aKMC for ionic diffusion: is the system too big?

[ce007]

Dear All,

In our group we look at lithium diffusion with atomistic potentials using NEB to get the activation barriers and MD to look at the diffusion.

We would like to use DFT to get the barriers but DFT-MD is not practical to look at the diffusion. Your aKMC code looks ideal for our purposes.

My question is; to get good statistics from the diffusion calculations we need a large cell 2000-5000 atoms. This is too large for VASP to deal with. Does aKMC need such a large cell to deal with the long range behaviour?

Many Thanks,

Chris Eames.

[graeme]

Even a few hundred atoms would be large for such a DFT calculation. But I would be very surprised if you needed such large cells to determine a converged energy barrier for Li diffusion. If you want to capture correlations in this large cell, you might want to run aKMC with your empirical (atomistic) potential or using a cluster expansion for the Li energetics. DFT could be used to look at a range of configurations as a test or fitting procedure for the empirical description. You can also run aKMC with DFT on the smaller cells to see if there are diffusion mechanisms that are not reproduced by your empirical potential and should be include in your KMC. If you do decide to try aKMC with your atomistic potential, we have a fairly new code for doing this which can make use of local, cluster, or distributed computational resources:
http://theory.cm.utexas.edu/henkelman/c ... rview.html

[ce007]

I had a look at the eOn code website and downloaded the source code but it isn't clear to me how it works.

Does it wrap around another code and just direct which structures to search with? If so how do I direct it to call the code? If not then how do I supply it with information about the material?

Thanks,

Chris.

[graeme]

It is a stand-alone code. There are some examples in the samples directory. For example, the samples/PtIsland contains parameters for an akmc simulation of a Pt island diffusing. You can see from this simulation what the input files look like, and how to define your simulation cell and atomic coordinates.

If you had a potential to describe your Li storage material, it would need to be added as in the client/potentials directory. We could help with this. We have also linked to the LAMMPS library which contains a wide variety of potentials.

It would certainly be some work to learn this tool. There is documentation at the link that I sent. It is also currently under active development so there is a small team ready to answer questions and sort out any problems that might arise.

[ce007]

Dear Graeme,

Thanks for the reply. I'll take a look at the examples and see how things work.

I have potentials for all of the two body and four body interactions in my material so I'll give these a try.

Thanks again,

Chris.

[ce007]

Yes you're right that it will take some work to learn how to use eon.

For example I can't find a description that tells me how to run the examples. I've tried ./init but this just gives me the config.ini and rectant.com files. If I copy the client executable to the run folder and type ./client it gives me a "problem loading parameters file" error.

So I've already become stuck....

[graeme]

Ah, perhaps we need some more detailed instructions. The code is called eon, and the executable is called eon.py in the main directory. So to run the sample, you can run "../../eon.py". You will also have to make the client before you do this, by running "make" in the client subdirectory, as described here: http://theory.cm.utexas.edu/henkelman/c ... stallation