[DFTB-Plus-User] Restarting SCC and SCC/MD

[Alathea Davies]

Hello!

I wanted to clarify my understanding of restarting an SCC or an SCC/MD calculation from DFTB+.

I am doing some benchmarking with an HPC cluster to check the optimum parallelization of DFTB+ for some larger simulations that I am running. This cluster will only allot me time in 12 hours increments, though, so I'll need to be able to restart a calculation if it times out in the middle of an SCC or MD iteration. My questions/current understanding is as follows:

  1.  For an SCC calculation (specifically Conjugate Gradient driver), the charges and geometry after a successful SCC convergence are written to the charges.bin and geo_end.gen files. For instance, if my calculation has a successful SCC convergence at geometry step 12, it will write these charges and the atom coordinates to the respective output files. If, however, my calculation times out in the middle of geometry step 13 and does not finish calculating the charges or positions of atoms, are the respective output files overwritten with the incomplete data, or are they only overwritten once SCC convergence has been reached? Would I be able to restart this calculation using the charges.bin and geo_end.gen files?

  2.  My understanding is similar for an SCC/MD calculation, although, I think there is a slight difference due to specifying the "MDRestartFrequency". In a similar scenario, if I am running an SCC/MD calculation with an MD restart frequency of 100, but my calculation times out at an MD step that is not a multiple of 100, have the charges.bin and geo_end.xyz positions and velocities been overwritten with the incomplete data, or are they complete up to the most recent MD step that is a multiple of 100? If I wanted to restart this type of calculation, I would need to take the per-atom velocities from the final step in the .xyz file and read those into the dftb_in.hsd script, as well as the final charges and final atomic coordinates. Am I understanding this correctly?

Thank you so much for your time and help! I greatly appreciate it!

~ Alathea (Ali) Davies


Graduate Research Assistant
Computational Design Research Group
University of Wyoming
Department of Chemistry
adavies2 at uwyo.edu

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.zfn.uni-bremen.de/pipermail/dftb-plus-user/attachments/20220329/c7d44279/attachment.htm>