The ABINIT Users Mailing List ( CLOSED )

["Andrew M. Rappe" <rappe@sas.upenn.edu>]

Dear Masayoshi,

I agree with everything you have said.
I can just add a few thoughts.

1. OPIUM oxygen pseudopotentials at 50 Ry are quite accurate, and you
may be able to get away with a bit lower cutoff (40 Ry) for ionic
solids.  For organic molecules, 50 Ry is needed due to short C-O and
H-O distances.

2. It is important to use semi-core electrons for both V and Y.  Was
this done?  (Then of course, transferability must be tested both for
the semicore electrons and the valence states.)

3. At least one of the phases of V2O5 is quite tricky to get right.  I
think that it's a layered solid, only held together by van der Waals
and weak dipole-dipole electrostatic forces along one direction.  So
getting the lattice constant right will be very difficult even for
V2O5, along the direction with no covalent bonding.

Andrew


On Fri, Mar 19, 2004 at 11:09:11AM +0900 or thereabouts, Masayoshi Mikami 
wrote:
> Hello,
> 
> I would like to ask some questions, based on the FAQ:
> http://www.abinit.org/ABINIT/Infos/FAQ.html#Difference
> 
> 1) The FULL input would be more helpful to grasp what you did.
>      (to avoid guessing what you did)
> 2) Which kind of pseudopotentials were adopted ?
>     I mean, the pseudopotentials were taken from the ABINIT WEB site ?
>     They were produced by a PP-making code (e.g. FHI98PP, OPIUM, ...) ?
>     If so, which electron configurations were adopted for the PP-making  
> ?
>    (which states were treated as core, which states as valence ?)
>     It might be a good chance to read some documents such as
>     ~ABINIT/Infos/Psp_infos/psp1.data . Literature survey may also  
> help...
> 3) In general, oxygen pseudopotential need high energy cut, such as  
> 60-70 Ry
>     for Troullier-Matins type. (the ecut will depend on the cut-off  
> radii.)
>     So, 30Ry-ecut results may deviate from plausible physical values.
> 4) How about simpler ** binary ** compounds to check the  
> pseudopotentiasl ?
>     For example, Y2O3, V2O5 ? (here, it might be better to keep the  
> charge states
>     of Y (3+) and V(5+)...)  Unfortunately, they might not be easy to  
> calculate
>     due to the size of the unit-cell... The primitive cell of Y2O3  
> needs 40 atoms;
>     the conventional cell has 80 atoms !)
>     I guess V-potential might be more difficult to treat than  
> Y-potential,
>     so V2O5 might be a good case to check the transferability of  
> V-potential ...
> 
> Best wishes,
> Masayoshi
> 
> On 2004/03/19, at 3:21, Aldo Humberto Romero wrote:
> 
> >
> >Dear Ab init users,
> >
> >I have been trying to do a calculation for YVO4 (with a cell very
> >similar to the Zircon structure, just published in PRB vol 63, 104305)
> >but using LDA and TM pseudo I am getting an error on the lattice  
> >parametrs
> >of almost 6%, which of course is unaceptable. I have changed the cutoff
> >(from
> >30Ry to 60Ry) and the situation get worst.. K point is not an issue
> >(I have used a Monkhorst-Pack mesh of 4x4x4  and 5x5x5 and the  
> >diffference
> >is less than 1 meV in the total energy...
> >
> >Any avice? something I should be aware of with respect to this
> >system?.
> >
> >Thanks in advance!
> >
> >-aldo.
> >
> >
> >----------------------------------------------------------------------- 
> >--
> >Prof. Aldo Humberto Romero   
> >IPICyT
> >Advanced Materials Department
> >Av. Venustiano Carranza 2425-A,
> >78270 San Luis Potosi, SLP, Mexico
> >
> >email:  aldo@ipicyt.edu.mx
> >Phone:  (52)-444-8-33-54-11
> >Fax:    (52)-444-8-33-54-12
> >http://www.ipicyt.edu.mx/
> >
> >