[lammps-users] 3+ elements in MEAM parameter file

Lammps Users,

I am trying to create a staged insertion simulation, where an Ag atom moves from one location to another by “growing” in the new spot. This involves modifying MEAM potential parameters to create multiple Ag atom types with varying contributions to the energy and/or size.

My problem is that I have 3 atom types in my simulation at the moment. Type 1=Ag, 2=Sn, 3=Ag3. For now, Ag and Ag3 are identical parameter wise, though they have different sections in the “library.meam” file. The initial setup is 1000 Sn atoms and 1 Ag atom. When I use an Ag3 atom the energy is different. Through some trial and error, I’ve found this is an issue with the MEAM parameter file–the file after the “library.meam” file containing parameters for alloyed reference structures. Within the file, I’m adhering to LAMMPS rules for setting X(I,J) variables, (e.g. Ec(1,2)= and Ec(2,3)=) but with the same values for both of these I get a different energy when using Ag vs. Ag3.

There is also a large variation for this system when you compile and link against “-limf” (Intel’s math library?) vs. NULL (which on my system defaults to GNU’s library). I only see this when I am using 3+ elements in my MEAM parameter file.

Michael Sellers

data.agint (36 KB)

library.meam (2.17 KB)

in.agmeam (613 Bytes)

grow.meam (255 Bytes)

Could this have something to do with the reference structure? L12 needs specific ordering I think. X_3-Y

Michael Sellers

Graduate Student, Chemical and Biological Engineering
University at Buffalo, The State University of New York

I'll let Greg answer this. It's beyond my level of MEAM understanding.

The fact that you get different answers with different math libs sounds
ominous. Is this an instant difference, or a cummulative long-time
difference (i.e. 2 simulations drift apart)?