[Pw_forum] Question for electron phonon coulping

[Stefano de Gironcoli]

One is interested in the limit for sigma going to zero with infinit
number of k points.
All other numbers are approximations to this results.
How can you judge whether they are good approximations ?
As usual by checking convergence.

Do a calculation with a given set of k points and calculate el-ph as a
function of sigma
Do again the same calculation increasing the number of k-points and the
same set of values for sigma.
You will notice that for sufficiently large values of sigma the two
curves will agrees. This means that
the k-points mesh is sufficiently accurate for sigmas larger than that
but may not be good for smaller values.
Repeating the calculation for even larger set of k-points (same set of
sigma) you will see that from a value
of sigma that is smaller that the previous one the most accurate
calculation (the one using more k-points)
will agree with the second-best (and with the first, less accurate,
calculation when these last two agree)
Trust only the curve in the region in which the calculation is converged
with respect to k-ponts summation.
For sufficiently large set of k-points the SAFE region will approach
enough the limit of sigma going to zero
that you can stop the procedure.

Stefano de Gironcoli

ma Yanming wrote:

>
> Dear All,
>
> I have a question concerning the electron-phonon coupling.
> I increase the parameter "nsig' to 40 in elphon.f90.
> So in my case I have 40 Gaussian Broadening steps. The calculated
> lambda(q) is negative number until 35 step, i.e., Gaussian Broadening:
> 0.35 Ry.
>
>> From step 36, the lambda(q) comes positive and increase gradully. Until 
>
> step 40 it continue increasing. It seems that it still hasn't
> converged. I am not clear the principle for this Gaussian broadening
> procedures. In my case, the Fermi energy is around 19.47 eV. I am
> wondering if it is reasonable if I take the positive number (in the
> final 40 step) as final lambda value. But at the step 40, the
> calculated lambda is at fermi energy 18.9 eV which is far away from
> 19.47eV.
>
> Any idea on the principle for Gaussian broadening procedures. Can I
> take the positive lambda after 40 step Gaussian broadening and by 35
> step negative number?
>
> THanks
>