11.4.1.4.4. dftgwbse

Compute electronic excitations using GW-BSE The following table contains the defaults input options for the calculator, The default OPTIONAL means this option is switched off, if no input is given. REQUIRED arguments have to be specified, otherwise an error is thrown.

Property Name	Default Value	Description	Valid Input
job_name	system	Input file name without extension, also used for intermediate files
dftpackage.name	xtp	Name of the DFT package	xtp orca
dftpackage.charge	0	Molecular charge	int
dftpackage.spin	1	Molecular multiplicity	int+
dftpackage.basisset	def2-tzvp	Basis set for MOs
dftpackage.auxbasisset	OPTIONAL	Auxiliary basis set for RI
dftpackage.externalfield	OPTIONAL	Field given in x y z components
dftpackage.executable	OPTIONAL	Path to executable for dftpackage
dftpackage.ecp	OPTIONAL	Effective Core Potentials for DFT Calculations
dftpackage.optimize	false	Perform a molecular geometry optimization	bool
dftpackage.functional	XC_HYB_GGA_XC_PBEH	Exchange correlation functional used. You can also specify an exchange and a correlation functional
dftpackage.scratch	/tmp/qmpackage	path to the scratch folder
dftpackage.polarization	false	Calculate polarisation	bool
dftpackage.dipole_spacing	0.1	Spacing for splitting multipoles into charges	float+
dftpackage.temporary_file	temp	Filename for temporary files without extension
dftpackage.convergence_tightness	tight	How accurate the dft self-consistency has to be	low normal tight verytight
dftpackage.cleanup		files to remove after the calculation. Specify the fileextensions
dftpackage.initial_guess	atom	Method to use to make initial guess, independent(electrons) or atom(densities) or previous calculation keyword orbfile or dimer_guess (combine two monomer .orb files, see dimer_guess_orbA/orbB below; closed-shell monomers work for restricted and unrestricted runs, open-shell monomers need the unrestricted path)	independent atom orbfile huckel huckel_dft dimer_guess
dftpackage.dimer_guess_orbA		Path to monomer A’s own, already-converged .orb file – only read when initial_guess=dimer_guess. Monomer A’s atoms must be an exact internal-geometry match (bond lengths/angles, NOT absolute position – translation/rotation between the standalone monomer and its placement in the dimer is expected and fine) for the first N_A atoms of this calculation’s own molecule.
dftpackage.dimer_guess_orbB		Same as dimer_guess_orbA, for monomer B – expected to match this calculation’s own remaining atoms (after monomer A’s own N_A atoms).
dftpackage.orca		orca specific keywords can be added here, where the xml tag corresponds to the option and the value to the option’s value
dftpackage.xtpdft.dft_in_dft.activeatoms		Indices of atoms in active region
dftpackage.xtpdft.dft_in_dft.threshold	0.4	Mulliken population above which orbital is considered in the active region
dftpackage.xtpdft.dft_in_dft.levelshift	10000.0	Levelshift energy for the projection operator
dftpackage.xtpdft.dft_in_dft.truncate_basis	false	Truncate the basis-set to active atoms only	bool
dftpackage.xtpdft.dft_in_dft.truncation_threshold	1e-4	Threshold to convert an inactive atom to a border atom	float+
dftpackage.xtpdft.screening_eps	1e-9	screening eps	float+
dftpackage.xtpdft.overlap_tolerance	1e-8	Eigenvalues of the AO overlap matrix below this value are removed from the orthogonalization (near linear dependencies; the removed directions become zero orbitals at the top of the spectrum). Raise to 1e-7 to 1e-6 if the log reports a nearly singular overlap and the SCF is unstable.	float+
dftpackage.xtpdft.ri_pair_threshold	1e-10	RI only: basis-function shell pairs whose three-centre integrals (P|ab) are bounded by less than this (Schwarz bound) are not stored. Reduces the memory of the stored RI tensor substantially for large molecules; energies change at the level of the threshold. 0 stores all pairs.	float+
dftpackage.xtpdft.fock_matrix_reset	5	how often the fock matrix is reset	int+
dftpackage.xtpdft.integration_grid	medium	vxc grid quality	xcoarse coarse medium fine xfine
dftpackage.xtpdft.force_uks_path	false	forcing UKS path	bool
dftpackage.xtpdft.polarizability	false	After the SCF, compute the static dipole polarizability tensor analytically (coupled-perturbed Kohn-Sham with the three dipole operators; closed-shell states, LDA/GGA/global hybrids) and store it in the orb (HDF5) file, e.g. as the target for molpol (qmpackage xtp).	bool
dftpackage.xtpdft.compute_forces	false	Compute analytic ground-state DFT nuclear forces (RI, and hybrid-exchange for UKS) and store them in the orb (HDF5) file. Adds real cost to every converged SCF, so this is opt-in.	bool
dftpackage.xtpdft.cdft.enabled	false	Enable a CDFT charge constraint on this calculation	bool
dftpackage.xtpdft.cdft.indices		Indices of atoms in the constrained fragment (0-based), using the same range syntax already used for diabatization.xml’s own fragment indices, e.g. ‘1 3 13:17’
dftpackage.xtpdft.cdft.charge	0.0	Target charge on the fragment, RELATIVE to its neutral reference state (the sum of the fragment atoms’ own nuclear charges) – e.g. +1.0 means one electron REMOVED from the fragment (a cation), -1.0 means one electron ADDED (an anion). Converted internally to an absolute target electron count once, at setup time.	float
dftpackage.xtpdft.cdft.initial_lambda	0.0	Initial guess for the Lagrange multiplier lambda, used as the center of the outer loop’s own bisection bracket	float
dftpackage.xtpdft.cdft.population_tolerance	1e-4	Outer-loop convergence criterion: the CDFT calculation is considered converged once the achieved fragment population is within this many electrons of the target	float+
dftpackage.xtpdft.cdft.max_iterations	50	Maximum number of outer (Lagrange-multiplier, bisection) iterations – each one is a full, warm-started inner SCF	int+
dftpackage.xtpdft.cdft.guess_strategy	warmstart	How each outer bisection trial’s own inner SCF is started, after the very first trial. ‘warmstart’ (default) reuses the immediately PRECEDING trial’s own converged MOs (initial_guess is forced to ‘orbfile’ internally) for every subsequent trial – usually helps, since consecutive lambda values are often close. ‘fresh’ instead re-uses the calculation’s own original, top-level initial_guess setting for every trial, never warm-starting from a previous lambda’s own MOs at all – worth trying if warm-starting itself seems to be part of a convergence difficulty (e.g. if consecutive lambda trials correspond to substantially different electronic structures, a previous trial’s own converged density could be a worse starting point than a fresh guess, not a better one).	warmstart fresh
dftpackage.xtpdft.convergence.energy	1E-7	DeltaE at which calculation is converged	float+
dftpackage.xtpdft.convergence.method	DIIS	Main method to use for convergence accelertation	DIIS mixing
dftpackage.xtpdft.convergence.DIIS_start	0.002	DIIS error at which DIIS takes over	float+
dftpackage.xtpdft.convergence.ADIIS_start	0.8	DIIS error at which ADIIS takes over	float+
dftpackage.xtpdft.convergence.DIIS_length	20	old hamiltonians to keep in history	int+
dftpackage.xtpdft.convergence.DIIS_maxout	false	if true remove the maximum element in the history if DIIS_length is exceeded.Otherwise, the oldest element is removed.	bool
dftpackage.xtpdft.convergence.levelshift	0.0	levelshift to apply to hamiltonian	float+
dftpackage.xtpdft.convergence.levelshift_end	0.2	DIIS error at levelshifting is disabled	float+
dftpackage.xtpdft.convergence.max_iterations	100	max iterations to use	int+
dftpackage.xtpdft.convergence.error	1e-7	convergence error	float+
dftpackage.xtpdft.convergence.mixing	0.7	mixing parameter for linear mixing of density matrices	float+
dftpackage.xtpdft.convergence.mixing_end	0.8	DIIS error below which mixing is disabled – independent of ADIIS_start (see ORCA’s own DampErr, kept independent of DIISStart for the same reason: difficult systems benefit from keeping mixing active well past the point where ADIIS/DIIS themselves start being tried, not exactly until that same point). Defaults to match ADIIS_start’s own default (0.8) for backward compatibility – existing configurations see identical behavior unless this is set explicitly.	float+
dftpackage.xtpdft.convergence.mixing_max	0.98	Ceiling mixing can adaptively ramp up toward as the SCF struggles (see consecutive ADIIS failures), rather than staying fixed at the base mixing value for an entire run. Matches ORCA’s own DampMax (default 0.98, confirmed directly from a real ORCA log’s own resolved SCF settings) – ORCA ramps its own damping factor from a base (DampFac, 0.7 by default) up to this ceiling only when actually needed, rather than paying the cost of heavy damping throughout.	float+
dftpackage.xtpdft.convergence.energy_reset	1.0	If the energy rises by more than this above the lowest energy reached so far, the (A)DIIS history is discarded and the SCF restarts, damped, from the lowest-energy density. 0 disables.	float+
dftpackage.xtpdft.convergence.second_order	auto	Second-order (trust-region augmented Hessian, TRAH) SCF steps. auto: switch to them when (A)DIIS fails repeatedly or the DIIS error stalls, then use them until convergence; always: use them from the first iteration; off: never. Exact analytic orbital Hessian for RKS and UKS (LDA/GGA and global hybrids; spin-polarized f_xc for UKS); for UKS with a functional without second derivatives in libxc, finite differences of the Fock matrix (two Fock builds per Hessian product).	auto always off
dftpackage.xtpdft.convergence.davidson_max_iter	50	Maximum number of micro-iterations (Hessian products) per second-order (TRAH) step.	int+
mpsfile	OPTIONAL	MPS file to read from with external multipoles for embedding
guess	OPTIONAL	File to read the wave function if the use guess flag is set to true
moguess	OPTIONAL	Orb file to read ONLY the MOs from (geometry still comes from job_name.xyz, or from guess if that is also set) – for warm-starting a geometry optimization step’s own SCF from a PREVIOUS step’s converged wavefunction at a different geometry, which guess itself cannot represent since it loads geometry and MOs together from the same file.
tasks	input,dft,parse,gwbse	task to compute	[guess input dft parse gwbse localize dft_in_dft]
gwbse.tasks	all	tasks to do	[gw singlets triplets excitons exciton_uks all]
gwbse.ranges	default	default: all levels in RPA, 1:2*HOMO in QP and all in BSE; other options: factor,explicit	default factor explicit full
gwbse.rpamax		only needed, if ranges is factor or explicit, number of levels in rpa
gwbse.qpmin		only needed, if ranges is factor or explicit, lowest MO to be used in GW
gwbse.qpmax		only needed, if ranges is factor or explicit, highest MO to be used in GW
gwbse.bsemin		only needed, if ranges is factor or explicit, lowest MO to be used in BSE
gwbse.bsemax		only needed, if ranges is factor or explicit, highest MO to be used in BSE
gwbse.qp_emax	0	cap on the GW window: highest QP level at most this many eV above the DFT LUMO (a degenerate set is kept whole); applied on top of ranges, 0 = off	float+
gwbse.qp_emin	0	cap on the GW window: lowest QP level at most this many eV below the DFT HOMO; 0 = off	float+
gwbse.bse_emax	0	cap on the BSE window: highest virtual level at most this many eV above the DFT LUMO; 0 = off	float+
gwbse.bse_emin	0	cap on the BSE window: lowest occupied level at most this many eV below the DFT HOMO; 0 = off	float+
gwbse.ignore_corelevels	none	exclude core MO level from calculation on RPA,GW or BSE level	RPA GW BSE none
gwbse.auxbasisset	OPTIONAL	Auxiliary basis set for RI, only used if DFT has no auxiliary set
gwbse.gw.mode	evGW	use single short (G0W0) or self-consistent GW (evGW)	evGW G0W0
gwbse.gw.scissor_shift	0.0	preshift unoccupied MOs by a constant for GW calculation	float
gwbse.gw.sigma_integrator	ppm	self-energy correlation integration method	ppm exact cda
gwbse.gw.eta	1e-3	small parameter eta of the Green’s function	float+
gwbse.gw.alpha	1e-3	parameter to smooth residue and integral calculation for the contour deformation technique	float
gwbse.gw.quadrature_scheme	legendre	If CDA is used for sigma integration this set the quadrature scheme to use	hermite laguerre legendre
gwbse.gw.quadrature_order	12	Quadrature order if CDA is used for sigma integration	8 10 12 14 16 18 20 40 100
gwbse.gw.qp_solver	grid	QP equation solve method	fixedpoint grid
gwbse.gw.qp_grid_search_mode	adaptive_with_dense_fallback	grid-search backend for QP root detection	adaptive dense adaptive_with_dense_fallback
gwbse.gw.qp_root_finder	brent	method to refine a sign-change interval of the QP equation to a root: brent (inverse quadratic interpolation with bisection safeguard, fewer Sigma evaluations) or bisection; both stop when the QP residual is below qp_sc_limit	bisection brent
gwbse.gw.qp_full_window_half_width	OPTIONAL	half-width of the full QP search window around the reference energy	float+
gwbse.gw.qp_dense_spacing	OPTIONAL	spacing of the dense QP scan used for robust sign-change detection	float+
gwbse.gw.qp_adaptive_shell_width	OPTIONAL	radial shell width of the adaptive QP scan	float+
gwbse.gw.qp_adaptive_shell_count	OPTIONAL	number of adaptive shell layers per side; if > 0 it overrides qp_adaptive_shell_width	int+
gwbse.gw.qp_grid_steps	OPTIONAL	DEPRECATED: legacy number of QP grid points; maps to qp_full_window_half_width and qp_adaptive_shell_width together with qp_grid_spacing	int+
gwbse.gw.qp_grid_spacing	OPTIONAL	DEPRECATED: legacy QP grid spacing; maps to qp_dense_spacing and contributes to qp_full_window_half_width together with qp_grid_steps	float+
gwbse.gw.qp_restrict_search	true	restrict occupied and virtual QP scans away from the zero-energy crossing region before retrying the full window	bool
gwbse.gw.qp_root_continuity	true	evGW: from the second iteration on keep each QP level on the root closest to its previous solution instead of choosing the root with the largest weight again (stabilises levels with two roots of comparable weight)	bool
gwbse.gw.core_exchange	true	with ignore_corelevels=RPA (frozen core): include the exchange with the core levels in Sigma_x, which Vxc also contains through the core density. Without it the occupied QP energies are shifted up by the missing core exchange (0.6 eV on average for a QM9 molecule)	bool
gwbse.gw.qp_root_tracking	false	evGW: from the second iteration on search each QP level only near its root of the previous iteration (window of a few times its last change, 0.01 to 0.1 Ha) and keep the closest accepted root unless its weight drops below half; otherwise, and for levels without such a root, the full search is used. After convergence one QP solve with the full search checks the roots, so the result is a fixed point of the full search. If a level has several such fixed points (strong satellites), tracking may end on another one than iterating with the full search does	bool
gwbse.gw.screening_update	adaptive	evGW: when the screening W is rebuilt. every: in every iteration. adaptive: the energies in G are iterated with W fixed and W is rebuilt once their last change is below screening_update_ratio times their change since W was built (or after screening_update_max_inner iterations); converged when the energies change by less than gw_sc_limit and differ from those of the last W by less than gw_sc_limit, the same evGW solution	every adaptive
gwbse.gw.screening_update_ratio	0.1	adaptive screening_update: rebuild W once the last change of the energies is below this fraction of their change since W was built	float+
gwbse.gw.screening_update_max_inner	10	adaptive screening_update: rebuild W at the latest after this many iterations with the same W	int+
gwbse.gw.qp_zero_margin	1e-6	margin below zero used as upper bound for occupied restricted scans	float+
gwbse.gw.qp_virtual_min_energy	-0.1	lower bound for virtual restricted scans	float
gwbse.gw.qp_sc_max_iter	100	maximum number of iterations for quasiparticle equation solution	int+
gwbse.gw.qp_sc_limit	1e-6	quasiparticle equation solver convergence	float+
gwbse.gw.sc_max_iter	50	Maximum number of iterations in eVGW	int+
gwbse.gw.mixing_order	3	Mixing of QP energies in evGW - 0: plain, 1: linear, >1: Anderson	int+
gwbse.gw.sc_limit	1e-5	evGW convergence criteria	float+
gwbse.gw.mixing_alpha	0.7	mixing alpha, also linear mixing	float+
gwbse.gw.do_qsgw	false	Run Quasiparticle self-consistent GW after evGW or G0W0	bool
gwbse.gw.qsgw_max_iterations	50	maximum number of iterations is QSGW	int+
gwbse.gw.qsgw_sc_limit	1e-5	QSGW convergence threshold on QP energies	float+
gwbse.gw.qsgw_max_virt_correction	0.5	QSGW: pertGW energy threshold for inclusion of virtual states	float+
gwbse.gw.qsgw_hartree_update	true	QSGW: update the Hartree potential with the QP density of the active occupied levels (false: keep J of the DFT density, the behaviour before)	bool
gwbse.gw.qsgw_regularization	eta	QSGW static self-energy: eta = symmetrised 1/2 Re[Sigma(e_p) + Sigma(e_q)] with the broadening eta; srg = similarity-renormalization-group form (Marie and Loos), Hermitian by construction, terms damped by 1 - exp(-s (D_p^2 + D_q^2)) with D the distance from their pole (exact and ppm Sigma)	eta srg
gwbse.gw.qsgw_srg_s	500	QSGW with qsgw_regularization=srg: flow parameter s; s = 1/(2 eta^2) matches the eta form to first order (s = 200 for eta = 0.05 Ha); Marie and Loos recommend 500 to 1000	float+
gwbse.gw.qsgw_hartree_screened	true	QSGW with environment screening: the Hartree update uses u = v + v_reac, so the dipoles induced in the environment follow the QP density (false: bare v, the environment keeps the response to the DFT density). No effect without environment screening or with qsgw_hartree_update=false	bool
gwbse.gw.out_of_window_shift	max	Shift of the RPA energies of levels outside the GW window (evGW, QSGW): max - largest correction of the occupied/virtual window levels; nearest - correction of the window boundary level; linear - linear fit of the corrections near the boundary, extrapolated at most over the fitted energy span	max nearest linear
gwbse.gw.out_of_window_fit_fraction	0.3	linear out_of_window_shift: fraction of the occupied/virtual window levels next to the boundary used in the fit (at least 2 levels)	float+
gwbse.gw.rebuild_3c_freq	0	rebuild the 3c integrals from scratch every this many evGW iterations; 0 (default) never. The integrals are only rotated by orthogonal matrices between iterations, so a rebuild resets rounding at the 1e-12 Ha level	int+
gwbse.gw.sigma_plot.states		plot sigma(omega) for the following states, e.g 1 3 5
gwbse.gw.sigma_plot.steps	201	points to plot, centred on the DFT energy of each state	int+
gwbse.gw.sigma_plot.spacing	1e-2	spacing of points units	float+
gwbse.gw.sigma_plot.filename	QPenergies_sigma.dat	File with the sigma plot data
gwbse.gw.sigma_dump.filename	sigma_dump.h5	HDF5 output file
gwbse.gw.sigma_dump.levels_emax	20	pole spectra and imaginary-axis Sigma_c only for levels up to this energy above the DFT LUMO; 0 = all QP levels	float+
gwbse.gw.sigma_dump.imag_points	24	number of Gauss-Legendre points on the imaginary axis	int+
gwbse.gw.sigma_dump.imag_scale	0.5	scale s of the imaginary grid nu = s(1+x)/(1-x)	float+
gwbse.gw.sigma_dump.spectral_min	-3.0	lower edge of the pole histogram (absolute energy)	float
gwbse.gw.sigma_dump.spectral_max	3.0	upper edge of the pole histogram (absolute energy)	float
gwbse.gw.sigma_dump.spectral_binwidth	0.02	bin width of the pole histogram	float+
gwbse.gw.sigma_dump.plot_steps	201	Sigma_c(omega) of all QP levels on this many points around their DFT energies (as sigma_plot, stored as float32); 0 = off	int+
gwbse.gw.sigma_dump.plot_spacing	0.01	spacing of the Sigma_c(omega) grid	float+
gwbse.bse.exctotal	10	maximum number of BSE states to calculate	int+
gwbse.bse.useTDA	false	use TDA for BSE	bool
gwbse.bse.dyn_screen_max_iter	0	maximum number of iterations for perturbative dynamical screening in BSE	int+
gwbse.bse.dyn_screen_tol	1e-5	convergence tolerance for perturbative dynamical screening in BSE	float+
gwbse.bse.davidson.correction	DPR	Davidson correction method	DPR OLSEN
gwbse.bse.davidson.tolerance	normal	Numerical tolerance	loose normal strict lapack
gwbse.bse.davidson.update	safe	how large the search space	min safe max
gwbse.bse.davidson.maxiter	50	max iterations	int+
gwbse.bse.use_Hqp_offdiag	false	Using symmetrized off-diagonal elements of QP Hamiltonian in BSE	bool
gwbse.bse.print_weight	0.5	print exciton WF composition weight larger than minimum	float+
gwbse.bse.fragments.fragment.indices	REQUIRED	indices of atoms in this fragment, e.g. 1 3 13:17
localize.max_iterations	10000	Maximum number of iterations for PM Localization	int+
localize.convergence_limit	1e-5	Convergence criteria for PM localization
localize.method	unitary-optimizer	Method for the localization optimization	[jacobi-sweeps unitary-optimizer]
logging_file	OPTIONAL	File to send logging data to.
archiveA	OPTIONAL	orbfile for moleculeA of guess
archiveB	OPTIONAL	orbfile for moleculeA of guess

Note

An xml file containing the defaults for the dftgwbse calculator can be created via -p dftgwbse -o FILENAME command line options `