[Pw_forum] Questions on spin-polarized calculations

[yhli]

Dear pwscf users,

I am trying to perform spin-polarized calculations using pwscf,
following the tutorial for LSDA found here:
http://www.fisica.uniud.it/~giannozz/QE-Tutorial. Unfortunately, I am
puzzled by several problems.

*Q1. **Can pwscf perform spin-polarized calculations using GGA functionals?*
I know this seems to be a stupid question, since spin-polarized GGA
calculation should be one of the basic capabilities of an /ab inition
/program. I checked the mannual for pwscf (INPUT_PW.html), and found
that "npsin=2" enables "LSDA". Also, in the tutorial mentioned above, it
reads "This approach goes under the name of Local Spin-Density
Approximation, or LSDA, even when the functional E xc is based not on
LDA but on GGA". I guess pwscf can do such calculations, but not convinced.

*Q2. Why the results of "fixed magnetization**" and "**Unconstrained
magnetization" are not consistent?*
In the tutorial I read that there are two approaches to optimizing the
magnetization. One is to vary the tot_magnetization mannually and to
find the minimum of total energy, while in the other approach the total
magnetization is determined during scf calculation by pwscf
automatically. I tried both approaches for bulk silicon and magnisium
oxide, which should be both non-magnetic, and found both approaches
predicted non-magnetic groud state. However, for a 2x2x2 super cell of
MgO dopped with one atom of Sc(scandium), the fixed magnetization
approach predicted the total magnetization should be 1 bohr, while
unstrained approach predicted it to be ~0.80. What's more, the value of
smearing also affects the total magnetization for the second approach.
Why are not they consistent? Which one should I trust?

*Q3. **How to specify starting_magnetization(i)?*
Are there any tricks to specify reasonable starting_magnetization for
different atomic species? Perhaps the magnet momentum of an isolated
atom is a good guess, but how to relate it to starting_magnetization? I
guess that they are related by the equation "starting_magnetization =
(nelec_spin_majority - nelec_spin_minority) / nelec_total", since for
all spin-up case starting_magnetization is 1.0 and for all spin-down
case it is -1.0. But I am not sure.

*Q4. Should total magnetization always be intergers?*
As mentioned in Q2, the total magnetization is fractional when
unstrained magnetization approach is used. Since each electron carries
one bohr of magneton, should the total magnetization always be integers?

All suggestions are appreciated.

Best,
Yunhai Li

Department of Physics, Southeast University
Nanjing, Jiangsu Province, P.R.C.


---
此电子邮件没有病毒和恶意软件，因为 avast! 防病毒保护处于活动状态。
http://www.avast.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20141027/b810eef4/attachment.html>