My Community

I am working on a system with a large number of atoms and electrons, which leads to lengthy self-consistent field (SCF) calculations. To reduce the computational time for Density of States (DOS) and band structure calculations, I am considering using the pre-converged CHGCAR and WAVECAR files from a previous SCF calculation.

Would setting NELM = 1 in the INCAR file for the DOS or band structure calculations be a valid approach to achieve faster convergence with fewer iterations, given that I am reusing the pre-converged charge density and wavefunctions? Are there any potential issues or best practices I should be aware of when applying this method for large systems?

Thank you for your guidance!

It's a good idea to use preconverged files. That should definitely speed up the calculations and I'm not aware of any issues specially for large systems.

What NELM=1 does it just sets the maximum number of electronic steps, cutting the convergence loop after the first step. So it does not accelarate convergence. It is the same as if you would lower the convergence criterum EDIFF.

Some tags you could consider to speed up your calculations:
LREAL = A #wiki/index.php/LREAL - projection operators are done in real space.
ALGO = Fast or VeryFast #wiki/index.php/ALGO - faster electronic solvers, but less robust, so please test if they work.
Minimize the required ENCUT and k mesh size on convergence tests of smaller systems.