This thread is locked.Only browsing is available.
Top Page > Browsing
Strange Mulliken Populations using new PAO in NEGF
Date: 2011/10/12 17:43
Name: Sarah Jones

Hi,

I have been using the new PAO/VPS for NEGF calculations and have found strange behaviour with the Mulliken Population. I am using Al7.0 and C5.0 PAO and the PBE functional.

After the NEGF converges (NormRD < 1.0e-06), the Mulliken Population gives a total of 1703.8 electrons, when the ideal neutral value is 1025. L0 and R0 regions are the worst affected, for example there are Al with a negative number of s electrons and over 40 p electrons, while some C have 1 electron and others 7. I have carried out the same calculation with the 06 PAO/VPS have not seen this kind of behaviour.

I would appreciate any help fixing this.

Thanks,

Sarah



メンテ
Page: [1]

Re: Strange Mulliken Populations using new PAO in NEGF ( No.1 )
Date: 2011/10/12 18:36
Name: T.Ozaki

Hi,

I guess that this is not a pseudopotential issue, but a local trap problem
in the SCF calculation as described in the paper (PRB 81, 035116 (2010)).
If your problem is the case, a little modification of code may improve
the situation. To do that, you can modify a source TRAN_DFT.c.
In a 214 line of TRAN_DFT.c, you may find

if (3<iter || SucceedReadingDMfile==1){

If you modify the number '3' to a larger number e.g. '15', the conventional
eigensolver is employed up to 14 SCF iterations. The treatment tends to
avoid such a local trap problem. (The number will be controllable through
the input file in the next version of OpenMX)

Also, a modest charge mixing by tuning SCF parameters may avoid the local
trap problem.

Regards,

TO
メンテ
Re: Strange Mulliken Populations using new PAO in NEGF ( No.2 )
Date: 2011/10/18 17:55
Name: Sarah Jones

Hi,

Thanks for your prompt reply.
I changed "if (3<iter || SucceedReadingDMfile==1){" to "if (15<iter || SucceedReadingDMfile==1){", but as soon as the NEGF starts I get similar results as before except that there are now Al atoms with a negative number of electrons:
Atom 87 MulP -0.3545-0.3545 sum -0.7089

I have tried the following for SCF mixing:

scf.Mixing.Type rmm-diisk
scf.Init.Mixing.Weight 0.001
scf.Min.Mixing.Weight 0.001
scf.Max.Mixing.Weight 0.2
scf.Kerker.factor 6
scf.Mixing.History 20
scf.Mixing.StartPulay 10
scf.Mixing.EveryPulay 5

scf.Mixing.Type rmm-diisk
scf.Init.Mixing.Weight 0.001
scf.Min.Mixing.Weight 0.001
scf.Max.Mixing.Weight 0.03
scf.Kerker.factor 6
scf.Mixing.History 20
scf.Mixing.StartPulay 10
scf.Mixing.EveryPulay 1

I am unsure what to change for the mixing parameters, as these have always worked with the older PAOs and any advice would be appreciated.

Thanks,

Sarah
メンテ

Page: [1]