How to find the global minimum...?

[Monu Joy]

Hi there

I have a general question: I would like to find out the global minimum of a guest molecule within the pore of a MOF. I know I can place the guest molecule in the pore and get the host-guest structure optimized, but if I would place the guest molecule in another position of the same pore, there also I can get the structure optimized. In this scenario how could I find the exact most probable location of the guest molecule within the MOF pore or its global minimum...?

Any help would be highly appreciated...

-Monu

[Fangyong Yan]

Hi, Monu,

You can run a molecular dynamics simulation at room temperature, for several picosecond, and then you can plot the radial distribution function, between your guest molecule and the MOF, by doing this you would have a general picture of the free energy surface of your guest molecule in the MOF. Then based on the free energy minima structures, you can start to optimize your guest-MOF complex structure. (the free energy surface may be different from the potential energy surface, but at least by doing MD simulation and calculating the free energy surface based on radial distribution function, you can have a basic idea about the free energy surface, which should not be quite different from potential energy surface).

Regards,

Fangyong

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/1852672747.3320.1610497192914%40mail.yahoo.com.

[Fangyong Yan]

Hi, Monu,

To explore the free energy surface better, you may need to start at a high temperature, and then annealing down to the room temperature. And for each temperature, you can plot the radial distribution function, for your guest molecule and your MOF atoms, for example, at 800 K, 700 K, ...., 300K. The reason starting at high temperature is that you can use heat to get your guest molecule out of the potentially trapped free energy minima, so you can explore the free energy minima more completely. 

Regards,

Fangyong

[Lucas Lodeiro]

Hi Monu and Fangyong,

I disagree a little bit with Fangyong because an AIMD calculation is not cheaper, and if the barriers between local minimas are high, you cannot see a structural change with the MD, at least in the scale time of ps. Firstly, I would try to optimize a couple of structures and measure the energy differences, this will give you a little insight about the barriers (the minimum energy for the barriers), If they are high, the AIMD probably cannot move from one local minima to another. If they are little, you can see the structural changes, but in this case, probably the global minima is not sufficient to modulate the properties of the system, and some other local minima have to be taken into account.

Other way is to generate with a script random positions of the molecule into the MOF and minimize them with a semiempirical formulation, as XTB (https://manual.cp2k.org/trunk/CP2K_INPUT/FORCE_EVAL/DFT/QS.html#METHOD), and rerun with the level of theory that you want just the low energy ones.

Regards - Lucas Lodeiro

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAJjmc%2BKviWAFmj5UmYqMWL1Z5wFpQZjdKkhGS%2B7CTp7-6jaLgg%40mail.gmail.com.

[Monu Joy]

Hi Lucas & Fangyong

Thanks for the help. I have performed two cell-opt calculations of guest@host where the following total energies respectively represent the guest at the center of the MOF pore and near to the pore surface -6235.818 and -6235.824. Do you guys think that this energy difference is reasonable...?

Also, do you by chance have any reference that shows local minimum geometry modulate the property of a system more effectively...? 

Thank you so much for your help!

-Monu

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAOFT4PKCKjc0vOGZUkbjsDEEqJM_w_UxYK-PC%2Bt-vnzZzLhrgg%40mail.gmail.com.

[Fangyong Yan]

Hi, Monu,

The difference energy is -6235.818 - (-6235.824) = 0.006 Hartree = 15.753 kj/mol. 

So if you use 800 K, the Boltzmann factor is: exp(-dE/R/T) = exp(-15.753*1000/8.314/800) = 0.09, which means you would have a probability of 0.09 to "jump" from the lower energy to higher energy, by the help of heat. 

So I think NVT at 800K, may be helpful for exploring the potential energy surface, even within 10ps. 

Regards,

Fangyong

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/1361749041.8962.1610759305502%40mail.yahoo.com.

[Anton S. Lytvynenko]

Dear Monu,

AFAIR CP2K has its established procedures for global minimum -- https://manual.cp2k.org/trunk/CP2K_INPUT/SWARM/GLOBAL_OPT.html

Did you try anything of this?

16.01.21 03:08, Monu Joy пише:

El mié, 13 ene 2021 a las 0:07, Fangyong Yan (<fyya...@gmail.com>) escribió:

To explore the free energy surface better, you may need to start at a high temperature, and then annealing down to the room temperature. And for each temperature, you can plot the radial distribution function, for your guest molecule and your MOF atoms, for example, at 800 K, 700 K, ...., 300K.

One should be aware that the vast majority of MOFs will be totally destroyed at such temperatures.

Yours,

Anton

[Monu Joy]

Thanks, Anton for pointing it out, and I will give it try. 

--

You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/aca37154-eb20-5115-6ff2-dba93ccd27fd%40gmail.com.

[Monu Joy]

Thanks, Fangyong for the suggestion.

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAJjmc%2BKyP-zV-p%2B%2BZtUkTgqoWQnovqnUvaVpG7NQby7a8KURgQ%40mail.gmail.com.

--

Thanks with Regards

[Sam Broderick]

Hi

In researching global optimization, I stumbled on a reply be Ole Schütt, who is the author of the SWARM part of CP2K. He suggested that the methods are somewhat older and to look at ASE. I think this is what I will look into.

Cheers

Sam

[rabdel]

Hi,

You can try with this approach: https://doi.org/10.1016/j.micromeso.2019.109885
Best regards,

Rabdel