![]() |
(git:27c966b)
|
Nevanlinna analytic continuation of the GW correlation self-energy from the imaginary to the real frequency axis, following J. Fei, C.-N. Yeh, E. Gull, PRL 126, 056402 (2021). Minus the correlation self-energy maps the upper half plane into itself. After a Moebius map onto the unit disk the interpolation becomes a Schur problem whose solution is causal by construction. The Schur recursion loses digits quickly and is therefore done in double-double arithmetic (about 31 significant digits, module double_double), so that no quadruple-precision support of the compiler is needed; the evaluation is stable in double precision. More...
Data Types | |
| type | gw_nevanlinna_env_type |
| Parameters of the quasiparticle solution with the Nevanlinna continuation; they are set in set_heuristic_parameters (gw_utils.F) and stay zero for the Padé continuation. More... | |
| type | nevanlinna_type |
| Schur parameters of one Nevanlinna interpolant. More... | |
Functions/Subroutines | |
| subroutine, public | nevanlinna_create (nev, x, y) |
| Builds the Nevanlinna interpolant of y(x). If the samples are not causal within the accuracy of the data, a small imaginary shift -i*delta*max|y| is added, with delta taken from a fixed ladder; nevsolvable is .FALSE. if no delta of the ladder works. | |
| subroutine, public | nevanlinna_evaluate (nev, x_val, func_val) |
| Evaluates the interpolant on the real axis. | |
| subroutine, public | nevanlinna_release (nev) |
| Releases the interpolant. | |
Nevanlinna analytic continuation of the GW correlation self-energy from the imaginary to the real frequency axis, following J. Fei, C.-N. Yeh, E. Gull, PRL 126, 056402 (2021). Minus the correlation self-energy maps the upper half plane into itself. After a Moebius map onto the unit disk the interpolation becomes a Schur problem whose solution is causal by construction. The Schur recursion loses digits quickly and is therefore done in double-double arithmetic (about 31 significant digits, module double_double), so that no quadruple-precision support of the compiler is needed; the evaluation is stable in double precision.
| subroutine, public gw_nevanlinna::nevanlinna_create | ( | type(nevanlinna_type), intent(out) | nev, |
| complex(kind=dp), dimension(:), intent(in) | x, | ||
| complex(kind=dp), dimension(:), intent(in) | y ) |
Builds the Nevanlinna interpolant of y(x). If the samples are not causal within the accuracy of the data, a small imaginary shift -i*delta*max|y| is added, with delta taken from a fixed ladder; nevsolvable is .FALSE. if no delta of the ladder works.
| nev | the interpolant |
| x | sample points, all in the upper or all in the lower half plane |
| y | values of the correlation self-energy at x |
Definition at line 88 of file gw_nevanlinna.F.
| subroutine, public gw_nevanlinna::nevanlinna_evaluate | ( | type(nevanlinna_type), intent(in) | nev, |
| real(kind=dp), intent(in) | x_val, | ||
| complex(kind=dp), intent(out) | func_val ) |
Evaluates the interpolant on the real axis.
| nev | the interpolant |
| x_val | real frequency |
| func_val | continued correlation self-energy at x_val |
Definition at line 185 of file gw_nevanlinna.F.
| subroutine, public gw_nevanlinna::nevanlinna_release | ( | type(nevanlinna_type), intent(inout) | nev | ) |
Releases the interpolant.
| nev | the interpolant |
Definition at line 233 of file gw_nevanlinna.F.