Posted by Francesco Aquilante on November 03, 2008 at 20:46:00:
In Reply to: Resolution of the Identity posted by Samir ZEIN on November 03, 2008 at 17:25:47:
Despite your negative impression about what Molcas can do in terms of "RI technique", as we stand this is the QC software that offers the greatest deal of flexibility wrt two-electron integral approximations. In version 7.0 you can use the Cholesky keyword in Seward and this will produce the same speed-ups of RI with no need to specify any auxiliary basis set and with controllable accuracy. In case you want to use RI tout-court, then it is 7.2 the version that you should use. In that case, we stronlgy recommend you to use the RICD option. Again, this does not require any additional input (except if you want to change the accuracy of the RI approximation), as the auxiliary basis set is generated on the fly (btw, the publication describing this "ab initio RI" is on J. Chem. Phys. 127, 2007, 114107). Finally, if you really want to resort to the old-fashioned way to do pre-cooked RI approximation, you can specify things like RIJ, RIJK or RIC but at the moment very few auxiliary basis set are ready-to-use and you would need to download them from the internet and add to the corresponding folder in basis_library...
: 1 - How can I precise the type of auxiliary basis set in my input file? (an example input with several atoms molecules would be of nice help).
Just write ONE and ONLY ONE of the following lines in Seward
Cholesky (7.0 or newer)
RICD (7.2 or newer)
: Comment:
: I think that the informations about the RI approximations listed in page 323 in the users manual could be completed by explaining that the RIJ is for the coulomb terms, RIK is for exchange terms (All these informations are available in the ORCA manual)... By the way, we cannot know anything about the RICD whose paper is submitted to ? journal. Finally the /molcas.7.0.i686/basis_library/jk_Basis/ directory was empty in our version of MOLCAS. Does this mean that the RIJK is currently not available (This is a second question).
Our philosophy is that users are (almost) always right. So, the information will be soon completed. Again, we believe that this RIJ, RIJK, ... auxiliary basis (not methods!) belong to the past century, when number crunching took over beauty and rigour of theory.
Best Regards,
francesco