Can you help me？
Please look at this discussion on how to compute the intermolecular potential energy. For a molecular system described by a type-I force field, the cohesive energy corresponds to the pairwise intermolecular energy [citation needed]. Including the contribution of long-range electrostatics is tricky, and, to me, contradictory advice has been given on the LAMMPS mailing list.
From the discussion I referred to, it seems that you can compute the intermolecular energy as a variable:
compute intra all group/group all pair yes kspace yes molecule intra
variable enb equal evdwl+ecoul+elong
variable inter equal v_enb-c_intra
Then, you normalise it by volume, molecules, or atoms, depending on your needs.
Could you please report your findings and any comparison with reference data? That would be very useful for the community (and for me to finish this work!).
1 Like
Thank you, I also read the original post and it was very well written! 
Hi Roscioni,
I do not think the scheme you posted is correct to compute the intermolecular potential energ (IMPE). Because this compute dose not exclude the kspace energy of the intermolecular.
compute intra all group/group all pair yes kspace yes molecule intra
So this inter is actually the pair energy of intermolecular, but the kspace energy of intermolecular is not considered. I augument it it’s going to be a big bias when the kspace energy palys an important role in some molecular systerm.
variable inter equal v_enb-c_intra
Hi @LTHorizon, thank you for the feedback. Could you provide an example that backs your claim? I have been thinking about this issue for quite a long time and the approach I suggested is for me consistent with the way the Coulomb potential energy is computed with a long-range solver. But I am happy to be proven wrong 
Hello, do you know how to use LAMMPS to calculate cohesive energy
As you can see from this discussion and the post linked in my first comment, the topic is still debated.
The “easy” way to solve it is to write the equations of the energy terms, including the long-range correction to the Coulomb energy, and cross-check those with the source code of compute group/group.
The poor man’s way is to find some robust reference data and reproduce them with LAMMPS using one of the many methods suggested. If you go this way, please report your findings in a new post.
hi，I discussed this in a new post