XTB
References: Grimme2017
Parameters needed to set up the xTB methods [Edit on GitHub]
Subsections
Keywords
Keyword descriptions
- CHECK_ATOMIC_CHARGES
Type: logical
Default: T
Lone keyword: T
Usage: CHECK_ATOMIC_CHARGES TDescription: Stop calculation if atomic charges are outside chemical range.
Mentions: ⭐Extended Tight Binding
- COULOMB_INTERACTION
Type: logical
Default: T
Lone keyword: T
Usage: COULOMB_INTERACTION TDescription: Use Coulomb interaction terms (electrostatics + TB3); for debug only
Mentions: ⭐Extended Tight Binding
- COULOMB_LR
Type: logical
Default: T
Lone keyword: T
Usage: COULOMB_LR TDescription: Use Coulomb LR (1/r) interaction terms; for debug only
Mentions: ⭐Extended Tight Binding
- DO_EWALD
Type: logical
Default: F
Lone keyword: T
Usage: DO_EWALDDescription: Deprecated debug override for the xTB Ewald treatment. By default the setting is inferred from SUBSYS/CELL/PERIODIC. If set to TRUE for CP2K-internal xTB, the DFT/POISSON section must use a periodic Poisson solver. For CP2K/tblite, the periodic long-range treatment is performed inside tblite.
Mentions: ⭐Extended Tight Binding
Warning
The keyword DO_EWALD is deprecated and may be removed in a future version.
Use SUBSYS/CELL/PERIODIC to select periodicity. DO_EWALD is retained only as a debug override for the CP2K-internal xTB Ewald path.
- DO_NONBONDED
Type: logical
Default: F
Lone keyword: T
Usage: DO_NONBONDED TDescription: Controls the computation of real-space (short-range) nonbonded interactions as correction to xTB.
Mentions: ⭐Extended Tight Binding
- EN_SHIFT_TYPE
Type: string
Default: Molecule
Usage: EN_SHIFT_TYPE [Select/Molecule/Crystal]Description: Shift function for electronegativity in EEQ method. [Select/Molecule/Crystal] Default Select from periodicity.
- EPS_PAIRPOTENTIAL
Type: real
Default: 1.00000000E-010
Usage: EPS_PAIRPOTENTIAL 1.0E-8Description: Accuracy for the repulsive pair potential.
- GFN_TYPE
Type: enum
Default: 1
Usage: GFN_TYPE (0|1|TBLITE)Description: Selects the xTB backend. GFN_TYPE 1 uses the CP2K-internal GFN1-xTB implementation. GFN_TYPE 0 uses the CP2K-internal GFN0-xTB implementation, which has no SCC variables to mix, so XTB/SCC_MIXER AUTO is treated as NONE. GFN_TYPE TBLITE uses the tblite backend and requires an XTB/TBLITE section, where the actual tblite method is selected.
Valid values:
0Use the CP2K-internal GFN0-xTB implementation.1Use the CP2K-internal GFN1-xTB implementation.TBLITEUse the CP2K/tblite backend; requires XTB/TBLITE.
Mentions: ⭐Extended Tight Binding
- HYDROGEN_STO_NG
Type: integer
Default: 4
Usage: HYDROGEN_STO_NG 3Description: Number of GTOs for Hydrogen basis expansion.
- SCC_MIXER
Type: enum
Default: AUTO
Usage: SCC_MIXER (AUTO|TBLITE|CP2K|NONE)Description: Selects the SCC variable mixer for xTB calculations. AUTO is treated as NONE for CP2K-internal GFN0-xTB, keeps the CP2K charge mixer for CP2K-internal GFN1-xTB, and uses tblite’s native SCC mixer for XTB/GFN_TYPE TBLITE. The SCC variables are the model populations and, where required by the selected method, atomic multipoles that feed the self-consistent xTB potential. TBLITE uses tblite’s modified Broyden mixer for those variables inside the CP2K SCF cycle. CP2K maps the variables onto CP2K’s charge-mixing infrastructure. When TBLITE is active, the native SCC convergence follows XTB/TBLITE/ACCURACY and the tblite-side SCC update limit follows TBLITE_MIXER/ITERATIONS. If SCC_MIXER is explicitly set to TBLITE, CP2K also sets DFT/SCF/MAX_SCF to TBLITE_MIXER/ITERATIONS. With SCC_MIXER AUTO or CP2K, MAX_SCF keeps its normal CP2K meaning. With QS/LS_SCF, explicit SCC_MIXER settings are reset to NONE internally and CP2K emits a warning; LS_SCF controls the density-matrix optimization. When CP2K is active, convergence follows the normal CP2K SCF controls, in particular DFT/SCF/EPS_SCF and DFT/SCF/MIXING. For GFN2/tblite, the CP2K mixer acts on the full SCC-variable set: shell charges plus atomic dipole and quadrupole variables. CP2K-internal GFN0-xTB has no SCC variables to mix, so AUTO and NONE are equivalent there and explicit CP2K or TBLITE mixer choices are reset to NONE. For CP2K-internal GFN1 and DFTB/DFTB3, the SCC variables are only charge or shell-charge like. The initial SCC variables are seeded from CP2K’s current density, so DFT/SCF/SCF_GUESS and DFT/WFN_RESTART_FILE_NAME define the starting point. NONE performs direct updates without mixing and is intended for debugging, since it can be unstable.
Valid values:
AUTODefault method-dependent choice.TBLITEUse the tblite modified Broyden SCC mixer.CP2KUse CP2K charge mixing for SCC variables.NONEDo not mix SCC variables; for debugging.
- SPIN_POLARISATION
Type: logical
Default: F
Lone keyword: T
Usage: SPIN_POLARISATION TDescription: Use the spin polarisation Hamiltonian for gfn1/2
- STO_NG
Type: integer
Default: 6
Usage: STO_NG 3Description: Provides the order of the Slater orbital expansion in GTOs.
- TB3_INTERACTION
Type: logical
Default: T
Lone keyword: T
Usage: TB3_INTERACTION TDescription: Use TB3 interaction terms; for debug only
Mentions: ⭐Extended Tight Binding
- USE_HALOGEN_CORRECTION
Type: logical
Default: T
Lone keyword: T
Usage: USE_HALOGEN_CORRECTION TDescription: Use XB interaction term
Mentions: ⭐Extended Tight Binding
- VARIATIONAL_DIPOLE
Type: logical
Default: F
Lone keyword: T
Usage: VARIATIONAL_DIPOLE TDescription: gfn0-xTB use dipole definition from energy derivative.
- VDW_POTENTIAL
Type: string
Usage: VDW_POTENTIAL typeDescription: vdW potential to be used: NONE, DFTD3, DFTD4. Defaults: DFTD3(gfn1), DFTD4(gfn0, gfn2).