Re: Generating topology files for crystalline materials

From: Axel Kohlmeyer (akohlmey_at_gmail.com)
Date: Sun Feb 10 2013 - 14:54:45 CST

On Sun, Feb 10, 2013 at 9:02 PM, FX <fxcoudert_at_gmail.com> wrote:
> Hi,
>
> Thanks for the help. (You're on the CPMD, VMD, NAMD lists… do you watch all the ".*MD" lists? :)

i've retired from CPMD several years ago. VMD/NAMD usually i only
respond when it refers to software that i'm maintaining or touches on
something that tickles my curiosity. my main effort these days is
actually geared towards LAMMPS.

>
>> the molfile plugin API now also supports writing out topology
>> information, so you can generate the topology information from VMD
>> scripting. this can be tedious and error prone, so it would be
>> beneficial to write some scripted tools for that. the topotools
>> plugin, for example provides a "middleware" that can automate a large
>> part of the process. check out:
>> https://sites.google.com/site/akohlmey/software/topotools
>
> If I "only" need to generate a list of bonds

same difference.

>> you'll have to come up with a smart way to add the bonds that wrap
>> around periodic boundaries
>
> Shoot. I was expecting this to be available in NAMD directly, not to have to reinvent it. Are there never bonds that wrap around the unit cell? I'm sure at least someone has simulated zeolites, fully flexible polymers, something like that...

this really is not a NAMD issue. NAMD just uses the bonding
information provided in the psf file and PBC treatment will make it
irrelevant whether bonds wrap around boundaries or not.

one strategy that i could imagine to work would be to first build the
bonding topology (in VMD) without taking PBC into account, e.g.
through "mol bondsrecalc" after setting suitable per-atom radii. and
then write a little script that can run a check for whether atoms have
all valences saturated and then look for suitable periodic neighbors
(via pbwithin) and add the missing bonds, write out the psf file and
feed it to NAMD.

>
>> and i would be curious to learn, if the
>> automatic derivation of topological angles and dihedrals will work
>> correctly in those cases (never had a chance to try it on a real world
>> problem).
>
> Well, the publication whose results I am trying to reproduce says simulations were run using NAMD, so it appears to work :)

i'm not talking about NAMD, but the topotools Tcl code.

>
>> it may be simpler to build the topology information direction, since
>> you know your structure and i assume there are some simple rules to
>> follow.
>
> I searched for "topology information direction" in the doc, tutorials, and pretty much anywhere in Google, but nothing turns up. What do you mean?

damn autocompletion. "direction" was supposed to be "directly".

> If it's a sort of "automatic topology" feature (like Gromacs or GULP has, where I can give a list of rules like "type N1 is any N atom linked to C, C and H", then yeah, I have simple rules I can feed it. How can I access it?

see above. for these kind of rules you could write a VMD script that
uses a loop over all atoms of type N1, then use pbwithin to find
suitable bond partners and compile a list of all such bonds and use
"topo setbondlist" to define all bonds and write out a psf file.

axel.

>
>
> Thanks again,
> FX Coudert
>

-- 
Dr. Axel Kohlmeyer  akohlmey_at_gmail.com  http://goo.gl/1wk0
International Centre for Theoretical Physics, Trieste. Italy.

This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:20:54 CST