The ABINIT Users Mailing List ( CLOSED )

[marc.torrent@cea.fr]

Aman,

In fact, your probably right... rc does not really affect the computation of 
vhnzc.

VH(nz+nc) is obtained as follows:
Decuce Vcore(r) from core-density with Poisson's equation, and then
VH(nz(r)+nc(r))=Vcore(r)-Z/r
Vcore has to be small far from nucleus ; that's the reason why we test that 
VH(nz+nc)=-Z/r above a given radius (We could directly test vcore, of course).

Core-density is given by N. Holzwarth's AtomPAW from 0 to "infinity" (a very 
large radius)... So, it actually is not affect by rc... and consequently 
Vhnzc is not.

I will change that "INFO" message in future versions of AtomPAW2Abinit, so as 
to make it less mysterious.

The second solution I suggested to you is still useable...
Try it, if you want (change vloc_meshsz_max in calc_vhnzc.f; default is 3000 
max.).. but 3000 points is a relatively high value for Abinit above which CPU 
time may be large. Change this value if you want a high accuracy; otherwise, 
leave it unchanged.


Marc


-----------------------------------------------------------------
# From: "Amancherla, Sundar (CORP, GEITC)" <Sundar.Amancherla@geind.ge.com>


Marc,

Thanks very much for the suggestions -

I tried with 3 rc values - 3.1, 3.3, 3.4

However, I do get the same message.
--------------
Info:
Max. value of mesh size for Vloc=Vhnzc(r) has been reached.
Vhnzc( 3000)= -Z/r+ -7.320E-06
Precision of Vloc in Abinit's psp file will be poor.
Possible action: increase sphere radius.
--------------

Does rc does not affect vhnzc ?

--
Aman


-----Original Message-----
From: marc.torrent@cea.fr [mailto:marc.torrent@cea.fr]
Sent: Friday, October 17, 2003 1:58 AM
To: forum@abinit.org
Subject: [abinit-forum] Re: atompaw2abinit


Hi,

This message is just an "informative" message ;
If PAW atomic data were perfect, then Vloc(R)=-Zv/R above a given radius R.
In "real computations" this is of course not the case...
but lower is the difference between Vloc and Z/R better is the result.

A good (empirical) precision is 10-7 or 10-8... but 10-6 is not too bad...

If you want to get better accuracy, two solutions:
1-simple but perhaps not possible: increase sphere radius (only if your
system allows it : spheres cannot overlap).
2-more complex: modify in the source code the radius above which Vloc is
taken as -Zv/r ; this parammeter is set by vloc_meshsz_max in subroutine
calc_vhnzc. You can increase this value but NOT TOO : this can increase
dramatically CPU time in Abinit.

An advice : leave your data unchanged (or increase vloc_meshsz_max just a
little)...



Marc Torrent



Note: what we call Vloc here actually is: Vhartree(nval(r)+ncore(r))



-----------------------------------------------------------------
# From: "Amancherla, Sundar (CORP, GEITC)" <Sundar.Amancherla@geind.ge.com>

I am trying to use atompaw to get a paw potential for Y. Prof Holzwarth was
kind enough to send me an input file for Y. The rc used is 3.0. I am getting
these messages after running atompaw2abinit -

--------------
Info:
Max. value of mesh size for Vloc=Vhnzc(r) has been reached.
Vhnzc( 3000)= -Z/r+ -7.320E-06
Precision of Vloc in Abinit's psp file will be poor.
Possible action: increase sphere radius.
--------------

Pls do let me know any comments / suggestions that you may have. Thanks very
much
--
Aman