4
1
mirror of https://github.com/pfloos/quack synced 2024-12-22 04:14:26 +01:00

more efficient ccG0W0

This commit is contained in:
Pierre-Francois Loos 2024-09-17 16:39:38 +02:00
parent 4fa5706b90
commit 8b7c715811
2 changed files with 80 additions and 117 deletions

View File

@ -166,7 +166,7 @@ subroutine RGW(dotest,doG0W0,doevGW,doqsGW,doufG0W0,doufGW,maxSCF,thresh,max_dii
if(doccG0W0) then
call wall_time(start_GW)
call ccRG0W0(maxSCF,thresh,nBas,nOrb,nC,nO,nV,nR,ERI_MO,ENuc,ERHF,eHF)
call ccRG0W0(maxSCF,thresh,max_diis,nBas,nOrb,nC,nO,nV,nR,ERI_MO,ENuc,ERHF,eHF)
! call ccRG0W0_mat(maxSCF,thresh,nBas,nOrb,nC,nO,nV,nR,ERI_MO,ENuc,ERHF,eHF)
call wall_time(end_GW)

View File

@ -1,4 +1,4 @@
subroutine ccRG0W0(maxSCF,thresh,nBas,nOrb,nC,nO,nV,nR,ERI,ENuc,ERHF,eHF)
subroutine ccRG0W0(maxSCF,thresh,max_diis,nBas,nOrb,nC,nO,nV,nR,ERI,ENuc,ERHF,eHF)
! CC-based GW module
@ -9,6 +9,7 @@ subroutine ccRG0W0(maxSCF,thresh,nBas,nOrb,nC,nO,nV,nR,ERI,ENuc,ERHF,eHF)
integer,intent(in) :: maxSCF
double precision,intent(in) :: thresh
integer,intent(in) :: max_diis
integer,intent(in) :: nBas
integer,intent(in) :: nOrb
@ -23,31 +24,27 @@ subroutine ccRG0W0(maxSCF,thresh,nBas,nOrb,nC,nO,nV,nR,ERI,ENuc,ERHF,eHF)
! Local variables
integer :: p,q
integer :: p,q,r,s
integer :: i,j,k,l
integer :: a,b,c,d
integer :: nSCF
double precision :: Conv
double precision,allocatable :: delta_2h1p(:,:,:)
double precision,allocatable :: delta_2p1h(:,:,:)
double precision,allocatable :: V_2h1p(:,:,:)
double precision,allocatable :: V_2p1h(:,:,:)
double precision,allocatable :: r_2h1p(:,:,:)
double precision,allocatable :: r_2p1h(:,:,:)
double precision,allocatable :: t_2h1p(:,:,:)
double precision,allocatable :: t_2p1h(:,:,:)
double precision :: x_2h1p
double precision :: x_2p1h
double precision :: x
double precision,allocatable :: eGW(:)
double precision,allocatable :: Z(:)
double precision,allocatable :: del(:,:,:)
double precision,allocatable :: res(:,:,:)
double precision,allocatable :: amp(:,:,:)
integer :: n_diis
double precision :: rcond
double precision,allocatable :: r_diis(:,:)
double precision,allocatable :: t_diis(:,:)
! Hello world
@ -59,15 +56,19 @@ subroutine ccRG0W0(maxSCF,thresh,nBas,nOrb,nC,nO,nV,nR,ERI,ENuc,ERHF,eHF)
! Form energy denominator and guess amplitudes
allocate(delta_2h1p(nO,nO,nV),delta_2p1h(nO,nV,nV))
allocate(V_2h1p(nO,nO,nV),V_2p1h(nO,nV,nV))
allocate(t_2h1p(nO,nO,nV),t_2p1h(nO,nV,nV))
allocate(r_2h1p(nO,nO,nV),r_2p1h(nO,nV,nV))
allocate(del(nOrb,nOrb,nOrb))
allocate(res(nOrb,nOrb,nOrb))
allocate(amp(nOrb,nOrb,nOrb))
allocate(eGW(nOrb),Z(nOrb))
allocate(r_diis(nOrb**3,max_diis))
allocate(t_diis(nOrb**3,max_diis))
! Initialization
eGW(:) = eHF(:)
Z(:) = 1d0
!-------------------------!
! Main loop over orbitals !
@ -79,65 +80,49 @@ subroutine ccRG0W0(maxSCF,thresh,nBas,nOrb,nC,nO,nV,nR,ERI,ENuc,ERHF,eHF)
Conv = 1d0
nSCF = 0
t_2h1p(:,:,:) = 0d0
t_2p1h(:,:,:) = 0d0
n_diis = 0
t_diis(:,:) = 0d0
r_diis(:,:) = 0d0
rcond = 0d0
amp(:,:,:) = 0d0
res(:,:,:) = 0d0
del(:,:,:) = huge(1d0)
! Compute energy differences
do i=nC+1,nO
do j=nC+1,nO
do a=1,nV-nR
do a=nO+1,nOrb-nR
delta_2h1p(i,j,a) = eHF(i) + eHF(j) - eHF(nO+a) - eHF(p)
del(i,j,a) = eHF(i) + eHF(j) - eHF(a) - eHF(p)
end do
end do
end do
do i=nC+1,nO
do a=1,nV-nR
do b=1,nV-nR
do a=nO+1,nOrb-nR
do b=nO+1,nOrb-nR
delta_2p1h(i,a,b) = eHF(nO+a) + eHF(nO+b) - eHF(i) - eHF(p)
del(b,a,i) = eHF(a) + eHF(b) - eHF(i) - eHF(p)
end do
end do
end do
! Compute V2h1p and V2p1h
do k=nC+1,nO
do l=nC+1,nO
do c=1,nV-nR
V_2h1p(k,l,c) = sqrt(2d0)*ERI(p,nO+c,k,l)
end do
end do
end do
do k=nC+1,nO
do c=1,nV-nR
do d=1,nV-nR
V_2p1h(k,c,d) = sqrt(2d0)*ERI(p,k,nO+d,nO+c)
end do
end do
end do
!----------------------!
! Loop over amplitudes !
!----------------------!
write(*,*)
write(*,*)'----------------------------------------------'
write(*,*)'| CC-based G0W0 calculation |'
write(*,*)'----------------------------------------------'
write(*,'(1X,A1,1X,A3,1X,A1,1X,A10,1X,A1,1X,A10,1X,A1,1X,A10,1X,A1,1X)') &
write(*,*)'-------------------------------------------------------------'
write(*,*)'| CC-based G0W0 calculation |'
write(*,*)'-------------------------------------------------------------'
write(*,'(1X,A1,1X,A3,1X,A1,1X,A15,1X,A1,1X,A15,1X,A1,1X,A15,1X,A1,1X)') &
'|','#','|','HF','|','G0W0','|','Conv','|'
write(*,*)'----------------------------------------------'
write(*,*)'-------------------------------------------------------------'
do while(Conv > thresh .and. nSCF < maxSCF)
@ -147,47 +132,33 @@ subroutine ccRG0W0(maxSCF,thresh,nBas,nOrb,nC,nO,nV,nR,ERI,ENuc,ERHF,eHF)
! Compute intermediates x_2h1p and x_2p1h
x_2h1p = 0d0
x = 0d0
do k=nC+1,nO
do l=nC+1,nO
do c=1,nV-nR
x_2h1p = x_2h1p + V_2h1p(k,l,c)*t_2h1p(k,l,c)
end do
end do
end do
x_2p1h = 0d0
do k=nC+1,nO
do c=1,nV-nR
do d=1,nV-nR
x_2p1h = x_2p1h + V_2p1h(k,c,d)*t_2p1h(k,c,d)
do q=nC+1,nOrb-nR
do r=nC+1,nOrb-nR
do s=nC+1,nOrb-nR
x = x + sqrt(2d0)*ERI(p,s,q,r)*amp(q,r,s)
end do
end do
end do
! Compute residual for 2h1p sector
do i=nC+1,nO
do j=nC+1,nO
do a=1,nV-nR
do a=nO+1,nOrb-nR
r_2h1p(i,j,a) = V_2h1p(i,j,a) + delta_2h1p(i,j,a)*t_2h1p(i,j,a)
res(i,j,a) = sqrt(2d0)*ERI(p,a,i,j) + (del(i,j,a) - x)*amp(i,j,a)
do k=nC+1,nO
do c=1,nV-nR
do c=nO+1,nOrb-nR
r_2h1p(i,j,a) = r_2h1p(i,j,a) - 2d0*ERI(j,nO+c,nO+a,k)*t_2h1p(i,k,c)
res(i,j,a) = res(i,j,a) - 2d0*ERI(j,c,a,k)*amp(i,k,c)
end do
end do
r_2h1p(i,j,a) = r_2h1p(i,j,a) - t_2h1p(i,j,a)*x_2h1p - t_2h1p(i,j,a)*x_2p1h
end do
end do
@ -196,70 +167,62 @@ subroutine ccRG0W0(maxSCF,thresh,nBas,nOrb,nC,nO,nV,nR,ERI,ENuc,ERHF,eHF)
! Compute residual for 2p1h sector
do i=nC+1,nO
do a=1,nV-nR
do b=1,nV-nR
do a=nO+1,nOrb-nR
do b=nO+1,nOrb-nR
r_2p1h(i,a,b) = V_2p1h(i,a,b) + delta_2p1h(i,a,b)*t_2p1h(i,a,b)
res(b,a,i) = sqrt(2d0)*ERI(p,i,b,a) + (del(b,a,i) - x)*amp(b,a,i)
do k=nC+1,nO
do c=1,nV-nR
do c=nO+1,nOrb-nR
r_2p1h(i,a,b) = r_2p1h(i,a,b) + 2d0*ERI(nO+a,k,i,nO+c)*t_2p1h(k,c,b)
res(b,a,i) = res(b,a,i) + 2d0*ERI(a,k,i,c)*amp(b,c,k)
end do
end do
r_2p1h(i,a,b) = r_2p1h(i,a,b) - t_2p1h(i,a,b)*x_2h1p - t_2p1h(i,a,b)*x_2p1h
end do
end do
end do
! Check convergence
Conv = max(maxval(abs(r_2h1p)),maxval(abs(r_2p1h)))
Conv = maxval(abs(res))
! Update amplitudes
amp(:,:,:) = amp(:,:,:) - res(:,:,:)/del(:,:,:)
t_2h1p(:,:,:) = t_2h1p(:,:,:) - r_2h1p(:,:,:)/delta_2h1p(:,:,:)
t_2p1h(:,:,:) = t_2p1h(:,:,:) - r_2p1h(:,:,:)/delta_2p1h(:,:,:)
! Compute self-energy
! DIIS extrapolation
if(max_diis > 1) then
n_diis = min(n_diis+1,max_diis)
call DIIS_extrapolation(rcond,nOrb**3,nOrb**3,n_diis,r_diis,t_diis,res,amp)
end if
! Compute quasiparticle energy
eGW(p) = eHF(p)
do q=nC+1,nOrb-nR
do r=nC+1,nOrb-nR
do s=nC+1,nOrb-nR
do i=nC+1,nO
do j=nC+1,nO
do a=1,nV-nR
eGW(p) = eGW(p) + V_2h1p(i,j,a)*t_2h1p(i,j,a)
eGW(p) = eGW(p) + sqrt(2d0)*ERI(p,s,q,r)*amp(q,r,s)
end do
end do
end do
do i=nC+1,nO
do a=1,nV-nR
do b=1,nV-nR
eGW(p) = eGW(p) + V_2p1h(i,a,b)*t_2p1h(i,a,b)
end do
end do
end do
! Renormalization factor
Z(:) = 1d0
! Dump results
write(*,'(1X,A1,1X,I3,1X,A1,1X,F10.6,1X,A1,1X,F10.6,1X,A1,1X,F10.6,1X,A1,1X)') &
write(*,'(1X,A1,1X,I3,1X,A1,1X,F15.10,1X,A1,1X,F15.10,1X,A1,1X,F15.10,1X,A1,1X)') &
'|',nSCF,'|',eHF(p)*HaToeV,'|',eGW(p)*HaToeV,'|',Conv,'|'
end do
write(*,*)'----------------------------------------------'
write(*,*)'-------------------------------------------------------------'
!------------------------------------------------------------------------
! End of SCF loop
!------------------------------------------------------------------------
@ -285,7 +248,7 @@ subroutine ccRG0W0(maxSCF,thresh,nBas,nOrb,nC,nO,nV,nR,ERI,ENuc,ERHF,eHF)
'|','#','|','e_HF (eV)','|','Sig_c (eV)','|','Z','|','e_QP (eV)','|'
write(*,*)'-------------------------------------------------------------------------------'
write(*,'(1X,A1,1X,I3,1X,A1,1X,F15.6,1X,A1,1X,F15.6,1X,A1,1X,F15.6,1X,A1,1X,F15.6,1X,A1,1X)') &
write(*,'(1X,A1,1X,I3,1X,A1,1X,F15.10,1X,A1,1X,F15.10,1X,A1,1X,F15.10,1X,A1,1X,F15.10,1X,A1,1X)') &
'|',p,'|',eHF(p)*HaToeV,'|',(eGW(p)-eHF(p))*HaToeV,'|',Z(p),'|',eGW(p)*HaToeV,'|'
write(*,*)'-------------------------------------------------------------------------------'