Abstract:
Two major challenges of numeric analytic continuation – restoring
the spectral density, $s(\omega)$, from corresponding Matsubara correlator,
$g(\tau)$ – are (i) producing the most smooth/featureless answer for
$s(\omega)$, without compromising the error bars on $g(\tau)$, and (ii)
quantifying possible deviations of the produced result from the actual
answer. We introduce the method of consistent constraints that solves both
problems.
Fulltext:
PDF file (316 kB)