4.8 SOC and noncollinear magnetism
These are unexecuted teaching inputs and starting settings to test. Original figures are schematics, not computed results. Use licensed VASP and PAW data, replace every placeholder, record the executable version and validate convergence.
4.8.1 Model, units and provenance
Use eV for energy, Å for length and eV/Å for force; 1 kbar = 0.1 GPa. State normalization per atom, molecule, primitive cell or simulation cell. Record PAW identifiers, release, ZVAL, ENMAX and permitted hashes; never redistribute POTCAR. SCF convergence addresses the chosen electronic problem; convergence of the target property requires separate tests.
Original schematic. Curves explain concepts; blank data areas await verified learner results. No calculation is claimed.
4.8.2 Worked case: procedure, interpretation and checks
Two different switches in the physics
Noncollinear magnetism allows a magnetization vector whose direction varies through the system. Spin–orbit coupling (SOC) ties spin to orbital and lattice degrees of freedom. A noncollinear calculation need not include SOC, and a nonmagnetic material may still have important SOC-induced band changes. Avoid the shortcut “SOC means the material is magnetic.”
Choose either a nonmagnetic semiconductor for SOC band splitting or a magnetic crystal for orientation-dependent energies. These are distinct exercises. Use the noncollinear executable vasp_ncl, suitable PAW datasets, a documented spin coordinate system, and much tighter convergence for small anisotropy energies than for a qualitative band plot. LSORBIT
A controlled starting point
First obtain a converged nonmagnetic reference density, then start a self-consistent spinor calculation. For a two-atom illustrative antialigned seed, the delta below sets moments along opposite spinor z directions. It does not describe a universal two-atom material:
MAGMOM has three components per atom here; its numerical moments are trial seeds in μB. For a nonmagnetic SOC calculation use an appropriate zero-moment initialization and verify the resulting state. For a noncollinear calculation without SOC, set LSORBIT false while retaining the spinor formalism. LSORBIT true activates the noncollinear machinery automatically; the explicit line in the example makes intent visible. LNONCOLLINEAR, MAGMOM
Do not retain collinear ISPIN=2 out of habit. The official documentation notes an error for the combination with MAGMOM and LNONCOLLINEAR in VASP 6.5.0. A previous magnetic density can carry a spin texture into a restart; changing MAGMOM alone does not necessarily reinitialize that texture. Starting different trial states from a nonmagnetic reference is easier to reason about.
Coordinate systems and interpretation
SAXIS defines the spinor basis. With the default z axis it coincides with the ordinary Cartesian convention; after changing it, spin components and orbital moments are not automatically expressed in the same coordinates. Transform them before comparing vectors. The local magnetic moments are also projection-dependent and need not sum exactly to a naive atom-only picture of the whole magnetization. SAXIS, LSORBIT output
ISYM=−1 is a conservative starting choice that avoids unintended symmetry restrictions but costs more. If exploiting symmetry later, test that the chosen operations preserve the intended magnetic state. Without SOC, a rigid global rotation should not change a spin-rotation-invariant energy apart from numerical error; with SOC it can. However, an unconstrained magnetization can rotate during SCF, so changing an initial direction alone does not prove that two final energies represent the desired axes. Check final moments and use an explicitly validated constrained-moment workflow if required.
Checks and exercise
For SOC band comparisons, retain geometry, functional, PAW, and k convention, and carefully account for spinor band counting. Band degeneracy depends on the actual symmetries: SOC does not generically split every band everywhere. For anisotropy, converge differences with mesh, occupation smearing, cutoff, and tolerance; sub-meV conclusions demand substantially stronger evidence than total energies printed to many decimals.
Exercise: compare a scalar-relativistic calculation and a self-consistent SOC calculation for one structure. Identify a symmetry-allowed splitting or a protected degeneracy without predicting its size in advance. For a magnetic extension, compare two verified final magnetization directions and report energy per chosen normalization, convergence uncertainty, and whether constraints were used. A band plot alone cannot establish a topological invariant.
4.8.3 Unexecuted inputs and analysis scaffolds
These are unexecuted teaching inputs and starting settings to test. Original figures are schematics, not computed results. Use licensed VASP and PAW data, replace every placeholder, record the executable version and validate convergence.
4.8.3.1 Input block 1
# Run with vasp_ncl; exactly two atoms in this illustrative seed
LNONCOLLINEAR = .TRUE.
LSORBIT = .TRUE.
SAXIS = 0 0 1
MAGMOM = 0 0 2 0 0 -2
GGA_COMPAT = .FALSE.
LASPH = .TRUE.
LORBIT = 11
LORBMOM = .TRUE.
ISYM = -1
NSW = 0
EDIFF = 1E-8
# Omit ISPIN=2. Choose restart flags for actual available files.
4.8.4 Related learning paths
- 1.1 Four input files, one physical question
- 1.2 A convergence laboratory with an error budget
- 4.7 Hybrid band gaps and the correct band workflow
- 7.1 Dielectric tensors, Born effective charges, and LO–TO splitting