Different diagonalization schemes that can be used for the iterative solution of the eigenvalue problem. More...

Functions/Subroutines
real(kind=dp) function	adiis_candidate_shift (scf_env)
	Fade candidate regularization through handover; never regularize a raw-map check.
subroutine, public	general_eigenproblem (scf_env, mos, matrix_ks, matrix_s, scf_control, scf_section, diis_step)
	the inner loop of scf, specific to diagonalization with S matrix basically, in goes the ks matrix out goes a new p matrix
subroutine, public	do_general_diag (scf_env, mos, matrix_ks, matrix_s, scf_control, scf_section, diis_step, probe)
	...
subroutine, public	do_general_diag_kp (matrix_ks, matrix_s, kpoints, scf_env, scf_control, update_p, diis_step, diis_error, qs_env, probe, added_mos_auto_grow, potential)
	Kpoint diagonalization routine Transforms matrices to kpoint, distributes kpoint groups, performs general diagonalization (no storgae of overlap decomposition), stores MOs, calculates occupation numbers, calculates density matrices in kpoint representation, transforms density matrices to real space.
subroutine, public	diag_kp_basic (matrix_ks, matrix_s, kpoints, fmwork)
	Kpoint diagonalization routine Transforms matrices to kpoint, distributes kpoint groups, performs general diagonalization (no storgae of overlap decomposition), stores MOs, calculates occupation numbers, calculates density matrices in kpoint representation, transforms density matrices to real space.
subroutine, public	do_scf_diag_subspace (qs_env, scf_env, subspace_env, mos, rho, ks_env, scf_section, scf_control)
	inner loop within MOS subspace, to refine occupation and density, before next diagonalization of the Hamiltonian
subroutine, public	diag_subspace_allocate (subspace_env, qs_env, mos)
	...
subroutine, public	do_special_diag (scf_env, mos, matrix_ks, scf_control, scf_section, diis_step)
	the inner loop of scf, specific to diagonalization without S matrix basically, in goes the ks matrix out goes a new p matrix
subroutine, public	do_ot_diag (scf_env, mos, matrix_ks, matrix_s, scf_control, scf_section, diis_step)
	the inner loop of scf, specific to iterative diagonalization using OT with S matrix; basically, in goes the ks matrix out goes a new p matrix
subroutine, public	do_ot_diag_kp (matrix_ks, matrix_s, matrix_t, kpoints, scf_env, scf_control, diis_step, added_mos_auto_grow)
	fixed-H OT diagonalization for real or complex K-point channels
subroutine, public	do_roks_diag (scf_env, mos, matrix_ks, matrix_s, scf_control, scf_section, diis_step, orthogonal_basis)
	Solve a set restricted open Kohn-Sham (ROKS) equations based on the alpha and beta Kohn-Sham matrices from unrestricted Kohn-Sham.
subroutine, public	do_block_krylov_diag (scf_env, mos, matrix_ks, scf_control, scf_section, check_moconv_only)
	iterative diagonalization using the block Krylov-space approach
subroutine, public	do_block_davidson_diag (qs_env, scf_env, mos, matrix_ks, matrix_s, scf_control, scf_section, check_moconv_only)
	iterative diagonalization using the block davidson space approach
subroutine, public	do_block_davidson_diag_kp (qs_env, scf_env, matrix_ks_kp, matrix_s_kp, kpoints, scf_control, added_mos_auto_grow)
	block-Davidson SCF step for complex K points, one channel per (local kpoint, spin); mirrors do_block_davidson_diag for the preconditioner scheduling and do_general_diag_kp for the H(k)/S(k) assembly and the density exit chain
subroutine, public	do_block_krylov_diag_kp (qs_env, scf_env, matrix_ks_kp, matrix_s_kp, kpoints, scf_control, check_moconv_only, added_mos_auto_grow)
	Block Krylov (Lanczos) refinement of the k-point MOS, the complex counterpart of do_block_krylov_diag: every channel (local kpoint, spin) is standardized with the Cholesky factor of S(k), refined by the complex block-Krylov kernel, and rotated back into the generalized problem.
subroutine, public	diag_kp_smat (matrix_s, kpoints, fmwork, lowdin_forces)
	Kpoint diagonalization routine Transforms matrices to kpoint, distributes kpoint groups, performs diagonalization.

Detailed Description

Different diagonalization schemes that can be used for the iterative solution of the eigenvalue problem.

History: started from routines previously located in the qs_scf module 05.2009

Function/Subroutine Documentation

◆ adiis_candidate_shift()

real(kind=dp) function qs_scf_diagonalization::adiis_candidate_shift ( type(qs_scf_env_type), intent(in) scf_env )

Fade candidate regularization through handover; never regularize a raw-map check.

Parameters

scf_env Current SCF state.

Returns: ...

Definition at line 218 of file qs_scf_diagonalization.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ general_eigenproblem()

subroutine, public qs_scf_diagonalization::general_eigenproblem	(	type(qs_scf_env_type), pointer	scf_env,
		type(mo_set_type), dimension(:), intent(in)	mos,
		type(dbcsr_p_type), dimension(:), pointer	matrix_ks,
		type(dbcsr_p_type), dimension(:), pointer	matrix_s,
		type(scf_control_type), pointer	scf_control,
		type(section_vals_type), pointer	scf_section,
		logical, intent(inout)	diis_step )

the inner loop of scf, specific to diagonalization with S matrix basically, in goes the ks matrix out goes a new p matrix

Parameters

scf_env	...
mos	...
matrix_ks	...
matrix_s	...
scf_control	...
scf_section	...
diis_step	...

History: 03.2006 created [Joost VandeVondele]

Definition at line 319 of file qs_scf_diagonalization.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ do_general_diag()

subroutine, public qs_scf_diagonalization::do_general_diag	(	type(qs_scf_env_type), pointer	scf_env,
		type(mo_set_type), dimension(:), intent(inout)	mos,
		type(dbcsr_p_type), dimension(:), pointer	matrix_ks,
		type(dbcsr_p_type), dimension(:), pointer	matrix_s,
		type(scf_control_type), pointer	scf_control,
		type(section_vals_type), pointer	scf_section,
		logical, intent(inout)	diis_step,
		type(hairy_probes_type), dimension(:), optional, pointer	probe )

...

Parameters

scf_env	...
mos	...
matrix_ks	...
matrix_s	...
scf_control	...
scf_section	...
diis_step	...
probe	...

Definition at line 633 of file qs_scf_diagonalization.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ do_general_diag_kp()

subroutine, public qs_scf_diagonalization::do_general_diag_kp	(	type(dbcsr_p_type), dimension(:, :), pointer	matrix_ks,
		type(dbcsr_p_type), dimension(:, :), pointer	matrix_s,
		type(kpoint_type), pointer	kpoints,
		type(qs_scf_env_type), pointer	scf_env,
		type(scf_control_type), pointer	scf_control,
		logical, intent(in)	update_p,
		logical, intent(inout)	diis_step,
		real(dp), intent(inout), optional	diis_error,
		type(qs_environment_type), optional, pointer	qs_env,
		type(hairy_probes_type), dimension(:), optional, pointer	probe,
		logical, intent(out), optional	added_mos_auto_grow,
		real(dp), dimension(:, :), intent(in), optional, target	potential )

Kpoint diagonalization routine Transforms matrices to kpoint, distributes kpoint groups, performs general diagonalization (no storgae of overlap decomposition), stores MOs, calculates occupation numbers, calculates density matrices in kpoint representation, transforms density matrices to real space.

Parameters

matrix_ks	Kohn-sham matrices (RS indices, global)
matrix_s	Overlap matrices (RS indices, global)
kpoints	Kpoint environment
scf_env	SCF environment
scf_control	SCF control variables
update_p	...
diis_step	...
diis_error	...
qs_env	...
probe	...
added_mos_auto_grow	true when the adaptive virtual-space buffer must grow
potential	Fixed Lowdin orbital potential for non-self-consistent k points.

History: 08.2014 created [JGH]

Definition at line 706 of file qs_scf_diagonalization.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ diag_kp_basic()

subroutine, public qs_scf_diagonalization::diag_kp_basic	(	type(dbcsr_p_type), dimension(:, :), pointer	matrix_ks,
		type(dbcsr_p_type), dimension(:, :), pointer	matrix_s,
		type(kpoint_type), pointer	kpoints,
		type(cp_fm_type), dimension(:), pointer	fmwork )

Kpoint diagonalization routine Transforms matrices to kpoint, distributes kpoint groups, performs general diagonalization (no storgae of overlap decomposition), stores MOs, calculates occupation numbers, calculates density matrices in kpoint representation, transforms density matrices to real space.

Parameters

matrix_ks	Kohn-sham matrices (RS indices, global)
matrix_s	Overlap matrices (RS indices, global)
kpoints	Kpoint environment
fmwork	global-layout FM template in the first array entry

History: 08.2014 created [JGH]

Definition at line 1580 of file qs_scf_diagonalization.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ do_scf_diag_subspace()

subroutine, public qs_scf_diagonalization::do_scf_diag_subspace	(	type(qs_environment_type), pointer	qs_env,
		type(qs_scf_env_type), pointer	scf_env,
		type(subspace_env_type), pointer	subspace_env,
		type(mo_set_type), dimension(:), intent(inout)	mos,
		type(qs_rho_type), pointer	rho,
		type(qs_ks_env_type), pointer	ks_env,
		type(section_vals_type), pointer	scf_section,
		type(scf_control_type), pointer	scf_control )

inner loop within MOS subspace, to refine occupation and density, before next diagonalization of the Hamiltonian

Parameters

qs_env	...
scf_env	...
subspace_env	...
mos	...
rho	...
ks_env	...
scf_section	...
scf_control	...

History: 09.2009 created [MI]

Note: it is assumed that when diagonalization is used, also some mixing procedure is active

Definition at line 1703 of file qs_scf_diagonalization.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ diag_subspace_allocate()

subroutine, public qs_scf_diagonalization::diag_subspace_allocate	(	type(subspace_env_type), pointer	subspace_env,
		type(qs_environment_type), pointer	qs_env,
		type(mo_set_type), dimension(:), intent(in)	mos )

...

Parameters

subspace_env	...
qs_env	...
mos	...

Definition at line 1918 of file qs_scf_diagonalization.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ do_special_diag()

subroutine, public qs_scf_diagonalization::do_special_diag	(	type(qs_scf_env_type), pointer	scf_env,
		type(mo_set_type), dimension(:), intent(inout)	mos,
		type(dbcsr_p_type), dimension(:), pointer	matrix_ks,
		type(scf_control_type), pointer	scf_control,
		type(section_vals_type), pointer	scf_section,
		logical, intent(inout)	diis_step )

the inner loop of scf, specific to diagonalization without S matrix basically, in goes the ks matrix out goes a new p matrix

Parameters

scf_env	...
mos	...
matrix_ks	...
scf_control	...
scf_section	...
diis_step	...

History: 03.2006 created [Joost VandeVondele]

Definition at line 1987 of file qs_scf_diagonalization.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ do_ot_diag()

subroutine, public qs_scf_diagonalization::do_ot_diag	(	type(qs_scf_env_type), pointer	scf_env,
		type(mo_set_type), dimension(:), intent(inout)	mos,
		type(dbcsr_p_type), dimension(:), pointer	matrix_ks,
		type(dbcsr_p_type), dimension(:), pointer	matrix_s,
		type(scf_control_type), pointer	scf_control,
		type(section_vals_type), pointer	scf_section,
		logical, intent(inout)	diis_step )

the inner loop of scf, specific to iterative diagonalization using OT with S matrix; basically, in goes the ks matrix out goes a new p matrix

Parameters

scf_env	...
mos	...
matrix_ks	...
matrix_s	...
scf_control	...
scf_section	...
diis_step	...

History: 10.2008 created [JGH]

Definition at line 2090 of file qs_scf_diagonalization.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ do_ot_diag_kp()

subroutine, public qs_scf_diagonalization::do_ot_diag_kp	(	type(dbcsr_p_type), dimension(:, :), pointer	matrix_ks,
		type(dbcsr_p_type), dimension(:, :), pointer	matrix_s,
		type(dbcsr_p_type), dimension(:, :), pointer	matrix_t,
		type(kpoint_type), pointer	kpoints,
		type(qs_scf_env_type), pointer	scf_env,
		type(scf_control_type), pointer	scf_control,
		logical, intent(inout)	diis_step,
		logical, intent(out)	added_mos_auto_grow )

fixed-H OT diagonalization for real or complex K-point channels

Parameters

matrix_ks	real-space Kohn-Sham matrices
matrix_s	real-space overlap matrices
matrix_t	...
kpoints	K-point environment
scf_env	SCF environment
scf_control	SCF controls
diis_step	set false; density mixing remains the outer accelerator
added_mos_auto_grow	request a larger adaptive virtual-space buffer

Definition at line 2190 of file qs_scf_diagonalization.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ do_roks_diag()

subroutine, public qs_scf_diagonalization::do_roks_diag	(	type(qs_scf_env_type), pointer	scf_env,
		type(mo_set_type), dimension(:), intent(in)	mos,
		type(dbcsr_p_type), dimension(:), pointer	matrix_ks,
		type(dbcsr_p_type), dimension(:), pointer	matrix_s,
		type(scf_control_type), pointer	scf_control,
		type(section_vals_type), pointer	scf_section,
		logical, intent(inout)	diis_step,
		logical, intent(in)	orthogonal_basis )

Solve a set restricted open Kohn-Sham (ROKS) equations based on the alpha and beta Kohn-Sham matrices from unrestricted Kohn-Sham.

Parameters

scf_env	...
mos	...
matrix_ks	...
matrix_s	...
scf_control	...
scf_section	...
diis_step	...
orthogonal_basis	...

History: 04.2006 created [MK] Revised (01.05.06,MK)

Note: this is only a high-spin ROKS.

Definition at line 2351 of file qs_scf_diagonalization.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ do_block_krylov_diag()

subroutine, public qs_scf_diagonalization::do_block_krylov_diag	(	type(qs_scf_env_type), pointer	scf_env,
		type(mo_set_type), dimension(:), intent(inout)	mos,
		type(dbcsr_p_type), dimension(:), pointer	matrix_ks,
		type(scf_control_type), pointer	scf_control,
		type(section_vals_type), pointer	scf_section,
		logical, intent(in), optional	check_moconv_only )

iterative diagonalization using the block Krylov-space approach

Parameters

scf_env	...
mos	...
matrix_ks	...
scf_control	...
scf_section	...
check_moconv_only	...

par History 05.2009 created [MI]

Definition at line 2626 of file qs_scf_diagonalization.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ do_block_davidson_diag()

subroutine, public qs_scf_diagonalization::do_block_davidson_diag	(	type(qs_environment_type), pointer	qs_env,
		type(qs_scf_env_type), pointer	scf_env,
		type(mo_set_type), dimension(:), intent(inout)	mos,
		type(dbcsr_p_type), dimension(:), pointer	matrix_ks,
		type(dbcsr_p_type), dimension(:), pointer	matrix_s,
		type(scf_control_type), pointer	scf_control,
		type(section_vals_type), pointer	scf_section,
		logical, intent(in), optional	check_moconv_only )

iterative diagonalization using the block davidson space approach

Parameters

qs_env	...
scf_env	...
mos	...
matrix_ks	...
matrix_s	...
scf_control	...
scf_section	...
check_moconv_only	...

par History 05.2011 created [MI]

Definition at line 2834 of file qs_scf_diagonalization.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ do_block_davidson_diag_kp()

subroutine, public qs_scf_diagonalization::do_block_davidson_diag_kp	(	type(qs_environment_type), pointer	qs_env,
		type(qs_scf_env_type), pointer	scf_env,
		type(dbcsr_p_type), dimension(:, :), pointer	matrix_ks_kp,
		type(dbcsr_p_type), dimension(:, :), pointer	matrix_s_kp,
		type(kpoint_type), pointer	kpoints,
		type(scf_control_type), pointer	scf_control,
		logical, intent(out), optional	added_mos_auto_grow )

block-Davidson SCF step for complex K points, one channel per (local kpoint, spin); mirrors do_block_davidson_diag for the preconditioner scheduling and do_general_diag_kp for the H(k)/S(k) assembly and the density exit chain

Parameters

qs_env	...
scf_env	...
matrix_ks_kp	real-space Kohn-Sham matrices, dbcsr image set
matrix_s_kp	real-space overlap matrix, dbcsr image set
kpoints	...
scf_control	...
added_mos_auto_grow	...

Definition at line 2953 of file qs_scf_diagonalization.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ do_block_krylov_diag_kp()

subroutine, public qs_scf_diagonalization::do_block_krylov_diag_kp	(	type(qs_environment_type), pointer	qs_env,
		type(qs_scf_env_type), pointer	scf_env,
		type(dbcsr_p_type), dimension(:, :), pointer	matrix_ks_kp,
		type(dbcsr_p_type), dimension(:, :), pointer	matrix_s_kp,
		type(kpoint_type), pointer	kpoints,
		type(scf_control_type), pointer	scf_control,
		logical, intent(in), optional	check_moconv_only,
		logical, intent(out), optional	added_mos_auto_grow )

Block Krylov (Lanczos) refinement of the k-point MOS, the complex counterpart of do_block_krylov_diag: every channel (local kpoint, spin) is standardized with the Cholesky factor of S(k), refined by the complex block-Krylov kernel, and rotated back into the generalized problem.

Parameters

qs_env	...
scf_env	...
matrix_ks_kp	real-space Kohn-Sham matrix, dbcsr image set
matrix_s_kp	real-space overlap matrix, dbcsr image set
kpoints	...
scf_control	...
check_moconv_only	only report the convergence statistics
added_mos_auto_grow	signals the outer loop that the virtual-space buffer must grow before this SCF step is redone

Definition at line 3182 of file qs_scf_diagonalization.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ diag_kp_smat()

subroutine, public qs_scf_diagonalization::diag_kp_smat	(	type(dbcsr_p_type), dimension(:, :), pointer	matrix_s,
		type(kpoint_type), pointer	kpoints,
		type(cp_fm_type), dimension(:)	fmwork,
		logical, intent(in), optional	lowdin_forces )

Kpoint diagonalization routine Transforms matrices to kpoint, distributes kpoint groups, performs diagonalization.

Parameters

matrix_s	Overlap matrices (RS indices, global)
kpoints	Kpoint environment
fmwork	full matrices distributed over all groups
lowdin_forces	retain the overlap spectrum for Lowdin force/stress derivatives

History: 02.2026 created [JGH]

Definition at line 3734 of file qs_scf_diagonalization.F.

Here is the call graph for this function:

Here is the caller graph for this function:

Functions/Subroutines

Detailed Description

Function/Subroutine Documentation

◆ adiis_candidate_shift()

◆ general_eigenproblem()

◆ do_general_diag()

◆ do_general_diag_kp()

◆ diag_kp_basic()

◆ do_scf_diag_subspace()

◆ diag_subspace_allocate()

◆ do_special_diag()

◆ do_ot_diag()

◆ do_ot_diag_kp()

◆ do_roks_diag()

◆ do_block_krylov_diag()

◆ do_block_davidson_diag()

◆ do_block_davidson_diag_kp()

◆ do_block_krylov_diag_kp()

◆ diag_kp_smat()