Incorrect potential energy of a Cu-C dimer with a "hybrid" potential

Hi, lammps developers and users,

I want to combine an EAM potential, LJ potential and Reaxff potential to describe the Cu-C system (the EAM potential for Cu-Cu interactions, the LJ potential for Cu-C interactions and the Reaxff potential for C-C interactions) using the “hybrid” method. To test the LJ potential, a Cu-C dimer with the distance of 3.3 Å was set in a box; however, I found that in that situation the potential energy of the Cu and C atoms are not equal and the total potential energy is not equal to that value we calculated manually (shown in Table 1). I don’t know why? All the input files for the test are also sent to you as attachments, would you please help me solve these problems ? I am looking forward to your reply, thank you very much.

Question-11.18.docx (20.5 KB)
CuC_dimer.dat (244 Bytes)
Cu_mishin1.eam1.alloy (690.0 KB)
ffield.reax.CHNi (5.8 KB) (879 Bytes)

You have a flawed model. Consistent use of ReaxFF parameterization usually requires to use it for all atoms. For your setup you are crucially missing the parameters for the charge equilibration of the non-ReaxFF atoms. A more consistent approach would be to use AIREBO instead (which can model Carbon quite accurately, if not better).

I don’t think the energy you see is incorrect. Most likely you are not considering the embedding energy of EAM and the Coulomb and atom/self energy of ReaxFF.

You can use the compute pair command to identify the contributions from the different pair styles.

1 Like

Thank you very much, in my opinion, the potential energies of the Cu-C dimer and the atoms in it are only determinated by the LJ potential, because the Cu-C interaction is described only by the LJ potential and there are no other same type atoms influencing the potential energy of the Cu and C atoms. According to your advice, we confirm that the EAM and the Reaxff potential also contribute to the potential energy of the Cu-C dimer, but we can not furtherly determine the contributions from which part of the EAM and Reaxff potential.

  • Opinions don’t count. What matters are the physics as described by the potential functions.
  • Those potential functions do have terms that result in a per-atom potential energy beyond the LJ contributions.
  • ReaxFF and EAM are manybody potentials, so you can not apply the logic for pair wise additive potentials. The compute pair command actually allows to determine the contributions from the individual terms of the ReaxFF potential.
  • The absolute value of the potential energy has no meaning in classical models anyway. What matters are potential energy differences.