[XtalOpt-devel] Lowest limit of total energy?

[ezurek at buffalo.edu]

Dear Atsushi,

thanks for the input files. I noticed a couple of things in them.

First of all, the volume of the cell of the strange structure is 468.87, 
whereas that of the alpha quartz is 107.14. Since you are using a double 
unit cell of SiO2 it would be best to restrict your volume in XtalOpt to 
about 2x107. Otherwise unphysical cells which are not close to the global 
min will be generated, and the algorithm will spend a lot of time looking 
at these species.

Another thing I noticed is that NSW=1 in the vasp INCAR file. That means 
you are doing a single point calculation, not a geometry optimization.

Whereas it is in principle possible to use the ideas of Darwinian instead 
of Lamarkian evolution (ie. in the former one does not optimize the 
structures which are taken to be parents of the next generation, whereas 
in the latter one does) it has been demonstrated by Woodley and Catlow 
that Lamarkian evolution finds the target structure faster in GAs/EAs:

Title: Structure prediction of titania phases: Implementation of Darwinian 
versus Lamarckian concepts in an Evolutionary Algorithm
Author(s): Woodley, SM; Catlow, CRA
Source: COMPUTATIONAL MATERIALS SCIENCE Volume: 45 Issue: 1 Pages: 84-95 
Published: 2009

I am right now optimizing the strange species. My guess is that during the 
optimization the volume will decrease and the energy will become more 
reasonable. The forces on the atoms in the initial structure are quite 
large.

So even though you cannot set the lowest min of total energy, I think 
that you will not need to do this if you use optimized structures to be 
the parents of the next generation.

Best Regards
Eva

> Dear Eva,
>
> Thank you for your quick reply. Probably only the geometrical
> constraints, this problem is not solved.
>
> Alpha-quartz is most stable at ambient condition, and as you wrote
> polymorphs found at ambient should have similar energies too.  The
> energy I got for the strange structure is calculated ~-4000 eV by
> VASP, however usual energy would be ~-150 eV. Just in case, I attach
> the files.
>
> Thank you,
>
> Togo
>
> On Wed, Feb 2, 2011 at 1:18 PM,  <ezurek at buffalo.edu> wrote:
>>
>> Dear Atsushi,
>>
>> first of all, thanks for contributing spglib to XtalOpt!
>>
>> As far as I know, there is no such option. The only thing we can limit right
>> now is the interatomic distances, unit cell angles and lengths but the
>> structures can optimize out of that range.
>>
>> I know that SiO2 has many different forms, and that in most of them there
>> are SiO4 tetrahedra. There are also amorphous forms. Do you know which form
>> has the lowest energy at 0 K? Or is there a bunch of structures which all
>> have quite similar energies? Are you sure that the energy computed with VASP
>> for these strange structures is too low?
>>
>> Best
>> Eva
>>
>>> Dear Xtalopt developers,
>>>
>>> I try to use Xtalopt with VASP. I use the Xtalopt obtained as,
>>>
>>> commit 7921a448b928d131bb19357cba85099c94a1da61
>>> Author: David C. Lonie <loniedavid at gmail.com>
>>> Date:   Thu Jan 20 16:28:41 2011 -0500
>>>
>>>
>>> My test system is Si6O12. Usually the structure is constructed with
>>> SiO4 tetrahadra linked by vertices. When Xtalopt finds this kind of
>>> structures, the total energy calculated by VASP becomes reasonable.
>>> Xtalopt sometimes finds strange arrangement of atoms, and
>>> unfortunately the total energy calculated by VASP becomes unusually
>>> extremely low (probably the crystal structure is strange enough beyond
>>> that VASP assumes). I want to kick those unusual structure out
>>> automatically, and am wondering how I can do that. For example, is it
>>> possible to set the lowest minimum of total energy in Xtalopt.
>>>
>>> Thank you,
>>>
>>> Togo
>>>
>>> --
>>> Atsushi Togo
>>> http://atztogo.users.sourceforge.net/
>>> atz.togo at gmail.com
>>> _______________________________________________
>>> XtalOpt-Devel mailing list
>>> XtalOpt-Devel at lists.openmolecules.net
>>>
>>> http://lists.openmolecules.net/listinfo.cgi/xtalopt-devel-openmolecules.net
>>>
>>>
>>
>> ---------------------------------------------------
>> Dr. Eva Zurek
>>
>> Assistant Professor
>> Department of Chemistry
>> University at Buffalo
>> State University of New York
>> 331 Natural Sciences Complex
>> Buffalo, NY 14260-3000
>>
>> email:  ezurek at buffalo.edu
>> phone:  +1-716-645-4332
>> fax:    +1-716-645-6963
>> www:    http://www.acsu.buffalo.edu/~ezurek/
>> ---------------------------------------------------
>>
>
>
>
> -- 
> Atsushi Togo
> http://atztogo.users.sourceforge.net/
> atz.togo at gmail.com
>

---------------------------------------------------
Dr. Eva Zurek

Assistant Professor
Department of Chemistry
University at Buffalo
State University of New York
331 Natural Sciences Complex
Buffalo, NY 14260-3000

email:  ezurek at buffalo.edu
phone:  +1-716-645-4332
fax:    +1-716-645-6963
www:    http://www.acsu.buffalo.edu/~ezurek/
---------------------------------------------------