[Pw_forum] PDOS output for spin-orbit calculation

[Henry J Seeley]

Dear Users,

I'm having some difficulty understanding the what the output files from
projwfc.x are actually showing. I've searched through the forum and
couldn't find a very informative answer regarding this question. The
projwfc.x input file documentation is not very helpful either, but
perhaps my understanding of total angular momentum eigenfunctions is the
limiting factor.

I currently have output files named (for example):
prefix.pdos_atm#1(Pb)_wfc#3(p_j1.5)

With data headers:
# E(eV) ldos(E) pdos(E)_1 pdos(E)_2 pdos(E)_3 pdos(E)_4

Now the first two columns are easy to understand, but the remaining four
don't make sense to me. I've gathered that they refer to different m_j
states, but they are labeled 1,2,3,4 and this confuses me. Also, #'s 1/4
are identical, as are 2/3. I'm assuming this is because I didn't specify
a starting magnetization.

What exactly do these four pdos atomic states represent?


Thank you very much,
Hank Seeley
Chemistry PhD student
University of Oregon

[Mohsen Modaresi]

Hi Hank,

In the end of projwfc input filediscription you can find enough usefull
information about the order of mi states.
http://www.quantum-espresso.org/wp-content/uploads/Doc/INPUT_PROJWFC.html

Also you should know in the presence of SOC the PDOS is some how more
complicated and spherical harminics are not eigen states of system and QE
uses total angular momentum (j), please see,
A. Dal Corso, A.M. Conte, Phys. Rev. B 71 (2005) 115106

Best,

Post by Henry J Seeley
Dear Users,
I'm having some difficulty understanding the what the output files from
projwfc.x are actually showing. I've searched through the forum and
couldn't find a very informative answer regarding this question. The
projwfc.x input file documentation is not very helpful either, but
perhaps my understanding of total angular momentum eigenfunctions is the
limiting factor.
prefix.pdos_atm#1(Pb)_wfc#3(p_j1.5)
# E(eV) ldos(E) pdos(E)_1 pdos(E)_2 pdos(E)_3 pdos(E)_4
Now the first two columns are easy to understand, but the remaining four
don't make sense to me. I've gathered that they refer to different m_j
states, but they are labeled 1,2,3,4 and this confuses me. Also, #'s 1/4
are identical, as are 2/3. I'm assuming this is because I didn't specify
a starting magnetization.
What exactly do these four pdos atomic states represent?
Thank you very much,
Hank Seeley
Chemistry PhD student
University of Oregon
_______________________________________________
Pw_forum mailing list
http://pwscf.org/mailman/listinfo/pw_forum

[Henry J Seeley]

Thank you for your swift reply.

I tried looking through the input file description, and I did not
understand exactly what I was reading. I was under the impression that
the section titled "Orbital Order" was only for non-spin orbit
calculations. I think my understanding of total angular momentum is
hurting me here.

Is there a way to project the wavefunctions I obtained from a spin-orbit
scf onto the classical AO states (px,py,pz) using projwfc.x?

Thank you,
Hank Seeley

Post by Mohsen Modaresi
Hi Hank,
In the end of projwfc input filediscription you can find enough
usefull information about the order of mi states.
http://www.quantum-espresso.org/wp-content/uploads/Doc/INPUT_PROJWFC.html
[2]
Also you should know in the presence of SOC the PDOS is some how more
complicated and spherical harminics are not eigen states of system and
QE uses total angular momentum (j), please see,
A. Dal Corso, A.M. Conte, Phys. Rev. B 71 (2005) 115106
Best,

Post by Henry J Seeley
Dear Users,
I'm having some difficulty understanding the what the output files from
projwfc.x are actually showing. I've searched through the forum and
couldn't find a very informative answer regarding this question. The
projwfc.x input file documentation is not very helpful either, but
perhaps my understanding of total angular momentum eigenfunctions is the
limiting factor.
prefix.pdos_atm#1(Pb)_wfc#3(p_j1.5)
# E(eV) ldos(E) pdos(E)_1 pdos(E)_2 pdos(E)_3 pdos(E)_4
Now the first two columns are easy to understand, but the remaining four
don't make sense to me. I've gathered that they refer to different m_j
states, but they are labeled 1,2,3,4 and this confuses me. Also, #'s 1/4
are identical, as are 2/3. I'm assuming this is because I didn't specify
a starting magnetization.
What exactly do these four pdos atomic states represent?
Thank you very much,
Hank Seeley
Chemistry PhD student
University of Oregon
_______________________________________________
Pw_forum mailing list
http://pwscf.org/mailman/listinfo/pw_forum [1]

--
Mohsen Modarresi,
PhD student of Solid State Physics, Ferdowsi University of Mashhad, Iran.
Phone [phone]
------
[1] http://pwscf.org/mailman/listinfo/pw_forum
[2]
http://www.quantum-espresso.org/wp-content/uploads/Doc/INPUT_PROJWFC.html
_______________________________________________
Pw_forum mailing list
http://pwscf.org/mailman/listinfo/pw_forum

[Henry J Seeley]

Dear Users,

I'm doing calculations to obtain the surface energy for a variety of
different crystal surfaces, so I'm running both bulk and slab
calculations of my systems.

I've been under the impression that I should be using k-grids of the
type NxNx1 for slab calculations, according to what I've read in
tutorials. But I've looked in the literature, and have found people
using both NxNxN and NxNx1 k-grids for their calculations. Which should
I be using? Is an NxNxN grid not always better?

Thank you for your time,
Hank Seeley
Chemistry PhD student
University of Oregon

[Nicola Marzari]

For a slab, in vacuum, there is no dispersion in the direction perpendicular
to the surface, so the NxNx1 sampling and the NxNxN sampling should be
identical.
(Of course the latter is N times more expensive, CPU-wise)

It they are not, it's because there is not enough vacuum.

Try it...
nicola

Post by Henry J Seeley
Dear Users,
I'm doing calculations to obtain the surface energy for a variety of
different crystal surfaces, so I'm running both bulk and slab
calculations of my systems.
I've been under the impression that I should be using k-grids of the
type NxNx1 for slab calculations, according to what I've read in
tutorials. But I've looked in the literature, and have found people
using both NxNxN and NxNx1 k-grids for their calculations. Which should
I be using? Is an NxNxN grid not always better?
Thank you for your time,
Hank Seeley
Chemistry PhD student
University of Oregon
_______________________________________________
Pw_forum mailing list
http://pwscf.org/mailman/listinfo/pw_forum

[Henry J Seeley]

Nicola,

Thank you very much!

-Hank

Post by Nicola Marzari
For a slab, in vacuum, there is no dispersion in the direction
perpendicular
to the surface, so the NxNx1 sampling and the NxNxN sampling should be
identical.
(Of course the latter is N times more expensive, CPU-wise)
It they are not, it's because there is not enough vacuum.
Try it...
nicola

Post by Henry J Seeley
Dear Users,
I'm doing calculations to obtain the surface energy for a variety of
different crystal surfaces, so I'm running both bulk and slab
calculations of my systems.
I've been under the impression that I should be using k-grids of the
type NxNx1 for slab calculations, according to what I've read in
tutorials. But I've looked in the literature, and have found people
using both NxNxN and NxNx1 k-grids for their calculations. Which should
I be using? Is an NxNxN grid not always better?
Thank you for your time,
Hank Seeley
Chemistry PhD student
University of Oregon
_______________________________________________
Pw_forum mailing list
http://pwscf.org/mailman/listinfo/pw_forum

[Henry J Seeley]

Dear Users,

I've been using pp.x to obtain ILDOS/PDOS/PSI^2 information, and I'm a
little surprised by how long it takes for the processing to finish. I've
used a variety of the pp.x features, and although each pp.x takes only a
couple minutes to finish, when running many of them to obtain specialty
information I find myself waiting days.
The processing gets stuck between the outputs:

" negative rho (up, down): 1.581E-03 0.000E+00"

and

" negative rho (up, down): 1.581E-03 0.000E+00".

I'm sorry if this isn't enough information to solve this issue, but I'm
hoping this is a simple fix for something I'm overlooking.

Thank you for your help,
Hank Seeley
PhD student, chemistry
University of Oregon

[Giovanni Cantele]

I would say that not enough information is included in your message.

What do you mean with “running many of them”? Maybe including several pp.x
executions in a single submission file for a queuing system? In the case, it might be
that the problem is not with pp.x but with the system where you are running it, or else
that you are using too many cpus for the execution that requires just one (for example
several weeks ago it was reported a similar problem running average.x on many cpus, that
was simply solved by running it on just one cpu).

Giovanni

Post by Henry J Seeley
Dear Users,
I've been using pp.x to obtain ILDOS/PDOS/PSI^2 information, and I'm a
little surprised by how long it takes for the processing to finish. I've
used a variety of the pp.x features, and although each pp.x takes only a
couple minutes to finish, when running many of them to obtain specialty
information I find myself waiting days.
" negative rho (up, down): 1.581E-03 0.000E+00"
and
" negative rho (up, down): 1.581E-03 0.000E+00".
I'm sorry if this isn't enough information to solve this issue, but I'm
hoping this is a simple fix for something I'm overlooking.
Thank you for your help,
Hank Seeley
PhD student, chemistry
University of Oregon
_______________________________________________
Pw_forum mailing list
http://pwscf.org/mailman/listinfo/pw_forum

[Henry J Seeley]

Thank you for the response,

By running many, I mean on the order of 100-1000 individual pp.x inputs
run in series from a looping script. I'm using ILDOS through many small
energy intervals to generate spatial maps of the LDOS for a system. I've
tried looking at PDOS boxes as an alternative, since I am interested in
getting better energy resolution than spatial resolution, but I had some
difficulty understanding how to generate boxes for the specific spatial
locations I was interested in. Maybe there is an easier way to
accomplish this?

I will try running on just one processor, perhaps that is the issue. I
will also consult with the administrators for the cluster I am using.

Thank you again,
Hank Seeley

Post by Giovanni Cantele
I would say that not enough information is included in your message.
What do you mean with “running many of them”? Maybe including several pp.x
executions in a single submission file for a queuing system? In the case, it might be
that the problem is not with pp.x but with the system where you are running it, or else
that you are using too many cpus for the execution that requires just one (for example
several weeks ago it was reported a similar problem running average.x on many cpus, that
was simply solved by running it on just one cpu).
Giovanni

Post by Henry J Seeley
Dear Users,
I've been using pp.x to obtain ILDOS/PDOS/PSI^2 information, and I'm a
little surprised by how long it takes for the processing to finish. I've
used a variety of the pp.x features, and although each pp.x takes only a
couple minutes to finish, when running many of them to obtain
specialty
information I find myself waiting days.
" negative rho (up, down): 1.581E-03 0.000E+00"
and
" negative rho (up, down): 1.581E-03 0.000E+00".
I'm sorry if this isn't enough information to solve this issue, but I'm
hoping this is a simple fix for something I'm overlooking.
Thank you for your help,
Hank Seeley
PhD student, chemistry
University of Oregon
_______________________________________________
Pw_forum mailing list
http://pwscf.org/mailman/listinfo/pw_forum

_______________________________________________
Pw_forum mailing list
http://pwscf.org/mailman/listinfo/pw_forum