PAW dataset for Niobium

ruslan · Post by **ruslan** » Tue Sep 07, 2010 5:46 pm

Hello, abinit users.

I am trying to create PAW dataset for Nb with five electros (4d4, 5s1 and 5p0 states handled in valence). Therefore i am using atompaw v. 2.3 with following setup:

Code: Select all

Nb 41
LDA-PW scalarrelativistic loggrid
5 5 4 0 0 0 ! f-orbital?
1 0 2   ! 1s2
2 0 2   ! 2s2
2 1 6   ! 2p6
3 0 2   ! 3s2
3 1 6   ! 3p6
3 2 10  ! 3d10
4 0 2   ! 4s2
4 1 6   ! 4p6
4 2 4   ! 4d4
5 0 1   ! 5s1
5 1 0   ! 5p0
0 0 0
c     ! 1s
c     ! 2s
c     ! 3s
c     ! 4s
v     ! 5s
c     ! 2p
c     ! 3p
c     ! 4p
v     ! 5p
c     ! 3d
v     ! 4d
2
2.45
y
3
n
y
-2.6
n
y
1.2
n
custom rrkj sinc
3 0 troulliermartins
2.35
2.19
2.3
2.35
1.78
2.35
0

Logarithmic derivatives seem to be on right place and projector functions have same order of magnitude as the wavefunctions. Nevertheless the produced dataset misbehaves due to high matrix values in RHOIJ:

Code: Select all

 *********** RHOIJ (atom   1) **********
 ********* *********   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000 ...
 ********* *********   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000 ...
   0.00000   0.00000 *********   0.00000   0.00000 *********   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000 ...
   0.00000   0.00000   0.00000 *********   0.00000   0.00000 *********   0.00000   0.00000   0.00000   0.00000   0.00000 ...
   0.00000   0.00000   0.00000   0.00000 *********   0.00000   0.00000 *********   0.00000   0.00000   0.00000   0.00000 ...
   0.00000   0.00000 *********   0.00000   0.00000 *********   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000 ...
   0.00000   0.00000   0.00000 *********   0.00000   0.00000 *********   0.00000   0.00000   0.00000   0.00000   0.00000 ...
   0.00000   0.00000   0.00000   0.00000 *********   0.00000   0.00000 *********   0.00000   0.00000   0.00000   0.00000 ...
   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000 *********   0.00000   0.00000   0.00000 ...
   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000 *********   0.00000   0.00000 ...
   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000 *********   0.00000 ...
   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000 ********* ...

for the following input

Code: Select all

# test ecut convergence
ndtset 10

# experimental bulk lattice constant of Nb
acell 3*3.30 angstrom

# convergence crit
toldfe 1.0d-7

# 1 atom per pc
natom 1
typat 1
znucl 41
xred 0 0 0

nband 9
# number of kpts
ngkpt 6 6 6

# bcc lattice
rprim
-0.5  0.5  0.5
 0.5 -0.5  0.5
 0.5  0.5 -0.5

nshiftk 1
shiftk 0.5 0.5 0.5

ecut: 10
pawecutdg: 10
ecut+ 2
pawecutdg+ 2

# methfessel-paxthon
occopt 6
tsmear 0.1 eV

prtden 0
prtwf 0
prteig 0

nstep 60

diemac 1000000

My question is: which quantity should i change to get rid of "high" projector values? I tried to change shape functions and projectors with no effect. Thank you in advance,

Ruslan Zinetullin

P.S. I tried to use PAW datasets provided by Prof. Zwanziger in http://www.abinit.org/downloads/PAW/CON ... Table.html which contain 13-electron setups (which i want to replace by 5-electron one). The logarithmic derivatives for s- and p-orbitals are not really superimposed at the discontinous points. Does it mean that this criterion is not as much important as the other ones?

jzwanzig · Post by **jzwanzig** » Tue Sep 07, 2010 8:14 pm

Hi,
I just tried to make Nb with only 5 electrons and failed in the same way you did--I think the "high" occupations are coming from the 4s and 4p states, which are actually quite close to the 5s and 5p states in Nb. I really doubt that it is possible to make an accurate Nb PAW set with only 5 electrons. I've attached my latest 13-electron version for you to try and modify if you like--

good luck,
Joe Zwanziger

Code: Select all

Nb  41 
LDA-PW scalarrelativistic loggrid 1201
5 5 4 0 0 0
4 2 4
5 0 1
5 1 0
0 0 0
c
c
c
v
v
c
c
v
v
c
v
2
2.6
y
0
n
n
y
2
n
custom rrkj
3 0 ultrasoft
2.6
2.6
2.6
2.6
2.6
2.6
2.6
1
4 2 4
5 0 1
5 1 0
0 0 0
0

ruslan · Post by **ruslan** » Tue Sep 21, 2010 7:41 pm

jzwanzig wrote:
Hi,
I just tried to make Nb with only 5 electrons and failed in the same way you did--I think the "high" occupations are coming from the 4s and 4p states, which are actually quite close to the 5s and 5p states in Nb. I really doubt that it is possible to make an accurate Nb PAW set with only 5 electrons. I've attached my latest 13-electron version for you to try and modify if you like--

good luck,
Joe Zwanziger

It took time to sort things out. As i already mentioned, i needed PAW dataset with relatively small rc. I ended up with following setup:

Code: Select all


Nb 41
LDA-PW loggrid 1500
5 5 4 0 0 0
4 2 4
5 0 1
5 1 0
0 0 0
c
c
c
v
v
c
c
v
v
c
v
2
2.3
n
n
y
1.5
n
vanderbilt
3 0 ultrasoft
2.0
2.0 ! 2.365
2.0
2.0
2.0
2.3
0
1
4 2 4
5 0 1
5 1 0
0 0 0
0

Physical properties seem ok: a0=3.248 angstrom compared to 3.304 gives 1.7% deviation and the bulk modulus is B=179.6 compared to 170 expt. I found also, that enabling scalarrelativistic handling increases B to 200 GPa, which seems odd.

Thank you for hints,
Ruslan Zinetullin

ABINIT Discussion Forums

PAW dataset for Niobium

PAW dataset for Niobium

Re: PAW dataset for Niobium

Re: PAW dataset for Niobium