Hello everyone,
To perform some calculations on the LME suffered by steels, I've tried to come up with a simple LJ potential to deal with the interaction of steel and liquid lead. I've obtained through CASTEP the diatomic potentials for all the elements, but the potential well values are too big when I actually make a LJ potential in LAMMPS with them. Trying to divide these values by the closest neighbours index or atoms per unit cell also gives either too big or too small values. How could I fix this?. I attach the diatomic potential for Ni, calculated using a cut-off energy of 900 eV for highest accuracy, which gives values of epsilon = 2.05 eV and sigma = 1.68 angstroms.
Sincerely yours,
Álvaro
Why not use EAM potential(s)? They are usually a pretty good fit for metals, much better than LJ.
Axel.
The system I'm simulating is made of an austenitic steel cell surrounded
by Pb-Bi eutectic. For the steel itself I'm using a EAM potential, but
need to also describe the interactions of Fe, Ni and Cr with Pb and Bi,
for which I chose LJ because the Pb-Bi will be in liquid phase and for
the sake of simplicity.
Álvaro
I cannot think of a simpler procedure than looking up and downloading existing potential parameters.
I would also like to point out that mixing and matching EAM and or other potential parameters using pair style hybrid is adding a significant inconsistency to your model as the atoms treated with LJ will not contribute to the embedding function that is crucial in EAM to represent the “metallic” behavior and thus will negatively impact the accuracy of the EAM part of the calculation. this also applies when you use multiple different EAM instances. like with all manybody interactions, they need to be represented by a single pair style instance to correctly represent the manybody nature. only when you have well separated phases, e.g. a metal surface interacting with a diamond cutting tool, you can expect the inherent error of the model to be limited to just the surface and be rather small, since the application will have that interaction be mostly represented by repulsion (which is predominantly a two-body interaction) and not attraction.
please also note that there are several conceptual problems with your approach to obtain LJ parameters.
-
the nature of the interaction between a pair of isolated metal atoms can be quite different from the bulk metal
-
a typical “closed shell” calculation may not be representative of the actual interaction. you may have a different configuration for the pair than for the isolated atoms. an instructive example would be a pair of hydrogen atoms. if you do an energy scan like you did show in your original email for hydrogen, you won’t get the correct dissociation energy with a simple DFT based energy scan. you will need a multi-configuration calculation, e.g. CASSCF in the simplest case to get a good representation for both short and long distances. transition metals can have additional problems to get the quantum calculations right which to discuss would be off-topic for this mailing list.
-
it is very difficult to fit a Lennard-Jones potential to experimental or quantum data since the shape of the curve is not represented well by the potential. a least squares fit or similar will be too dependent on which and how many points you choose for the fit. a better strategy if often to fit to a Morse potential, where you have the minimum and dissociation energy (if computed correctly) as fixed parameters and only need to fit the width parameter, which is best done by looking at the gradient around the minimum (since you will need the best accuracy there anyway). obtaining lennard-jones parameters from morse parameters is much simpler. for the aforementioned pair of hydrogen atoms (because of the covalent nature), a Morse potential is an extremely good fit for the shape of the potential including dissociation.
Axel.
Thank you very much. I have checked the NIST repository but there was no available potentials for Bi or Pb-Bi systems, only for Pb. Still, how could I mix the EAM for Pb-Bi and the one I use for the steel?
Álvaro
If you use plain EAM potentials, LAMMPS will do the mixing for you. If you use eam/fs or eam/alloy, you will have to do the mixing yourself.
there are examples in the literature for how to do that. the LAMMPS distribution has a tool that can generate mixed potentials from per atom settings in the tools/eam_database/ folder. having explicitly mixed potentials offers the opportunity to adjust the cross-terms for an improved representation versus the automatically mixed interactions, but with access to the corresponding input parameters it should be possible to generate a custom potential file for your needs. in any case, I would expect such mixing of potentials to produce superior accuracy over an EAM/LJ hybrid model, however well computed and fitted.
axel.
Hi
there exist Pb and Bi EAM potentials in the literature and in LAMMPS potentials see NIST repository. Bi EAM has been parametrized in liquid state.
See for all potentials
27. X. Zhou, R. Johnson, H. Wadley. Misfit-energy-increasing dislocations in vapor-deposited CoFe/NiFe multilayers. Phys. Rev. B 69 (2004) 144113.
28. D. K. Belashchenko. Computer Simulation of the Properties of Liquid Metals: Gallium, Lead, and Bismuth. Russ. J. Phys. Chem. A 86 (2012) 779-790.
29. Y. Gao, G. Raos, C. Cavallotti, M. Takahashi. Molecular Dynamics Simulation on Physical Properties of Liquid Lead, Bismuth and Lead-bismuth Eutectic (LBE). Procedia Engineering 157 (2016) 214 – 221.
Pascal Brault
DR CNRS
GREMI UMR7344
CNRS-Université d’Orléans
The system I'm simulating is made of an austenitic steel cell surrounded
by Pb-Bi eutectic. For the steel itself I'm using a EAM potential, but
need to also describe the interactions of Fe, Ni and Cr with Pb and Bi,
for which I chose LJ because the Pb-Bi will be in liquid phase and for
the sake of simplicity.
Álvaro
The reason why I decided to start calculating on my own the diatomic potentials for all the species involved in my simulation is because this is my graduate's final thesis and thought an approach from the bottom would be more interesting. Also, there are not many references on the parameters of FeNiCr with Pb-Bi systems, and, as Zhou's potential for these type of steels already fits my needs, understood that a simple LJ for the liquid metals would be the quickest option. I didn't thought the contribution to the inaccuracy of using a LJ together with an EAM alloy would be so relevant, so I'll take it in account, but, as the results will be compared to experimental and ab-initio results, I think all of this can lead to at least some conclusions.
Thank you for your help,
Álvaro
The reason why I decided to start calculating on my own the diatomic potentials for all the species involved in my simulation is because this is my graduate's final thesis and thought an approach from the bottom would be more interesting. Also, there are not many references on the parameters of FeNiCr with Pb-Bi systems, and, as Zhou's potential for these type of steels already fits my needs, understood that a simple LJ for the liquid metals would be the quickest option. I didn't thought the contribution to the inaccuracy of using a LJ together with an EAM alloy would be so relevant, so I'll take it in account, but, as the results will be compared to experimental and ab-initio results, I think all of this can lead to at least some conclusions.
Thank you for your help,
Álvaro
I'll explore all possibilities and see which one offers me better results, as my time is limited and don't know if I'll have enough time to do a proper fitting for the whole system.
Álvaro
It is my personal observation (based on personal experience and many exchanges with people over the years) that there are no shortcuts in research. so when people say they make choices in the interest of saving time, their chances of ending up wasting their time are usually high.
It is rather straightforward to verify the impact of EAM vs. EAM+LJ by taking a small/simple system for which you do have a consistent set of EAM parameters and then just replace one of them with an LJ potential of your own making and compare the results.
you are - of course - free to make your own choices. after all that is what makes doing research attractive.
and the details of that is quite off-topic for this list, anyway, so I’ll have to tell myself to shut up about it and let you do what you think is your best option.
axel.
Thank you very much for your help, Axel. I'll take in account your experience and opinion.
Álvaro
