quantum-espresso/Modules/w1gauss.f90

95 lines
2.5 KiB
Fortran

!
! Copyright (C) 2001 PWSCF group
! This file is distributed under the terms of the
! GNU General Public License. See the file `License'
! in the root directory of the present distribution,
! or http://www.gnu.org/copyleft/gpl.txt .
!
!-----------------------------------------------------------------------
function w1gauss (x, n)
!-----------------------------------------------------------------------
!
! w1gauss(x,n) = \int_{-\infty}^x y delta(y) dy
! where delta(x) is the current approximation for the delta function,
! as obtained from w0gauss(x,n)
!
! --> (n>=0) : Methfessel-Paxton case
!
! --> (n=-1): Cold smearing (Marzari-Vanderbilt-DeVita-Payne)
! w1gauss = 1/sqrt(2*pi)*(x-1/sqrt(2))*exp(-(x-1/sqrt(2))**2)
!
! --> (n=-99): Fermi-Dirac case. In this case w1gauss corresponds
! to the negative of the electronic entropy.
!
!
USE kinds, ONLY : DP
USE constants, ONLY : pi
implicit none
real(DP) :: w1gauss, x
! output: the value of the function
! input: the point where to compute the function
integer :: n
! input: the order of the smearing function
!
! here the local variables
!
real(DP) :: a, hp, arg, hpm1, hd, f, onemf, xp
! the coefficients a_n
! the hermite function
! the argument of the exponential
! the hermite function
! the hermite function
! Fermi-Dirac occupation number
! 1 - f
! auxiliary variable (cold smearing)
integer :: i, ni
! counter on n values
! counter on 2n values
! Fermi-Dirac smearing
if (n.eq. - 99) then
if (abs (x) .le.36.0) then
f = 1.0d0 / (1.0d0 + exp ( - x) )
onemf = 1.0d0 - f
w1gauss = f * log (f) + onemf * log (onemf)
! in order to avoid problems for large values of x
else
! neglect w1gauss when abs(w1gauss) < 1.0d-14
w1gauss = 0.0d0
endif
return
endif
! Cold smearing
if (n.eq. - 1) then
xp = x - 1.0d0 / sqrt (2.0d0)
arg = min (200.d0, xp**2)
w1gauss = 1.0d0 / sqrt (2.0d0 * pi) * xp * exp ( - arg)
return
endif
! Methfessel-Paxton
arg = min (200.d0, x**2)
w1gauss = - 0.5d0 * exp ( - arg) / sqrt (pi)
if (n.eq.0) return
hd = 0.d0
hp = exp ( - arg)
ni = 0
a = 1.d0 / sqrt (pi)
do i = 1, n
hd = 2.0d0 * x * hp - 2.0d0 * DBLE (ni) * hd
ni = ni + 1
hpm1 = hp
hp = 2.0d0 * x * hd-2.0d0 * DBLE (ni) * hp
ni = ni + 1
a = - a / (DBLE (i) * 4.0d0)
w1gauss = w1gauss - a * (0.5d0 * hp + DBLE (ni) * hpm1)
enddo
return
end function w1gauss