NTEMPER: Difference between revisions
mNo edit summary |
No edit summary |
||
| Line 11: | Line 11: | ||
Furthermore, no temperature swapping is attempted for {{TAG|NTEMPER}}=0. This allows many independent calculations to be performed in directories 01, 02, 03, ... which can be useful when performing many similar calculations for small systems or using machine-learned force fields on many-core machines (strong scaling is sometimes not particularly good for small systems). | Furthermore, no temperature swapping is attempted for {{TAG|NTEMPER}}=0. This allows many independent calculations to be performed in directories 01, 02, 03, ... which can be useful when performing many similar calculations for small systems or using machine-learned force fields on many-core machines (strong scaling is sometimes not particularly good for small systems). | ||
A final note is in place. | A final note is in place. If INCAR files exist in the subdirectories, the tag {{TAG|NTEMPER}} must be set in all {{TAG|INCAR}} files of the subdirectories 01, 02, 03, and {{TAG|NTEMPER}} must be identical. Failure to observe this rule can lead to unexpected behavior. | ||
== Related tags and articles == | == Related tags and articles == | ||
{{TAG|IMAGES}}, | {{TAG|IMAGES}}, | ||
Revision as of 17:01, 12 March 2024
NTEMPER = [integer]
Default: NTEMPER = 200
Description: NTEMPER specifies how often temperature swaps are attempted during parallel tempering. The flag must be used in combination with IMAGES and LTEMPER.
NTEMPER must be a positive integer or zero. Swaps are on average attempted after NTEMPER molecular dynamics steps. Specifically, a random number between [0,1[ is drawn. The next temperature swaps are attempted after NTEMPER steps. Obviously, for NTEMPER=1 swaps are attempted at every MD step. In the present code version, temperature swaps are alternatively performed between images 1,2 and 3,4 and 5,6 etc., or upon the next swap between images 2,3 and 4,5 and 6,7 etc. This implies that many swaps are attempted at the same time.
Furthermore, no temperature swapping is attempted for NTEMPER=0. This allows many independent calculations to be performed in directories 01, 02, 03, ... which can be useful when performing many similar calculations for small systems or using machine-learned force fields on many-core machines (strong scaling is sometimes not particularly good for small systems).
A final note is in place. If INCAR files exist in the subdirectories, the tag NTEMPER must be set in all INCAR files of the subdirectories 01, 02, 03, and NTEMPER must be identical. Failure to observe this rule can lead to unexpected behavior.