[gmx-users] Minimization stops without reaching the requested precision Fmax < 1000

[Mahdi Bagherpoor]

Dear Gromacs Users,

I am trying to add the tetrahedral zinc atom dummy model into CHARMM36
force field in Gromacs. In my system, this dummy zinc interact with four
Cystein residues (CYM).  The parameters of the zinc are available in CHARMM
format, which I used to use it in NAMD package. After adding appropriate
parameters, when I do energy minimization, I get this error:
========================

Energy minimization has stopped, but the forces have not converged to the
requested precision Fmax < 1000 (which may not be possible for your system).
...
Steepest Descents converged to machine precision in 1045 steps,
but did not reach the requested Fmax < 1000.
Potential Energy  = -1.5922652e+06
Maximum force     =  3.2106902e+07 on atom 3526
Norm of force     =  1.4857810e+05

============================
Atom 3526 is dummy atom of zinc ion. The dummy parameters added in
ffbonded.itp file are:
---------- ffbonded.itp

;zinc dummy model, bonds
      HK       ZD     1   0.09000000    451872.00
      HK       HK     1   0.14700000    451872.00
;zinc dummy model, angles
      HK       ZD       HK     5   109.500000   460.240000   0.00000000
    0.00
      HK       HK       HK     5    60.000000   460.240000   0.00000000
    0.00
      HK       HK       ZD     5    35.250000   460.240000   0.00000000
    0.00
;zinc dummy model, dihedrals
      ZD       HK       HK       HK     9    35.300000     0.000000     2
      HK       ZD       HK       HK     9   120.000000     0.000000     2
      HK       HK       HK       HK     9    70.000000     0.000000     2

----------- ffnonbonded.itp

;zinc dummy model
   ZD    30    61.380000    0.000  A  0.310000000000  0.00001
   HK     1     1.008000    0.500  A  0.000000000000  0.00000

=================================================
In contrast to Gromacs, when I do minimization in NAMD, there is no such an
error in minimization. I should mention that I am using single precision
platform of Gromacs.
I would appreciate if you could share me in case you have some ideas about
this problem.

Cheers,
Mahdi