[Pw_forum] smoothing PDOS issue (QE 5.0.2)

[David Grifith]

Hi

I usually use the following setting for DOS calculations and the resulting
DOS graphs are so smooth.

scf:
occupation='smearing'
smearing='gaussian'
degauss=0.001

nscf:
occupations='tetrahedra'

&dos:
DeltaE=0.1


but it seems that in the recent version of q/e the "smoothing" has been
removed and using the following setting my PDOS graphs are not smooth
anymore.

&projwfc
DeltaE=0.01 , ngauss=0 , degauss=0.001

when I set smoothing=0.3 as I did in the old versions of q/e the recent
version ignores the smoothing procedure and does nothing to smooth the PDOS
graph.

I would appreciate if anyone helps me to make PDOS graphs as smooth as DOS
ones. Many Thanks

[Paolo Giannozzi]

Post by David Grifith
when I set smoothing=0.3 as I did in the old versions of q/e the
recent version ignores the smoothing procedure and does nothing to
smooth the PDOS graph.
I would appreciate if anyone helps me to make PDOS graphs as smooth as
DOS ones.

Not sure what you define as "smooth". The DOS in a true crystal is far
from smooth: it has singularities and you need the tetrahedron method
to have a good DOS. With gaussian broadening, you may need a relatively
large broadening or else you will get a series of spikes, but then
you miss the singularities. dos.x can calculate DOS using tetrahedra
(this is what it should do according to your input data). projwfc.x
cannot calculate PDOS with tetrahedra, if I remember correctly. More
exactly: it could, provided somebody take the existing code that
calculates the PDOS for phonons (it should be routine dos_gam in
matdyn.f90), adapts it to electrons.

Anyway: without a simple test (with outputs) and a definition of
what are the "old" and "recent" versions, it is hard to say more.
I do not remember any recent change in the DOS-producing codes
that may explain the phenomenology you describe.

Paolo

[David Grifith]

Dear Paolo

thanks for your reply.

at the end of the following document it was mentioned that

" Obsolete variables, ignored:
io_choice
smoothing "


http://www.quantum-espresso.org/wp-content/uploads/Doc/INPUT_PROJWFC.html

in the following you can see my input files:

*scf:*
&control
calculation='scf',
restart_mode='restart',
prefix='B',
pseudo_dir = '../pseudo/',
outdir='../tmp/',
/
&system
ibrav= 0, celldm(1)=2.75899995, nat= 2, ntyp= 1,
ecutwfc = 40, ecutrho = 400, occupations='smearing',
smearing='gaussian', degauss=0.001,
nspin=2, starting_magnetization(1)=1
/
&electrons
conv_thr = 1.0e-10
mixing_beta = 0.3
/
CELL_PARAMETERS
1.50000000 0.86600000 0.00000000
1.50000000 -0.86600000 0.00000000
0.00000000 0.00000000 8.00000000
ATOMIC_SPECIES
C 12.0107 C.pbe-rrkjus.UPF
ATOMIC_POSITIONS {alat}
C 0.000000000 0.000000000 0.000000000
C 1.997267588 0.000001338 0.000000000
K_POINTS {automatic}
4 4 1 0 0 0


*nscf:*
&control
calculation='nscf',
restart_mode='restart',
prefix='B',
pseudo_dir = '../pseudo/',
outdir='../tmp/',
wf_collect=.TRUE.
/
&system
ibrav= 0, celldm(1)=2.75899995, nat= 2, ntyp= 1,
ecutwfc = 40, ecutrho = 400, occupations='tetrahedra',
nspin=2, starting_magnetization(1)=1
/
&electrons
conv_thr = 1.0e-10
mixing_beta = 0.3
/
CELL_PARAMETERS
1.50000000 0.86600000 0.00000000
1.50000000 -0.86600000 0.00000000
0.00000000 0.00000000 8.00000000
ATOMIC_SPECIES
C 12.0107 C.pbe-rrkjus.UPF
ATOMIC_POSITIONS {alat}
C 0.000000000 0.000000000 0.000000000
C 1.997267588 0.000001338 0.000000000
K_POINTS {automatic}
20 20 1 0 0 0

*dos:*
&dos
outdir='../tmp/'
prefix='B'
fildos='B.dos',
Emin=-40.0, Emax=20.0, DeltaE=0.1,
/

*pdos:*
&projwfc
outdir='../tmp/'
prefix='B'
Emin=-40.0, Emax=20.0, DeltaE=0.1
ngauss=0, degauss=0.001
/

overall the only way that I can make my PDOS graphs as smooth as DOS ones
is reducing degauss value to 0.01 ! but in some cases that we study the
proper value for degauss is 0.001 and reducing it we may miss something.
mightn't it ?
We are going to compare a large sets of PDOS and DOS graphs and prefer not
to compare one graph with lots of singularities with the others that are
too smooth. I appreciate your advice in advance. Many Thanks
--
Sincerely Yours
David G.
JCU

* when I set smoothing=0.3 as I did in the old versions of q/e the*>* recent version ignores the smoothing procedure and does nothing to*>* smooth the PDOS graph. *>* *>* I would appreciate if anyone helps me to make PDOS graphs as smooth as*>* DOS ones.*

*
*

on *Sat Apr 27 16:35:42 CEST 2013, Paolo Giannozzi wrote: * *
** *
Not sure what you define as "smooth". The DOS in a true crystal is far
from smooth: it has singularities and you need the tetrahedron method
to have a good DOS. With gaussian broadening, you may need a relatively
large broadening or else you will get a series of spikes, but then
you miss the singularities. dos.x can calculate DOS using tetrahedra
(this is what it should do according to your input data). projwfc.x
cannot calculate PDOS with tetrahedra, if I remember correctly. More
exactly: it could, provided somebody take the existing code that
calculates the PDOS for phonons (it should be routine dos_gam in
matdyn.f90), adapts it to electrons.

Anyway: without a simple test (with outputs) and a definition of
what are the "old" and "recent" versions, it is hard to say more.
I do not remember any recent change in the DOS-producing codes
that may explain the phenomenology you describe.

Paolo

[Paolo Giannozzi]

Post by David Grifith
io_choice
smoothing "

very obsolete: removed on 30 June 2004. "Smoothing" was in eV,
"degauss" is in Ry, so in case you are used to a jurassic
version of QE, you just need to set degauss=smoothing/13.6058
in order to have the same results as before.

Post by David Grifith
overall the only way that I can make my PDOS graphs as smooth as DOS
ones is reducing degauss value to 0.01 ! but in some cases that we
study the proper value for degauss is 0.001 and reducing it we may
miss something. mightn't it ?

your definition of "reducing" is not the same I use: "reducing" =
"making it smaller". Anyway: the broadening you need for scf
calculations is not necessarily the same you need for PDOS.
You can safely specify a different broadening for PDOS. It
will affect only the way your pictures look like.

Post by David Grifith
We are going to compare a large sets of PDOS and DOS graphs and prefer
not to compare one graph with lots of singularities with the others
that are too smooth. I appreciate your advice in advance.

My advice is to look at the attached notes, look at the dos_gam routine
in PHonon/PH/matdyn.f90, implement PDOS with tetrahedra.

P.
--
Paolo Giannozzi, Dept. Chemistry&Physics&Environment,
Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
Phone [phone], fax [phone]
-------------- next part --------------
A non-text attachment was scrubbed...
Name: tetra.pdf
Type: application/pdf
Size: 44169 bytes
Desc: not available
Url : http://pwscf.org/pipermail/pw_forum/attachments/20130429/9383e812/attachment.pdf