[gmx-users] confusion on NPT MD simulation

[Ming Tang]

Dear gromacs experts,

I tried to equilibrate a triple helix using NPT, and the .mdp file is like this:

DEFINE           =  -DPOSRES
integrator       =  md
dt               =  0.0009
nsteps           =  1000000
nstxout          =  0
nstvout          =  0
nstlog           =  10000
nstxtcout        =  10000
xtc-precision    =  10
cutoff-scheme    =  verlet
coulombtype      =  reaction-field-zero
coulomb-modifier =  potential-shift-verlet
rcoulomb-switch  =  0.8
rcoulomb         =  1.4
epsilon_r        =  15
vdw-modifier     =  force-switch
rvdw-switch      =  0.8
rvdw             =  1.4
tcoupl           =  v-rescale
tc-grps          =  Protein non-protein
tau-t            =  1.0 1.0
ref-t            =  310 310
Pcoupl           =  parrinello-Rahman
Pcoupltype       =  isotropic
tau-p            =  5
compressibility  =  3e-4
ref-p            =  1.0
refcoord_scaling =  all
pbc              =  xyz
freezegrps       =  C
freezedim        =  Y Y Y

C is a group containing atoms of both the first and the last residues in all the three chains.
But after 18ns' simulation, I found the one end of the triple helix moved. Could you help to tell me whether this is normal or there is something wrong with my simulation?
Then, I tried to use berendsen for both tcoupl and pcoupl. It turned out that both ends were fixed.  I need to keep the original geometry shape of the triple helix, but berendsen algorithm is believed to be inaccurate.

Thanks.