[gmx-users] Some Issues on Substrates/PBC
André Farias de Moura
moura at ufscar.br
Wed Aug 6 20:00:30 CEST 2008
if you are not interested in the substrate dynamics, then you could try
the freeze group options (freezegrps and freezedim). I have used these
options for graphite and it worked just fine.
> Dear Listmembers,
> This post is quite lengthy because it contains several different problems
> I came across during my simulations and that I cannot figure out up to
> now. I
> included some links to pictures to give you some visualization and would
> glad for every suggestion or hint even to part of my problems.
> Generally for my simulations I need substrates of silicon oxide and
> (each one in different simulations). The substrate fills the simulation
> completely in x-y direction so it should look like an infinite surface
> considering the periodic boundary conditions.
> The simulation with silicon oxide is working fine but I experienced some
> strange thing though: When the .gro file contains negative coordinates and
> generate a .tpr for 4 nodes the structure is messed up totally after
> minimization (although it converges and gives no errors). I guess that the
> breaking up in four parts is somehow connected to the 4 nodes, because
> generating and running a 1 node .tpr solves this problem and the structure
> keeps intact after em - see pictures here: ( http://tinyurl.com/59zfda ).
> I shift the coordinates in the .gro with editconf so that every coordinate
> positive the structure can be em with both 1 or 4 node without problems.
> For the main md I have to stick with 1 node anyway because I get a "Shake
> block crossing node boundaries" error when I try to use 4 nodes. I found
> advise to use less nodes in the mailing list archives and as I said it
> for me, too, but is there any other way to get around this? For the test
> I'm doing now 1 node is fine, but for the real simulations later on I
> like to use more nodes because the system will get bigger...
> With the graphite I have another problem: Again I have to stick with only
> node, otherwise the structure is broken after energy minimization (with 4
> nodes http://tinyurl.com/5eqngz ). Additionally here the energy
> is not converging even after 2500 steps, forces keep in the range of 10^4
> I can still do a md with this file that is not exploding (md with the
> structure without any energy minimization will result in an exploding
> as one might expect). After energy minimization the graphite sheets (four
> in the simulation) are bended up/down correspondently to the connected
> periodic boundary at the edges of the simulation box (box is triclinic,
> http://tinyurl.com/5m72ke ). When I run an md with this as starting
> without position restraints on the carbon atoms, the sheets cleave
> from the edges and then reunite in a curled conformation within 10ps (
> http://tinyurl.com/5j4l3z ).
> If I put position restraints on all carbon atoms the graphite keeps flat,
> of course I still have the up/down bending at the box edges, which I also
> to get rid off to have a real "infinite" graphite surface.
> First I thought that this was maybe a problem of my graphite structure not
> fitting into my simulation box (hence the curling of the surface) but I
> checked the bonding lengths and b0 in the topology, the unit cell vectors
> box vectors thoroughly and everything seems to be right. If I copy my
> simulation box and shift it by one of the box vectors I get a perfect
> continuation without any deviation on the boundaries, so everything seems
> it should be for the starting structure.
> I also simulated the same graphite sheets within a bigger box, so that
> carbons could not feel their counterparts over the periodic boundaries.
> gave no distortion at the edges after em and the em converged, but of
> in this way I cannot get an continous surface.
> Even stranger is, that after em the whole structure coordinates are
> into the neighbouring periodic cell! For a bigger system with some
> additionally molecules that should absorb to the graphite surface the
> sheets even ended up in different periodic images of the simulation box
> em ( http://tinyurl.com/5uaxfa ). Maybe that's more a visualization
> because the following md simulation shows, that the molecules adsorb to
> surface nevertheless (as one would suggest from pbc). Of course I can
> reconfigure my .gro for visualization by substracting or adding the box
> vectors untill all atoms end up inside the 'real' simulation box, but is
> really supposed to behave that way, or is there something definitely wrong
> with my structure?
> I saw this "shifting" of the whole substrate into another periodic image
> the simulation box also in the simulation with the silicon substrate: it
> 'jumped' into the periodic cell and then jumped back one frame later.
> The other present (small) molecules stayed in the real simulation box all
> time. I guess that this is maybe a problem caused by the substrate
> filling up the whole simulation cell (at least in xy), but is there a way
> getting Gromacs to always write out coordinates that lie within the real
> simulation cell instead of an periodic image?
> And is there a way of defining bonds over the periodic boundaries? When I
> tried this it always seemed as if Gromacs interpreted these bonds as
> overstretched bonds (that are stretched over the whole simulation box
> instead of just reaching the "short way" over the periodic boundary)
> crippled structures ( http://tinyurl.com/54uvs4 ).
> All simulations used the ffG43b1 force field. Cutoffs of 1.2 and an Rlist
> 1.4. Coulomb was shifted from 0.0 and VdW from 0.9. Timestep was 2 fs and
> = xyz. The boxes were large in z direction to generate xy pbc. I can also
> the topologies online if someone would like to look at them.
> I hope anyone can give me some advice because although I tried a lot of
> right now I feel not like coming close to a solution for my problems. :(
> Best regards,
> gmx-users mailing list gmx-users at gromacs.org
> 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
Prof. Dr. André Farias de Moura
Departamento de Química
Universidade Federal de São Carlos
São Carlos - SP - Brasil
More information about the gromacs.org_gmx-users