LSPECTRALGW: Difference between revisions

Revision as of 18:13, 13 March 2017

LSPECTRALGW = .FALSE. | .TRUE.

Default: LSPECTRALGW

= .FALSE.

{{{3}}}

Description: LSPECTRALGW specifies to use the spectral method for calculating the self-energy.

If LSPECTRALGW = .TRUE. is set, the imaginary part of the self-energy $\Sigma (\omega )=GW$ is calculated from the imaginary part of screened potential $W(\omega )$ by shifting the poles of $W$ by $\pm \epsilon$ , where $\epsilon$ are the poles of the Green's function $G$ . Generally, LSPECTRALGW affects the compute time very little. QP energies also hardly change when LSPECTRALGW is modified. However, LSPECTRALGW = .TRUE. is usually is slightly more robust, and should be selected for molecules and other systems with flat bands. One the other hand, LSPECTRALGW = .TRUE. seems to converge slightly slower, as the complex shift CSHIFT is decreased.

Related Tags and Sections

LSPECTRAL

Example Calculations using this Tag

Contents

@@ Line 1: / Line 1: @@
 {{TAGDEF|LSPECTRALGW|.FALSE. {{!}} .TRUE.}}
-{{DEF|LSPECTRALGW|.TRUE.|if {{TAG|NOMEGA}}>2}}
+{{DEF|LSPECTRALGW|.FALSE.}}
-Description: {{TAG|LSPECTRALGW}} specifies to use the spectral method.
+Description: {{TAG|LSPECTRALGW}} specifies to use the spectral method for calculating the self-energy.
 ----
-If {{TAG|LSPECTRALGW}} = .TRUE. is set, the imaginary part of the independent particle polarizability <math>\chi_{\mathbf{q}}^0 (\mathbf{G}, \mathbf{G}', \omega)</math> is calculated first, and afterwards the full independent particle polarizability is determined using a Kramers-Kronig (or Hilbert) transform. This reduces the computational work load by almost a factor {{TAG|NOMEGA}}/2. The downside of the coin is that the response function must be kept in memory for all considered frequencies, which can cause excessive memory requirements. VASP therefore distributes the dielectric functions among the available compute nodes.
+If {{TAG|LSPECTRALGW}} = .TRUE. is set, the imaginary part of the self-energy <math>\Sigma(\omega)= G W</math> is calculated from the imaginary part of screened potential <math>W(\omega)</math> by shifting the poles of
+<math>W </math> by <math> \pm \epsilon </math>, where <math> \epsilon </math> are the poles of the Green's function <math> G</math>.
+Generally, {{TAG|LSPECTRALGW}} affects the compute time very little. QP energies also hardly
+change when {{TAG|LSPECTRALGW}}  is modified.
+However, {{TAG|LSPECTRALGW}} = .TRUE. is usually is slightly more robust,
+and should be selected for molecules and other systems with flat bands.
+One the other hand, {{TAG|LSPECTRALGW}} = .TRUE. seems to converge slightly slower,
+as the complex shift {{TAG|CSHIFT}} is decreased.
-A similar trick is used when the QP-shifts are calculated. In general it is strongly recommended to set {{TAG|LSPECTRALGW}} = .TRUE., except if memory requirements are too excessive.
 == Related Tags and Sections ==