[molpro-user] Technical Question

Martin Spenke martinspenke at yahoo.de
Mon May 18 20:08:57 BST 2015


Dear Gerald, 
many many thanks, i was falsely assuming that  integral transformation is only done for the active space.I need FCIDUMP for a subsequent dmrg calculation, but before dumping the integrals, i also need to perform orbital localization and rotation.
Any way i already downloaded your wmme code. I may come back to you again.Best wishesMartin 


     Gerald Knizia <knizia at theochem.uni-stuttgart.de> schrieb am 20:46 Montag, 18.Mai 2015:
   

 Dear Martin,
FCIDUMP is not meant for actual calculations in this way. It is more of
a debugging & experimenting tool, meant for cases in which the full
integral transformation is trivial(!) compared to what you intend to do
with it afterwards (that is, high-order coupled cluster, FCI-like
treatments, stochastic methods, etc). It is not for routine
calculations.

Note also that fci computes the FULL integral transformation, not only
the active space integrals, as you seem to imply. The only orbitals
which get removed are core orbitals, but non-core closed and virtual
orbitals stay in.

If you need more complex stuff, you could try to hack up your own
integrals. E.g., from fitting integrals via wmme (which is a program
available at my homepage and also comes with some python code for
reading Molpro xml files).

Best wishes,
Gerald



On Sat, 2015-05-16 at 17:27 +0000, Martin Spenke wrote:
> Thanks a lot Muammar, 
> 
> 
> now i have another problem, i try to dump the 2-electron integrals by
> FCIDUMP keyword for 110 atoms, but regardless of using gdirect or
> not, 
> it takes me huge huge amount of memory, any possible way to circumvent
> this ?
> I have 512 GB ram available and let it run on 16 processors, but it
> seems it is not enough to obtain FCIDUMP file.
> 
> 
> number of electrons : 6 
> considered occupied states : 3
> considered unoccupied states : 3
> The molecule has many orbitals but i am only interested in homo-lumo
> region, 
> i do not figure it out why molpro needs such a huge amount of memory
> for dumping 
> FCIDUMP file for this relatively small active space ?????
> 
> 
> Best wishes
> Martin
> 
> 
> 
> 
> Muammar El Khatib <muammarelkhatib at gmail.com> schrieb am 19:07
> Samstag, 16.Mai 2015:
> 
> 
> 
> Hi Martin,
> 
> On 14/05/15 21:19, Martin Spenke wrote:
> > Unfortunately my tmp file increase to some 200-300 GB, how can i
> avoid that
> > molpro writes the 2-electron integrals into the disk
> 
> 
> Check the GDIRECT directive in the manual:
> 
> https://www.molpro.net/info/2012.1/doc/manual/node131.html
> 
> 
> Regards,
> 
> -- 
> Muammar El Khatib.
> http://muammar.me | http://proyectociencia.org
> 
> 
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user
> 
> 
> 
> 
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user




  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20150518/79b007a7/attachment-0001.html>


More information about the Molpro-user mailing list