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.

% mv Test.FChk dna.Fchk<br />

run gaussiantoqmc . . . and follow the prompts<br />

% mv dna.qmc gwfn.data<br />

run casino.<br />

The code should automatically detect what sort <strong>of</strong> gaussian job it is and give you the opportunity<br />

to construct an excited-state wave function if applicable.<br />

It can deal with the following sorts <strong>of</strong> calculation:<br />

• HF and DFT ground states, open and closed shell.<br />

• CIS excited states, open and closed shell.<br />

• CASSCF states. Getting gaussian to output these can be problematic for large calculations.<br />

It is possible that gaussiantoqmc will get confused if you use some combination <strong>of</strong> IOps other<br />

than those described in Sec. 8.6.3.<br />

• Time-dependent HF (TD-HF) or DFT (TD-DFT) excited states.<br />

If the user chooses to output a CIS or TD-DFT wave function, they are given the option <strong>of</strong> resumming<br />

it. The wave function must also be resummed if the user wishes to analyse its composition. As<br />

distributed, gaussiantoqmc will not do this analysis but if you wish to switch this option on then the<br />

flag analyse cis in cis data.f90 must be set to T and gaussiantoqmc recompiled.<br />

With this flag set the code evaluates the percentage contribution <strong>of</strong> each single excitation to the CIS<br />

expansion. Degenerate virtual and occupied orbitals are identified and their contributions summed.<br />

The final output takes the form <strong>of</strong> files called fromi j.dat where i–j indicates a range <strong>of</strong> degenerate<br />

occupied orbitals (if i = j then i is a nondegenerate orbital). These files detail all excitations out<br />

<strong>of</strong> the specified orbitals along with a percentage giving their contribution to the CIS expansion as a<br />

whole. The sum <strong>of</strong> the percentages (final column) from each <strong>of</strong> the fromi j.dat should be 100 if<br />

everything is working OK.<br />

8.6.2 Other features <strong>of</strong> GAUSSIANTOQMC<br />

The code also contains some crude normalization and plotting routines that are really just debugging<br />

aids. By setting the flag test=T in gaussiantoqmc.f90 and recompiling, the user is given the option<br />

<strong>of</strong> plotting and testing the normalization <strong>of</strong> individual molecular orbitals. The axis along which the<br />

plotting is done is set in wfn construct.f90 and this must be hacked if the user wishes to change<br />

things.<br />

8.6.3 Getting GAUSSIAN to do what you want<br />

In principle, gaussian can do an awful lot <strong>of</strong> things. In fact, some <strong>of</strong> these things seem to require<br />

magical incantations. These will be described in this section, broken down into the different calculations<br />

to which they apply. The comments on g98 refer to revision A9 and may depend on which<br />

version is used.<br />

General bits and pieces<br />

Some points to note:<br />

• Both g94 and g98 appear to have formatting errors when printing out the Gaussians used in<br />

the ECP expansion—large exponent values are replaced with stars. This does not affect the<br />

subsequent calculation.<br />

• g94’s ECP (pseudopotential) package will not accept expansions containing more than 13 Gaussians<br />

per angular-momentum channel.<br />

• g98’s ECP integral package crashes when one attempts to do a large (both in terms <strong>of</strong> basis<br />

and ECP expansion) calculation with symmetry switched on. The solution to this is to switch<br />

symmetry <strong>of</strong>f using ‘Nosymm’.<br />

107

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

Saved successfully!

Ooh no, something went wrong!