This thread is locked.Only browsing is available.
Top Page > Browsing
Electronic Temperature in openmx
Date: 2013/06/08 05:26
Name: Ruibin Liu

Hi,

How does the 'Electronic temperature' come into calculation in openmx? From the manual, we know that SCF calculation could get converged more quickly if we set a higher Electronic temperature. The default value is 300k. What does this value mean for electrons? Do you have any tips for setting a proper temperature?

Another question, if we want to calculate the doublet, triplet or quartet state, what kinds of keywords should be specified? And, as we can find out from many examples, the up and down spins of H are both set to be 0.5. what is the difference compared to the case if we set one H to be 0 and 1, another H to be 1 and 0?


Best,
Ruibin
メンテ
Page: [1]

Re: Electronic Temperature in openmx ( No.1 )
Date: 2013/06/09 08:18
Name: T. Ozaki

Hi,

> How does the 'Electronic temperature' come into calculation in openmx?

The electronic temperature comes into the Fermi-Dirac function which determines
the occupation number of each state. As you can imagine, the function with the zero
temperature is a step function which may lead to a sudden exchange of occupied states
during the SCF iteration especially for small gap systems and metals, resulting in
an instability of the SCF convergence. On the other hand, one can see that a higher
temperature avoids such a situation, and also electrons are excited to originally
unoccupied states which have more nodes in the wave functions. The excitation shortens
correlation length in the density matrix, resulting in the easier SCF convergence.


> if we want to calculate the doublet, triplet or quartet state, what kinds of
> keywords should be specified?

I know that other codes such as Gaussian, GAMESS, and Dmol support the functionality.
However, OpenMX does not support the functionality.


> the up and down spins of H are both set to be 0.5. what is the difference compared
> to the case if we set one H to be 0 and 1, another H to be 1 and 0?

To explore possible magnetic states, one can control an initial spin state for each
atom, which allows us to start the SCF iteration from non-magnetic, antiferro-, ferri-,
ferro-, and even non-colliear magnetic states. How many magnetic states appear strongly
depends on the degree of freedom in spin space of the system you are interested in.
By comparing the total energies of magnetic states, one can determine the possible
ground state.


Best regards,

TO
メンテ
Re: Electronic Temperature in openmx ( No.2 )
Date: 2013/06/13 08:29
Name: Ruibin Liu

Hi To,

Thanks a lot!

We can expect that the HOMO level of a semiconductor nanocrystal with 500 electrons to be 250 without considering splitting states when the electronic temperature is 0 K. If we set a high electronic temperature like 500 K for the same system, it is possible that the HOMO level is one state lower or higher than 250, not just because of degeneracy. What does it mean for a lower or higher HOMO state? And here is an example:

Number of States = 1344.00000000000000
HOMO = 671
Eigenvalues
*
*
*
670 -0.12630226803230 -0.12630226803230
671 -0.12624946934887 -0.12624946934887
672 -0.12572610208562 -0.12572610208562
673 -0.12442818012200 -0.12442818012200
*
*
*


Best,
Ruibin


メンテ
Re: Electronic Temperature in openmx ( No.3 )
Date: 2013/06/13 09:26
Name: T. Ozaki

Hi,

> it is possible that the HOMO level is one state lower or higher than 250,
> not just because of degeneracy.

Yes, it is. But this happens due to (near) degeneracy.
When we calculate a cluster system with no spin polarization, the HOMO level should
be exactly the half of the number of electrons. If not, this simply implies that there
must be (near) degenerate states near HOMOs and/or LUMOs, and one should be aware that
the HOMO level at 0 K (no temperature smearing) is the half of the number of electrons
for an isolated cluster system with no spin polarization.

In case of bulk metals where many bands intersect with the chemical potential, the
determination of the highest occupied level at each k-point is not trivial anymore,
thus OpenMX uses such a way in determining the 'HOMO' level.

Regards,

TO
メンテ
Re: Electronic Temperature in openmx ( No.4 )
Date: 2013/06/14 23:57
Name: Ruibin Liu

Hi TO,

Thanks again.

As for my example, the HOMO level is 671 and its eigen energy is near that of 670, not 672. If we output MOs, does it mean that the file for homo0_0.cube file is for level 671, homo0_1.cube for 670, lumo0_0.cube for 672, and so on?

If so, let's consider a scenario that we have the same system but a weak external electric field is added to it. The result shows that the HOMO level is now 673. Is it meaningful if we compare the HOMO orbitals from homo0_0.cube files of these two cases?


Best,
Ruibin
メンテ
Re: Electronic Temperature in openmx ( No.5 )
Date: 2013/06/16 12:00
Name: T. Ozaki

Hi

> If we output MOs, does it mean that the file for homo0_0.cube file is for level 671,
> homo0_1.cube for 670, lumo0_0.cube for 672, and so on?

Yes, it does. In order to make sure if the assignment is correct or not, you can
check the eigenvalue of each state. The eigenvalue of each state and the chemical potential
can be found in the first and second lines in the cube file, respectively.
By comparing those with the eigenvalues of all the states in *.out, you can easily confirm
what states you are looking.

> If so, let's consider a scenario that we have the same system but a weak external electric
> field is added to it. The result shows that the HOMO level is now 673. Is it meaningful if
> we compare the HOMO orbitals from homo0_0.cube files of these two cases?

As explained above, once you are sure what they are, such a comparison will be meaningful.

Regards,

TO
メンテ

Page: [1]