[Pw_forum] How to choose the suitable lambdas from el-ph output ?

[Malgorzata Wierzbowska]

On Fri, 2 Feb 2007, [big5] 陳 少華 wrote:

  Hi Max,
  as for the choice of the broadening (in two delta functions,
  in lambda definition) one should take as small region around
  the Fermi energy as possible.
  And "possible" means a compromise with the size of the k-mesh.
  It is usually about Gaussian Broadening = 0.02 or 0.03 Ry
  (if I have more time and k-points then could be 0.01).

  I observed the behaviour as follows:
  values of DOS and lambda, going from gaussian broadening = 0.05
  towards 0.01, first change continuously and then drop
  down because there is not enough k-point sampling.

  The range for the broadening 0.01 - 0.1, which is implemented now,
  is a bit too wide and the step 0.01 is too big.
  Maybe we should change it into: starting 0.006, step 0.004, last 0.04.

  The Fermi energy in scf can be calculated with tetrahedron method
  or again with gaussian broadening (as it is for lambda)
  so maybe this is not the best choice to compare....
  But thank you for an idea! I never thought this way before but...
  maybe you could look which DOS is closest to the experimental value
  in which you "trust"..... and.... then maybe your  choice will be
  between 0.02 and 0.03 ??

  More save and suitable for predictive calculations is, however,
  testing the k-mesh versus lambda(broadening) and DOS(broadening) curves.

   Gosia



> Dear members,
> When I calculated the el-ph interaction ,the output
> file shows ten different lambda datas from Gaussian
> Broadening  0.010 to 0.1 Ry as below :
>
> electron-phonon interaction  ...
>
>     Gaussian Broadening:   0.010 Ry, ngauss=   1
>     DOS =  1.571120 states/spin/Ry/Unit Cell at Ef=
> 8.320255 eV
>     double delta at Ef = 21.385257
>     lambda( 1)=  0.051882   gamma=  0.731974 GHz
>     lambda( 2)= -0.071021   gamma= -1.275961 GHz
>     lambda( 3)=  1.701719   gamma= 75.615803 GHz
>                            ..
>                            ..
>                            ..
>  Gaussian Broadening:   0.100 Ry, ngauss=   1
>     DOS =  2.674519 states/spin/Ry/Unit Cell at Ef=
> 8.288962 eV
>     double delta at Ef = 17.639171
>     lambda( 1)=  0.052070   gamma=  1.250555 GHz
>     lambda( 2)=  0.181461   gamma=  5.549722 GHz
>     lambda( 3)=  0.372624   gamma= 28.185890 GHz
>
> My question is how to choose the suitable one of them
> ?
> By now I choose the one which its Ef is the most
> neariest from scf calculted Ef.Does it make sense ?
>
> Best Regards
> max
>
>
>
> ___________________________________________________
> 您的生活即時通 － 溝通、娛樂、生活、工作一次搞定！
> http://messenger.yahoo.com.tw/
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>