I would appreciate your advice on an issue I encountered when using an SQS generated with mcsqs.
1. System and SQS details
-
Parent srtcture: Co3O4 spinel
-
Supercell: 112 atoms (48 cations, 64 O)
-
Metal fractions: Co 0.46875, Ti 0.125, Ge 0.125, Nb 0.125, Sb 0.0625, Ta 0.03125, W 0.0625
-
mcsqs.log output :
Objective_function = -1.985059- Largest value in
Correlations_mismatch≈ 0.08 (all < 0.1)
The POSCAR was produced with sqs2poscar.
2. Problem observed in VASP
In VASP, the calculation results yields an unphysically low total energy (less than −10,000 eV). Because the identical INCAR settings converge without issue for pristine Co₃O₄, I suspect the problem lies in the generated geometry rather than in the VASP input parameters.
3. Questions
Have you ever encountered mcsqs‑generated structures exhibiting issues—such as unphysically short interatomic distances, incorrect lattice vectors, or neglect of element–element interactions—that cause DFT calculations to diverge or produce nonsensical energies?
Alternatively, in your experience, do such convergence failures stem more from the intrinsic compositional complexity of multi‑component oxides than from structural flaws in the SQS itself?
Any guidance or troubleshooting suggestions would be greatly appreciated.
Thank you for your time and for developing such a valuable toolkit.