[gmx-users] Re: compressing a system
Justin A. Lemkul
jalemkul at vt.edu
Sun Mar 20 23:44:43 CET 2011
> On 20 March 2011 15:23, Justin A. Lemkul <jalemkul at vt.edu
> <mailto:jalemkul at vt.edu>> wrote:
> Moeed wrote:
> Dear experts,
> I am trying to build up a polymer in hexane system by increasing
> the density and use this for FE calculation. After PR step, my
> NVT and NPT trailes failed. Initially I used to get LINCS and
> 1-4 warnings (even for NVT) which were not because of flawed
> topology file. It turned out that simulations crashed just
> because of using -np > 4. But still even with this -np, NPT did
> not work which made me to swtich to berendsen from parrinelo
> rahman scheme. As I approched the desired density again
> simulation crashed, so I used
> trjconv -s .tpr -f .trr -o frame2300.gro -dump 2300
> to extract one of the frames before crash then I did another NVT
> to equilibrate.
> mpirun -np 4 mdrun_mpi -deffnm PE60-110Hex-NPT-frame2300_md
> after around 1 ns I get the error below ( mdp file is also
> included). I described above since I encountered the situation
> where root cause of problem was not topology and just the
> computational issue ( I mean -np), I am just curious if the same
> thing applies here. So I have two concerns:
> 1- Crash at the early stages of run ( topology is working fine
> for a single hexane, so this can not be the reason). only works
> only for -np =<4.
> 2- Crash when compressing because chain crashes on itself. ( I
> have no control on this because chains takes different shapes
> and becomes extended)
> can you please comment on the proper settings for compressing
> the system.
> more details: There is only one polyethylene chain ( 60 units)
> in 110 hexane ( box size 20 nm at the beginning) The chain is
> not convoluted and has a little extended shape that crashes on
> itself for long runs.
> This is likely the problem. The compression of the system is
> occurring too fast for the system to adequately respond. If the PE
> chain is fully extended, any significant change in the box size
> (which is mostly a function of the solvent, in this case, hexane)
> will case PE to interact via PBC and collapse.
> Mark: Why aren't you using genbox on your polymer starting
> configuration and an equilibrated box of hexane of the right
> me: Hi, I need to work with polymer 15 Wt% system, so I manually
> calculated how many hexane I need to get this concentration. So
> for a single PE chain of 60 units I need 110 hexane. Here is the
> The treatment of your system is also problematic. You need a
> certain composition, but a straight PE chain with only 110 hexane
> will probably respond very drastically to any non-equilibrium
> conditions. You need to generate a more sensible configuration for
> the PE chain such that it does not collide with itself. I would
> suspect that a more compact PE chain in conjunction with a hexane
> box with a density closer to your target would (1) eliminate the
> need for compression and (2) be more stable.
> Thank you.
> 1- So You mean I should compress a chain separately and make a hexane
> box and combine these two? How is this possible to insert a solute in a
> prepared solvent box?
Likely you shouldn't have to compress anything. If your PE + hexane system
crashes because you're squeezing it, a PE-only system will do the same. A
configuration of PE generated from an in vacuo simulation, or one resulting from
a simulation in a larger box of hexane should suffice.
You would construct a system like you would almost any protein/water system -
take your PE, define a box for it, and fill it with hexane using genbox.
> 2- As I compress the system (NPT equilibrated to density of 0.62 g/l)
> chain takes a realistic shape, which is not extended. NVT on this
> configuration sometimes cause the chain to a more straight chain.Now I
> extracted a frame from NPT a little before equilibration (density 0.5,
> and actually this is the density I want) and equilibrated by NVT and no
> crash happened. As I asked in previous message I only want to assure if
> I can extract a frame ( even density is not equilibrated) followed by
> NVT to equilibbrate and use this as production runs.?
I see no problem with that. Equilibrate until you reach the conditions you
want, then proceed with MD under the statistical mechanical ensemble that makes
the most sense. That's the same protocol that just about everyone uses for
Justin A. Lemkul
ICTAS Doctoral Scholar
Department of Biochemistry
jalemkul[at]vt.edu | (540) 231-9080
More information about the gromacs.org_gmx-users