epsilon.log for high band counts

Submitted by gkedziora on Wed, 08/30/2017 - 12:42


Bug reports

I was studying the convergence of epsilon inverse of GaAs using the method recommended in the README file for epsilon. That is I looking at the value of epsilon inverse for (G_max, G_max) in epsilon.log approaching 1.0 and the off-diagonal values approaching zero. But for large band counts, I do not get the (G_max,G_max). I can provide inputs and outputs upon request by email. Here is the head of epsilon.log:

Epsilon code, Complex version, run on 18-Aug-2017 at 15:38:55 -0400

Running MPI version (parallel)
Running with 72 MPI task(s)
Computing the static inverse dielectric matrix

We are using matrix communication scheme

We are communicating via MPI

We are using no truncation

Highest occupied band (unshifted grid) = 9
Valence max (unshifted grid) = 7.125371 eV
Conduction min (unshifted grid) = 8.263621 eV
Middle energy (unshifted grid) = 7.694496 eV
Fermi energy (unshifted grid) = 7.694496 eV

Highest occupied band (shifted grid) = 9
Running with 1 valence pools
Number of conduction bands per processor: 12
Number of valence bands per processor: 9

Memory available: 3386.1 MB per PE
Memory required for execution: 1872.1 MB per PE
Memory required for vcoul: .0 MB per PE

Cell Volume = 0.304609241E+03

Number of electrons per unit cell (from ifmax) = 18.000000
Number of electrons per unit cell (from occupations) = 18.000000
Plasma Frequency = 1.723454 Ry

- Screened Coulomb cutoff: 30.000 Ry
- Total number of bands: 868
- Number of q-points: 22