[gmx-users] force evaluation

Florian Dommert dommert at fias.uni-frankfurt.de
Tue Jan 6 09:59:57 CET 2009


  since I am also need direct access to the forces and play around a
  little bit I realized that this is not that easy. However if you
  set "nstfout=1", you should get the forces in every step.
  But I have another question, regarding the direct part of the coulomb forces.
  Since we have worked out an error estimation for the SPME algorithm,
  that allows us to estimate the absolute error introduced by the
  interpolation and the cutoff. We wrote a little tool to test the
  estimate and it works very well. But comparing the forces calculated by
  this tool and Gromacs yields a root mean square difference of the
  forces of 10^-4. That is a very big difference, because the precision
  of the forces should be 10^-12. Now the question arises, which forces
  are "wrong" ? Since I talk about the forces of the first step before
  any integration, only the force calculation can introduce the
  differences and I can not imagine, that the compiler optimizations are
  responsible. Moreover the comparison was done, using the Ewald sum, so
  there is no FFT involved, that can cause the differences.
  The next point I thought about were the tabulated potentials, but
  turning this of in the source code yielded even a larger difference.

  The problem is that the error estimate of Kolafa and Perram (
  http://dx.doi.org/10.1080/08927029208049126 ) for the
  Ewald sum underestimates the error introduced by Gromacs, while our
  tool fits to the estimate.

  Can anybody imagine the source of difference ?
  Furthermore I have found no possiblity to turn of the tabulation of the
  coulombic forces in the mdp options, did I miss it or is there no
  possibility ?
  Finally it would be very interested in a possiblity to switch off the
  vdw interactions easily. So far I used an appropiate topology files with C6 and
  C12 set to zero. Is there an easier solution as using a table filled
  with zeros ?



* Mark Abraham <Mark.Abraham at anu.edu.au> [2009-01-06 08:45:05 +1100]:

> Silvia Crivelli wrote:
>> Hello,
>> I need to calculate the forces (derivatives) for every atom in a protein
>> as this is being locally minimized using mdrun.
>> When I run mdrun I can get the maximum force F-max and the atom on which 
>> this
>> force is evaluated printed in the log file (for example,
>> "F-max             =  2.20323e+03 on atom 2641 ".)
>> I need to be able to compute the force value on every atom of the protein.
>> How can I do this with Gromacs?
>> I set the parameter nstfout = 10 in my .mdp file but this doesn't seem to 
>> make a difference.

> "nstfout = 10" writes the per-atom forces to the trajectory file every 10 
> steps, which you can verify with gmxcheck and/or gmxdump. I haven't verified 
> that these actually get written for energy minimization, and they probably 
> wouldn't get written at all if your EM doesn't get to 10 steps....
> Mark
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at http://www.gromacs.org/search before posting!
> Please don't post (un)subscribe requests to the list. Use the www interface 
> or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php

Florian Dommert

Computational and Theoretical Softmatter & Biophysics group

Frankfurt Institute for Advanced Studies
Johann-Wolfgang-Goethe University

Ruth-Moufang-Str. 1
60438 Frankfurt am Main

Phone: +49(0)69 / 798 - 47529
Fax:   +49(0)69 / 798 - 47611

EMail: dommert at fias.uni-frankfurt.de
Home: http://fias.uni-frankfurt.de/~simbio/Florian_Dommert
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 194 bytes
Desc: not available
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20090106/1a10cba9/attachment.sig>

More information about the gromacs.org_gmx-users mailing list