Restarting path integral MD simulation (PIGLET)

[Unmesh Mondal]

Hello everyone,

I am running a path integral MD simulation using GLE thermostat (PIGLET) with RPMD real time propagator. Every time I restart the calculation, I get sudden jumps in the the different quantities (Temp, potential energy,virial kinetic energy and conserved quantity).

I have incorporated the following tags for restart:

SCF_GUESS RESTART

and

&EXT_RESTART
  EXTERNAL_FILE ./TPA-1.restart
&END EXT_RESTART

and "WFN_RESTART_FILE_NAME" is commented out to use the default as in path integral simulation generates "2n" (n=number of replicas) number of different restart wavefunction files.

Can someone help me out with this issue? Attaching the input file(pimd_glet.in), the file(test-energy-1.dat) for the above mentioned quantities along with a figure for the same.  

Thanks

Unmesh Mondal

IISER Pune, India

[Thomas Kühne]

Dear Unmesh, 

you may simply try tightening the convergence values, e.g. 

EPS_DEFAULT 1.0E-12

EPS_SCF 1.0E-7 

Best, 

Thomas

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/73e67e03-89bd-415a-95ca-aba8c4dde7eb%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
<Figure_1.png><pimd_glet.in><test-energy-1.dat>

==============================

Thomas D. Kühne

Dynamics of Condensed Matter

Chair of Theoretical Chemistry

University of Paderborn

Warburger Str. 100

D-33098 Paderborn

Germany

tdku...@mail.upb.de

+49/(0)5251/60-5726

[Unmesh Mondal]

Dear Thomas,

Thank you for your reply. But,

EPS_DEFAULT 1.0E-12

EPS_SCF 1.0E-7

needs to be tightened to reduce drift in constant of motion which I will certainly incorporate . Additionally, what bothers me is the sudden jump at the

restart point which dies off in 20-30 steps. I am attaching the zoomed in figure highlighting the area of concern. I don't experience the same with classical nuclei AIMD.

Please do correct me if I have misunderstood the suggestion or any concept.

Regards,

Unmesh

[Christoph Schran]

Dear Unmesh,

which version of CP2K are you using? There has been a bug fix relatively recently (https://github.com/cp2k/cp2k/pull/359) for the computation of the conserved quantity and the primitive estimator in PINT runs. This fix will be in the upcoming new release v7.1.

How exactly are you restarting your runs? In the plots you attached, it looks like the thermostat would be reinitialized.

Best, Christoph

[Unmesh Mondal]

Thank You for your reply Christoph and apology for the late response.

>which version of CP2K are you using? There has been a bug fix relatively recently (https://github.com/cp2k/cp2k/pull/359) for the computation of the >conserved quantity and the primitive estimator in PINT runs. This fix will be in the upcoming new release v7.1.

I am using Cp2k 7.0 development version.  Yes, I will give the new version a try and check if the problem is solved.

>How exactly are you restarting your runs? In the plots you attached, it looks like the thermostat would be reinitialized.

Yes, I also assume the reinitialization of the thermostat takes few steps to equilibrate hence the sudden jumps (in the plots). I do not observe any sudden jumps when the simulation is run continuously without any restart. Regarding the restart, I have the wavefunction restart to use the default and an external restart file which the code prints at every step. Appending relevant part of  my initial mail: 

>I have incorporated the following tags for restart:

>SCF_GUESS RESTART

>and

>&EXT_RESTART
>  EXTERNAL_FILE ./TPA-1.restart
>&END EXT_RESTART

>and "WFN_RESTART_FILE_NAME" is commented out to use the default as  path integral simulation generates "2n" (n=number of replicas) number >of different restart wavefunction files.

Please let me know if you any extra information is needed.

Thanks 

Unmesh

 

--
You received this message because you are subscribed to a topic in the Google Groups "cp2k" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/cp2k/o19yhdkl0q8/unsubscribe.
To unsubscribe from this group and all its topics, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/8ce15b5f-89ee-44ac-af57-58f2476aac58%40googlegroups.com.