QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: Dipankar Saha on February 22, 2014, 12:00

Title: distance optimization
Post by: Dipankar Saha on February 22, 2014, 12:00

 :)

Title: Re: OptimizeGeometry /distance optimization
Post by: Dipankar Saha on April 28, 2015, 20:50

For the optimizer methods, LBFGS and qusiNewton .... is there any distinction, that we should keep in mind while using any of those??

Regards_
Dipankar  :)

Title: Re: distance optimization
Post by: Anders Blom on May 1, 2015, 22:44

LBFGS is a lot better. Uses less memory and converges better. Probably we will delete the QuasiNewton method later on.

Title: Re: distance optimization
Post by: Umberto Martinez on May 4, 2015, 11:44

VASP and ATK are very much different due the use of different methods.
Please have a look at the corresponding reference manuals.

1. In ATK you can specify different boundary conditions along the three directions. In this case along the C direction perpendicular to the surface.
Hence, you do not need the dipole correction in the middle of the vacuum region which you can apply in VASP (I assume this is what you refer to).
have a look at the work function tutorials here, for example  http://quantumwise.com/publications/tutorials/itemlist/category/98-surface-science
Large vacuum region is still a good idea though.

2. you can not really compare the plane-wave energy cut-off with the density mesh cut-off in ATK.

Title: Re: OptimizeGeometry /distance optimization
Post by: Dipankar Saha on May 5, 2015, 15:08

The fig. attached shows....geometry optimization of a Silicon nanowire... (@ ATK tutorial) ; where we find that there are constraints on several atoms on both the ends. Besides, the "constrain cell" is checked...for x, y and z.
_what does it mean when we constrain cells in different directions....??

Regards_
Dipankar

Title: Re: distance optimization
Post by: Umberto Martinez on May 7, 2015, 11:07

It means you do not allow your cell to relax to to optimize the stress.
Only forces will be optimized. For the atoms you do not constrain, of course.

Title: Re: OptimizeGeometry /distance optimization
Post by: Dipankar Saha on May 7, 2015, 15:28

Okayy...!!! ........ /Thanks again Umberto....  :)

Title: Re: OptimizeGeometry /distance optimization
Post by: Dipankar Saha on May 7, 2015, 15:33

Besides, I wanted to know_
for any optimized struct.  ....what does the radial distribution interpret (for given t_max, t_min)??

Title: Re: distance optimization
Post by: Umberto Martinez on May 8, 2015, 10:30

These parameters are used if you analyze a molecular dynamics simulation and they specify the range in time you want to analyze.
They are not used when you analyze a single configuration (which is not possible in ATK 2014 by the way)

Title: Re: OptimizeGeometry /distance optimization
Post by: Dipankar Saha on May 11, 2015, 12:01

In ATK....does the algorithm for finding the strain between two different materials.... follow the same procedure as it is described  in the following paper.....?

http://arxiv.org/pdf/1308.0969v1.pdf

Title: Re: distance optimization
Post by: Anders Blom on May 11, 2015, 13:01

Yes

Title: Re: distance optimization
Post by: Dipankar Saha on May 11, 2015, 14:08

Dr. Anders Blom,

Thank u.....  :)