Copper - Borazines potential

Good morning,

I am new to LAMMPS (and in general MD) and I would like to simulate the clustering of some borazine molecules on Copper (111). Following the suggestions on the forum, I have searched for some papers doing the same thing finding just: https://onlinelibrary.wiley.com/doi/abs/10.1002/anie.201300948
In the paper, the potential used is said to be: F. Iori et al, GolP: An atomistic force-field to describe the interaction of proteins with Au(111) surfaces in water, J. Comput. Chem. 30, 1465 (2009). However, it seems to me that in this paper the only interaction parametrized are only N and S, in contrast with C/B/N interactions that I should have.

Am I missing something on the potential construction on LAMMPS? Does someone have some other experience with this kind of system and have some suggestion for me? Thanks!

Laura Caputo

Laura,

if you are new to MD, then you should not start with such a project. This requires an experienced MD simulation practitioner. Such a person would have seen that the MD simulations described in the publication you cite are using not very high accuracy models. 1) rather than using specific parameterizations for borazine compounds, they use the UFF generic force field (you can do it, but its accuracy is limited) and 2) they use a rather simplified model to represent the “metallic” character of the surface, which is called the “image charge model”. For this model you just represent the (nearly ideal) polarizability of the metal surface by adding “mirror charges” after defining a material specific “mirror plane” position. In the given example the authors argue that the adsorption energy of copper is similar to that of gold and thus they use the same parameters. Also they don’t let that metal surface atoms relax/reconstruct.

This is clearly not a model with the level of detail you are expecting.

Please note that LAMMPS currently does not ship with an implementation of the image charge model. This is quite tricky to do correctly for a parallel MD code using domain decomposition as LAMMPS does. If I remember correctly, there is an external package that has an implementation, but that dos not work with MPI.

Hello,

Thanks for the fast reply. I understand now some of the models used. Which could be a first ‘nearly-tutorial’ step that I could perform to start having some MD experience? Could a simplified optimization of a simpler molecule on Copper be better (or I should avoid the substrate for now?)?

What you are asking is a topic for a discussion with your adviser or a designated tutor (e.g. a collaborator of your adviser if your adviser does not have sufficient experience) and not really a topic for this forum.

That said, I think your idea of looking for a simpler model and gaining experience by reproducing published results is a very good one. As I mentioned, there are going to be two challenges: 1) the accurate modeling of your molecules. That is best tested for the pure molecules or in solution but without a substrate. and 2) the modeling of the metal surface. An important question here is whether you want to have mobile substrate atoms or frozen ones. Also you need to look for alternatives to using the image charge model (using a model like ReaxFF could be an option, but that is a different (proverbial) can of worms) or locate the external, serial implementation and test that.

I would proceed in 3 steps :
- first, make a simulation with only the molecules (structures with parameters compatible with LAMMPS can be downloaded from the atb : https://atb.uq.edu.au/)
- second, make a simulation with only the solid phase
- third, make a hybrid simulation liquid + solid (typically with hybrid pair coeff and different thermostating for the solid and the liquid, which requires a bit of practice)

Best,
Simon

I think a frozen substrate could be sufficient for our simulations. I will test my molecule alone and start gaining some experience.

Thanks Simon for your reply.

I understand the step-by-step process. If I understood correctly your answer, I can use potentials tested for the molecules for step 1, potential tested for Copper in step 2 and then I should search a way to hybridize the simulation. For the last step, should I use since the beginning a pair potential from the beginning?

It depends of your system. Some solid surface use the same pair style as molecules (like the classical lj/cut/coul/long), and in this case the hybridization is simple. If your solid uses a different pair style, then it is slightly more difficult (but still rather simple dont worry).

Try to make step 1 and step 2 works, and then you can come back here and we can discuss the specifics of the hybridization.

Thank you very much for your support! I will try to compute step one and two and come back here.

To polarize the copper substrate (if you’re happy with it staying static), you can use the ELECTRODE package in the latest stable version of LAMMPS. Assign the copper particles into two contiguous groups of roughly equal size, set the potential difference between them to zero, and run fix electrode/conp with the two groups as electrodes and electroneutrality turned on (symm yes).

Then the copper particles will charge or discharge as required with the total system charge remaining zero. Note that this only guarantees a (mostly) physical simulation of the metal polarization process. Validating the result of the overall simulation is still something that needs to be done.