14.09.2014 Views

CASINO manual - Theory of Condensed Matter

CASINO manual - Theory of Condensed Matter

CASINO manual - Theory of Condensed Matter

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

1<br />

START SET 1<br />

Spherical harmonic l,m<br />

0 0<br />

Expansion order N_h<br />

3<br />

Spin dep (0->uu=dd=ud; 1->uu=dd/=ud; 2->uu/=dd/=ud)<br />

-1<br />

Cut<strong>of</strong>f (a.u.) ; Optimizable (0=NO; 1=YES)<br />

0.d0 1<br />

Parameter values ; Optimizable (0=NO; 1=YES)<br />

END SET 1<br />

END H TERM<br />

where use is made <strong>of</strong> spin-triplet dependencies as defined in the custom striplet dep input block.<br />

Notes:<br />

• There are 6 types <strong>of</strong> term available: u (isotropic electron–electron terms), χ (isotropic<br />

electron–nucleus terms), f (isotropic electron–electron–nucleus terms), p (plane-wave expansions<br />

in electron–electron separation), q (plane-wave expansions in electron position) and W /H<br />

(isotropic, homogeneous three-electron term). All <strong>of</strong> these terms are optional: e.g., omitting all<br />

the lines from ‘START Q TERM’ to ‘END Q TERM’ will give a Jastrow factor with no q terms.<br />

In many cases (particularly in the presence <strong>of</strong> pseudopotentials) only u and χ terms are needed,<br />

and this is good since higher-order terms such as f-functions can become very expensive in large<br />

systems.<br />

• The ‘truncation order’ should be either 2 or 3. If it is <strong>of</strong> value C then the Jastrow factor is<br />

C times differentiable everywhere; it must therefore be at least 2 for the kinetic-energy integrand<br />

to be well-defined. If it is 3 then the local energy is continuous in configuration space<br />

(assuming the orbitals are smooth). This is not strictly required, and leads to the loss <strong>of</strong> some<br />

variational freedom, but it makes the numerical optimization <strong>of</strong> cut<strong>of</strong>f lengths easier. In general,<br />

we recommend setting C = 3.<br />

• In a future release, the u and χ terms may be made anisotropic; there are place-holders for this<br />

in the Jastrow data set. At present, however, both must be isotropic. There should only be one<br />

set <strong>of</strong> u terms, and the spherical harmonic l and m values should therefore be set to 0.<br />

• The ‘expansion order’ <strong>of</strong> u determines the number <strong>of</strong> parameters. Typically it is given a value<br />

between 4 and 8.<br />

• The ‘spin-dep params’ line allows the user to specify whether the same u-parameters are to<br />

be used for parallel- and antiparallel-spin electron-pairs. If the value is set to 0 then the same<br />

parameters are used for parallel and antiparallel pairs (this option should not be used in general);<br />

if the value is set to 1 then different parameters are used for parallel and antiparallel pairs; if<br />

it is set to 2 then different parameters are used for up-spin, down-spin and opposite-spin pairs:<br />

this is useful for spin-polarized systems. 12<br />

• The cut<strong>of</strong>f length is given a default value if it is set to 0. Note that it is possible to optimize<br />

the cut<strong>of</strong>f length using variance minimization. This can be useful, because the choice <strong>of</strong> cut<strong>of</strong>f<br />

length has a significant effect on the optimized energy and variance. Unfortunately, optimizing<br />

the cut<strong>of</strong>f length is numerically difficult: variance minimization will take many more iterations<br />

to converge if the cut<strong>of</strong>f is optimizable. Setting the truncation order to 3 can help somewhat.<br />

• It is possible to specify exactly which <strong>of</strong> the expansion coefficients (‘parameter values’) are<br />

optimizable and which are not. One does not need to specify all <strong>of</strong> the expansion coefficients:<br />

any that are not listed are assumed to be zero. If all <strong>of</strong> the parameter values in all <strong>of</strong> the<br />

Jastrow terms are blank, as is the case in the example given above, then the following default<br />

behaviour will be used: if one is studying a homogeneous system, or a 3D-periodic system<br />

or is using energy minimisation, a simple default for u will be chosen that satisfies the Kato<br />

12 When particles other than electrons are present in the systems, introduced via the particles block input keyword,<br />

these definitions change a little. Information on the particle and particle-pair splittings are given near the beginning <strong>of</strong><br />

the out file.<br />

62

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!