Vladimir,
Thanks alot! Your comment cleared alot of confusions in my mind.
Haitao
Vladimir Timochevski wrote:
Hi, Haitao
Everything is OK with your calculations, don't worry. In your
pseudopotential Cd 4d-states are already included as valence states,
so the code generated DZ orbitals for them. When you use standard DZP
you always get DZ for all valence states and SZ for polarization
orbitals.
I think that my Zn-example was misleading in this sense. Ning's
question was more how to deal with the situation when you need to have
several orbitals in the basis set with the same angular momentum. Then
you can not use a standard notation (DZP), but have to explicitly
describe them in the PAO.basis block. A good example would be the As
atom, where you can either put 3d10-states in pseudopotential, and
then use a standard notation (in case of DZP your polarization orbital
will be a 4d-type orbital), or explicitly include it as a valence
state, but then, if you also want to include 4d to improve the
description of conduction states, you will have 2 functions with the
same l-value. So, Ning was asking what to use for these low-lying
3d-states? DZ or SZ? The answer is "... depends on each particular
case". I would say that the lower is this semicore state in energy,
the "closer" it is to atomic-like state, and the higher is the
probability that SZ will describe it correctly.
Vladimir.
Haitao Liu wrote:
Hi,
Reading the previous emails remind me checking a number of
calculations I did. For my case, I generate Cd pseudopotential with
the following configuration: 4d10 5s2 and core correction. I then
used standard DZP to generate basis (no %block PAO.basis section) in
my calculation
When I look at the output, I found that the program generated DZ
basis sets for 4d, 5s, and SZ for 5p, which is the polarization
orbital (see below). Are those automatically generated basis not good
for 4d? Is there anything I miss? I'd appreciate any comment.
part of the output:
-----------------------------------------------------------------------------------------------
atom: SANKEY-TYPE ORBITALS:
atom: Selected multiple-zeta basis: split
SPLIT: Orbitals with angular momentum L= 0
SPLIT: Basis orbitals for state 5s
SPLIT: PAO cut-off radius determined from an
SPLIT: energy shift= 0.005000 Ry
izeta = 1
lambda = 1.000000
rc = 7.417129
energy = -0.410139
kinetic = 0.337270
potential(screened) = -0.747408
potential(ionic) = -9.152469
izeta = 2
rmatch = 5.997186
splitnorm = 0.150000
energy = -0.348542
kinetic = 0.552560
potential(screened) = -0.901102
potential(ionic) = -9.990383
SPLIT: Orbitals with angular momentum L= 2
SPLIT: Basis orbitals for state 4d
SPLIT: PAO cut-off radius determined from an
SPLIT: energy shift= 0.005000 Ry
izeta = 1
lambda = 1.000000
rc = 4.387621
energy = -0.860629
kinetic = 6.859800
potential(screened) = -7.720429
potential(ionic) = -20.962464
izeta = 2
rmatch = 2.348500
splitnorm = 0.150000
energy = -0.443697
kinetic = 9.457328
potential(screened) = -9.901025
potential(ionic) = -24.055945
POLgen: Perturbative polarization orbital with L= 1
POLgen: Polarization orbital for state 5s
izeta = 1
rc = 7.417129
energy = -0.083172
kinetic = 0.610108
potential(screened) = -0.693280
potential(ionic) = -8.465104
atom: Total number of Sankey-type orbitals: 15
atm_pop: Valence configuration(local Pseudopot. screening):
5s( 2.00)
5p( 0.00)
4d(10.00)
Vna: chval, zval: 12.00000 12.00000
Vna: Cut-off radius for the neutral-atom potential: 7.417129
-----------------------------------------------------------------
end of output
Haitao
Vladimir Timochevski wrote:
Yes, Andrei is absolutely right: Zn is not a good example of using
SZ for semicore - it's d-states overlap with s-band, and therefore
should be more considered as "valence". SZ should be OK for such
elements as As or Ga, where these states are located deeper in
energy, and therefore are more localized. But in any case, as
Andrei pointed out, everything depends on what you would like to
obtain, and you should always check your particular desired quantity
for convergence. (I saw results of big supercell calculations, where
people freeze Zn 3d, putting it in pseudopotential. However, I would
not recomment that ...)
Thanks for your comment, Andrei!
Vladimir.
Andrei Postnikov wrote:
On Fri, 26 Jan 2007, Vladimir Timochevski wrote:
| Usually semicore states are quite localized, and from my
experience a SZ-type
| of orbitals is sufficient for them. So, in case of Zn (3d10
semicore states)
| put something like
| %block PAO.Basis
| Zn 2
| n=4 0 2 P
| 0.0 0.0
| 1.0 1.0
| n=3 2 1
| 0.0
| 1.0
| %endblock PAO.Basis
| This will give you 2 4s-orbitals, 3 4p-orbitals, and 5
3d-orbitals. Cutoff
| radii will be generated automatically.
Dear Vladimir:
This is indeed a quite risky advise... of course the final decision
depends on
what you are doing and level of accuracy you need. You'll probably
get the position of Zn3d band right with SZ, but lattice parameter,
elastic propeties, and phonons could be quite off. 3d of Zn is
hardly a good semicore, it is really too close and contribute
substantially to bondings.
Best regards,
Andrei
