[gmx-users] harmonic angle potential

[Michael Patra]

Dear all,

I wounder whether the harmonic angle potential is implemented correctly in
Gromacs. Since the potential is supposed to depend only on the angle, the
computed forces should only try to change that angle. The Gromacs angle 
potential, however, seems to also try to change the bond length. This is seen
when then bond potential is switched off. (For technical reasons, perhaps some
bug, mdrun crashes when evaluating an angle when there are no bonds, so one has
to introduce bonds with a vanishing potential.) One will then see that the
particles connected by the angle are pushed away from each other.

At the end of this message, you will find a short script that will reproduce
this behaviour with Gromacs 3.1.4.

Yours faithfully,

- Michael Patra


#!/bin/sh
cat > grompp.mdp << \EOF
integrator = md
nsteps = 10000
EOF
cat > topol.top << \EOF
[ defaults ]
  1             1               no              1.0     1.0

[ atomtypes ]
    C     1.000      0.000     A  0.0 0.0    ; no vdW interaction
 
[ angletypes ]
 C C C 1 120.0 1.0

[ bondtypes ]
 C C 1 0.5 0.0               ; dummy bond without potential to make mdrun happy
  
[ nonbond_params ]
    C C  1 0.0 0.0

[ moleculetype ]
Carbonstuff 0

[ atoms ]
1       C     1       POL     P1      1       
2       C     1       POL     P2      2       
3       C     1       POL     P3      3       

[ bonds ]
1 2 1
2 3 1

[ angles ]
1 2 3 1

[ system ]
Test of angle routines

[ molecules ]
Carbonstuff 1
EOF
cat > conf.gro << \EOF
Test system
    3
    1POL     P1    1   5.000   4.000   5.000
    1POL     P2    2   5.000   5.000   5.000
    1POL     P3    3   4.000   5.000   5.000
  10.00000  10.00000  10.00000
EOF
grompp
mdrun -v
ngmx