[Pw_forum] How does PDOS projection define xyz directions

[Sai Kumar Ramadugu]

Dear QE Users,


I'm confused about how the projwfc.x does the projection to produce PDOS
plots.

I'm running the calculations of bulk magnetite Fe3O4. It has two ways to
set up the

structure: cubic (ibrav = 1, with 56 atoms) cell and hexagonal (ibrav = 4,
with 42

atoms) cell. I did PDOS calculation for both of them and checked different
d-orbital

components (eg: dz2 and dx2-y2, and t2g: dxy, dxz and dyz) for octahedral
Fe. I found

the five different d-orbital distributions are very different for the same
type of Fe.


The PDOS of different d-orbitals of octahedral Fe in Fe3O4 cubic cell shows
perfect,

classical textbook-style octahedral crystal field splitting, where dz2 and
dx2-y2 are

degenerate and dxy, dxz and dyz are degenerate. And their energy levels
also match

theoretical bonding, nonbonding and antibonding model.


However, the PDOS of different d-orbitals of octahedral Fe in Fe3O4
hexagonal cell

is much more complicated than the cubic one. None of the d-orbitals are
degenerate.

Although I can tell that their total d-orbital distribution must be the
same with the one

we got from the cubic cell, the "dz2, dx2-y2, dxy, dxz and dyz" here are
NOT the "real

ones" that we expected to see as those in cubic case. For example, a peak
that is

supposed to be only contributed by dz2 or dx2-y2 (eg) orbital, is a mixture
of all the five

d-orbitals. So now we can't get the correct d-orbital eg and t2g
occupancies in

hexagonal cell.


I'm thinking why different cell type can make such big difference in PDOS
distributions.

I guess the "x, y, z" directions defined by projwfc.x in PDOS projection is
"cell vector-

dependent". In cubic Fe3O4, all the six Fe-O bonds around octahedral Fe are
parallel

to one of the a, b and c cell vectors. So the "x, y, z" in PDOS projection
happen to be the

same with the "x, y, z" in Fe crystal field splitting. But in hexagonal
Fe3O4, all of the

octahedral Fe-O bonds are off the cell vectors. So the default "x, y, z" in
projection are no

longer the same as those that we are looking for on octahedral Fe sites.


So now I'm thinking how to change the way the projwfc.x does the
projection. Because

Fe3O4 is just a test for us. We are actually focusing on hematite Fe2O3,
which can only

be presented by hexagonal (or rhombohedral) cell where all the Fe-O are off
the cell

vectors. I found that there is something wrong with the eg and t2g PDOS but
don't know

how to correct it. There seems to be a file called ".../flib/ylmr2.f90". Is
this the one

controlling projection directions? Or is there any other way that we can
let projwfc.x does

the PDOS projection along the directions that we really want?


I am attaching the figures of pdos calculations for cubic Fe3O4 and
hexagonal Fe3O4.


Any suggestions are welcome.


Thank you,


----------------------------

Sai Ramadugu

University of Iowa
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20141229/f65b1982/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: mag_pdos_hex.png
Type: image/png
Size: 29655 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20141229/f65b1982/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: mag_pdos_cub.png
Type: image/png
Size: 20286 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20141229/f65b1982/attachment-0001.png>