Slow Growth with ML Check

[jess_white]

Hi,

I am just double checking the way which I am doing the Slow Growth is correct. In the examples the Andersen thermostat is used, however I am hoping to use the Langevin thermostat to keep it consistent with my machine learning prior. So I probably should explain why I am doing the slow growth method. Essentially a main point of this study is to see the dissociation of Mg/Ca from the diopside mineral surface. So I can created an ICONST where I am increasing the distance between the Mg and O atoms. I am also confused by the INCREM tag - what is the difference between it being negative and being positive. I did some tests myself and came to the conclusion that if the INCREM tag is positive it means that I am increasing the distance between the two elements selected in the ICONST. I also noticed that even if I have the temperature set as 300K it spikes up to 1000K which makes me think I need to change perhaps some parameters such as the LANGEVIN_GAMMA. I have also increased the time step to be 1fs - is that pushing it too much?

Just overall I just want to see if the calculations I am doing is justified etc. Sorry for the overload of information. Please let me know if you need more specific information :)

Thanks,
Jess

[jess_white]

Here is the files :) Deleted the ML_AB and ML_ABN as was too big.

[jonathan_lahnsteiner2]

Dear Jess White,

For the slow growth approach we have a bunch of tutorial pages on our vasp wiki. Please take a look at the following Tutorial pages. The tutorials are analogous to what you want to do and contain plenty of information to do your double checking.
https://www.vasp.at/wiki/index.php/Slow-growth_approach
https://vasp.at/wiki/Nuclephile_Substitution_CH3Cl_-_SG
https://www.vasp.at/tutorials/latest/tr ... tes/part3/
Also take a look at the INCREM- wiki page. It states that the tag controls the transformation velocity. So a negative value would mean that the distance decreases.

Regarding the thermostat: I was checking your OSZICAR file which shows that you are starting your simulation from a structure far from equilibrium. In the MD steps you have supplied your system is still cooling down from a starting structure around 1000K to the 300K set in your INCAR file (and has not reached the desired temperature yet). So you should definitely pre-equilibrate this structure before conducting any slow growth approach. Maybe this page on molecular dynamics can be helpful for some background information:
https://www.vasp.at/tutorials/latest/md/

If you have further questions please contact us again.

All the best Jonathan