This thread is locked.Only browsing is available.
Top Page > Browsing
KS determinant
Date: 2013/05/30 08:32
Name: Ruibin Liu   <rainbee1988@gmail.com>

Hi,

Can we get the Kohn-Sham determinant for constrained and non-constrained DFT?


Regards,
Ruibin
メンテ
Page: [1]

Re: KS determinant ( No.1 )
Date: 2013/05/30 12:11
Name: T. Ozaki

Hi,

There is no direct keyword enabling us to get the KS determinant.
However, one might be able to construct it by using the one particle KS
eigenstates calculated by OpenMX.

Regards,

TO
メンテ
Re: KS determinant ( No.2 )
Date: 2013/06/01 08:02
Name: Ruibin Liu  <rainbee1988@gmail.com>

Hi TO,

Thanks for replying. Since we cannot get KS eigenstates by using O(N) methods only, we cannot get KS determinant even in your suggesting method, right?

Suppose we are performing 'cluster' calculation for a single molecule with 100 valence electrons, what kinds of keywords should we specify to get the eigenstates for making the KS determinant?

By the way, how can we explain that the position of HOMO is sometimes smaller than half of number of states, while sometimes equal to or larger than?
1. Chemical Potential (Hartree) = -0.18529881613709
Number of States = 4052.00000000000000
HOMO = 2025

2.Chemical Potential (Hartree) = -0.22543578783091
Number of States = 4104.00000000000000
HOMO = 2052


3.Chemical Potential (Hartree) = -0.14780448978233
Number of States = 4076.00000000000000
HOMO = 2041


Best,
Ruibin
メンテ
Re: KS determinant ( No.3 )
Date: 2013/06/04 10:27
Name: T. Ozaki

Hi,

> Since we cannot get KS eigenstates by using O(N) methods only, we cannot get KS
> determinant even in your suggesting method, right?

The O(N) method does not calculate eigenstates but local density of states.
So, it would be difficult to construct the KS determinant within the O(N) method.

> Suppose we are performing 'cluster' calculation for a single molecule with 100
> valence electrons, what kinds of keywords should we specify to get the eigenstates
> for making the KS determinant?

Using the keyword 'HS.fileout'

HS.fileout on # on|off, default=off

one can get the KS Hamiltonian and overlap matrix. By diagonalizing the KS eq. with
these matrices, the KS eigenstates are obtained.

Also, by setting the keyword 'level.of.fileout=2', the LCAO coefficients are obtained.
See also the manual:
http://www.openmx-square.org/openmx_man3.7/node80.html

If you want to get value of the eigenstates at each grid point in real space, you can
output cube files of the eigenstates. See also the manual:
http://www.openmx-square.org/openmx_man3.7/node22.html


> By the way, how can we explain that the position of HOMO is sometimes smaller
> than half of number of states, while sometimes equal to or larger than?

As for how to determine the HOMO level, see the following discussions:
http://www.openmx-square.org/forum/patio.cgi?mode=view&no=995
http://www.openmx-square.org/forum/patio.cgi?mode=view&no=975

Such an situation may happen if there are nearly degenerate states near the Fermi level.

Regards,

TO
メンテ
Re: KS determinant ( No.4 )
Date: 2013/06/04 04:20
Name: Ruibin Liu

Hi TO,

Thanks a lot for the two replies! They really help me.

Best,
Ruibin
メンテ
Re: KS determinant ( No.5 )
Date: 2013/06/29 08:18
Name: Ruibin Liu

Hi TO,

Questions about KS determinant from O(N) methods again.

There is a section '23 Large-scale calculations' in the openmx manual which says we can combine O(N) method and conventional diagonalization method by using the former self-consistent charge density, so that we can obtain full wave functions. Is this method reliable and efficient? Because I didn't find any papers using this approach.

Another question, can we perform constraint DFT (Voorhis et.al., J. Chem. Phys. 125, 164105 (2006)) calculation in openmx?


Regards,
Ruibin




メンテ
Re: KS determinant ( No.6 )
Date: 2013/07/01 15:32
Name: T. Ozaki

Hi,

> Because I didn't find any papers using this approach.

As for the issue, please take a look at the page 10 in PRB 74, 245101(2006).

> Another question, can we perform constraint DFT (Voorhis et.al., J. Chem. Phys. 125, 164105
> (2006)) calculation in openmx?

There is no such a functionality in OpenMX.

Regards,

TO
メンテ

Page: [1]