Coulombic energy comparsion between same model with different bond terms

LAMMPS LAMMPS General Discussion
1

Dear developers and users,

I am comparing the same MOF model with different bond energy terms.

in the first case, there are 4 types of LJ parameters and 6 types of bond terms. the energy calculation was implemented in LAMMPS and also validated using GULP. Both the interatomic terms and coulombic energy are well-matched in the two software.

in the second case, I only removed one bond term (i.e., Cu-Cu bond term that in reality does not exist), and kept others the same as in the first case. The interatomic terms were well-matched with the one from GULP, but the coulombic energy differs by 873.68 KJ/mol between the two programs. and the one from gulp is similar to the first case, which I think makes sense since both the framework and charges were kept the same for the 2 cases.

here is the energy comparison for 2 cases by LAMMPS, the coul energy differs by ca. 200 kcal/mol when the framework and charges are almost same:

image
image1100Ă—268 16.2 KB

so I think the problem probably occurs in my lammps input file. Only removing one bond term caused such a large difference in coulombic energy makes me confused.

here is the comparison between lammps and gulp for 2 cases:
4 LJ+6 bonds:

image
image1277Ă—262 21.2 KB

4 LJ+5 bonds:

image
image1217Ă—266 21 KB

can anyone spot what’s wrong with my calculation? thank you in advance.
4lj+6bonds.Cu-Cu (45.8 KB)
4lj+5bonds.no_Cu-Cu (60.0 KB)
in.4lj+5bonds (569 Bytes)
in.4lj+6bonds (568 Bytes)

Best regards,
wang

2

You have a different total number of bonds in the two data files: 792 vs 768.
With the default settings for “special_bonds” all 1-2, 1-3, and 1-4 pairs in the bond topology are excluded from non-bonded interactions, i.e. LJ and Coulomb. More bonds leads to more excluded non-bonded interactions.

This is very typical for most molecular force fields (there are differences in the 1-4 terms, some force fields keep them with a scaling factor). Only simplified models like bead-spring models exclude only the bonded terms, and some coarse grain models fully include non-bonded terms (which makes their bond terms more difficult to parameterize since you need to subtract out the nonbonded contribution). The Coulomb interactions are usually much stronger than LJ terms, so you would see a larger difference there.

1 Like