Skip to content

Relax atoms and cells with distinct criteria

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.

Choose physical degrees of freedom

calculation='relax' moves atoms at fixed lattice; 'vc-relax' permits selected cell changes. SCF does neither. Constraints follow the physics: an epitaxial film cannot be treated as free bulk just because full relaxation is easy to request.

Copy the complete first input into 03-relax/si.relax.in, retaining cell/cards and replacing preliminary cutoffs/mesh with tested values. Make these edits rather than treating the snippets as standalone files:

# Add to existing &CONTROL and change calculation to 'relax':
  nstep = 100,
  etot_conv_thr = 1.0d-5,
  forc_conv_thr = 1.0d-4,

&IONS
  ion_dynamics = 'bfgs',
/

Place IONS after ELECTRONS before cards. etot_conv_thr is Ry and forc_conv_thr Ry/bohr; 1d-4 Ry/bohr is approximately 0.00257 eV/Å. These geometry criteria differ from conv_thr electronic stopping. A relaxation solves multiple electronic problems; force thresholds below electronic noise are meaningless. Reaching nstep is not success.

An intentionally constrained volume relaxation

For cubic Si use the same full input, set calculation='vc-relax', keep IONS and add:

&CELL
  cell_dynamics = 'bfgs',
  press = 0.0,
  press_conv_thr = 0.5,
  cell_dofree = 'volume',
/

Pressure and pressure tolerance are kbar, with 1 kbar=0.1 GPa. cell_dofree='volume' keeps shape while changing volume, a deliberate cubic example constraint rather than universal prescription. Broader freedom must match the physical model and supported release. Official structural workflow; PW parameters.

Full stress, not just its trace

Read tensor and pressure using printed units/sign convention. For unconstrained zero-pressure bulk, shear and anisotropic normal components matter. Small mean pressure may hide opposing diagonal stresses. Under physical constraints, nonzero constrained components can be expected; assess only degrees of freedom actually allowed to relax. Symmetric Si atoms may have zero force at a nonoptimal volume, so atomic force alone cannot prove zero pressure.

Verify the final model in a new SCF

Confirm convergence rather than nstep/wall time exhaustion. Read the last complete final cell and coordinates with units; do not combine final positions with original lattice. Create a fresh fixed-geometry SCF with clean scratch, then recheck forces and stress using converged/stronger settings. Save constraints and structure provenance. Later paths must be generated for the actual cell, not copied from another orientation.

Exercise

Compare fixed-cell and volume optimization starting from the same geometry. Predict allowed changes before running, then report actual evidence separately. Include step count, criteria, final volume, force/stress and static verification. Blank planned tables are appropriate until execution. PARADIM teaching collection supplies independent exercises.

Previous：Converge the quantity you intend to publish · Next：Bands and DOS: different sampling tasks