Calculation of grain boundary energy

Dear all lammps-users:

As the title mentioned, my problem is about calculation of grain boundary energy. Firstly, I constructed a bicrystal system as the example of lammps tutorial ( . Second, “displace_atoms”, “minimize”, “fix box/relax” and “minimize”
commands are used to calculate the grain boundary energy.But my result is
different from Sansoz(F. Sansoz, JF. Molinari. Acta Mater 2005; 53:1931) w
who used the lammps get the result. And I compared with Sansoz’s dump file
, I find the number of atoms and site of atoms are same. I don’t know how to
solve the problem. (The appendixes are my in file using different modeling methods).
Thank you very much!

in (2.47 KB)

in.classmodel (2.74 KB)

So you calculated an energy with LAMMPS and someone else calculated a
different energy. There are a myriad reasons why that might be, but
two big ones are:

1. Different geometry
2. Different potential

A good way to eliminate 2 from further consideration is to first match
a result of Sansoz et al using a simpler geometry, such as the
ground-state crystal. After that, you can focus on 1, which for grain
boundaries, is pretty complicated.

Dear Thompson:

Thank you for your reply. About your suggestions I think that the grain boundary energy is a intrinsic quality of grain boundary. So the result should be similar in different geometries and the different potentials. But in my previous result, my calculation result was 138 mJ/m^2, and the result of Sanzos is 309 mJ/m^2.(It’s my problem that I don’t say it clearly.)

Yesterday, Mr Kim told me the lattice parameter of Cu was not 3.615, it should be sqrt(2)*2.555 (2.555 was used in fitting MEAM potential parameters). And my fomula to calculate grain boundary energy had a problem. In my previous viewpoint, a grain boundary includes two planes, so I divided 4 in my fomula. But a grain bounday actually includes one plane. So it should divide 2.

Today, I adopt your suggestion, I use EAM potential which is same as Sanzos, the result is 316 mJ/m^2.

Then I have a question want to consult with everybody. Which is better, using two types or only one type in generatting grain boundary?
In the end, I would like to hereby express to you my heartfelt thanks!
Sincerely yours,
Tong Ke

Good. It sounds like the biggest issue was the factor of 2, and the
switching to Sanzos' potential also helped. The remaining difference
(309 versus 316) could be due to a variety of factors, including
differences in the potential and the geometry. When I say geometry,
that includes things like slightly different lattice constants. Once
again, you could check the potential by comparing with Sanzos' paper
for things that are less complicated than grain boundary energy, such
as fully relaxed lattice constant and cohesive energy.

If your question is about whether to use 1 or 2 LAMMPS atom types,
then I think in principle it does not matter. It is probably easier to
analyze the results and visualize if you use two atom types.