LFOCKACE: Difference between revisions
No edit summary |
No edit summary |
||
| Line 4: | Line 4: | ||
Description: {{TAG|LFOCKACE}} determines whether the Adaptively Compressed Exchange Operator is used.{{cite|linlin:jctc:2016}} | Description: {{TAG|LFOCKACE}} determines whether the Adaptively Compressed Exchange Operator is used.{{cite|linlin:jctc:2016}} | ||
*N.B.:Available for CPU and OpenACC version of VASP.6 when compiled with [[Precompiler flags|- | *N.B.:Available for CPU and OpenACC version of VASP.6 when compiled with [[Precompiler flags|-Dfock_dblbuf]]. | ||
---- | ---- | ||
For {{TAGBL|LFOCKACE}}=.TRUE. the Cholesky decomposition <math>X=LL^\dagger</math> of the Fock exchange matrix <math>X_{ij} = \langle \tilde\psi_i \mid \tilde V_X \mid \tilde\psi_j \rangle </math> is calculated and the adaptively compressed exchange operator <math>\tilde V_{ACE} = -\sum_i \mid \tilde X_i \rangle \langle \tilde X_i \mid </math> is used for the action of the Fock exchange on the pseudo orbitals. This method can be used for [[Hartree-Fock_and_HF/DFT_hybrid_functionals|hybrid functionals]] in combination with the Davidson algorithm ({{TAGBL|ALGO}}=Normal) to save a factor of <math>\approx 3</math> in computation time. | For {{TAGBL|LFOCKACE}}=.TRUE. the Cholesky decomposition <math>X=LL^\dagger</math> of the Fock exchange matrix <math>X_{ij} = \langle \tilde\psi_i \mid \tilde V_X \mid \tilde\psi_j \rangle </math> is calculated and the adaptively compressed exchange operator <math>\tilde V_{ACE} = -\sum_i \mid \tilde X_i \rangle \langle \tilde X_i \mid </math> is used for the action of the Fock exchange on the pseudo orbitals. This method can be used for [[Hartree-Fock_and_HF/DFT_hybrid_functionals|hybrid functionals]] in combination with the Davidson algorithm ({{TAGBL|ALGO}}=Normal) to save a factor of <math>\approx 3</math> in computation time. | ||
Revision as of 14:04, 16 December 2019
LFOCKACE = .TRUE. | .FALSE.
Default: LFOCKACE = .TRUE.
| Default: LFOCKACE | = .TRUE. | for VASP.6 |
| = N/A | for VASP.5.X and older |
Description: LFOCKACE determines whether the Adaptively Compressed Exchange Operator is used.[1]
- N.B.:Available for CPU and OpenACC version of VASP.6 when compiled with -Dfock_dblbuf.
For LFOCKACE=.TRUE. the Cholesky decomposition [math]\displaystyle{ X=LL^\dagger }[/math] of the Fock exchange matrix [math]\displaystyle{ X_{ij} = \langle \tilde\psi_i \mid \tilde V_X \mid \tilde\psi_j \rangle }[/math] is calculated and the adaptively compressed exchange operator [math]\displaystyle{ \tilde V_{ACE} = -\sum_i \mid \tilde X_i \rangle \langle \tilde X_i \mid }[/math] is used for the action of the Fock exchange on the pseudo orbitals. This method can be used for hybrid functionals in combination with the Davidson algorithm (ALGO=Normal) to save a factor of [math]\displaystyle{ \approx 3 }[/math] in computation time.
For LFOCKACE=.TRUE. the conventional orbital representation is used.
Related Tags and Sections
AEXX, AEXX, AGGAX, AGGAC, LHFCALC, hybrid functionals, settings for specific hybrid functionals