[Wien] Interface relaxation

[L. D. Marks]

With respect to optimizing the bulk, this means optimize the lattice 
parameter with the particular functional that you are using (LDA/PBE). In 
general this is a few percent different from experiment.

I would do things slightly differently. (N.B., I'm biased, but the FAQ is 
not too bad according to others.)

1) Optimize the bulk with RMT's 5-10% smaller than setrmt. When doing this 
find a reasonable RKMAX, as well as a well converged one.

2) Create a symmetric slab with 10-20 a.u. of space between the surfaces, 
perhaps a total size of 30-50 a.u. (Keep it as small as reasonable.) I use 
cryscon and excel for this, transferring positions as cif format which is 
reasonably simple. Do this by hand and use a good atom viewer (I use 
Atoms); don't trust computer codes to really get it right. By hand, you 
may want to adjust the adsorbate positions to a reasonable first guess 
(which is probably wrong but that's life).

3) Optimize the positions (all, not some subset) with an RKMAX about 
0.5-1.0 smaller than in 1) and the RMT's from 1). Positions seem to be 
much less sensitive than other things to the RKMAX convergence. Depending 
upon your surface, you might need to reduce the RMT's and use clminter. 
During the optimization. I do not recommend fixing atoms, this really does 
not save you much time in a quasi-Newton method and can lead to other 
problems. I tend to use 2-3 k-points, 1 k-point (unless I have no choice) 
can behave badly.

4) Increase RKMAX and rerun until it is converged or I run out of computer 
resources. (Save the previous run, then just increase RKMAX by about 0.25 
until you are happy or have lost patience.)

5) As appropriate, add more bulk and/or more surface. In practice, adding 
a bulk unit cell in the center with your bulk lattice parameter and 
shifting all the positions (excel) works well.

On Tue, 11 Jul 2006, morteza rafiee wrote:

> Dear prof. Andersen,
> Thank you for your soon reply, and your valuable
> comment.
> My bulk has no internal parameter. Do you mean then
> that I have to optimize the lattice parameter here?
> I could not realize your second step where you
> disscused "building a symmetric film with the
> distances from the optimized bulk calculation, and
> optimize it perpendicular to the film only", and I
> will be thankful to you if give us more discription
> about "distances" and "perpendicular". How it could be
> possible to restrict the relaxation only in the
> perpendicular direction to the film.
> We belive that your discription is exactly what we are
> looking for it. But unfortunatly we could not get it
> as we do not have any experiences in this subject.
> Please assume that we are very new user.
>
> Best regards,
> M. Rafiee.
>
>
> Dear Mr. Rafiee,
>
> my strategy for this is:
>
> 1) Optimize the "bulk" volume of the substrate - this
> is fast.
>
> 2) build a symmetric film (with your deposited layers
> on both sides of
> the substrate) with the distances from the optimized
> bulk calculation,
> and optimize it perpendicular to the film only.
>
> Step 2 takes some time, but in general you should
> start with thin films
> (1 substrate layer and the deposited layers), then
> repeat for thicker
> substrates. At some point you will see that the
> distances between the
> central layers does not change anymore. Then your film
> is optimized.
>
> Best regards,
> Torsten Andersen.
>
>
> morteza rafiee wrote:
>> Dear wien user's,
>> I would like to deposit layers of atoms on a
>> substrate.
>> Obviously relaxation is time consuming, thus I
>> appreciate if someone let us know which of the
>> following methods are more appropriate taking cpu
> time
>> into account.
>> 1. Constructing the structure file only for the
>> substrate and then relax it. After calculating the
>> relaxed atomic positions of the substrate using
> mini,
>> depositing the atoms on the relaxed substrate, and
>> then as a second time try to relax the atomic
>> positions of the deposited layers.
>>
>> 2.Generating both substrate and the deposited layers
>> on it from scratch and then relax all the atoms of
>> layers and substrate simultaneously.
>>
>> It seems that (we have not any experiences in this
>> respect) one can save the cpu time using the first
>> method due to the fact that the number of atoms that
>> must be positioned in their relaxed locations is
>> smaller than the case that we let be relaxed all the
>> atomic positions of the layers and substrate
>> simultaneously as described in the second method.
>>
>> Your,
>> M. Rafiee
>>
>> __________________________________________________
>> Do You Yahoo!?
>> Tired of spam?  Yahoo! Mail has the best spam
> protection around
>> http://mail.yahoo.com
>> _______________________________________________
>> Wien mailing list
>> Wien at zeus.theochem.tuwien.ac.at
>>
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>>
>
> -- 
> Dr. Torsten Andersen        TA-web:
> http://deep.at/myspace/
> AG Hübner,
> Department of Physics, Kaiserslautern University
> http://cmt.physik.uni-kl.de
> http://www.physik.uni-kl.de/
>
>
> __________________________________________________
> Do You Yahoo!?
> Tired of spam?  Yahoo! Mail has the best spam
> protection around
> http://mail.yahoo.com
>
> __________________________________________________
> Do You Yahoo!?
> Tired of spam?  Yahoo! Mail has the best spam protection around
> http://mail.yahoo.com
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>

Note: if you have an old email address for me, please note that "nwu" has
been changed to "northwestern".
-----------------------------------------------
Laurence Marks
Department of Materials Science and Engineering
MSE Rm 2036 Cook Hall
2225 N Campus Drive
Northwestern University
Evanston, IL 60208, USA
Tel: (847) 491-3996 Fax: (847) 491-7820
email: L-marks at northwestern dot edu
http://www.numis.northwestern.edu
-----------------------------------------------