LANCZOSTHR: Difference between revisions
m (Tal moved page Construction:LANCZOSTHR to LANCZOSTHR) |
No edit summary |
||
| Line 6: | Line 6: | ||
::<math> | ::<math> | ||
\mathrm{RMS}[\epsilon] = \sqrt{\sum_{j=1}^N\left[\epsilon_{i}(\omega_j)-\epsilon_{i+1}(\omega_j)\right]^2} | \mathrm{RMS}[\epsilon] = \sqrt{\sum_{j=1}^N\frac{1}{N}\left[\epsilon_{i}(\omega_j)-\epsilon_{i+1}(\omega_j)\right]^2} | ||
</math> | </math> | ||
Revision as of 15:25, 18 December 2024
LANCZOSTHR = [real]
Default: LANCZOSTHR = [math]\displaystyle{ 10^{-3} }[/math]
Description: LANCZOSTHR is used by the BSE Lanczos algorithm to stop the iterative procedure, once the dielectric function has reached numerical convergence.
The difference between the dielectric function at two consecutive iterations, [math]\displaystyle{ i }[/math] and [math]\displaystyle{ i+1 }[/math], is computed as root-mean-square over the frequency grid
- [math]\displaystyle{ \mathrm{RMS}[\epsilon] = \sqrt{\sum_{j=1}^N\frac{1}{N}\left[\epsilon_{i}(\omega_j)-\epsilon_{i+1}(\omega_j)\right]^2} }[/math]
and once [math]\displaystyle{ \mathrm{RMS}[\epsilon]\lt = }[/math]LANCZOSTHR the iterative algorithm stops.
Related tag and articles
BSE, BSE calculations, Bethe-Salpeter equations