quantum-espresso/GWW/pw4gww/wannier_bse.f90

460 lines
15 KiB
Fortran

!
! Copyright (C) 2001-2013 Quantum ESPRESSO 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 .
!
!
! Modified by Joshua Elliott November 2020 as JDE
!
! This subroutine computes the overlap between Wannier orbitals Ovv' and
! computes the (v*w_v*w_v')(r) term for each vv' such that Ovv'>s_bse, and
! writes to disk Ovv' and (v*w_v*w_v')(r) ,and z_beta_v_v'=v*phi_beta*wv*wv'
subroutine wannier_bse(ispin,w_wfcs,o_mat)
USE io_global, ONLY : stdout, ionode, ionode_id
USE io_files, ONLY : prefix, tmp_dir, diropn
USE kinds, ONLY : DP
USE wannier_gw, ONLY : num_nbndv,dual_bse,s_bse,l_truncated_coulomb,truncation_radius,vg_q,&
max_ngm,numw_prod, pmat_type, l_no_GW_just_screening, l_no_GW_bare_coulomb,& ! JDE
no_GW_cg_maxit, no_GW_cg_threshold, ewvc ! JDE
USE fft_custom_gwl
USE wvfct, ONLY : npwx, npw, nbnd, nbndx ! JDE
USE gvecw, ONLY : ecutwfc
USE mp_pools, ONLY : intra_pool_comm
USE mp_world, ONLY : mpime, nproc, world_comm
USE mp, ONLY : mp_sum, mp_bcast ! JDE
USE gvect
USE wavefunctions, ONLY : psic
USE constants, ONLY : e2, fpi
USE cell_base, ONLY: tpiba,tpiba2,omega
USE mp_wave, ONLY : mergewf,splitwf
implicit none
INTEGER, EXTERNAL :: find_free_unit
INTEGER, intent(in) :: ispin
COMPLEX(kind=DP), intent(in) :: w_wfcs(npw,num_nbndv(ispin))
REAL(kind=DP), intent(out) :: o_mat(num_nbndv(ispin),num_nbndv(ispin))
TYPE(fft_cus) :: fc
COMPLEX(kind=DP), allocatable :: w_wfcs_t(:,:)
REAL(kind=DP), ALLOCATABLE :: w_wfcs_r(:,:)
REAL(kind=DP), ALLOCATABLE :: w_wfcs_2(:,:)
REAL(kind=DP), ALLOCATABLE :: ww_prod(:)
COMPLEX(kind=DP), allocatable :: ww_prodg(:),ww_prodg2(:)
COMPLEX(kind=DP), ALLOCATABLE :: evc_g(:)
COMPLEX(kind=DP), ALLOCATABLE :: p_basis(:,:)
REAL(kind=DP), ALLOCATABLE :: fac(:)
REAL(kind=DP), ALLOCATABLE :: z(:)
INTEGER, ALLOCATABLE ::iww(:)
INTEGER :: ii,ig, iv,jv,np,np_max,i
INTEGER :: iunu, iungprod,iunz,iuni
REAL(kind=DP) :: qq
LOGICAL :: exst
logical :: debug, debug_operator ! JDE
integer :: iundebug
! JDE
INTEGER :: wannier_product_unit, loop_wannier_prod, &
wannier_write_product_unit, wwwdebug, iunwww, &
start_loop_wannier
EXTERNAL :: operator_1_vp, operator_debug
COMPLEX(kind=DP), allocatable :: ww_read_prodg2(:), ww_solved_prodg2(:)
REAL(KIND=dp), ALLOCATABLE :: overlap(:,:), diagonal(:), e(:)
COMPLEX(KIND=dp), ALLOCATABLE :: evc_jsh(:,:)
COMPLEX(KIND=dp), DIMENSION(:,:), ALLOCATABLE :: x, b, u, h, Ah, evc, pu
INTEGER :: cg_iter
LOGICAL :: cgsolve_conv
! JDE
fc%ecutt=ecutwfc
fc%dual_t=dual_bse
debug=.true.
debug_operator=.false.
! FFT the wannier function to r-space (dual grid)
write(stdout,*) 'Call initialize_fft_custom'
call initialize_fft_custom(fc)
allocate(w_wfcs_t(fc%npwt,num_nbndv(ispin)))
allocate( evc_g(fc%ngmt_g ) )
allocate(w_wfcs_r(fc%nrxxt,num_nbndv(ispin)))
allocate(w_wfcs_2(fc%nrxxt,num_nbndv(ispin)))
if(fc%dual_t==4.d0) then
w_wfcs_t(1:fc%npwt,1:num_nbndv(ispin))= w_wfcs(1:fc%npwt,1:num_nbndv(ispin))
else
do ii=1, num_nbndv(ispin)
call mergewf(w_wfcs(:,ii),evc_g,npw,ig_l2g,mpime,nproc,ionode_id,intra_pool_comm)
call splitwf(w_wfcs_t(:,ii),evc_g,fc%npwt,fc%ig_l2gt,mpime,nproc,ionode_id,intra_pool_comm)
enddo
endif
do ii=1,num_nbndv(ispin),2
psic(1:fc%nrxxt)=(0.d0,0.d0)
if (ii==num_nbndv(ispin)) then
psic(fc%nlt(1:fc%npwt)) = w_wfcs_t(1:fc%npwt,ii)
psic(fc%nltm(1:fc%npwt)) = CONJG( w_wfcs_t(1:fc%npwt,ii) )
else
psic(fc%nlt(1:fc%npwt))=w_wfcs_t(1:fc%npwt,ii)+(0.d0,1.d0)*w_wfcs_t(1:fc%npwt,ii+1)
psic(fc%nltm(1:fc%npwt)) =CONJG(w_wfcs_t(1:fc%npwt,ii))+(0.d0,1.d0)*CONJG(w_wfcs_t(1:fc%npwt,ii+1))
endif
CALL cft3t( fc, psic, fc%nr1t, fc%nr2t, fc%nr3t, fc%nrx1t, fc%nrx2t, fc%nrx3t, 2 )
w_wfcs_r(1:fc%nrxxt,ii)= DBLE(psic(1:fc%nrxxt))
if(ii/=num_nbndv(ispin)) w_wfcs_r(1:fc%nrxxt,ii+1)= DIMAG(psic(1:fc%nrxxt))
w_wfcs_2(1:fc%nrxxt,ii)=w_wfcs_r(1:fc%nrxxt,ii)**2
if(ii/=num_nbndv(ispin)) w_wfcs_2(1:fc%nrxxt,ii+1)=w_wfcs_r(1:fc%nrxxt,ii+1)**2
enddo
! compute the overlap matrix o_vv'
call dgemm('T','N',num_nbndv(ispin),num_nbndv(ispin),fc%nrxxt,1.d0, w_wfcs_2, &
& fc%nrxxt,w_wfcs_2, fc%nrxxt, 0.d0, o_mat,num_nbndv(ispin))
call mp_sum(o_mat,world_comm)
o_mat(1:num_nbndv(ispin),1:num_nbndv(ispin))= &
& o_mat(1:num_nbndv(ispin),1:num_nbndv(ispin))/(fc%nr1t*fc%nr2t*fc%nr3t)
do iv=1,num_nbndv(ispin)
do jv=1,num_nbndv(ispin)
write(stdout,*) 'iv,jv,o_mat',iv,jv,o_mat(iv,jv)
enddo
enddo
FLUSH(stdout)
! write it on disk
if(ionode) then
iunu = find_free_unit()
if (ispin==1) open(unit=iunu,file=trim(tmp_dir)//trim(prefix)//'.wbse1',status='unknown',form='unformatted')
if (ispin==2) open(unit=iunu,file=trim(tmp_dir)//trim(prefix)//'.wbse2',status='unknown',form='unformatted')
write(iunu) num_nbndv(ispin)
write(iunu) s_bse
do ii=1,num_nbndv(ispin)
write(iunu) o_mat(1:num_nbndv(ispin),ii)
enddo
close(iunu)
endif
! if Ovv'> s_bse compute in G-space the v*w_v*w_v' product
iungprod = find_free_unit()
if (ispin==1) CALL diropn( iungprod, 'vww_bse1.',npw*2, exst)
if (ispin==2) CALL diropn( iungprod, 'vww_bse2.',npw*2, exst)
if(debug) then
iundebug = find_free_unit()
open(iundebug,file='vww_pw4gww.dat')
endif
! compute V(G)
allocate(fac(npw))
if(l_truncated_coulomb) then
do ig=1,npw
qq = g(1,ig)**2.d0 + g(2,ig)**2.d0 + g(3,ig)**2.d0
if (qq > 1.d-8) then
fac(ig)=(e2*fpi/(tpiba2*qq))*(1.d0-dcos(dsqrt(qq)*truncation_radius*tpiba))
else
fac(ig)=e2*fpi*(truncation_radius**2.d0/2.d0)
endif
enddo
fac(:)=fac(:)/omega
else
fac(:)=0.d0
fac(1:npw)=vg_q(1:npw)
endif
allocate(ww_prod(fc%nrxxt))
allocate(ww_prodg(fc%npwt))
allocate(ww_prodg2(npw))
ii=0
do iv=1, num_nbndv(ispin)
do jv=1, num_nbndv(ispin)
if (o_mat(iv,jv)>=s_bse) then
ii=ii+1
ww_prod(1:fc%nrxxt)= w_wfcs_r(1:fc%nrxxt,iv)* w_wfcs_r(1:fc%nrxxt,jv)
psic(1:fc%nrxxt)=ww_prod(1:fc%nrxxt)
CALL cft3t(fc, psic, fc%nr1t, fc%nr2t, fc%nr3t, fc%nrx1t, fc%nrx2t, fc%nrx3t, -2 )
ww_prodg(1:fc%npwt) = psic(fc%nlt(1:fc%npwt))
if(fc%dual_t==4.d0) then
ww_prodg2(:)=ww_prodg(:)
else
call mergewf(ww_prodg,evc_g,fc%npwt,fc%ig_l2gt,mpime,nproc,ionode_id,intra_pool_comm)
call splitwf(ww_prodg2,evc_g,npw,ig_l2g,mpime,nproc,ionode_id,intra_pool_comm)
endif
ww_prodg2(1:npw)=ww_prodg2(1:npw)*fac(1:npw)
call davcio(ww_prodg2,npw*2,iungprod,ii,1)
if(debug) then
if(ionode) then
write(iundebug,*) npw
do i=1,npw
write(iundebug,*) ww_prodg2(i)
enddo
endif
endif
endif
enddo
enddo
write(stdout,*) 'bse ii found=',ii
write(stdout,*) 'max_ngm=',max_ngm
write(stdout,*) 'npw=',npw
FLUSH(stdout)
close(iungprod)
! JDE start
! Obtain polarization basis
! Iterative method
IF (l_no_GW_just_screening) THEN
allocate(ww_read_prodg2(npw))
allocate(ww_solved_prodg2(npw))
if (debug) then
wwwdebug = find_free_unit()
open(wwwdebug,file='Www_pw4gww.dat')
endif
! Open files for wannier products
wannier_product_unit = find_free_unit()
IF (ispin .EQ. 1) CALL diropn( wannier_product_unit, 'vww_bse1.',npw*2, exst)
IF (ispin .EQ. 2) CALL diropn( wannier_product_unit, 'vww_bse2.',npw*2, exst)
wannier_write_product_unit = find_free_unit()
IF (ispin .EQ. 1) CALL diropn( wannier_write_product_unit, 'Www_bse1.',npw*2, exst)
IF (ispin .EQ. 2) CALL diropn( wannier_write_product_unit, 'Www_bse2.',npw*2, exst)
! Allocate variables needed for cgsolve routine
IF (.NOT. l_no_GW_bare_coulomb) THEN
ALLOCATE(overlap(nbndx,1))
ALLOCATE(diagonal(npw))
ALLOCATE(evc_jsh(npw,1))
ALLOCATE(e(1))
ALLOCATE(b(npw,1))
ALLOCATE(u(npw,1))
ALLOCATE(h(npw,1))
ALLOCATE(Ah(npw,1))
ALLOCATE(pu(npw,1))
evc_jsh= 0.d0
diagonal= 0.d0
overlap= 0.d0
! BE VERY CAREFUL PMAT_TYPE IS A GLOBAL VARIABLE
WRITE(stdout,*) 'Note: we have changed pmat_type on the fly!'
pmat_type = 0
! BE VERY CAREFUL PMAT_TYPE IS A GLOBAL VARIABLE
END IF
INQUIRE(FILE=TRIM(tmp_dir)//TRIM(prefix)//'.restart_Www_stat', EXIST=exst)
IF (exst) THEN
IF (ionode) THEN
iunwww = find_free_unit()
OPEN(UNIT=iunwww,FILE=TRIM(tmp_dir)//TRIM(prefix)//'.restart_Www_stat')
READ(iunwww,*) start_loop_wannier
CLOSE(iunwww)
END IF
start_loop_wannier = start_loop_wannier + 1
CALL mp_bcast(start_loop_wannier, ionode_id, world_comm)
ELSE
start_loop_wannier = 1
END IF
DO loop_wannier_prod = start_loop_wannier, ii
CALL davcio(ww_read_prodg2,npw*2,wannier_product_unit,loop_wannier_prod,-1)
ww_solved_prodg2(:) = 0.d0
! NOTE: read Vww :: do not need to apply Coulomb operator
!ww_read_prodg2(1:npw) = ww_read_prodg2(1:npw) * fac(1:npw)
IF (.NOT. l_no_GW_bare_coulomb) THEN
IF (debug_operator) THEN
WRITE(*,*) 'before cg'
CALL cgsolve(operator_debug, npw, evc_jsh, npw, 1, overlap, &
1, .TRUE., .FALSE., diagonal, &
.FALSE.,e,ww_read_prodg2,u,h,Ah,pu,no_GW_cg_maxit,no_GW_cg_threshold,cg_iter,ww_solved_prodg2 )
ELSE
WRITE(stdout,*) 'CGSOLVE ww PAIR:', loop_wannier_prod
cgsolve_conv = .false.
DO WHILE (.NOT. cgsolve_conv)
overlap=0.d0
diagonal=0.d0
evc_jsh=0.d0
CALL cgsolve(operator_1_vp, npw, evc_jsh, npw, 1, overlap, 1, .TRUE., .FALSE., diagonal, .FALSE.,&
e, ww_read_prodg2, u, h, Ah, pu, no_GW_cg_maxit, no_GW_cg_threshold, cg_iter, ww_solved_prodg2 )
IF (cg_iter .LT. no_GW_cg_maxit) THEN
cgsolve_conv = .true.
END IF
END DO
END IF
ELSE
ww_solved_prodg2(1:npw) = ww_read_prodg2(1:npw)
END IF
CALL davcio(ww_solved_prodg2,npw*2,wannier_write_product_unit,loop_wannier_prod,1)
if(debug) then
if(ionode) then
write(wwwdebug,*) npw
do i=1,npw
write(wwwdebug,*) ww_solved_prodg2(i)
enddo
endif
endif
IF (ionode) THEN ! Store the last wannier product index
iunwww = find_free_unit()
OPEN(UNIT=iunwww,FILE=TRIM(tmp_dir)//TRIM(prefix)//'.restart_Www_stat')
WRITE(iunwww,*) loop_wannier_prod
CLOSE(iunwww)
END IF
END DO
IF (.NOT. l_no_GW_bare_coulomb) THEN
DEALLOCATE(overlap, diagonal, e, b, u, h, Ah, pu)
END IF
close(wannier_product_unit)
close(wannier_write_product_unit)
if (debug) close(wwwdebug)
write(stdout,*) 'files closed'
! Or read from file
ELSE
allocate(p_basis(npw,numw_prod))
CALL diropn( iungprod, 'wiwjwfc_red', npw*2, exst )
do ii=1,numw_prod
call davcio(p_basis(:,ii),npw*2,iungprod,ii,-1)
p_basis(1:npw,ii)=p_basis(1:npw,ii)*fac(1:npw)
enddo
close(iungprod)
END IF
! JDE end
! maximum number of non-zero overlap
np_max=0
do iv=1, num_nbndv(ispin)
np=0
do jv=1, num_nbndv(ispin)
if (o_mat(iv,jv)>=s_bse) np=np+1
enddo
if (np>np_max) np_max=np
enddo
IF (.NOT. l_no_GW_just_screening) THEN ! JDE
if(ionode) then
iunz = find_free_unit()
if (ispin==1) open(unit=iunz,file=trim(tmp_dir)//trim(prefix)//'.zbse1',status='unknown',form='unformatted')
if (ispin==2) open(unit=iunz,file=trim(tmp_dir)//trim(prefix)//'.zbse2',status='unknown',form='unformatted')
write(iunz) num_nbndv(ispin)
write(iunz) s_bse
write (iunz) np_max
write (iunz) numw_prod
endif
allocate(z(numw_prod))
z(1:numw_prod)=0.d0
do iv=1, num_nbndv(ispin)
do jv=1, num_nbndv(ispin)
if (o_mat(jv,iv)>=s_bse) then
ww_prod(1:fc%nrxxt)= w_wfcs_r(1:fc%nrxxt,iv)* w_wfcs_r(1:fc%nrxxt,jv)
psic(1:fc%nrxxt)=dcmplx(ww_prod(1:fc%nrxxt),0.d0)
CALL cft3t(fc, psic, fc%nr1t, fc%nr2t, fc%nr3t, fc%nrx1t, fc%nrx2t, fc%nrx3t, -2 )
ww_prodg(1:fc%npwt) = psic(fc%nlt(1:fc%npwt))
call mergewf(ww_prodg,evc_g,fc%npwt,fc%ig_l2gt,mpime,nproc,ionode_id,intra_pool_comm)
call splitwf(ww_prodg2,evc_g,npw,ig_l2g,mpime,nproc,ionode_id,intra_pool_comm)
call dgemm('T','N',numw_prod,1,2*npw,2.d0,p_basis,&
&2*npw,ww_prodg2,2*npw,0.d0,z,numw_prod)
if(gstart==2) then
do ii=1,numw_prod
z(ii)=z(ii)-dble(p_basis(1,ii)*conjg(ww_prodg2(1)))
enddo
endif
call mp_sum(z,world_comm)
if(ionode) then
write(iunz) z
endif
endif
enddo
enddo
close(iunz)
END IF ! JDE
allocate(iww(np_max))
! in file iwwbse1 we write, for each iv, the set of jv for which
! o_mat(iv,jv)>=s_bse
if(ionode) then
iuni = find_free_unit()
if (ispin==1) open(unit=iuni,file=trim(tmp_dir)//trim(prefix)//'.iwwbse1',status='unknown',form='unformatted')
if (ispin==2) open(unit=iuni,file=trim(tmp_dir)//trim(prefix)//'.iwwbse2',status='unknown',form='unformatted')
write(iuni) num_nbndv(ispin)
write(iuni) s_bse
write (iuni) np_max
endif
do iv=1, num_nbndv(ispin)
ii=0
iww(1:np_max)=0
do jv=1, num_nbndv(ispin)
if (o_mat(jv,iv)>=s_bse) then
ii=ii+1
iww(ii)=jv
endif
enddo
if(ionode) write(iuni) iww
enddo
if (ionode) close(iuni) ! JDE fixes a seg fault when fcheck=bounds
if(debug) close(iundebug)
deallocate(iww)
IF (.NOT. l_no_GW_just_screening) THEN ! JDE
deallocate(z)
deallocate(p_basis)
END IF ! JDE
deallocate(ww_prod)
deallocate(ww_prodg)
deallocate(ww_prodg2)
deallocate(fac)
call deallocate_fft_custom(fc)
deallocate(w_wfcs_t)
deallocate(w_wfcs_r)
deallocate(w_wfcs_2)
deallocate(evc_g)
end subroutine