[gmx-users] LINCS warnings at high temperature md run

[Sidong Tu]

Dear all,

I was trying to simulate lysozyme and polymer in water model tip3p using
charmm36 force field at 500K on Gromacs 5.1.2. The software ran on GPU and
the system size was about 60,000 atoms. I have done equilibrium with
Berendsen(B) and Parrinello-Rahman(PR), and then md. The thermal coupling I
used was v-rescale.
I always got a LINCS warning and my job stopped during the md. Here's the
parameter I have used:
Stage  Position_Constraint  Pcoupl  gen_vel  LINCS       dt
nsteps
1         Protein & polymer    B           yes     all bonds    0.002
200,000
2         Protein                    B           yes     all bonds
0.002    200,000
3         None                       B           yes     all bonds
0.002    200,000
4         None                       PR         no       all bonds
0.002    50,000
5         None                       PR         no       h bonds
0.002    1.25e+8

At stage 5, the md run stopped at 149 steps and showed relative constraint
deviation after LINCS.

I have searched for this warning and there is an explanation that it might
be introduced by the bad equilibrium. Then I added a equilibrium stage with
Berendsen pressure coupling after stage 3 with LINCS constraints of h bonds
for 50,000 steps without position constraints.

This time there was still warings with LINCS and it stopped at step 79026
in md:

Step 79026, time 158.052 (ps)  LINCS WARNING
relative constraint deviation after LINCS:
rms 15.598468, max 1553.834961 (between atoms 15966 and 15968)
bonds that rotated more than 30 degrees:
 atom 1 atom 2  angle  previous, current, constraint length
  15941  15943   40.7    0.2082   0.1109      0.1111
  15945  15946   75.6    0.2551   7.7014      0.1111
  15945  15947   90.0    0.2299   4.2042      0.1111
  15949  15950   67.8    0.1543   0.1150      0.1111
  15949  15951   86.2    0.1658   8.9758      0.1111
  15966  15967   90.0    0.1111   0.2587      0.1111
  15966  15968  143.1    0.1111 172.7422      0.1111
  15970  15971   90.0    0.1111   0.1405      0.1111
Wrote pdb files with previous and current coordinates

The simulation worked well with dt=0.001 without LINCS. But I still want to
find if there is a way to simulate at high temperature(500K) with dt=0.002
so that I can use less steps?


-- 
Sidong Tu
Department of Materials Science and Engineering
Clemson University
Phone: (213)265-6768