Jump to content

Requests for technical support from the VASP team should be posted in the VASP Forum.

LMAXFOCKAE: Difference between revisions

From VASP Wiki
Csheldon (talk | contribs)
No edit summary
 
(2 intermediate revisions by 2 users not shown)
Line 1: Line 1:
#REDIRECT [[NMAXFOCKAE and LMAXFOCKAE]]
{{TAGDEF|LMAXFOCKAE|[integer]}}
{{TAGDEF|LMAXFOCKAE|[integer]}}
{{DEF|LMAXFOCKAE|-1|for Hartree-Fock and hybrid functionals |  4 | for post DFT methods}}


Description: {{TAG|NMAXFOCKAE}} and {{TAG|LMAXFOCKAE}} determine whether
{{DEF|LMAXFOCKAE|-1|DFT, Hartree-Fock|4|post-DFT methods}}
the overlap densities in the Fock exchange and correlated wave function methods are accurately reconstructed on the plane wave grid. This flag generally only applies to the Fock-exchange part as well as many-body
post DFT methods (GW, RPA, MP2, etc.).


Description: Specifies the maximum spherical angular momentum quantum number (<math>L</math>) up to which augmentation charges are accurately restored on the plane-wave grid. This parameter controls the quality of shape restoration for the [[Projector-augmented-wave_formalism|PAW method]].
----
----
In the PAW method, the difference between the charge density of the all-electron partial waves <math>\phi_\beta</math> and
In the [[Projector-augmented-wave_formalism|PAW method]], the difference between the charge density of the all-electron partial waves <math>\phi_n</math> and the pseudo partial waves <math>\tilde \phi_n</math>
the pseudo partial waves <math>\tilde \phi_\beta</math>
 
<math>
<math>
Q_{\alpha\beta}(r)= \phi^*_\alpha(r)\phi_\beta(r)  - \tilde \phi^*_\alpha(r)\tilde \phi_\beta(r)
Q_{nm}({\mathbf r})= \phi^*_n({\mathbf r})\phi_m({\mathbf r})  - \tilde \phi^*_n({\mathbf r})\tilde \phi_m({\mathbf r})
</math>
</math>
is usually restored on spherical grids centered at each atom
(one-center terms inside the PAW spheres).  To describe long range electrostatic effect, the ''moments'' of the differences of the all-electron and pseudo charge density
also need to added to the pseudo density on the plane wave grid. This is done up to a certain ''l'' quantum number.
These augmentation charges exactly restore the moments of the all-electron density on the plane wave
grid. For the charge densities used in the Hartree and DFT term,
this augmentation is done exactly up to the maximum ''l'' quantum number required by the POTCAR files,
whereas for the Fock exchange,  the augmentation on the plane wave grid is controlled by {{TAG|LMAXFOCK}}.


For the RPA, GW, and most post DFT methods, the one-center terms are, however, presently
is usually treated on spherical grids centered at each atom
not implemented. Depending on the material, this can cause sizable errors
(one-center terms inside the PAW spheres, see [[Projector-augmented-wave_formalism|PAW method]] method and more detailed description for {{TAG|LMAXFOCK}}). To describe long-range electrostatic effects, the ''moments'' of the differences of the all-electron and pseudo charge density also need to be added on the plane wave grid (compensation density, see [[Projector-augmented-wave_formalism|PAW method]]). These compensation charges exactly restore the moments of the all-electron density on the plane wave grid. For the Fock exchange the maximum ''L'' quantum number up to which the augmentation is done is controlled by {{TAG|LMAXFOCK}}.
in particular for 3d and (to a lesser extent) 2p, 4d and 5d elements.
To correct for this error, an alternative treatment is implemented
on the plane wave grid. This  allows to restore the all-electron densities accurately on the plane wave grid
instead of the one-center grids by specifying the flags {{TAG|LMAXFOCKAE}} and {{TAG|NMAXFOCKAE}}.


To achieve this improved treatment on the plane wave grid, <math> Q_{\alpha\beta}(r) </math> is Fourier transformed  to reciprocal space <math> Q_{\alpha\beta}(q) </math> and then expanded
For [[GW approximation of Hedin's equations|GW]], [[ACFDT/RPA calculations|RPA]], and [[:Category:Many-body perturbation theory|post-DFT]] methods, the one-center terms are, however, presently not implemented, which can cause sizable errors, particularly for 3''d'' and (to a lesser extent) 2''p'', 4''d'', and 5''d'' elements. To correct for this error, an alternative treatment is implemented on the plane-wave grid for the exchange and correlation contributions, allowing for accurate restoration of the all-electron densities.  
in a set of orthogonal functions localized at each atomic site. These augmentation charges
are then added to the pseudo charge densities on the plane wave grid.


This improved treatment — termed ''shape restoration'' — is available for exchange as well as many-body correlation contributions. <!--For exchange, the exact one-center terms are also implemented (cf. {{TAG|LFOCKSTD}}); this means shape restoration is not required and should not change the results.--> More details on shape restoration are explained in Ref. {{cite|shishkin:prb:2006}} and {{cite|unzog:prb:2022}}.


For {{TAG|LMAXFOCKAE}}=-1 (the default for DFT and Hartree-Fock calculations), only the moments of the all-electron charge densities are restored on the plane wave grid. This setting is exact for Hartree-Fock
Shape restoration allows restoration of the all-electron densities accurately already on the plane-wave grid by specifying the tags {{TAG|LMAXFOCKAE}} and {{TAG|NMAXFOCKAE}}. This accurate augmentation charge <math>\hat n</math> is calculated for the total orbital and angular momentum quantum numbers <math>L</math> (up to {{TAG|LMAXFOCKAE}}) and <math>M</math> using shape-restoring radial functions <math>\Delta g_L(r)</math> with coefficients <math>c^L_{nm}</math> to be determined:
since the one-center terms are implemented.


If {{TAG|LMAXFOCKAE}} is set to values larger than -1 (and {{TAG|NMAXFOCKAE}}=1), not only the moments of the all-electron charge density are restored, but also the all-electron charge density is restored up to a typical plane wave energy of 140 eV. This setting yields very accurate results for post DFT methods (MP2, RPA, GW, etc.) for most sp bonded materials.  {{TAG|LMAXFOCKAE}} is used to specify the maximum spherical (l) quantum number up
<math>
to which this more accurate treatment is used. The default is {{TAG|LMAXFOCKAE}}=4 for post DFT methods.
\hat n=\sum_{LM\atop n,m}\rho_{LM}\!\left[q^{LM}_{nm}Y_{LM}(\Omega)+c^L_{nm}\,\Delta g_L(r)Y_{LM}(\Omega)\right]
If no accurate augmentation is desired by the user, simply set {{TAG|LMAXFOCKAE}}=-1 in the INCAR file.
</math>


If {{TAG|LMAXFOCKAE}} is set to values larger than -1 and {{TAG|NMAXFOCKAE}}=2, the charge density is restored accurately on the plane wave grid up to a typical plane wave energy of 380 eV. As before,  {{TAG|LMAXFOCKAE}} can be used to specify the maximum spherical (l) quantum number up
where <math>Y_{LM}(\Omega)</math> are the spherical harmonics and <math>q^{LM}_{nm}</math> are the moments of the charge difference.
to which this more accurate treatment is used.  {{TAG|NMAXFOCKAE}}=2 yields very accurate results for
post DFT methods (MP2, RPA, GW) even for difficult 3d elements. For RPA and MP2 total energy calculations, differences between {{TAG|NMAXFOCKAE}}=1 and {{TAG|NMAXFOCKAE}}=2 are usually tiny for total energy differences. Since the absolute correlation energies might change, it is vital to use the same setting for
{{TAG|NMAXFOCKAE}} and {{TAG|LMAXFOCKAE}}, if energy differences are calculated.
For GW calculations, increasing  {{TAG|NMAXFOCKAE}} from 1 to 2 might change QP energies by 100-200 meV for 3d and late 4d and 5d elements.


{{TAG|NMAXFOCKAE}} controls how accurately this compensation is performed by specifying the maximum expansion order of spherical Bessel functions <math>j_L</math> with carefully chosen coefficients <math>\alpha_{\beta L}</math> and <math>q_{\beta L}</math> (cf. Appendix of Ref. {{cite|unzog:prb:2022}} for details).


The setting for {{TAG|LMAXFOCKAE}} should be also considered carefully. Generally, it suffices to set {{TAG|LMAXFOCKAE}} to twice the maximum ''l'' quantum number found in the {{FILE|POTCAR}} file.
<math>
For instance for sp elements, {{TAG|LMAXFOCKAE}} = 2 suffices. For d elements, {{TAG|LMAXFOCKAE}} = 4 suffices
\;\Delta g_L(r)=\sum_{\beta=1}^{NMAXFOCKAE}\alpha_{\beta L}\, j_L\!\left(q_{\beta L}\, r\right)
(a d electron can create charge densities with ''l''-quantum numbers up to 4), whereas for f elements, users
</math>
should test whether  {{TAG|LMAXFOCKAE}} = 6 is required.
{{NB|mind|This tag generally only applies to the Fock-exchange and [[:Category:Many-body perturbation theory|many-body perturbation theory]] (MBPT) (GW, RPA, [[MP2]], etc). The default for DFT and Hartree-Fock, {{TAG|LMAXFOCKAE|-1}}, restores only the moments of the all-electron charge densities on the plane-wave grid. For Hartree-Fock, the setting is very precise since the one-center terms are implemented in radial grids as for DFT.
}}


In summary, useful manual settings of {{TAG|NMAXFOCKAE}}  and {{TAG|LMAXFOCKAE}} are:
== Recommendations ==
* {{TAG|LMAXFOCKAE}}=-1, to switch off the accurate augmentation altogether
=== Element-Dependent ===
* {{TAG|LMAXFOCKAE}}=4 (or larger) to force an accurate treatment for the charge augmentation on the plane wave grid (can be selected even in Hartree-Fock type calculations).
The setting for {{TAG|LMAXFOCKAE}} should be chosen based on the elements in the calculation:
* {{TAG|NMAXFOCKAE}}=2, to select the very accurate augmentation on the plane wave grid. Please check whether the VASP default setting for {{TAG|LMAXFOCKAE}} suffices (OUTCAR file).
* '''''s'' and ''p'' elements''': {{TAG|LMAXFOCKAE|2}}
* '''''d'' elements''': {{TAG|LMAXFOCKAE|4}} (a ''d'' electron can create charge densities with L-quantum numbers up to 4)
* '''''f'' elements''': {{TAG|LMAXFOCKAE|6}} may be required - test for each case


The general rule is to set {{TAG|LMAXFOCKAE}} to approximately twice the maximum ''l'' quantum number found in the {{FILE|POTCAR}} file.
=== General ===
* Setting {{TAG|LMAXFOCKAE|4}} (or larger) forces an accurate treatment for charge augmentation on the plane-wave grid. This can be selected even in Hartree-Fock-type calculations, but causes some additional noise.
*If no accurate augmentation is desired set {{TAG|LMAXFOCKAE|-1}}, which will reduce the noise in the correlation energies.
* For DFT calculations, the exact one-center terms are implemented. This means shape restoration is not required and should not change the results. For post-DFT calculations, shape restoration significantly improves accuracy.
=== Relationship with NMAXFOCKAE ===
{{TAG|LMAXFOCKAE}} and {{TAG|NMAXFOCKAE}} are strongly coupled. It is vital to use consistent settings for both when comparing energy differences.
* {{TAG|NMAXFOCKAE|1}} and {{TAG|LMAXFOCKAE|0|op=≥}}: Charge density is restored up to approximately 150 eV (controlled by {{TAG|QMAXFOCKAE}}).
* {{TAG|NMAXFOCKAE|2}} and {{TAG|LMAXFOCKAE|0|op=≥}}: Charge density is restored up to approximately 400 eV.
{{NB|important|{{TAG|NMAXFOCKAE|2}} is for very accurate augmentation on the plane-wave grid. Use {{TAG|NMAXFOCKAE|2}} only with care, as it can result in very noisy data for coarse FFT grids.|:}}


== Related tags and articles ==
== Related tags and articles ==
{{TAG|LMAXFOCK}}, {{TAG|NMAXFOCKAE}}, {{TAG|QMAXFOCKAE}}, {{TAG|LFOCKAEDFT}}
{{TAG|NMAXFOCKAE}}, {{TAG|LMAXFOCK}}, {{TAG|QMAXFOCKAE}}, {{TAG|LFOCKAEDFT}}, {{TAG|LFOCKSTD}}
----
 
{{sc|LMAXFOCKAE|Examples|Examples that use this tag}}
 
== References ==


[[Category:INCAR tag]][[Category:Hybrids]][[Category:GW]]
[[Category:INCAR tag]]
[[Category:ACFDT]]
[[Category:Low-scaling GW and RPA]]
[[Category:Exchange-correlation functionals]]
[[Category:Hybrid functionals]]
[[Category:Many-body perturbation theory]]
[[Category:GW]]

Latest revision as of 13:46, 22 July 2026

LMAXFOCKAE = [integer] 

Default: LMAXFOCKAE = -1 DFT, Hartree-Fock
= 4 post-DFT methods

Description: Specifies the maximum spherical angular momentum quantum number ([math]\displaystyle{ L }[/math]) up to which augmentation charges are accurately restored on the plane-wave grid. This parameter controls the quality of shape restoration for the PAW method.


In the PAW method, the difference between the charge density of the all-electron partial waves [math]\displaystyle{ \phi_n }[/math] and the pseudo partial waves [math]\displaystyle{ \tilde \phi_n }[/math]

[math]\displaystyle{ Q_{nm}({\mathbf r})= \phi^*_n({\mathbf r})\phi_m({\mathbf r}) - \tilde \phi^*_n({\mathbf r})\tilde \phi_m({\mathbf r}) }[/math]

is usually treated on spherical grids centered at each atom (one-center terms inside the PAW spheres, see PAW method method and more detailed description for LMAXFOCK). To describe long-range electrostatic effects, the moments of the differences of the all-electron and pseudo charge density also need to be added on the plane wave grid (compensation density, see PAW method). These compensation charges exactly restore the moments of the all-electron density on the plane wave grid. For the Fock exchange the maximum L quantum number up to which the augmentation is done is controlled by LMAXFOCK.

For GW, RPA, and post-DFT methods, the one-center terms are, however, presently not implemented, which can cause sizable errors, particularly for 3d and (to a lesser extent) 2p, 4d, and 5d elements. To correct for this error, an alternative treatment is implemented on the plane-wave grid for the exchange and correlation contributions, allowing for accurate restoration of the all-electron densities.

This improved treatment — termed shape restoration — is available for exchange as well as many-body correlation contributions. More details on shape restoration are explained in Ref. [1] and [2].

Shape restoration allows restoration of the all-electron densities accurately already on the plane-wave grid by specifying the tags LMAXFOCKAE and NMAXFOCKAE. This accurate augmentation charge [math]\displaystyle{ \hat n }[/math] is calculated for the total orbital and angular momentum quantum numbers [math]\displaystyle{ L }[/math] (up to LMAXFOCKAE) and [math]\displaystyle{ M }[/math] using shape-restoring radial functions [math]\displaystyle{ \Delta g_L(r) }[/math] with coefficients [math]\displaystyle{ c^L_{nm} }[/math] to be determined:

[math]\displaystyle{ \hat n=\sum_{LM\atop n,m}\rho_{LM}\!\left[q^{LM}_{nm}Y_{LM}(\Omega)+c^L_{nm}\,\Delta g_L(r)Y_{LM}(\Omega)\right] }[/math]

where [math]\displaystyle{ Y_{LM}(\Omega) }[/math] are the spherical harmonics and [math]\displaystyle{ q^{LM}_{nm} }[/math] are the moments of the charge difference.

NMAXFOCKAE controls how accurately this compensation is performed by specifying the maximum expansion order of spherical Bessel functions [math]\displaystyle{ j_L }[/math] with carefully chosen coefficients [math]\displaystyle{ \alpha_{\beta L} }[/math] and [math]\displaystyle{ q_{\beta L} }[/math] (cf. Appendix of Ref. [2] for details).

[math]\displaystyle{ \;\Delta g_L(r)=\sum_{\beta=1}^{NMAXFOCKAE}\alpha_{\beta L}\, j_L\!\left(q_{\beta L}\, r\right) }[/math]

Recommendations

Element-Dependent

The setting for LMAXFOCKAE should be chosen based on the elements in the calculation:

  • s and p elements: LMAXFOCKAE = 2
  • d elements: LMAXFOCKAE = 4 (a d electron can create charge densities with L-quantum numbers up to 4)
  • f elements: LMAXFOCKAE = 6 may be required - test for each case

The general rule is to set LMAXFOCKAE to approximately twice the maximum l quantum number found in the POTCAR file.

General

  • Setting LMAXFOCKAE = 4 (or larger) forces an accurate treatment for charge augmentation on the plane-wave grid. This can be selected even in Hartree-Fock-type calculations, but causes some additional noise.
  • If no accurate augmentation is desired set LMAXFOCKAE = -1, which will reduce the noise in the correlation energies.
  • For DFT calculations, the exact one-center terms are implemented. This means shape restoration is not required and should not change the results. For post-DFT calculations, shape restoration significantly improves accuracy.

Relationship with NMAXFOCKAE

LMAXFOCKAE and NMAXFOCKAE are strongly coupled. It is vital to use consistent settings for both when comparing energy differences.

Related tags and articles

NMAXFOCKAE, LMAXFOCK, QMAXFOCKAE, LFOCKAEDFT, LFOCKSTD

Examples that use this tag

References