Bonded atoms far aprt in Coarse-grained simulation

Yes. There is a singularity in the dihedral forces whenever the bond angle between any 3 of them approaches 180 degrees. So you cannot use traditional dihedral forces in this case (even if you use a small timestep). The “spherical” dihedral style was designed to solve this problem. For details see:

Note: If the equilibrium rest angle is close to 180 degrees, you may need to use a more complicated expansion than the simple 3-term example shown on that web page, but any periodic function of phi, theta1, and theta2 can be approximated using this dihedral style, and you do not need to define separate angle interactions between atoms 1,2,3 and 2,3,4. (However you must insure that the bond distances always remain reasonably large, preferably > 1 Angstroms.)


P.S. You will have to make sure you compiled LAMMPS with the USER-MISC package installed. For details about that, see: