[lammps-users] MEAM potential for Fe-C

Dear lammps-users

We have implemented a calculation on Fe-C system using the MEAM potentials proposed by Lee
in Acta Mat. 54, 701-711 (2006) for the Fe-C interaction and the Fe-Fe interaction
proposed in Phys Rev B 64, 184102 (2001) and the C-C interaction proposed in
CALPHAD 29, 7 (2005).

However, we could not calculate the correct cohesive energy for the reference structure
using LAMMPS-21Oct2010. The correct value should be -6 eV, and we get -1.2 eV.
We are afraid that we made some mistakes in creating the MEAM input.

Below we are attaching the input files (input file for lammps, library.meam, FeC.meam, atomic coordinates data)
and the output file (log.lammps).

Could anyone of you help us to clarify this? … Thank you.


Griselda Garcia

Dear Prof. Griselda Garcia and lammps-users

I am the author of the Fe-C MEAM potential. I am not a good user of the lammps system.

However, as I know, the second nearest-neighbor (2NN) MEAM potential parameters are not transferable to the lammps-meam code yet,

especially for the case of alloy systems.

This is simply because 2NN MEAM formalism for alloy systems is partly different from that of the lammps-meam.

If anyone want, I can provide my own MD code.

As a matter of fact, the correct cohesive energy for the reference structure (L12-Fe3C) is -4.11 (-4.290.75-7.37.25 + 0.95) eV.

I am sorry to those who are not interested in this discussion.

Sincerely yours,

Byeong-Joo Lee

Hi Griselda,

I’ve been working with Byeong-Joo Lee to get to the bottom of some differences between his implementation of second-neighbor MEAM and the one in lammps. I’ve nearly finished this, but there’s still a bit of cleanup to do; this should get released soon.

However, none of these differences should affect the energy you compute in the reference system; you should always get the correct energy you specified in Ec(1,2). I haven’t run your input files, but one thing I see is that you don’t seem to have second neighbor calculations turned on for the Fe-C interaction itself; you need nn2(1,2)=1. That alone may fix your problem with the reference energy.

Also, be very careful with the entry of Cmin and Cmax. The ordering in lammps is Cmin(i,j,k), where i and j are the interacting pair and k is the “screening” atom between them. Other codes and papers might list these in the order (i,k,j), and this difference can cause confusion. I suspect you may have these in the wrong order, since you specify both Cmin(1,2,1) and Cmin(2,1,1), which in lammps notation are the same thing. You should only need to enter the value for Cmin(1,2,1). Take a look at the “pair_style meam” documentation page to make sure this makes sense.

But again, I don’t expect that you’ll fully match the Lee paper data until those formulation changes get released.



Greg Wagner

Sandia National Laboratories

P.O. Box 969 MS 9409 Livermore, CA 94551

Tel: (925) 294-2180 Fax: (925) 294-3410

Email: gjwagne@…3…