IWAVPR: Difference between revisions
No edit summary |
(Major update to explain that IWAVPR < 10 is depreciated. Also defaults have been corrected) |
||
| Line 1: | Line 1: | ||
{{TAGDEF|IWAVPR|0 {{!}} 1 {{!}} 2 {{!}} 3 {{!}} 10 {{!}} 11 {{!}} 12 {{!}} 13}} | {{TAGDEF|IWAVPR|0 {{!}} 1 {{!}} 2 {{!}} 3 {{!}} 10 {{!}} 11 {{!}} 12 {{!}} 13}} | ||
{{DEF|IWAVPR| | {{DEF|IWAVPR|12|if {{TAG|IBRION}}{{=}}0 (MD) and 11 (relaxation, on-the-fly machine learning MD)|0|else (static calculation)}} | ||
Description: {{TAG|IWAVPR}} determines how orbitals and/or charge densities | Description: {{TAG|IWAVPR}} determines how orbitals and/or charge densities | ||
| Line 7: | Line 7: | ||
---- | ---- | ||
For {{TAG|IWAVPR}}<10, the file {{TAG|TMPCAR}} is used to store old orbitals that | |||
are required for the prediction. | are required for the prediction. This setting is depreciated, and not supported by the MPI version. | ||
The recommended settings are {{TAG|IWAVPR}}>10. In this case, the prediction is | |||
are stored in the main memory | performed without an external file {{TAG|TMPCAR}} (i.e. all required arrays | ||
are stored in the main memory). | |||
The following options are available for {{TAG|IWAVPR}}: | |||
*{{TAG|IWAVPR}}=11 | *{{TAG|IWAVPR}}=0 no extrapolation, usually not preferable for first-principles molecular dynamics simulations or relaxations of the ions into the groundstate. | ||
*{{TAG|IWAVPR}}=1|11 Simple extrapolation of the charge density using atomic charge densities (eq. (9.8) in thesis G. Kresse). This switch is convenient for geometry optimizations (ionic relaxation and volume/cell shape with the conjugate gradient or Quasi-Newton methods, i.e. {{TAG|IBRION}}=1,2,3 etc.) | |||
*{{TAG|IWAVPR}}=2|12 A second-order extrapolation for the orbitals and the charge density (eq. (9.9) in thesis G. Kresse) is performed. This results in superior performance for first-principles molecular-dynamics simulations. It might cause instabilities during on-the-fly learning, so the default is 11 in this case. | |||
<!--However, the following warning may occur: <code>Information: wavefunction orthogonal band xxx..</code>. This is a sign of band crossing and --> | <!--However, the following warning may occur: <code>Information: wavefunction orthogonal band xxx..</code>. This is a sign of band crossing and --> | ||
*{{TAG|IWAVPR}}=3|13 In this case a second order extrapolation for the orbitals, and a | *{{TAG|IWAVPR}}=3|13 In this case a second-order extrapolation for the orbitals, and a simple extrapolation of the charge density using atomic charge densities is done. This is a mixture between {{TAG|IWAVPR}}=1 and 2, however, it is usually worse than {{TAG|IWAVPR}}=2. | ||
:Mind: We don't | :Mind: We don't encourage this setting. | ||
== Related tags and articles == | == Related tags and articles == | ||
Latest revision as of 10:56, 14 February 2025
IWAVPR = 0 | 1 | 2 | 3 | 10 | 11 | 12 | 13
| Default: IWAVPR | = 12 | if IBRION=0 (MD) and 11 (relaxation, on-the-fly machine learning MD) |
| = 0 | else (static calculation) |
Description: IWAVPR determines how orbitals and/or charge densities are extrapolated from one ionic configuration to the next configuration.
For IWAVPR<10, the file TMPCAR is used to store old orbitals that are required for the prediction. This setting is depreciated, and not supported by the MPI version. The recommended settings are IWAVPR>10. In this case, the prediction is performed without an external file TMPCAR (i.e. all required arrays are stored in the main memory).
The following options are available for IWAVPR:
- IWAVPR=0 no extrapolation, usually not preferable for first-principles molecular dynamics simulations or relaxations of the ions into the groundstate.
- IWAVPR=1|11 Simple extrapolation of the charge density using atomic charge densities (eq. (9.8) in thesis G. Kresse). This switch is convenient for geometry optimizations (ionic relaxation and volume/cell shape with the conjugate gradient or Quasi-Newton methods, i.e. IBRION=1,2,3 etc.)
- IWAVPR=2|12 A second-order extrapolation for the orbitals and the charge density (eq. (9.9) in thesis G. Kresse) is performed. This results in superior performance for first-principles molecular-dynamics simulations. It might cause instabilities during on-the-fly learning, so the default is 11 in this case.
- IWAVPR=3|13 In this case a second-order extrapolation for the orbitals, and a simple extrapolation of the charge density using atomic charge densities is done. This is a mixture between IWAVPR=1 and 2, however, it is usually worse than IWAVPR=2.
- Mind: We don't encourage this setting.