The ABINIT Users Mailing List ( CLOSED )

[Masayoshi Mikami <mmikami@yk.rim.or.jp>]

Dear Xu ZhiPing:

a) From the manual:
http://www.abinit.org/ABINIT/Infos_v4.2/varrlx.htm#optcell
At the present stage, we cannot optimize both acell(3) and 
angdeg(3) at the same time. But, if you might know any 
appropriate range of acell(3) and angdeg(3) beforehand 
(e.g. you know some experimental results ...), you might 
want to run 2-dimensional minimum search manually ... 
I mean, You could have a total energy contour map with 
respect to angdeg(3) and acell(3), like this :

angdeg(3)
^
|
|
|
0------->acell(3)
   --- a range for acell(3) (and angdeg(3))to be checked...
(multi-DATASET idea may be helpful to run such jobs ? 
 kindly see the help documents/Test_* files in ABINIT.) 

In general cases (ordinary inorganic systems), 
I guess total energy may be more dependent on acell, 
rather than on angdeg. So, if you have an estimation of 
angdeg(3), you might fix it and run somejobs by changing 
acell(3) first. Then you find acell(3) to minimize 
the total energy. Then check the variation of the total 
energy with respect to angdeg(3). 
You might notice the variation might be rather smaller 
than that with respect to acell. 
Still, you should have the 2-d contour map (in the above) 
to confirm the energy minimum.

b) nonlocal interaction (e.g. van der Waals)
For "rigorous" treatment of van der Waals, you might want to 
read related papers, for example, 
 "Accurate density functionals: Approaches using the
 adiabatic-connection fluctuation-dissipation theorem" 
Martin Fuchs and Xavier Gonze, Phys. Rev. B65, 235109 (2002),
and references therein... Try also the related tests available 
in ABINIT.

But I do not know whether such treatment can practically deal
with your targets or not... At present, I think that there are
no "rigorous and convenient" Exc functionals for van der Waals, 
for larger systems ... 

By chance, LDA may reproduce sort of van der Waals, for example,
in graphite/fullerene systems (i.e. lattice constant may be anyhow
reproduced with rather high energy cut-off), but it seems very 
"empirical"... Who knows the reason ? In fact, I also noticed 
that GGA cannot not reproduce the van der Waals in graphite and
C60 fcc solid. 
So we need to pile up our experience for "difficult" systems ...

Hope this helps,
Masayoshi

P.S. Please note that:
1) e-mails containing "two byte characters"(like chinese characters)
may induce  in "Subject". So, two byte characters might be
avoided, if one would not like to see "" in Subject.
2) Different question could be split from the first post, 
for someone to post follows.
3) Please give an appropriate "Title" to your post. Thanks.

From: <xzp02@mails.tsinghua.edu.cn>
Subject: [abinit-forum]  
Date: Sat, 13 Dec 2003 10:38:51 +0800
Message-ID: 
<271282920.19440@mails.tsinghua.edu.cn>,<002e01c3c122$415e3e10$62246fa6@ismserver>

> Dear ABINITers:
>     I am trying to relax a supercell, how can I constraint acell(1),
> acell(2), angdeg(1) and angdeg(2) and force the cell relax only with
> acell(3) and angdeg(3) (angle between rprim(1) and rprim(2))? I have
> browsed the optcell item, but I do not know how to deal with that. Can
> anyone give me some advice?
>    Another question, how can I take into account the nonlocal
> interaction(such as van der Waal interaction) convictively? Thanks.
> 
> Best Regards.
> 
> ZP
> 
> -----------------------------------------------
> Xu ZhiPing
> Ph.D Candidate
> Room: 3405, Phone: 86-10-627-83814
> Department of Mechanics,
> Qinghua University, Beijing, CHINA
> -----------------------------------------------