GW
Parameters influencing GW calculations on molecules, see also ‘Electronic band structure from GW’, https://manual.cp2k.org/trunk/methods/properties/bandstructure_gw.html. [Edit on GitHub]
Subsections
Keywords
Keyword descriptions
- SECTION_PARAMETERS
Type: logical
Default: F
Lone keyword: T
Usage: &GW .TRUE.Description: Activates GW calculations.
- ANALYTIC_CONTINUATION
Type: enum
Default: PADE
Usage: ANALYTIC_CONTINUATIONDescription: Defines which type of analytic continuation for the self energy is used
Valid values:
TWO_POLEUse ‘two-pole’ model.PADEUse Pade approximation.
- CORR_MOS_OCC
Type: integer
Default: 10
Aliases: CORR_OCC
Usage: CORR_OCC 3Description: Number of occupied MOs whose energies are corrected in GW. Counting beginning from HOMO, e.g. 3 corrected occ. MOs correspond to correction of HOMO, HOMO-1 and HOMO-2. Numerical effort and storage of RI-G0W0 increase linearly with this number. In case you want to correct all occ. MOs, insert either a negative number or a number larger than the number of occ. MOs. Invoking CORR_MOS_OCC -2 together with a BSE cutoff, sets a sufficiently large CORR_MOS_OCC for the given BSE cutoff deduced from DFT eigenvalues.
- CORR_MOS_VIRT
Type: integer
Default: 10
Aliases: CORR_VIRT
Usage: CORR_VIRT 3Description: Number of virtual MOs whose energies are corrected by GW. Counting beginning from LUMO, e.g. 3 corrected occ. MOs correspond to correction of LUMO, LUMO+1 and LUMO+2. Numerical effort and storage of RI-G0W0 increase linearly with this number. In case you want to correct all virt. MOs, insert either a negative number or a number larger than the number of virt. MOs. Invoking CORR_MOS_VIRT -2 together with a BSE cutoff, sets a sufficiently large CORR_MOS_VIRT for the given BSE cutoff deduced from DFT eigenvalues.
- CROSSING_SEARCH
Type: enum
Default: NEWTON
Usage: CROSSING_SEARCH Z_SHOTDescription: Determines, how the self_energy is evaluated on the real axis.
Valid values:
Z_SHOTCalculate the derivative of Sigma and out of it Z. Then extrapolate using Z.NEWTONMake a Newton-Raphson fix point iteration.BISECTIONMake a bisection fix point iteration.
- EPS_ITER
Type: real
Default: 1.36000000E-003 [eV]
Usage: EPS_ITER 0.00005Description: Target accuracy for the eigenvalue self-consistency. If the G0W0 HOMO-LUMO gap differs by less than the target accuracy during the iteration, the eigenvalue self-consistency cycle stops. Unit: Hartree.
- EV_GW_ITER
Type: integer
Default: 1
Usage: EV_GW_ITER 3Description: Maximum number of iterations for eigenvalue self-consistency cycle. The computational effort of GW scales linearly with this number. In case of choosing GW_SELF_CONSISTENCY EVGW, the code sets EV_GW_ITER 10.
- FERMI_LEVEL_OFFSET
Type: real
Default: 2.00000000E-002
Usage: FERMI_LEVEL_OFFSET 1.0E-2Description: Fermi level for occ. orbitals: e_HOMO + FERMI_LEVEL_OFFSET; Fermi level for virt. orbitals: e_LUMO - FERMI_LEVEL_OFFSET. In case e_homo + FERMI_LEVEL_OFFSET < e_lumo - FERMI_LEVEL_OFFSET, we set Fermi level = (e_HOMO+e_LUMO)/2. For cubic-scaling GW, the Fermi level is always equal to (e_HOMO+e_LUMO)/2 regardless of FERMI_LEVEL_OFFSET.
- GAMMA_ONLY_SIGMA
Type: logical
Default: F
Aliases: GAMMA
Lone keyword: T
Usage: GAMMA TRUEDescription: If true, the correlation self-energy is only computed at the Gamma point. The Gamma point itself is obtained by averaging over all kpoints of the DFT mesh.
- HEDIN_SHIFT
Type: logical
Default: F
Lone keyword: T
Usage: HEDIN_SHIFTDescription: If true, use Hedin’s shift in G0W0, evGW and evGW0 (aka scGW0). Details see in Li et al. JCTC 18, 7570 (2022), Figure 1. G0W0 with Hedin’s shift should give similar GW eigenvalues as evGW0; at a lower computational cost.
- IC_CORR_LIST
Type: real[ ]
Default: -2.72113839E+001 [eV]
Usage: IC_CORR_LIST… Description: List of image charge correction from a previous calculation to be applied in G0W0 or evGW. Keyword is active, if the first entry is positive (since IC corrections are positive occupied MOs. The start corresponds to the first corrected GW level.
- IC_CORR_LIST_BETA
Type: real[ ]
Default: -2.72113839E+001 [eV]
Usage: IC_CORR_LIST_BETA… Description: IC_CORR_LIST for beta spins in case of open shell calculation.
- IMAGE_CHARGE_MODEL
Type: logical
Default: F
Aliases: IC
Lone keyword: T
Usage: IC TRUEDescription: If true, an image charge model is applied to mimic the renormalization of electronic levels of a molecule at a metallic surface. For this calculation, the molecule has to be reflected on the desired xy image plane. The coordinates of the reflected molecule have to be added to the coord file as ghost atoms. For the ghost atoms, identical basis sets the normal atoms have to be used.
- KPOINTS_SELF_ENERGY
Type: integer[3]
Default: 0 0 0
Repeatable: yes
Usage: KPOINTS_SELF_ENERGY nx ny nzDescription: Specify number of k-points for the k-point grid of the self-energy. Internally, a Monkhorst-Pack grid is used. A dense k-point grid may be necessary to compute an accurate density of state from GW. Large self-energy k-meshes do not cost much more computation time.
- NPARAM_PADE
Type: integer
Default: 16
Usage: NPARAM_PADE 16Description: Number of parameters for the Pade approximation when using the latter for the analytic continuation of the self energy. 16 parameters (corresponding to 8 poles) are are recommended.
- NUMB_POLES
Type: integer
Default: 2
Usage: NUMB_POLES 2Description: Number of poles for the fitting. Usually, two poles are sufficient.
- OMEGA_MAX_FIT
Type: real
Default: 7.34996000E-001
Usage: OMEGA_MAX_FIT 0.5Description: Determines fitting range for the self-energy on the imaginary axis: [0, OMEGA_MAX_FIT] for virt orbitals, [-OMEGA_MAX_FIT,0] for occ orbitals. Unit: Hartree. Default: 0.734996 H = 20 eV.
- PERIODIC_CORRECTION
Type: logical
Default: F
Lone keyword: T
Usage: PERIODIC_CORRECTIONDescription: If true, the periodic correction scheme is used employing k-points. Method is not recommended to use, use instead PERIODIC_LOW_SCALING which much more accurate than the periodic correction.
- PRINT_EXX
Type: enum
Default: FALSE
Usage: PRINT_EXX TRUEDescription: Print exchange self-energy minus exchange correlation potential for Gamma-only calculation (PRINT). For a GW calculation with k-points we use this output as exchange self-energy (READ). This is a temporary solution because the hybrid MPI/OMP parallelization in the HFX by Manuel Guidon conflicts with the parallelization in low-scaling GW k-points which is most efficient with maximum number of MPI tasks and minimum number of OMP threads. For HFX by M. Guidon, the density matrix is fully replicated on every MPI rank which necessitates a high number of OMP threads per MPI rank for large systems to prevent out of memory. Such a high number of OMP threads would slow down the GW calculation severely. Therefore, it was decided to temporarily divide the GW k-point calculation in a Gamma-only HF calculation with high number of OMP threads to prevent out of memory and a GW k-point calculation with 1 OMP thread per MPI rank reading the previousHF output.
Valid values:
TRUEPlease, put TRUE for Gamma only calculation to get the exchange self-energy. If ‘SIGMA_X’ and the corresponding values for the exchange-energy are written, the writing has been successfulFALSEFALSE is needed if you want to do nothing here.READPlease, put READ for the k-point GW calculation to read the exact exchange. You have to provide an output file including the exact exchange. This file has to be named ‘exx.dat’.SKIP_FOR_REGTESTSKIP_FOR_REGTEST is only used for the GW k-point regtest where no exchange self-energy is computed.
- PRINT_SELF_ENERGY
Type: logical
Default: F
Lone keyword: T
Usage: PRINT_SELF_ENERGYDescription: If true, print the self-energy for all levels for real energy together with the straight line to see the quasiparticle energy as intersection. In addition, prints the self-energy for imaginary frequencies together with the Pade fit.
- REGULARIZATION_MINIMAX
Type: real
Default: 0.00000000E+000
Usage: REGULARIZATION_MINIMAX 1.0E-6Description: Tikhonov regularization for computing weights of the Fourier transform from imaginary time to imaginary frequency and vice versa. Needed for large minimax grids with 20 or more points and a small range.
- RI_SIGMA_X
Type: logical
Default: T
Lone keyword: T
Usage: RI_SIGMA_XDescription: If true, the exchange self-energy is calculated approximatively with RI. If false, the Hartree-Fock implementation in CP2K is used.
- SC_GW0_ITER
Type: integer
Default: 1
Usage: SC_GW0_ITER 3Description: Maximum number of iterations for GW0 self-consistency cycle. The computational effort of GW is not much affected by the number of scGW0 cycles. In case of choosing GW_SELF_CONSISTENCY EVGW0, the code sets SC_GW0_ITER 10.
- SELF_CONSISTENCY
Type: enum
Default: G0W0
Usage: SELF_CONSISTENCY evGW0Description: Decide the level of self-consistency of eigenvalues (= quasiparticle energies = single-electron energies) in GW. Updates of Kohn-Sham orbitals (for example qsGW) are not implemented. For details which type of eigenvalue self-consistency might be good, please consult Golze, Dvorak, Rinke, Front. Chem. 2019.
Valid values:
G0W0Use DFT eigenvalues; not update.EVGW0Update DFT eigenvalues in G, not in W.EVGWUpdate DFT eigenvalues in G and W.
- SOC
Type: enum
Default: NONE
Usage: SOCDescription: Calculate the spin-orbit splitting of the eigenvalues/band structure using the spin-orbit part of the GTH pseudos parametrized in Hartwigsen, Goedecker, Hutter, Phys. Rev. B 58, 3641 (1998), Eq. 19, parameters in Table I.
Valid values:
NONENo SOC.LDAUse parameters from LDA (PADE) pseudopotential.PBEUse parameters from PBE pseudopotential.
- SOC_ENERGY_WINDOW
Type: real
Default: 5.00000000E+001 [eV]
Usage: SOC_ENERGY_WINDOW 20.0_eVDescription: For perturbative SOC calculation, only take frontier levels in an energy window [E_HOMO - SOC_ENERGY_WINDOW/2 , E_LUMO + SOC_ENERGY_WINDOW/2 into account for the diagonalization of H^GW,SOC.
- UPDATE_XC_ENERGY
Type: logical
Default: F
Lone keyword: T
Usage: UPDATE_XC_ENERGYDescription: If true, the Hartree-Fock and RPA total energy are printed and the total energy is corrected using exact exchange and the RPA correlation energy.