From chemistry-request@www.ccl.net Fri Sep 4 12:32:11 1998 Received: from aitana.cpd.ua.es (aitana.cpd.ua.es [193.145.233.5]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with ESMTP id MAA09584 Fri, 4 Sep 1998 12:32:07 -0400 (EDT) Received: from fisic1.ua.es (fisic1.ua.es [193.145.230.65]) by aitana.cpd.ua.es (AIX4.2/UCB 8.7/8.7) with SMTP id SAA35906 for ; Fri, 4 Sep 1998 18:25:40 +0200 (METDST) Received: from localhost by fisic1.ua.es; (5.65/1.1.8.2/18Oct95-0850PM) id AA03032; Fri, 4 Sep 1998 18:28:26 +0200 Sender: juancar@fisic1.ua.es Message-Id: <35F0152A.2F1C@fisic1.ua.es> Date: Fri, 04 Sep 1998 18:28:26 +0200 From: Juan Carlos Sancho X-Mailer: Mozilla 3.04Gold (X11; I; OSF1 V3.2 alpha) Mime-Version: 1.0 To: chemistry@www.ccl.net Subject: Re:CCL:CCL:Multideterminantal DFT Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit In DFT the electron density is the key function because the HK theorem establishes that the ground state exact energy is provided through an universal functional. KS theory makes practicable DFT affording its calculation using a single Slater determinant subject to the condition of providing the exact density. Posteriorly the need to include the on-top pair density has always appeared in situations when quasi-degenerations are present. In these cases the density functional deficiencies have been solved using this on-top density constructed from a minimal number of Slater determinant neccesary to take into account this degeneracy. As a possible solution it has been proposed years ago, to write the density as a sum of terms with a non-integer occupation number associated with the KS orbitals. A funtional like this was used for the first time in DFT by Lie and Clementi (1974). If we select, for the Ks partition of the energy, an exchange potential similar to the HF one, we have an unique way of defining the remaining term (correlation potential) and most of the published correlation potentials have been obtained using the above definition of the exchange potential. We can look for correlation functionals depending of the polarization concept (Perdew 86) or not (Lee-Yang-Parr) but, in all cases, the potential energy curves are not reproduced although the correct dissociation limit is reached in an unrestricted KS calculation. And, why this potentials provide good results whenever the wavefunction is well approximated by just one Slater determinant?. I think that, when the exact energy requires a many-determinant wavefunction, a partition of the energy must be made with the many-determinant contribution completely included into the exchange-correlation term. The DFT potentials published so far have not been obtained considering this depedence. Certainly the deficiency appears in significative systems. For instance, in the PEC of fluorine molecule we can see how UKS does not provide the shape of the curve at intermediate distance. If we use a two-determinant wavefunction (GVB with perfect pairing) and evaluating later the correlation energy by means of the Colle-Salvetti functional, using the GVB two-body density matrix previously obtained, we recover the shape of the curve and the results improves greatly the UKS one !. What comes out from this analysis ?, How we can recognize the non-local corrections needed for the functionals ?, Dynamical vs. non-dynamical correlation ? ... It would be a great pleasure to find answers to this questions between all, mixing experiences and opinions. Thanks, ---------------------------------------------------------------------- J.C. Sancho-Garcia Dpto. Quimica-Fisica Universidad de Alicante Apartado 99, 03080 Alicante SPAIN juancar@fisic1.ua.es ----------------------------------------------------------------------