Lattice distortion with dynamics method in EON-AKMC

[lokiloki]

Hello,

I am running Eon-akmc simulations on a metallic system. We are trying to simulate interstitial diffusion in a perfect lattice (bcc). I am using the dynamics method for saddle searches and performing dynamics at an elevated temperature of around 1200 K. The Config file is attached here.

During AKMC moves the barrier between the states (for interstitial jumps between two minima) should have a value comparable to 0.8 eV (based on known activation energy barriers in this system). However, we are finding several lower barrier values of around 0.3 - 0.45 eV. Apparently, EON uses these lower barrier values to move the interstitial atoms, rather than the known diffusion barrier values of 0.8 eV. In the visualisation of different states, I found some distortions in the lattice at the boundaries and the interstitial atoms which moved were generally found at the surface (most likely, corresponding to the lower barriers of 0.3 – 0.45 eV).

To me, surface distortions seem to suggest a finite-sized system. Is this true? Does EON assume periodic boundary conditions by default? If not, how can we implement/change boundary conditions for the dynamics run? The snapshot for different states are also attached here.

Would someone be able to help me resolve these issues? Is there a possibility to restrict the surface distortions in the metal matrix during dynamics? Or preventing or discarding the lower barrier value while making akmc moves?

I am attaching the snapshots of different states where you may be able to see the distortions at the surface also.

Any help regarding the issue will be appreciated.

Thank you

Regards,

Lokesh Soni
PhD student
IIT Bombay, INDIA

[graeme]

Hi Lokesh,

First, great to see that you are able to use the AKMC method in EON. You are absolutely correct in that AKMC will generally choose low barrier events with high rates over higher barrier events with low rates. It can be frustrating when the system has low barrier processes, which you may not really care about, when there are higher barrier processes that you do care about. An important point, however, is that AKMC is designed specifically to show you what the system would do, even if it is not what you want or think it should do. In you case, I can't really tell from the three images what the mechanism of the 0.3-0.4 eV barriers are. I see a shift of all the atoms, but I would be very surprised if that could happen with a barrier as low as 0.4 eV. I would also be concerned, as you asked, that any process which involves the entire system is an artifact of the periodic boundary conditions. On that point, yes, EON is written to use periodic boundaries. Isolated systems will need sufficient vacuum to isolate periodic images. Systems can also be constrained with frozen atoms.

I would actually be interested in looking into this further. If you are willing to post your calculation, I can take a closer look.

One final point: there are some tricks to deal with low barriers, but apart from coarse graining (which is exact), they can mess up the correct dynamics. My vote is to understand the dynamics in your system and see if the events make sense or if something in the setup needs to be changed. It is better to do this before thinking about ways to get the dynamics that you want.

Graeme

[lokiloki]

Dear graeme,

Thank you very much for the reply.

As you mentioned, I have constrained only the interstitial atoms to move during AKMC bydisplace_type_list command. I suspect that the distortion in the lattice is due to the dynamics run only. However, the interstitial atoms which were moved during generally found at the lattice boundaries only. I am attaching a part of the eon-dynamics during the AKMC run to better visualise the problem.

With reference to a previous thread - http://theory.cm.utexas.edu/henkelman/f ... hp?t=11664 I have also tried to implement the max_rate command to my system in order to restrict any other faster events. I have uncommented the max_rate implementation in akmcstates.py and akmcstatelist.py in the source code as shown below.

Although the log does state that the event has exceeded the max_rate limit for low barrier values, but it still chooses it for making an akmc move. I think the code does not omit the higher rates, it just assigns the maximum allowed rate as the process rate to such events.
I am also attaching the screenshots of the config file and dynamics.txt generated during the EON-AKMC run.

One more query that I have, what does the keyword “product” in the process table signifies?

Any help regarding the issue will be appreciated.

Thank you

Regards,

Lokesh Soni
PhD student
IIT Bombay, INDIA

[graeme]

Hi Lokesh,

It's hard for me to tell what is going on from the movie. The movement that you indicate could (in my eye) just be atoms moving across the periodic boundary conditions. Well, maybe there is a small diffusion event for the interstitial at the top-left of the simulation. I would like to see (i) what is the potential and (ii) what are the low barrier processes. It think this is a better place to start than jumping to the max_rate tricks, which will give you only approximate dynamics.

I know these calculations can be large, but if you can .tar.gz it, then see if you can either post it here or provide a link from your machines or a cloud service.

The -1 in the product column indicates that it is an unknown product state. If you have a positive integer there, it will indicate the (known) state into which the process is going to.