Question on PDF simulation

[Leon Wang]

Hi, Bachir,

Thanks for your help and I got some results. However, I still have following questions regarding to PDF simulations.

1. When I built a hollandite model, I usually get a small model which match the stoichiometric values, and then build a large supercell. Here, there is a potential problem. In those papers, I saw Ba and Cs share some sites. In another word, those sites (symmetric equivalent ones) may be occupied by Ba or Cs. However, Ba (or Cs) take all symmetric equivalent sites. Let's say, siteA (x0,y0,z), siteB(x_1,y_1,z), site C(x_2,y_2,z), ....site N (x_(n-1), y_(n-1), z).  Here, you might notice that z has different values.  So, sites with same x_i and y_i but different z, will be occupied by the same element. In my current models, Ba may take site A, B, C...H, and Cs may take site I...N. In a real structure, Ba might take some sites from A to N, and Cs take the rest of sites from A to N. Please refer the attached figure. In your example, it does not show the swap when performing PDF simulations. Do I need to add swapList to exchange Ba and Cs to make sure the refined structure is close to the real structure?

2. Is it possible for fullrmc automatically to swap and remove some Ba and Cs atoms? To my limited knowledge of fullrmc, it consider each site is fully occupied by one atom. If I add swapList and removeList in run.py, meanwhile I set the stoichiometric values for each element, is it possible for fullrmc automatically to swap and remove atoms to match the stoichiometric values? I do not want manually set Ba and Cs position because my setting may not be reasonable.

Any further suggestion would be highly appreciated.

Appreciatively,

Leon

[Bachir Aoun]

Helo Leon,

Swapping is the mechanism that you should use to emulate occupancy ratios. Even though implementing occupancy shouldn't difficult to implement but It's not supported by fullrmc for good reasons.

In reality occupancy ratio is a hack or a patch to account for atoms replacements in small box simulations.

fullrmc is a stochastic big box model implementation and therefore there is no need to have occupancy ratios because you will be able to account for those statistically in your simulation.

You must create groups of single atoms of your choice and create the swapping lists with other atoms in the system.

swapping is a move generator implementation and everything will happen automatically for you upon engine runtime.

one thing that might be useful to implement is elements transmutation (dynamically changing elements type) but i haven't tackled this yet. this can be for future releases.  

regards 

[Robert Koch]

Hi Everyone,

I just wanted to add that, knowing Leon's structure, the sites he is discussing are 66% occupied by vacancies in small box models, with a varying Cs/Ba ratio in the remaining 34%. Creating an explicit supercell by randomly removing Cs/Ba to match 66% vacancy freezes in random vacancy ordering, and, as the two Cs/Ba/Vacancy sites are less than 1 Angstrom apart in the small box models, can lead to nonphysical pair distances.

There presumably is a (strong) correlation in the occupancy of the two Cs/Ba/Vacancy sites, such that if one is occupied by Cs/Ba, the adjacent one is not occupied (or is occupied by a vacancy). 

This can be handled either in the supercell creation process (ensuring Cs/Ba-Cs/Ba pairs are always  > 1 Angstrom) or by introducing a dummy species with no scattering into the RMC. This correlation cannot cannot be captured with small box models, but could be captured with RMC. I'd recommend reviewing the post on vacancies from a few weeks ago:

https://groups.google.com/forum/?utm_medium=poem&utm_source=footer#!topic/fullrmc/VE9eIohcubo

Regards,

Rob

[Leon Wang]

Thanks for your response.

Hope everyone stay safely.

Leon

[Leon Wang]

Hi, Rob,

Thanks for your reply and reference. I am trying to understand your conversation and test them on my side.

Best,

Leon

[Bachir Aoun]

Rob,

I appreciate you answering. This forum is meant for everyone to take a stab at responding whenever they feel like they can comment and answer a particular question. 

So far it was me doing all the responses. I happy to seeing that this started to be a real stack-overflow like Q&A forum the way i intended it. 

You all stay safe 

regards 

[Robert Koch]

We all gotta share the knowledge!

Cheers,

Rob