[gmx-users] Re: Re: thermostate and frozen atoms

m b mic0404 at yahoo.com
Tue Jul 29 04:19:00 CEST 2003

> > dear all
> > if I freeze atoms grompp wouldn't
> > let me use a thermostate (Nose Hoover) only
> > for the non-frozen atoms
> > If I include the frozen atoms as a seperate
> > group for temperature scaling my system explodes.
> Which part explodes ? Did you minimize ? Any bad
> contact ? Any equilibration (with constraints) ?
> XAviere

I simulate a polymer above a crystal slab.
the initial structure is perfectly alright.
When I use the first of the two input scripts
given below (md.mdp 1, where only a small part of the
crystal slab (the "frozen" group) is frozen) 
the system behaves well, both the LJ and the 
Coulomb energy are negative at t=0.
If I use the second one (md.mdp 2, where the
entire crystal slab is frozen and only the
polymer is allowed to move) both ener.edr and md.log
say (for the SAME input structure and topology)
that the kinetic energy is zero and the temperature
is nan (from t=0 until the simlation crashes
after about 150 steps)
Here the potential energy of the initial structure
is negative as well and stays so until the end
while the intra-molecular energy terms increase
steadily. After a short while the angles and bonds 
become heavily distorted (also to be seen with
ngmx) and then lincs freaks out and the system blows

(The same thing happens if I use only one
tc-grp, namely the whole system or if I
set the target temperature for the 
frozen group to zero)

A while ago there was a discussion in this
group about freezing atoms in combination with
pressure coupling with the conclusion that this
was probably not a good idea. However, temperature
coupling (of the non-frozen part) and frozen atoms
should be alright if, for now, we forget about some
"physical" inconsistencies mentioned in another
But apparently it does NOT work, or there is
something I am missing here ...
The root of the problem might be that i HAVE to
include the frozen group as tc-group (otherwise
grompp gives me an error)

another point: in a mail in the archive
somebody states that if one freezes a group
of atoms one has also got to zero the COM motion of
this group otherwise the group would move
as a whole (concerted, a "flying ice-cube")
is this true ?? it seems somewhat counter-intuitive
that atoms can still move, albeit in a concerted 
fashion, after having frozen their movements
explicitely in ALL directions (freezedim = Y Y Y)

looking forward to any comments!

md.mdp 1, only small part of the crystal frozen ...

title                =  flexible xtal
cpp                  =  /usr/bin/cpp
integrator           =  md
dt                   =  0.002
nsteps               =  1000
nstcomm              =  1000
comm_grps            =  xtal
nstxout              =  1
nstvout              =  0
nstfout              =  0
nstlog               =  10
nstenergy            =  10
constraints          =  hbonds
constraint_algorithm =  lincs
unconstrained_start  =  no
lincs_order          =  4
energygrps           =  xtal p01 p02 p03 p04 p05 p06
p07 p08 p09 p10
freezegrps           =  frozen
freezedim            =  Y Y Y
pbc                  =  xyz
coulombtype          =  shift
vdwtype              =  shift
rcoulomb_switch      =  1.3
rvdw_switch          =  1.3
rcoulomb             =  1.5
rvdw                 =  1.5
rlist                =  1.3
nstlist              =  10
Tcoupl               =  nose-hoover
tc-grps              =  xtal polymer
tau_t                =  0.1 0.02
ref_t                =  300 300
Pcoupl               =  no
gen_vel              =  yes
gen_temp             =  300.0
gen_seed             =  1799

md.mdp 2, the whole crystal is frozen ...

as above, but ...

comm_grps            =  xtal

(the "xtal" group is the whole crystal slab, the
"frozen" group is a small portion of this slab)

Do you Yahoo!?
Yahoo! SiteBuilder - Free, easy-to-use web site design software

More information about the gromacs.org_gmx-users mailing list