lower level routines for linear scaling SCF More...

Functions/Subroutines
subroutine, public	ls_scf_init_matrix_s (matrix_s, ls_scf_env)
	initialize S matrix related properties (sqrt, inverse...) Might be factored-out since this seems common code with the other SCF.

subroutine, public	compute_matrix_preconditioner (matrix_s, preconditioner_type, ls_mstruct, matrix_bs_sqrt, matrix_bs_sqrt_inv, threshold, order, eps_lanczos, max_iter_lanczos)
	compute for a block positive definite matrix s (bs) the sqrt(bs) and inv(sqrt(bs))

subroutine, public	apply_matrix_preconditioner (matrix, direction, matrix_bs_sqrt, matrix_bs_sqrt_inv)
	apply a preconditioner either forward (precondition) inv(sqrt(bs)) * A * inv(sqrt(bs)) backward (restore to old form) sqrt(bs) * A * sqrt(bs)

subroutine, public	density_matrix_sign (matrix_p, mu, fixed_mu, sign_method, sign_order, matrix_ks, matrix_s, matrix_s_inv, nelectron, threshold, sign_symmetric, submatrix_sign_method, matrix_s_sqrt_inv)
	compute the density matrix with a trace that is close to nelectron. take a mu as input, and improve by bisection as needed.

subroutine, public	density_matrix_sign_fixed_mu (matrix_p, trace, mu, sign_method, sign_order, matrix_ks, matrix_s, matrix_s_inv, threshold, sign_symmetric, submatrix_sign_method, matrix_s_sqrt_inv)
	for a fixed mu, compute the corresponding density matrix and its trace

subroutine, public	density_matrix_trs4 (matrix_p, matrix_ks, matrix_s_sqrt_inv, nelectron, threshold, e_homo, e_lumo, e_mu, dynamic_threshold, matrix_ks_deviation, max_iter_lanczos, eps_lanczos, converged, iounit)
	compute the density matrix using a trace-resetting algorithm

subroutine, public	density_matrix_tc2 (matrix_p, matrix_ks, matrix_s_sqrt_inv, nelectron, threshold, e_homo, e_lumo, non_monotonic, eps_lanczos, max_iter_lanczos, iounit)
	compute the density matrix using a non monotonic trace conserving algorithm based on SIAM DOI. 10.1137/130911585. 2014.04 created [Jonathan Mullin]

subroutine, public	compute_homo_lumo (matrix_k, matrix_p, eps_min, eps_max, threshold, max_iter_lanczos, eps_lanczos, homo, lumo, unit_nr)
	compute the homo and lumo given a KS matrix and a density matrix in the orthonormalized basis and the eps_min and eps_max, min and max eigenvalue of the ks matrix

Detailed Description

lower level routines for linear scaling SCF

History: 2010.10 created [Joost VandeVondele]

Author: Joost VandeVondele

Function/Subroutine Documentation

◆ ls_scf_init_matrix_s()

subroutine, public dm_ls_scf_methods::ls_scf_init_matrix_s	(	type(dbcsr_type)	matrix_s,
		type(ls_scf_env_type)	ls_scf_env
	)

initialize S matrix related properties (sqrt, inverse...) Might be factored-out since this seems common code with the other SCF.

Parameters

matrix_s	...
ls_scf_env	...

History: 2010.10 created [Joost VandeVondele]

Author: Joost VandeVondele

Definition at line 73 of file dm_ls_scf_methods.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ compute_matrix_preconditioner()

subroutine, public dm_ls_scf_methods::compute_matrix_preconditioner	(	type(dbcsr_type), intent(inout)	matrix_s,
		integer	preconditioner_type,
		type(ls_mstruct_type)	ls_mstruct,
		type(dbcsr_type), intent(inout)	matrix_bs_sqrt,
		type(dbcsr_type), intent(inout)	matrix_bs_sqrt_inv,
		real(kind=dp)	threshold,
		integer	order,
		real(kind=dp)	eps_lanczos,
		integer	max_iter_lanczos
	)

compute for a block positive definite matrix s (bs) the sqrt(bs) and inv(sqrt(bs))

Parameters

matrix_s	...
preconditioner_type	...
ls_mstruct	...
matrix_bs_sqrt	...
matrix_bs_sqrt_inv	...
threshold	...
order	...
eps_lanczos	...
max_iter_lanczos	...

History: 2010.10 created [Joost VandeVondele]

Author: Joost VandeVondele

Definition at line 216 of file dm_ls_scf_methods.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ apply_matrix_preconditioner()

subroutine, public dm_ls_scf_methods::apply_matrix_preconditioner	(	type(dbcsr_type), intent(inout)	matrix,
		character(len=*)	direction,
		type(dbcsr_type), intent(inout)	matrix_bs_sqrt,
		type(dbcsr_type), intent(inout)	matrix_bs_sqrt_inv
	)

apply a preconditioner either forward (precondition) inv(sqrt(bs)) * A * inv(sqrt(bs)) backward (restore to old form) sqrt(bs) * A * sqrt(bs)

Parameters

matrix	...
direction	...
matrix_bs_sqrt	...
matrix_bs_sqrt_inv	...

History: 2010.10 created [Joost VandeVondele]

Author: Joost VandeVondele

Definition at line 318 of file dm_ls_scf_methods.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ density_matrix_sign()

subroutine, public dm_ls_scf_methods::density_matrix_sign	(	type(dbcsr_type), intent(inout)	matrix_p,
		real(kind=dp), intent(inout)	mu,
		logical	fixed_mu,
		integer	sign_method,
		integer	sign_order,
		type(dbcsr_type), intent(inout)	matrix_ks,
		type(dbcsr_type), intent(inout)	matrix_s,
		type(dbcsr_type), intent(inout)	matrix_s_inv,
		integer, intent(in)	nelectron,
		real(kind=dp), intent(in)	threshold,
		logical, optional	sign_symmetric,
		integer, optional	submatrix_sign_method,
		type(dbcsr_type), intent(in), optional	matrix_s_sqrt_inv
	)

compute the density matrix with a trace that is close to nelectron. take a mu as input, and improve by bisection as needed.

Parameters

matrix_p	...
mu	...
fixed_mu	...
sign_method	...
sign_order	...
matrix_ks	...
matrix_s	...
matrix_s_inv	...
nelectron	...
threshold	...
sign_symmetric	...
submatrix_sign_method	...
matrix_s_sqrt_inv	...

History: 2010.10 created [Joost VandeVondele] 2020.07 support for methods with internal mu adjustment [Michael Lass]

Author: Joost VandeVondele

Definition at line 374 of file dm_ls_scf_methods.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ density_matrix_sign_fixed_mu()

subroutine, public dm_ls_scf_methods::density_matrix_sign_fixed_mu	(	type(dbcsr_type), intent(inout)	matrix_p,
		real(kind=dp), intent(out)	trace,
		real(kind=dp), intent(inout)	mu,
		integer	sign_method,
		integer	sign_order,
		type(dbcsr_type), intent(inout)	matrix_ks,
		type(dbcsr_type), intent(inout)	matrix_s,
		type(dbcsr_type), intent(inout)	matrix_s_inv,
		real(kind=dp), intent(in)	threshold,
		logical	sign_symmetric,
		integer	submatrix_sign_method,
		type(dbcsr_type), intent(in), optional	matrix_s_sqrt_inv
	)

for a fixed mu, compute the corresponding density matrix and its trace

Parameters

matrix_p	...
trace	...
mu	...
sign_method	...
sign_order	...
matrix_ks	...
matrix_s	...
matrix_s_inv	...
threshold	...
sign_symmetric	...
submatrix_sign_method	...
matrix_s_sqrt_inv	...

History: 2010.10 created [Joost VandeVondele]

Author: Joost VandeVondele

Definition at line 484 of file dm_ls_scf_methods.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ density_matrix_trs4()

subroutine, public dm_ls_scf_methods::density_matrix_trs4	(	type(dbcsr_type), intent(inout)	matrix_p,
		type(dbcsr_type), intent(in)	matrix_ks,
		type(dbcsr_type), intent(in)	matrix_s_sqrt_inv,
		integer, intent(in)	nelectron,
		real(kind=dp), intent(in)	threshold,
		real(kind=dp), intent(inout)	e_homo,
		real(kind=dp), intent(inout)	e_lumo,
		real(kind=dp), intent(inout)	e_mu,
		logical, intent(in), optional	dynamic_threshold,
		type(dbcsr_type), intent(inout), optional	matrix_ks_deviation,
		integer, intent(in)	max_iter_lanczos,
		real(kind=dp), intent(in)	eps_lanczos,
		logical, intent(out), optional	converged,
		integer, intent(in), optional	iounit
	)

compute the density matrix using a trace-resetting algorithm

Parameters

matrix_p	...
matrix_ks	...
matrix_s_sqrt_inv	...
nelectron	...
threshold	...
e_homo	...
e_lumo	...
e_mu	...
dynamic_threshold	...
matrix_ks_deviation	...
max_iter_lanczos	...
eps_lanczos	...
converged	...
iounit	...

History: 2012.06 created [Florian Thoele]

Author: Florian Thoele

Definition at line 725 of file dm_ls_scf_methods.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ density_matrix_tc2()

subroutine, public dm_ls_scf_methods::density_matrix_tc2	(	type(dbcsr_type), intent(inout)	matrix_p,
		type(dbcsr_type), intent(in)	matrix_ks,
		type(dbcsr_type), intent(in)	matrix_s_sqrt_inv,
		integer, intent(in)	nelectron,
		real(kind=dp), intent(in)	threshold,
		real(kind=dp), intent(inout)	e_homo,
		real(kind=dp), intent(inout)	e_lumo,
		logical, intent(in), optional	non_monotonic,
		real(kind=dp), intent(in)	eps_lanczos,
		integer, intent(in)	max_iter_lanczos,
		integer, intent(in), optional	iounit
	)

compute the density matrix using a non monotonic trace conserving algorithm based on SIAM DOI. 10.1137/130911585. 2014.04 created [Jonathan Mullin]

Parameters

matrix_p	...
matrix_ks	...
matrix_s_sqrt_inv	...
nelectron	...
threshold	...
e_homo	...
e_lumo	...
non_monotonic	...
eps_lanczos	...
max_iter_lanczos	...
iounit	...

Author: Jonathan Mullin

Definition at line 1027 of file dm_ls_scf_methods.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ compute_homo_lumo()

subroutine, public dm_ls_scf_methods::compute_homo_lumo	(	type(dbcsr_type)	matrix_k,
		type(dbcsr_type)	matrix_p,
		real(kind=dp)	eps_min,
		real(kind=dp)	eps_max,
		real(kind=dp)	threshold,
		integer, intent(in)	max_iter_lanczos,
		real(kind=dp), intent(in)	eps_lanczos,
		real(kind=dp)	homo,
		real(kind=dp)	lumo,
		integer	unit_nr
	)

compute the homo and lumo given a KS matrix and a density matrix in the orthonormalized basis and the eps_min and eps_max, min and max eigenvalue of the ks matrix

Parameters

matrix_k	...
matrix_p	...
eps_min	...
eps_max	...
threshold	...
max_iter_lanczos	...
eps_lanczos	...
homo	...
lumo	...
unit_nr	...

History: 2012.06 created [Florian Thoele]

Author: Florian Thoele

Definition at line 1298 of file dm_ls_scf_methods.F.

Here is the call graph for this function:

Here is the caller graph for this function:

Functions/Subroutines

Detailed Description

Function/Subroutine Documentation

◆ ls_scf_init_matrix_s()

◆ compute_matrix_preconditioner()

◆ apply_matrix_preconditioner()

◆ density_matrix_sign()

◆ density_matrix_sign_fixed_mu()

◆ density_matrix_trs4()

◆ density_matrix_tc2()

◆ compute_homo_lumo()