[Wannier] Wannier90-VASP: Dipole moment of isolated H20

[Nicola Marzari]

Thanks Maurice for the update!

Two comments: the WFs you want should always be constructed from the 4 
lowest occupied states (of course, you can include empty states for
different "disentanglement" purposes - but let's say here you are
focused on the usual water WFs).

Then, the results should not be dependent on having a VASP calculation 
use empty states or not to converge to the ground state (depending on 
circumstances, those could help or not the speed of the convergence,
but for a crystalline insulator or a molecule with filled shells the
filled manifold at self-consistency should be identical no matter how 
many empty states you use).

				nicola


On 02/10/2013 19:28, Maurice de Koning wrote:
> Hi Nicola,
>
> This has been a while, but I recently started looking again into the problem I had using the Wannier90-VASP interface in
> reproducing the WC's for the isolated water molecule as in the paper by Silvestrelli and Parrinello (JCP 111 3572,1999) .
>
> Initially I was not able to reproduce the WCs, even using exactly the same calculation parameters as in Silvestrelli's paper. Now I have
> found out that the results are quite sensitive to the number of bands that you include in the KS calculation and in the posterior Wannierization.
> By default VASP always includes a certain number of empty bands, unless you explicitly tell it not do so. In my first calculations, the Wannier optimization
> always included a number of such empty bands and the results for the 4 WCs associated with the covalent bonds and the lone pairs did not agree with those
> reported by Silvestrelli. Now, if you force VASP to work only with the 4 occupied bands and Wannierize only these, the results  then do agree with those of Silvestrelli.
>
> So it seems, at least in this case, that one should only consider the occupied bands in the calculation.
>
> Cheers,
>
> Maurice
>
>>
>>
>> Dear Maurice,
>>
>>
>> not sure if this ever had a follw up. Note that the Wannier formalism
>> uses a reciprocal space representation to descrive <r> and <r^2> - so
>> quantities like the spreads (very much so) and the centers (much less)
>> depend on k-point sampling (or one the size of the unit cell, if using gamma sampling) much mroe that one usually expects - you could have
>> used a k-point mesh fine enough to converge the electronic structure
>> (or a supercell largge enough), but not enough to converge the Wannier-related quantities.
>>
>> For your water molecule, you should be able to calculate its dipole in
>> real space just as a charge integral in space, or using a Wannier code -
>> and you could plot convergence of these quantities.
>>
>> Some comments on this matter should be also in an old PRB of Umari and
>> Pasquarello.
>>
>> All the best,
>> 	
>> 		nicola
>>
>>
>>
>>
>>
>> On 26/06/2012 19:25, Maurice de Koning wrote:
>>> Hello all,
>>>
>>> We have started using the Wannier 90 interface to VASP and as a first test case we are trying to
>>> reproduce the Wannier centers for the isolated water molecule as in the paper by Silvestrelli and Parrinello (JCP 111 3572,1999).
>>>
>>> Using the PAW approach, the PBE xc functional, a cubic periodic box with side 10.6 Angstrom, a plane-wave kinetic energy cut-off of 1000 eV and Gamma-point sampling, the
>>> relaxed geometry of the water molecule is identical to that reported by Silvestrelli, with an O-H distance of 0.972 Angstroms and an internal angle of 104.4 degrees.
>>>
>>> However, when using the Wannier90 interface to VASP, the Wannier centers that we obtain do not coincide with those reported in Silvestrelli. The position of the Wannier
>>> centers corresponding to the lone pairs are reasonable. We find distances of 0.306 Angstroms from the oxygen ion, and an internal angle of 120.8 degrees, which agrees reasonably well with the  values of 0.300 Angstrom and 126.4 degrees reported by Silvestrelli.
>>>
>>> The WFCs associated with the electrons in the covalent bonds, however, are very different. We find a distance from the oxygen ion of 0.47 Angstrom versus a value of 0.53 Angstrom reported by Silvestrilli. These differences lead to quite a different electric dipole moment.
>>>
>>> Any suggestions as to what might be the reason for these discrepancies?
>>>
>>> Cheers,
>>>
>>> Maurice
>>> _______________________________________________
>>> Wannier mailing list
>>> Wannier at quantum-espresso.org
>>> http://www.democritos.it/mailman/listinfo/wannier
>>>
>>
>>
>> --
>>
>> ----------------------------------------------------------------------
>> Prof Nicola Marzari, Chair of Theory and Simulation of Materials, EPFL
>


-- 

----------------------------------------------------------------------
Prof Nicola Marzari, Chair of Theory and Simulation of Materials, EPFL