[gmx-users] Re: question about non-orthongonal box vectors in pdc

[Tsjerk Wassenaar]

Hi,

I know the box correction in ns.c (which is done in a fairly neat way, but
can still fail for obscure reasons, which shouldn't happen ever), but I
meant also during the setup, if Gromacs is fed a really skewed box. In that
case a lattice reduction could be performed to get the optimal box
representation for further processing. This will usually involve both
shifting lattice vectors as well as performing a rotation. Now that's not
yet available in Gromacs :) In this way you can always obtain a unit cell
definition for which the projection of one vector on another does not exceed
+/- half the size of that vector. According to the gromacs rules for
triclinic boxes, this is still allowed.

I agree that this will still not make PME perfect, since in any case there
are still deviations from a radially uniform spreading of charge.

Tsjerk

On 3/30/06, Erik Lindahl <lindahl at sbc.su.se> wrote:
>
> Hi,
> Box correction is already done automatically in Gromacs (ns.c) - what I
> meant here was "fairly skewed compared to a rectangular box", e.g. 30
> degrees. Since the charge spreading has to be done on a triclinic grid in
> that case it is no longer radially uniform, which _could_ affect accuracy a
> bit (i.e., you might need to user higher interpolation order).
>
> Cheers,
>
> Erik
>
>
>
> On Mar 30, 2006, at 2:05 PM, Tsjerk Wassenaar wrote:
>
>
> Hi Erik,
>
> Actually a severely skewed box is basically an imaginary concept. For
> normal MD dimensions it is always possible to perform a simple
> transformation to transform a box into one which complies to the Gromacs
> rules in an optimal way. That is, if the box vector lengths are around the
> same order of magnitude. Maybe a routine for this could be added to gromacs
> (editconf/genbox/grompp) to fix skewed boxes, and to optimize the use of
> PME. I added a routine to my version of pbc.c which could provide a basis.
> This does not change the system, but only uses other, equally valid, lattice
> vectors. Let me know if you're interested.
>
> Tsjerk
>
> On 3/30/06, Erik Lindahl <lindahl at sbc.su.se> wrote:
> >
> > Hi,
> > As Tsjerk wrote we always work with the triclinic representation in the
> > code, which fills up space nicely. Due to rounding it might not be a
> > _perfect_ representation of e.g. a hexagonal box, truncated octahedron,
> > rhombic dodecahedron, etc., but that shouldn't affect the physics of the
> > triclinic box.
> >
> > The reason I first asked about boxes is that the PME grid too has to be
> > triclinic for such boxes, and I don't think anybody has done any detail PME
> > accuracy analysis in the case of severely skewed boxes.
> >
> > Cheers,
> >
> > Erik
> >
> > On Mar 30, 2006, at 9:43 AM, Tsjerk Wassenaar wrote:
> >
> > Hi Michael,
> >
> > It will in any case still fill up space in a periodic way. That can be
> > done with an infinite number of triclinic boxes, of which the formal
> > (perfectly regular) truncated octahedron is only one. Besides, the shift
> > vectors are used to calculate the distances between particles, determine the
> > neighbour lists, etc. If the coordinates are only stored with a certain
> > precision, there is no point in having lattice vectors (box vectors)
> > available with higher precision. You may want to read a bit on periodic
> > boundary conditions from Henk Bekker (notably his 1996 JCC paper).
> >
> > By the way.., don't use the truncated octahedron (unless it's for
> > comparison), try a rhombic dodecahedron instead. It's smaller and most
> > likely better in terms of the simulation results.
> >
> > Cheers,
> >
> > Tsjerk
> >
> > On 3/30/06, Michael Shirts <mrshirts at gmail.com > wrote:
> > >
> > > That would be pbc (periodic boundary conditions), not pdc (who knows
> > > what that is), of course . . .
> > >
> > > On 3/30/06, Michael Shirts <mrshirts at gmail.com> wrote:
> > > > So, to get a truncated octahedron, or some other shape, you use
> > > > certain ratios for the box vectors as detailed in the manual.  But
> > > the
> > > > box vectors in the .gro file are truncated to a certain precision,
> > > so
> > > > don't exactly meet these ratios.  What is the consequence of not
> > > > exactly meeting these ratios?  It won't precisely fill space in a
> > > > periodic way -- so does it mess up the pbc in some way?  Anyone
> > > know?
> > > >  Any experiences?  Should I make sure the ratios are correct up to
> > > > machine precision?  Is there a way to handle this more robustly?
> > > >
> > > > Thanks,
> > > > Michael Shirts
> > > > Research Fellow
> > > > Columbia University
> > > >
> > > _______________________________________________
> > > gmx-users mailing list    gmx-users at gromacs.org
> > > http://www.gromacs.org/mailman/listinfo/gmx-users
> > > 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
> > >
> >
> >
> >
> > --
> >
> > Tsjerk A. Wassenaar, M.Sc.
> > Groningen Biomolecular Sciences and Biotechnology Institute (GBB)
> > Dept. of Biophysical Chemistry
> > University of Groningen
> > Nijenborgh 4
> > 9747AG Groningen, The Netherlands
> > +31 50 363 4336
> > _______________________________________________
> > gmx-users mailing list    gmx-users at gromacs.org
> > http://www.gromacs.org/mailman/listinfo/gmx-users
> > 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
> >
> >
> > _______________________________________________
> > gmx-users mailing list    gmx-users at gromacs.org
> > http://www.gromacs.org/mailman/listinfo/gmx-users
> > 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
> >
> >
>
>
> --
>
> Tsjerk A. Wassenaar, M.Sc.
> Groningen Biomolecular Sciences and Biotechnology Institute (GBB)
> Dept. of Biophysical Chemistry
> University of Groningen
> Nijenborgh 4
> 9747AG Groningen, The Netherlands
> +31 50 363 4336
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> 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
>
>
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> 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
>
>


--

Tsjerk A. Wassenaar, M.Sc.
Groningen Biomolecular Sciences and Biotechnology Institute (GBB)
Dept. of Biophysical Chemistry
University of Groningen
Nijenborgh 4
9747AG Groningen, The Netherlands
+31 50 363 4336
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20060330/4bff296c/attachment.html>