A fast and precise DFT wavelet code

SCF loop input parameters

A fast and precise DFT wavelet code
(Redirected from Input.mix)
Jump to: navigation, search

The "mix:" block in input.yaml contains the parameters required to use a density (or potential) mixing scheme for the self consistent field loop. This loop may imply a smearing of the occupancies for finite temperature DFT calculations. The old input scheme used a separate file named input.mix.

Example

mix:
   iscf: 7          # Mixing parameter.
   itrpmax: 50      # Maximum number of diagonalisation iterations.
   rpnrm_cv: 1.e-11 # Stop criterion on the residue of potential or density.
   norbsempty: 100  # number of extra orbitals ( unoccupied ).
   tel: 1.e-3       # Electronic temperature.
   alphamix: 0.95   # Proportion of old density kept. Must be larger than 1/2.
   alphadiis: 2.    # Multiplying factors for the electronic DIIS.

Example of an input.mix file (old input scheme)

7             # Mixing scheme type
50            # Maximum number of diagonalisation iterations
1.e-11        # Stop criterion on the residue of potential or density
100 1e-3 1    # No. of additional bands, elec. temperature, smearing method
0.95 2.0      # Multiplying factors for the mixing and the electronic DIIS

Entry descriptions

Descriptions of the keys:

  • iscf: An integer giving the mixing scheme and the mixing target. It follows the ABINIT convention. For values lower than 10, the potential is mixed, while for values greater than 10, the density is mixed. A recommended option is Pulay mixing (7 or 17 for potential or charge mixing, respectively)
  • itrpmax: Maximum number of repeats for the mixing scheme. Each iteration implies an optimization of the wavefunctions at constant potential according to the settings in input.dft.
  • rpnrm_cv: The stop criterion on the residue of the potential or the density. Typical values are in the order of 1e-10 to 1e-12.
  • norbsempty: The maximum number of additional (initially unoccupied) bands considered. If a larger value is given, The maximum of orbitals available via the input guess procedure will be used automatically
  • Tel: The electronic temperature in energy units (Ha).
  • occopt: The smearing function to determine the occupation. The conventions differ from those used by ABINIT. 1 = Gaussian, 2 = Fermi, 3 = Marzari (bumb minimization), 4 = Marzari (monotonic tail), 5 = Methfessel and Paxton.
  • alphamix : The closer this value gets to one, the more cautious will be the mixing of the charge or potential. In the case of straight mixing, it corresponds to Failed to parse (PNG conversion failed; check for correct installation of latex and dvipng (or dvips + gs + convert)): \alpha
according to Failed to parse (PNG conversion failed; check for correct installation of latex and dvipng (or dvips + gs + convert)):  \rho_{new} ~\leftarrow ~\alpha~\rho_{old} + (1-\alpha)~\rho_{new}

. alphamix must be larger than one half. A typical choice for metals is 0.95.

  • alphadiis : The DIIS multiplier used for preconditioning the wavefunctions at constant potential.

Related DFT variables

When performing a calculation with a mixing scheme as described above, it is important to adjust some DFT settings accordingly in the "dft:" block (or in the file input.dft for the old input scheme).

  • itermax: Few iterations are used in the inner loop to optimize the wavefunctions at constant potential, typically 2 to 4.
  • nrepmax: An additional repeat loop with diagonalisation does not make sense, thus nrepmax should be kept at 1.
  • idsx: The DIIS history length should be set to zero.
  • gnrm_cv: This value should be kept small, e.g. 1e-8, such that the inner loop will do itermax iterations and not less.
Personal tools