[gmx-users] L-BFGS energy minimization wrong!

[=?GB2312?B?U3SopnBoYW5lIFSopmxldGNoqKZh?=]

zzhwise1 a =A8=A6crit :
> hi everyone !
>    i have ask some questions about energy minimization before ,but=20
> still,there are problems !this time it show:
> -------------------------------------------------------                =
   =20
> Program mdrun, VERSION 3.3.1                                           =
   =20
> Source code file: nsgrid.c, line: 226                                  =
   =20
>                                                                        =
   =20
> Range checking error:                                                  =
   =20
> Explanation: During neighborsearching, we assign each particle to a gri=
d  =20
> based on its coordinates. If your system contains collisions or paramet=
er =20
> errors that give particles very high velocities you might end up with s=
ome=20
> coordinates being +-Infinity or NaN (not-a-number). Obviously, we canno=
t  =20
> put these on a grid, so this is usually where we detect those errors.  =
   =20
> Make sure your system is properly energy-minimized and that the potenti=
al =20
> energy seems reasonable before trying again.                           =
   =20
>                                                                        =
   =20
> Variable ci has value -2147483648. It should have been within [ 0 .. 25=
0 ]=20
> Please report this to the mailing list (gmx-users at gromacs.org=20
> <mailto:gmx-users at gromacs.org>)            =20
> -------------------------------------------------------                =
   =20
>                                                                        =
   =20
> and my gro has no problem!my mdp is :
> title               =3D lbfilm, 300K               =20
> cpp                 =3D  /bin/cpp                  =20
> define              =3D                            =20
> constraints         =3D  none                      =20
> integrator          =3D  l-bfgs                    =20
> ;dt                  =3D  0.002 ; ps !             =20
> nsteps              =3D  5000 ; total 60 ns.       =20
> nstlog              =3D  1                         =20
> nstenergy           =3D  1                         =20
> emtol               =3D  10.0                      =20
> emstep              =3D  0.01                      =20
> ;nstcgsteep          =3D  10                       =20
> niter               =3D  20                        =20
> nbfgscorr           =3D  10                        =20
> nstlist             =3D  1                         =20
> ns_type             =3D  grid                      =20
> pbc                 =3D  xyz                       =20
> coulombtype         =3D  PME                       =20
> rlist               =3D  1.1                       =20
> rcoulomb            =3D  1.1                       =20
> vdwtype             =3D switch                     =20
> rvdw-switch         =3D 0.9                        =20
> rvdw                =3D 1.0                        =20
> Dispcorr            =3D EnerPres                   =20
>                                                  =20
> my system is two parell monolayes  and the distance is 1.5 nm !
> could any one give me any advice?
> thanks!
> =20

Hi,

I would avoid the switch function for calculating Van Der Wall
interactions since it seems not appropriate (see Van Der Spoel and van
Maaren, JCTC, 2006, 2, 1-11).

Try using regular vdw radius of 1.4 nm.

I would use:

coulombtype         =3D  PME
rlist               =3D  0.9
rcoulomb            =3D  0.9
rvdw                =3D  1.4


Cheers,

St=A8=A6phane

--=20
St=A8=A6phane T=A8=A6letch=A8=A6a, PhD.                  http://www.stele=
tch.org
Unit=A8=A6 Math=A8=A6matique Informatique et G=A8=A6nome http://migale.jo=
uy.inra.fr/mig
INRA, Domaine de Vilvert                  T=A8=A6l : (33) 134 652 891
78352 Jouy-en-Josas cedex, France         Fax : (33) 134 652 901