Yambo Community Forum

The data I refer to here has the labels X_Q_1, X_Q_2, etc.

Dear Christian,
the database are in netcdf format, you can either dump the file (ncdump) or better use a python script. Here I attach you a script to read a static database, it should be generalized for the ndb.pp which has 2 frequencies.
Note, the matrix elements are complex, 7675362 means 3837681 entries i.e. a matrix 1959x1959. This should be the dimension of your block if I did not any mistake.

The format should be the following e.g. :
float X_Q_1(D_0000000002, D_0000000005, D_0000000005, D_0000000002)
(N_frequencies, N_g, N_g', Re/Im)

Why is the fit actually done for an imaginary frequency?

Because the eps should be a smooth function in the imaginary axis. This is the Godby-Needs model that seems to be the most reliable. You can check some literature on the plasmon-pole models


Best,
Daniele

Dear Daniele,

Thanks, I got the script working. As there seems to be no way to get the poles and residuals from the code, I assume the best way to check if the PPA makes sense is to do a full frequency calculation and then fit the model to the two frequencies by hand.


Best,

Christian

Dear Christian,
the residuals and poles can be calculated from the values of eps^-1 it is 2 equation in 2 variable:
you can find the expression in the Yambo cheatsheet (Eq.9):
http://www.yambo-code.org/wiki/index.ph ... heatsheets

Full frequency calculation is in principle more accurate and will tell you if the PPA is suitable, but it is very slow in converging with respect to the number of frequencies.

Best,
Daniele