import>Ajs1 at 15:41, 20 October 2008

2008-10-20T15:41:33Z

← Older revision		Revision as of 15:41, 20 October 2008
Line 1:		Line 1:
			[[ CamCASP ]] => [[ CamCASP/Notes \| Notes ]] => Rotating integrals

	==1. Rotating integrals==		==1. Rotating integrals==

import>Ajs1 at 14:22, 20 October 2008

2008-10-20T14:22:52Z

New page

==1. Rotating integrals==

CamCASP allows the User fine control over integrals and the density-fitting. This can be used to save time (and also debug) during certain types of calculations. So, if you need to evaluate integrals like the density-overlap integral for many dimer configurations, here are two ways of doing it:

* Redoing the density-fitting for each dimer configuration
<pre>
Set DF
Redo-df-on-rotation
End

Begin DF-ints
Int Type OVRL Desc XAXB Print
End
Begin E1exch
NO-E1exch
Overlap integrals without constraints
End

Edit CD
Rotate by 90 degrees about 0.0 0.0 1.0
Show geometry and MOs
Update MOs & Geom
Show geometry & MOs
End

Begin DF-ints
Int Type OVRL Desc XAXB Print
End
Begin E1exch
NO-E1exch
Overlap integrals without constraints
End
</pre>

This works as follows:

# We first tell the code to repeat density-fitting even if all that has happened is that a molecule has been rotated.
# Next you could perform the density-fitting for the two molecules, but it will be automatically done when the '''DF-INTS''' is called.
# Use the '''DF-INT'''egral module to calculate the integral. Here is it the overlap integral needed to calculate the density-overlap. The density-overlap is a special kind of integral that is not calculated in this module.
# Use the '''E1EXCH''' module to calculate the density-overlap integrals. The '''NO-E1exch''' command prevents the calculation of the first-order exchange energy which is calculated by default.
# Now rotate one of the molecules (CD). The '''UPDATE MOs & GEOM''' command must be used to update the molecule geometry '''and''' the molecular orbitals.
# You could re-do the density-fitting for the rotated molecule (CD), but it will be automatically done by '''DF-INTS'''...
# Recalculate the overlap integrals, but this time include the '''ROTATE''' option that will rotate them so as to be consistent with the previously calculated density-fitting solution.
# And repeat the calculation of the denisty-overlap integrals.

This is a wasteful procedure. A better way is as follows...

* Doing the density-fitting only once & rotating integrals
<pre>
Set DF
Redo-df-on-rotation False
End

Begin DF-ints
Int Type OVRL Desc XAXB Print
End
Begin E1exch
NO-E1exch
Overlap integrals without constraints
End

Edit CD
Rotate by 90 degrees about 0.0 0.0 1.0
Show geometry and MOs
Update
Show geometry & MOs
End

Begin DF-ints
Int Type OVRL Desc XAXB Print Rotate
End
Begin E1exch
NO-E1exch
Overlap integrals without constraints
End

</pre>

This works as follows:

# We first stop the code from repeating density-fitting if all that has happened is that a molecule has been rotated.
# Next you could perform the density-fitting for the two molecules, but it will be automatically done when the '''DF-INTS''' is called.
# Use the '''DF-INT'''egral module to calculate the integral. Here is it the overlap integral needed to calculate the density-overlap. The density-overlap is a special kind of integral that is not calculated in this module.
# Use the '''E1EXCH''' module to calculate the density-overlap integrals. The '''NO-E1exch''' command prevents the calculation of the first-order exchange energy which is calculated by default.
# Now rotate one of the molecules (CD). The '''UPDATE''' command must be used to update the molecule geometry. But do not update the molecular orbitals (i.e., don't use '''UPDATE MOs''').
# Recalculate the overlap integrals, but this time include the '''ROTATE''' option that will rotate them so as to be consistent with the previously calculated density-fitting solution.
# And repeat the calculation of the denisty-overlap integrals.
--[[User:am592|alston]] 18:39, 17 October 2008 (BST)

CamCASP/Notes/1 - Revision history

import>Ajs1 at 15:41, 20 October 2008

import>Ajs1 at 14:22, 20 October 2008