[gmx-users] Re: compressing the system
vvchaban at gmail.com
Thu Jul 29 07:38:48 CEST 2010
> I tried to get the same results as you got for density and total energy
> starting from the 8 chains is the box. I used the revised mdp file and since
> PME option takes much time I used shift instead for a moment to verify if I
> can get the density you obtained. Also I used constraints for bonds. density
> I get after 1800ps is 850 kg/m3 while what you got is a slightly different
> (after 1000ps density varies around 880!). Also the total energy you have is
> bout 2300 whereas I get somethong around 12000!. I dont know why they are so
Different methods for electrostatics give very different absolute
energies as a result. C'est la vie...
> In the second trial I used "constraints = none" just the "shift"
> for electrostatics is not the same as the mdp file you suggested. again
> results are not the same for total energy (20000 Kj/mol). Below is the
> average values...also simulation crashes at 1550 ps and the last density I
> am getting is 834! still less than what you got. dont really know why
> results are not the same.
Because the methods are not identical. Please read about
electrostatics treatment to understand this issue.
>The reason simulation crashes could be due to
> dt=0.002 while removing constraints?
It can be due to C-H fast oscillations, for example. Maybe it is a
good idea to try SHAKE for these bonds if the time-step is so critical
> 1- What I see is that the system is getting compressed to a size of around 3
> 3 3 nm is the first 20ps abruptly and in the remainder of time system
> reaches equilibrium... This is very different approach than what I was
> trying doing before. I used berendsen scheme for pressure coupling and box
> reduced is size slowly but I had difficulty reaching the size I want and
> most of the simulations crashed. Does this mean berenden is not suitable for
> compressing system for all systems?
I did not ever observe that. Please also note that I used
semiisotropic pressure coupling for your system because of its
specific [rather anisotropic] structure with long molecules. I suppose
this option should provide more realistic density in your particular
> 2- You said I dont need NVT at all. But if I want to study the system I get
> from NPT giving the density I want, can I not do NVT for a fixed system size
> and calcualte energies or other quantities I am after?
Of course, you can. This is the only way to have a constant density.
More information about the gromacs.org_gmx-users