Computing FES from multiple RESTART files

[Diana Marlén Castañeda Bagatella]

Hello everyone,
I hope you are doing well.

I am running calculations in a cluster that constrains them to 3 days per job; after these 3 days, I need to rerun the simulation from their last checkpoint.

CP2K generates a restart file during the simulation process, which is used by the graph.popt program (included in the CP2K module) to save the free energy surface (FES) path in a file called fes.dat. Likewise, the same restart file can be used to continue the simulation from its last interruption (checkpoint).

The inconvenience happens when the simulation is interrupted as the restart file only saves the simulation progress during the 3-day wall time. Hence, graph.popt can only save the fes path of the last piece of simulation. To avoid losing information, I copy the restart of the interrupted simulation to a new folder and continue the simulation in this new folder. However, graph.popt can read only one restart file, and I cannot find any way to create a restart file containing the information of the simulation (which is split into several pieces) and make it readable for graph.popt.

I would greatly appreciate it if someone could advise me on how to compute the FES of the whole simulation properly.

I wish you a great day, and thanks for your time,
Diana.

[Marcella Iannuzzi]

Dear Diana, 

I am not sure that I fully understand your problem. 

I suppose you are running metadynamics. If not, please provide more information.

If you properly restart the simulation, restarting also the metadynamics if this is the case, the new restart file contains the whole information.

The graph code can be used to evaluate the FES at any step using the last generated restart file.

Regards

Marcella

[David M. Sherman]

Perhaps I'm misreading your post, but I have a horrible feeling that, every 3 days, you are simply restarting the whole metadynamics simulation from the very beginning!  As such, you will never get a converged FES.

Just to clarify: you simply need to restart the job using the the *.restart file from the previous 3 day run as the input file for the next 3 day run (that's why it's called a "restart" file...).  Cp2k reads a *.restart file as an input file.  Maybe that's obvious; if so, forgive me.  Note, however, that the restart file (as it evolves) will keep accumulating all of the HILLS that were generated during the whole metadynamics run, not just the HILLS generated over the past 3 days.  There is no need to move things to a new folder (directory); again, you simply restart the job using the *.restart file as the input file.  You don't need the previous *.restart files to generate the FES. Indeed, the latest *.restart file actually contains a record of the time-evolution of the FES that you can generate using the -span argument to the "graph" program. The only thing that you need to be careful about is if your job stops during an update of the restart file.  That's why you should always have your input file generate a few *restart.bak files so that, if necessary,  you can copy the last *restart.bak-1 to *.restart.