Recently, I have been using the Gay-Berne potential to simulate the mechanical and wetting properties of clay particles, and two articles are referred, J. Chem. Phys. 140, 154309 (2014) and Applied Surface Science 532 (2020) 147423. The Gay-berne potential parameters in J. Chem. Phys. 140, 154309 (2014) are obtained by fitting the free energy curve between two clay particles, which is computed by FEP method. The Gay-berne potential parameters are listed as figure1 below. And I set the force field parameters in LAMMPS as “pair_style gayberne 1 1 2 125” and “pair_coeff 1 1 65.53 5.5 0.4283 0.4283 0.003656 0 0 0 125”. The shape of the ellipsoid particle a = b = 10.412 nm, c = 0.962 nm. The initial structure is relaxed by “fix nve/asphere” with timestep 40fs. And a “fix npt/asphere” is followed. The simulation details can be found in the input file, which has been uploaded. The simulation results obtained using the aforementioned simulation settings are unphysical and unreasonable. Some ellipsoidal particles are intersecting with each other and not exhibiting any repulsive forces as shown in figure2. I have tried many methods to fix these unreasonable results, but none of them have been effective. This issue has been troubling me for a long time, and I would greatly appreciate any guidance from fellow colleagues. Your assistance would be highly appreciated.

There are various problems with your input script, but the most important is how the parameters of the GB potential are defined. From the table you posted, I would have used

I have no time to read the papers you suggested, but the first step is to compare the results of new simulations with the published ones.
Could you verify what is the effect of the pressure on the properties of the platelets? I have started a thread on the computation of pressure for atom_style ellipsoid, it would be good to report the outcome of the attached input file, where I have made a few changes for you to review. test02.in (1.9 KB)

Thank you for your patient and professional response. My question has been resolved satisfactorily. Please forgive me for bothering you again. I am still very confused about the gayberne potential. The explanation of the epsilon parameter in the LAMMPS manual is “well depth (energy units)”. The explanation of epsilon_a, epsilon_b, and epsilon_c is “relative well depth of type I for side-to-side/face-to-face/end-to-end interactions”. I am very confused about the reference depth for the relative well depth mentioned here. Is it relative to the value of epsilon (absolute_epsilon_a/epsilon)? The article provides the depth of the free energy potential well of face-to-face and edge-to-edge interactions, as shown in Figure 4. According to the information given by the free energy barrier, the depth of the face-to-face potential well is approximately 580 kcal/mol, and the depth of the side-to-side potential well is approximately 78 kcal/mol. Should I set epsilon to 580 kcal/mol, epsilon_a and epsilon_b to 78/580, epsilon_c to 1? Am I understanding it correctly?

Thank you once again for your patient response and guidance.

Yes, correct. I personally find the notation for the pair_coeff to be cumbersome, but perhaps this choice could have originated for the need to mix different pair styles. Anyway, my advice is to define the IJ pair coefficients only for I=J, as you did. And remember that the \epsilon_a,\epsilon_b,\epsilon_c are the relative well depth energies.

PS I am not a professor. Just a random guy with an internet connection

Thank you for your enthusiasm and patience. My question has been resolved. Furthermore, you will undoubtedly become an exceptional professor in the future. I wish you all the best.

Please share the results of the simulations on platelets. The question of how to compute the pressure for finite-size particles has popped out repeatedly. The more comparisons we have with independent simulations (especially Monte-Carlo), the better we can make a case for settling this topic. Cheers!

With the patient guidance of Hothello, I obtained seemingly correct results in platelet simulation. Now, I am attaching the modified input file, as well as the obtained LAMMPS output results for everyone’s reference. Of course, there are still many issues involved, and I welcome criticism and corrections.