Firefly and PC GAMESS-related discussion club


 
Learn how to ask questions correctly  
 
 
We are NATO-free zone
 



Re^2: Output of the gradients during CI optimization

Julia Endicott
julia.endicott@mail.utoronto.ca


Alex,

Thank you for having reply so quickly.

1) About the first question: I am already calculating the gradients that I need during a CI optimisation run ("runtyp=optimize") at each steps. What I would like is to write them at along the optimisation run (without extra calculation) for example in the output file, or in the PUNCH file.
I haven't find yet a keyword for this type of output, and I didn't find these gradients in any file (I only found the composite gradient in the output file).
Does such a keyword exist? Or do I need to restart each point of the optimisation as a "single point" calculation with "runtyp=gradient"?

2) About the second question: I could consider to use penalty function number 1, "B*ln(1.0 + (Δ/A)^2)", with a really large B factor, but I didn't find how to modify the parameters A and B. How can I change them?

Regards,

Julia


On Tue May 28 '13 1:00am, Alex Granovsky wrote
----------------------------------------------
>Hi Julia,

>>1) I assume that firefly needs the gradients of the states involved in the CI search in order to build the penalty function gradient. Is there any possibility to output these gradients?

>Are you interested in gradients for SA-MCSCF?

>You can find documented here:

>http://classic.chem.msu.su/gran/gamess/readme.htmlhttp://classic.chem.msu.su/gran/gamess/readme.html

>In particular:

>http://classic.chem.msu.su/gran/gamess/ss-gradients.htmlhttp://classic.chem.msu.su/gran/gamess/ss-gradients.html

>For SA-MCSCF, you just need to provide istate and jstate and run runtyp=gradient run.

>>2) My CI optimization is converging to a minimum in the intersection space but the degeneracy is not as small as I would like. Is there any possibility to run an optimization of the energy difference only?

>You can use Lagrange multiplier based optimization
>(see http://classic.chem.msu.su/gran/gamess/Locating_Conical_Intersections_and_Interstate_Crossings_with_Firefly.htmlhttp://classic.chem.msu.su/gran/gamess/Locating_Conical_Intersections_and_Interstate_Crossings_with_Firefly.html for details).
>Alternatively, you can provide non-default value of parameters of
>a penalty function to make it more sensitive to energy difference.

>Hope this helps.

>Kind regards,
>Alex Granovsky
>
>
>On Mon May 27 '13 10:31pm, Julia Endicott wrote
>-----------------------------------------------
>>I have two questions related to conical intersection (CI) optimizations.

>>1) I assume that firefly needs the gradients of the states involved in the CI search in order to build the penalty function gradient. Is there any possibility to output these gradients?

>>2) My CI optimization is converging to a minimum in the intersection space but the degeneracy is not as small as I would like. Is there any possibility to run an optimization of the energy difference only?

>>Thank you in advance,

>>Julia


[ Previous ] [ Next ] [ Index ]           Tue May 28 '13 4:14pm
[ Reply ] [ Edit ] [ Delete ]           This message read 1025 times