[gmx-users] program crashes when -table option used

[Dongsheng Zhang]

Dear GMX users,


I have a coarse-grained model whose nonbonded potentials are tabulated,
so I have to use user defined potential in gromacs. With Yang Ye's great
help, I can implement the new force field. However, the program always
crashed because of Large VCM(group rest). I have done a test by run the
program in two different ways. My new force field uses gromos FF format,
so I used c6 and c12 parameters.


First I ran the program in a normal way, which means I use LJ for
regular VDWs and 1-4 interactions. The program ran fine.

Second I copied table6-12.xvg to my working directory and rename it to a
proper name, then I used this table as my user-defined potential for
VDWs and 1-4 interactions. 

I expect the above two ways give me very similar results. This is true
for the first 51 steps. In step 52, the second run suddenly crashed
because of Large VCM(group rest).

The details are as following:

The differences in two mdp files:

59c59
< vdw-type                 = user  ;cut-off
---
> vdw-type                 = cut-off;  user  ;cut-off

The information from two log files:

full2.log (first run):

           Step           Time         Lambda
             51        0.10200        0.00000

   Energies (kJ/mol)
           Bond          Angle          LJ-14     Coulomb-14        LJ
(SR)
    3.02207e+02    3.00007e+02   -3.73112e+01    0.00000e+00   -2.27638e
+02
   Coulomb (SR)      Potential    Kinetic En.   Total Energy
Temperature
    0.00000e+00    3.37264e+02    1.01670e+03    1.35397e+03    3.09963e
+02
 Pressure (bar)
    4.78614e+00

           Step           Time         Lambda
             52        0.10400        0.00000

   Energies (kJ/mol)
           Bond          Angle          LJ-14     Coulomb-14        LJ
(SR)
    3.00485e+02    2.99534e+02   -3.73372e+01    0.00000e+00   -2.27861e
+02
   Coulomb (SR)      Potential    Kinetic En.   Total Energy
Temperature
    0.00000e+00    3.34820e+02    1.01849e+03    1.35331e+03    3.10508e
+02
 Pressure (bar)
    4.73457e+00

           Step           Time         Lambda
             53        0.10600        0.00000

   Energies (kJ/mol)
           Bond          Angle          LJ-14     Coulomb-14        LJ
(SR)
    2.98871e+02    2.99071e+02   -3.73627e+01    0.00000e+00   -2.28077e
+02
   Coulomb (SR)      Potential    Kinetic En.   Total Energy
Temperature
    0.00000e+00    3.32503e+02    1.02011e+03    1.35262e+03    3.11004e
+02
 Pressure (bar)
    4.67219e+00




full.log (second run):

           Step           Time         Lambda
             51        0.10200        0.00000

   Energies (kJ/mol)
           Bond          Angle          LJ-14     Coulomb-14        LJ
(SR)
    3.02206e+02    3.00007e+02   -3.73112e+01    0.00000e+00   -2.27636e
+02
   Coulomb (SR)      Potential    Kinetic En.   Total Energy
Temperature
    0.00000e+00    3.37266e+02    1.01670e+03    1.35397e+03    3.09963e
+02
 Pressure (bar)
    4.78604e+00

           Step           Time         Lambda
             52        0.10400        0.00000

Large VCM(group rest): -362781422043267072.00000,
46998104838242304.00000, -95810352322707456.00000, ekin-cm:          inf
   Energies (kJ/mol)
           Bond          Angle          LJ-14     Coulomb-14        LJ
(SR)
    3.00484e+02    2.99534e+02   -9.09769e+31    0.00000e+00   -2.27860e
+02
   Coulomb (SR)      Potential    Kinetic En.   Total Energy
Temperature
    0.00000e+00   -9.09769e+31            inf            inf
inf
 Pressure (bar)
            inf

           Step           Time         Lambda
             53        0.10600        0.00000


Compare step 51 in two cases, every item has very similar value. In step
52, the second run give me very large value of LJ14. This causes the
problem. However I compare the output xtc files at step 52 i those twp
runs. They are identical. It really surprises me that LJ-14 is so large
in the second run.


Any suggest to find what's wrong will be highly appreciated. If you need
more information, please let me know.

All the best!

Dongsheng