry tutorial.

Measure, then independently verify

Measure both O–H lengths and H–O–H angle with units. Transfer sufficient digits to a fresh ENERGY_FORCE input at identical settings. This distinguishes optimizer stopping from small final forces. Save initial/final geometry, trajectory and new single-point evidence, not merely a viewer image.

Vacuum, crystal and slab are different physical models

For water retain MT/nonperiodic treatment, hold internal geometry fixed and test centered boxes 18,22,26 Å. Increase all atomic coordinates by half the box-length increase to recenter without altering distances. Compare target energy differences and forces, not just reported cell length. Grid discretization changes can produce small nonmonotonic effects.

A 3D crystal instead needs physical lattice and compatible periodic CELL/POISSON XYZ settings. Changing flags on one water does not create liquid: it creates a periodic array at the chosen density. Slabs need appropriate dimensional electrostatics/vacuum, sometimes dipole treatment, with solver-supported periodicity. Cell reference.

GEO_OPT is fixed-cell positions; CELL_OPT adds stress, pressure and lattice degrees. Never optimize isolated-molecule vacuum volume. Bulk equilibrium volumes need stress convergence with grid, basis and sampling; atomic force convergence alone is not enough.

Exercise

Record initial/final lengths and angle, complete optimizer status, largest final component, fresh force verification and box test. Keep measurements blank until executed. Explain why periodic water arrays, isolated water and liquid water are distinct models.

---

Previous：SCF, grids and basis: separate tests · Next：Short MD with honest diagnostics