Differences

This shows you the differences between two versions of the page.

--- howto:md [2015/05/11 18:39] – [MD Convergence] 130.60.136.225
+++ howto:md [2021/07/20 19:33] – [How to run Molecular Dynamics] tkuehne
@@ Line 2: / Line 2: @@
 ====== How to run Molecular Dynamics ======
-Molecular dynamics a good method to perform thermodynamical averages, and to look at dynamical properties. It also a good starting point for many other more advanced techniques.
+Molecular dynamics is a good method to perform thermodynamical averages, and to look at dynamical properties. It is also a good starting point for many other more advanced techniques.
 During an MD one expects the density to change more or less continuously, thus it is possible to use the solutions of the previous steps to create the new initial guess for the density (and wavefunctions for OT).
 Indeed cp2k can use different extrapolations techniques (''FORCE_EVAL%DFT%QS'').
-For MD a good guess is ''PS'' with an ''EXTRAPOLATION_ORDER'' of 3. If you are doing something where from one step to the other there is little continuity, (geometry optimization, montecarlo), then something with little continuity like ''LINEAR_PS'' or just the previous density (''USE_PREV_P'') might be better. If you change the particle number then a ''USE_GUESS'' that restarts from scratch with the restart method choosen in ''FORCE_EVAL%DFT%SCF%SCF_GUESS''. The ''ASPC'' extrapolation is used in connection with langevin and a reduced SCF convergence (just on SCF step) and hystory restart to simulate big systems.
+For MD a good guess is ''PS'' with an ''EXTRAPOLATION_ORDER'' of 3. If you are doing something where from one step to the other there is little continuity, (geometry optimization, montecarlo), then something with little continuity like ''LINEAR_PS'' or just the previous density (''USE_PREV_P'') might be better. If you change the particle number then a ''USE_GUESS'' that restarts from scratch with the restart method choosen in ''FORCE_EVAL%DFT%SCF%SCF_GUESS''. The ''ASPC'' extrapolation is used in connection with langevin and a reduced SCF convergence (just on SCF step) and history restart to simulate big systems.
 ===== Trajectory =====