[Pw_forum] What's wrong in my calculation of epsilon of fcc Au?

[gbliu]

Dear all,

   Now, I know that the bad results of dielectric function relate the 
pseudopotential(PP)!
   The PP I used is from abinit by fhi2upf.
   After that, I tested another qso PP, which is from 
http://fpmd.ucdavis.edu/potentials/index.htm and
http://eslab.ucdavis.edu/software/index.htm . Using this PP, the profile 
of the dielectric function curve is similar with the result from wien2k. 
But, this PP gives out somewhat bad band structures!

   So, it confuses me that a PP which can give good band structure 
results in bad dielectric function, while a PP which gives somewhat bad 
band structure results in a dielectric function comparable with wien2k. 
Can anyone give some explanations? (see attachments)

   Another question aslo confuses me as mentioned in the original email: 
what's the meaning of the "plasmon frequency" given by epsilon.x? And 
why it depends on the input parameter intrasmear and even wmax?

Best regards!

                    Liu Guibin
                Postdoctoral Fellow,
                Department of Physics,
                The University of Hong Kong
                gbliu at hku.hk


于 2011/7/7 17:39, gbliu 写道:
> Dear all,
>
>       I want to calculate dielectric function of carbon nanotube, but 
> I have no experiences of calculating dielectric function before. 
> Hence, I decide to start with a simple case of fcc gold, because there 
> are referable results from wien2k (wien2k is capable of simple case 
> such as fcc Au, but too resource consumptive to deal with large cells 
> such as CNT).
>       Because epsilon.x doesn't support USPP, the UPF file of Au is 
> got by converting fhi file from abinit using fhi2upf.x. I calculate 
> fcc Au with a=7.71bohr and 15x15x15 k-mesh using both pwscf(v4.3) and 
> wien2k. First I check the band structures and find that the results 
> from both softwares agree well (see attachments). But the results of 
> epsilon from pwscf is not satisfactory as the results from wien2k:
>
>       (1) Figures imag_eps and real_eps are the compares with only 
> interband contribution. (note that for pwscf, I subtract the intraband 
> contribution [-omega_p^2/omega^2] to get only interband contribution 
> for compares). We can see the differences are notable. What's the 
> reason? BTW, I use a homogeneous k-mesh also 15x15x15 generated by 
> myself for epsilon calculation (see input file fccAu.nscf4eps.in).
> ------input for epsilon.x--------
> &inputpp
>      outdir='./work'
>      prefix='fccAu'
>      calculation='eps'
> /
> &energy_grid
>      smeartype='gauss'
>      intersmear=0.1d0
>      intrasmear=0d0
>      wmax=20d0
>      wmin=0d0
>      nw=600
>      shift=0d0
> /
> --------------------------------
>
>      (2) As for the intraband contribution, wien2k gives the plasma 
> frequency 9.0075eV which agrees well with experimental value (ref: 
> http://wave-scattering.com/drudefit.html), but the result from 
> epsilon.x is puzzling.
> Specifically, plasma frequency given in the output of epsilon.x is 
> dependent on the input parameters:
> <1>. intrasmear=0,   wmax=20d0  -->   The bulk xx plasmon frequency 
> [eV] is:    14.877926200
> <2>. intrasmear=0,   wmax=10d0  -->   The bulk xx plasmon frequency 
> [eV] is:    7.625822195
> <3>. intrasmear=0,   wmax=6d0   -->   The bulk xx plasmon frequency 
> [eV] is:    3.156554381
> <4>. intrasmear=0.1, wmax=20d0  -->   The bulk xx plasmon frequency 
> [eV] is:    15.766403193
> <5>. intrasmear=0.1, wmax=10d0  -->   The bulk xx plasmon frequency 
> [eV] is:    9.322251787
> <6>. intrasmear=0.1, wmax=20d0  -->   The bulk xx plasmon frequency 
> [eV] is:    6.263713572
> However, none of these values is the one got by fitting the results 
> using Drude model [ 1 - omega_p^2/(omega^2+i*omega*gamma), gamma is 
> the intrasmear here ]. By fitting the results using Drude model, I get 
> omega_p=5.528769eV, for all the cases above (<1>~<6>). But this 
> omega_p is far from the experimental value (~9 eV)! What's wrong? And 
> what are the meanings of the values given in <1>~<6> ?
>
>       How can I get more accurate results like the ones from wien2k?
>       Please help me if you know something about this, thanks in advance!
>
> Best regards!
>
>                    Liu Guibin
>                Postdoctoral Fellow,
>                Department of Physics,
>                The University of Hong Kong
>                gbliu at hku.hk
>
>
>
>
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20110710/3f69bcf6/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: band.png
Type: image/png
Size: 86619 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20110710/3f69bcf6/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: imag_eps1.png
Type: image/png
Size: 15858 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20110710/3f69bcf6/attachment-0001.png>