Summary: DFT on openshell-Systems

From: Frank Schaper <hein' at \`chemie.uni-konstanz.de>
Subject: Summary: DFT on openshell-Systems
Date: Sun, 23 Jun 2002 19:21:45 +0200
 Dear Netters,
 thanx for the fast response to:
 "Deng, Jun" <jdeng' at \`ppg.com>,
 "Per-Ola Norrby" <pon' at \`mail.kemi.dtu.dk>,
 Stefan Grimme <grimmes' at \`uni-muenster.de>,
 Tomasz Borowski <borowski' at \`chemia.uj.edu.pl>,
 Sorana Ionescu <soranaio' at \`yahoo.com>,
 Antonia de Noronha <noronha' at \`dedalus>
 and Frank Neese <neese' at \`mpi-muelheim.mpg.de>.
 Here is requested summary:
 - Is a strong dependence on the DFT-functional a "normal" behavior for
 open-shell systems?
 The answer was a general "yes" from all experts. Sometimes even
 orbital
 orders differ with the chosen functional. The local density approximation
 (VWN-functional) was regarded a bad choice anyway.  In general, DFT seems
 to prefer lower spin states in contrast to HF-methods.
 - Is there any way to take a shortcut for geometry optimizations? E.g.
 calculating the geometry closed-shell and using single-point energies for
 different open-shell spin states.
 - Is a geometry optimization mandatory for every spin state (possible in
 my case S=1, 3 and 5) or are single-point calculations sufficient? I'm
 very much afraid they are not.
 Summarizing the answers to both questions, it seems to be necessary to
 optimize every geometry with the functional used for the energy
 calculations. Geometries should also be optimized for every spin state
 (that was a silly question in any case).
 - Can the calculated energies of the same complex with different spin
 states be compared directly or is the spin-pairing-energy missing to some
 extend?
 Here the opinions seem to differ a little bit. As far as I have
 understood, energies of different spin states can be compared, i.e. the
 spin pairing energy is included in the calculated energies.  On the other
 hand, single-determinant DFT-methods seem to have problems describing
 lower spin states and the relative energies (though they can be compared
 in principle) are rather qualitative up to wrong. Multi-configuration
 methods were proposed to give better results.
 Thanks again for the responses.
 Hein
 --
 Frank (Hein) Schaper - AG Brintzinger - Universitaet Konstanz
 email : Frank.Schaper' at \`uni-konstanz.de - Frank' at \`Schaper.com
         mail' at \`FrankSchaper.de
 web   : www.FrankSchaper.de
 s-mail : Frank Schaper - AG Brintzinger - Universitaetsstr. 10
          78434 Konstanz