3.8 Atomic and cell relaxation
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.
3.8.1 Decide which variables may move
A fixed-cell relaxation changes atomic positions; a cell relaxation also changes chosen lattice degrees of freedom. These solve different physical problems. Use the converged cutoff, mesh, potentials, occupations and electronic settings from the previous lesson. The patch below replaces the corresponding tags in that full INCAR; each tag must appear once. It is not a standalone input and has not been run.
IBRION=2 selects conjugate-gradient structure optimization; NSW=100 is the maximum ionic step count, not an achievement of convergence. ISIF=2 permits ionic motion while holding cell shape and volume fixed. POTIM=0.5 is a trial step-control setting here, not a 0.5-fs molecular-dynamics timestep. Starting geometry and stiffness can require a different value. IBRION; POTIM; ISIF.
The negative EDIFFG=-0.01 sets a force stopping criterion: force norms must fall below 0.01 eV/Å. Positive EDIFFG instead uses energy change in eV; changing the sign changes the meaning, not simply the severity. EDIFF=1E-7 remains an electronic threshold. Choose sufficiently small electronic noise before demanding small forces. A force criterion alone is not a complete cell-stress acceptance rule. EDIFFG; EDIFF.
3.8.2 A symmetric example can teach the wrong lesson
Ideal diamond Si has symmetry-compatible coordinates. Atomic forces can already vanish at an incorrect lattice parameter, so a fixed-cell ionic run may finish without showing much movement. That does not prove the cell volume is optimal. For a deliberate ionic test, make a separate slightly displaced structure and explain its symmetry treatment. Preserve the exact initial structure rather than quietly editing the published baseline.
3.8.3 Cell optimization and basis stress
For a zero-pressure bulk exercise, changing the patch to ISIF=3 allows atomic positions, shape and volume to vary. This is a more general relaxation than a constrained cubic volume scan; justify the permitted degrees of freedom. A substrate-bound film or deliberately constrained model needs a different choice. New options, such as ISIF=8, require a sufficiently new version; they are not assumptions of this basic course. ISIF.
Finite basis errors can create artificial Pulay stress and shift the apparent equilibrium volume. Energy convergence at one geometry does not by itself establish accurate stress throughout a changing cell. Use a more generous cutoff for volume changes and test the final cell with stronger settings. An initial trial around 1.3 times the largest ENMAX is a useful documented starting scale, not a universal proof of stress convergence. Check the precision guidance for your installed version rather than treating PREC=Accurate as a guarantee. Pulay stress; PREC.
Separately establish that the optimizer found a minimum in the current numerical model and that the numerical model is accurate enough. Read the full stress tensor, cell shape, volume, force norms, symmetry and final SCF status. In VASP positive pressure-like diagonal stress indicates compression. At zero external pressure, declare and check a residual-stress tolerance; a small average can conceal anisotropic components. Use the printed kilobar convention consistently: 1 kbar = 0.1 GPa.
3.8.4 Fresh static verification
Inspect CONTCAR before copying it to a new directory as POSCAR. Retain the old inputs and optimization log. Set NSW=0, IBRION=-1 and perform a consistent high-precision static run with the final geometry. Starting fresh using ISTART=0, ICHARG=2 makes provenance straightforward. Recheck forces and stress with converged or stronger settings. This static run becomes a clear parent for subsequent bands/DOS.
An alternative is to scale the lattice vectors uniformly through several volumes around the initial geometry and compute a converged energy–volume curve. State whether internal positions were relaxed at each volume. If a fitted minimum lies outside the sampled range, extend the scan rather than reporting an extrapolated optimum as established. Volume relaxation.
3.8.5 Exercise and deliverable
Report initial/final volume, maximum residual force, stress tensor, ionic step count, final electronic state and the stronger-setting static check. Reaching NSW or the scheduler wall time is not optimization completion. If unable to run, submit a planned acceptance table with blank results and the distinction between ionic and cell degrees of freedom.
Previous：SCF, cutoff and k-mesh convergence · Next：Bands and density of states