This thread is locked.Only browsing is available.
Top Page > Browsing
Difference Charge Density
Date: 2010/12/10 00:03
Name: Gabriel Greene   <gabriel.greene@tyndall.ie>

Dear Dr. Ozaki,

I am reading through the technical notes for OpenMX, specifically "Total Energy and Forces: Ver. 1.2" and "Charge Mixing Methods: Ver. 1.0".

In regard to the kerker mixing scheme, can you explain the reasoning behind defining a difference charge density, dn(r), before fourier transforming and mixing in reciprocal space? I am wondering, why not just fourier transform n(r) directly and mix n(q) before back-transforming to get n(r) again..?

Thanks for the help,

Gabriel Greene
メンテ
Page: [1]

Re: Difference Charge Density ( No.1 )
Date: 2010/12/10 15:30
Name: T.Ozaki

Hi,

> why not just fourier transform n(r) directly and mix n(q) before back-transforming
> to get n(r) again..?

At least there are two evident reasons to do that.

(i)
Since the Hartree potential by the difference charge density is much shallower
than that by the original charge density, the numerical integration for constructing
the Hamiltonian matrix elements can be performed by a coarser regular mesh.

(ii)
The rearrangement of coulombic energy terms by Eq. 19 allows us to avoid the Ewald-type
sum for the sum of core repulsion energy, and to make the most of majority energy
contributions short range.

> In the Charge Mixing Methods: Ver.1.0, under Section 3, "Kerker mixing in
> momentum space", the kerker factor is defined as
> w(q) = |q|^2 / (|q|^2 + q0^2)
> where q0 = gamma*|qmin|

gamma is an adjustable parameter which controls the degree of suppression for
charge sloshing, smaller less suppression, larger more suppression.
In OpenMX gamma does correspond to scf.Kerker.factor.

Regards,

TO
メンテ

Page: [1]