 Re: VPS and PAO files of Pt ( No.1 ) |
- Date: 2005/10/11 20:58
- Name: T.Ozaki
- Hi,
> 2. I have tried to calculate a VPS file of Pt but failed with the following errors;
In ADPACK, the logarithmic derivative of wave function for
the pseudopotential is calculated by an iterative method.
Then, sometimes we tend to meet such a difficulty as you met
in getting convergece of the iteration. However, since the order of
"dif" is around 0.00001, the calculated logarithmic derivative may not
deviate largely from the completely converged one.
Also, the error is not related to the pseudopotential generation itself,
but the calculation of logarithmic derivatives. So, it is not a serious
error.
Best regards,
T.Ozaki
|
| Re: VPS and PAO files of Pt ( No.2 ) |
- Date: 2005/10/12 17:12
- Name: Watari, Noriko <noriko_watari@mhi.co.jp>
- How can user control the precision of potential (or atomi orbital)?
I have calculated Pt atom also with ADF. The following list is the comparision of
adpack and ADF.
ADPACK ADF
<ALL> **** Energies of atom ****
<ALL> Ekin = 20234.229049464226 (Hartree) 21113.6392
<ALL> EHart = 7159.888612251748 (Hartree) 7187.0281
<ALL> Exc = -249.181654064797 (Hartree) -355.6975
<ALL> Eec = -45413.783984607995 (Hartree) -46390.2085
<ALL> Etot = -18268.847976956818 (Hartree) -18445.2387
<ALL> Eeigen = -11177.134808382989 (Hartree) -11376.7765
(ADF is a commercial software and established reputation with calculation for
catalytic system, so these numbers are reliable to a certain extent.)
Aren't these differences a problem?
Don't the logarithmic derivative errors affect these energies?
|
| Re: VPS and PAO files of Pt ( No.3 ) |
- Date: 2005/10/13 12:47
- Name: T.Ozaki
- Hi,
> How can user control the precision of potential (or atomi orbital)?
The computational accuracy can be controlled by the following
three keywords: grid.xmin, grid.xmax, and grid.num.
In ADPACK, the radial differential equation is solved from
rmin = exp(grid.xmin) and rmax = exp(grid.xmax), and two radial
wave functions are merged at a matching point.
You may see how these energy components depend on the three parameters.
In addition, if you want to know how these parameters are related to
physical properties when the pseudopotential is used in calculations
for bulks and molecules, you can check in a systematic way.
By the way, did you solve the atomic Kohn-Sham equation scalar
relativistically or full relativistically ? You may find that
the energy components strongly depend on the way of relativistic
treatment. For example, see a database provided by the NIST
http://physics.nist.gov/PhysRefData/DFTdata/
> Aren't these differences a problem?
In my experiences, if the result converges at some level,
it may not affect strongly results in applications.
Rather than this issue, how one construct pseudopotentials is
the matter one has to seriously consider to perform reliable
pseudopotential calculations.
> Don't the logarithmic derivative errors affect these energies?
No, it is not related to these energies. You can easily check
by changing a parameter log.deri.RadF.calc=on|off.
Best regards,
T.Ozaki
 |
| Re: VPS and PAO files of Pt ( No.4 ) |
- Date: 2005/10/21 15:07
- Name: Watari, Noriko <noriko_watari@mhi.co.jp>
- Thank you. The information on atomic data of NIST is very useful for me.
|
|
This thread is locked.Only browsing is available.
VPS and PAO files of Pt