Tabulated Dihedral Potential

GROMACS version: 2019.6
GROMACS modification: Yes/No
Here post your question
I perform coarse-grained simulations of a small isotactic polypropylene system of 40 chains and 50 beads per chain. For the dihedral potential, I am using a tabulated potential that has been extracted from united atom simulations.
As I read in the manual, the function type for dihedrals in the top file is 8 with table number 0 and k equals 1. Moreover, I supply a table file table_d0.xvg to mdrun which includes x,f(x),-f '(x).
The simulation runs properly (NPT for 10ns). Moreover, the dihedral distribution is similar to the one from the united atom simulation. The dihedral energy stabilizes after some steps
Though, I receive a warning in the output file: " For the 359 non-zero entries for table 0 in table_d0.xvg the forces deviate on average 119% from minus the numerical derivative of the potential"
Any insights for this warning?
Thank you in advance.
table_d0.xvg (15.5 KB)

The problem is that your potential and derivative are very noisy. I don’t know how you computed the derivative. This leads too large noise on the forces which is bad. You should always smooth a potential that has been obtained from noisy statistics and then take the gradient of the smoothed potential.