CECAM Workshop on
Exchange-Correlation Energy Functionals:
Assessment and Prospects
June 10-14, 2002
Gustavo E. Scuseria
Department of Chemistry
Rice University, Houston, Texas, USA
Andreas Savin
CNRS
Universite Pierre et Marie Curie, Paris, France
David J. Tozer
Department of Chemistry
University of Durham, Durham, UK
CECAM
Ecole Normale Superieure de Lyon,
46, Allee d'Italie,
69364 Lyon CEDEX 07 - FRANCE
Scientific Motivation
Density Functional Theory (DFT) has become the "de facto"
workhorse of electronic structure theory. Both computational
quantum chemistry and condensed matter physics extensively used DFT
for the theoretical prediction of molecular and materials properties:
there is widespread consensus among electronic structure practitioners
that DFT offers a very high "value" in terms of its accuracy to
computational cost ratio.
Even though DFT is in principle an exact theory, its practical accuracy
is limited by the accuracy of the so-called exchange-correlation
functional (Exc), which introduces quantum mechanical many-body
electronic effects into the model. Most terms in the total energy functional
can be written as simple expressions. Unfortunately, Exc is not precisely known.
Over the years, DFT practitioners have developed a number of approximations
to the yet unknown exact Exc functional. Among others, these efforts
include functionals based on properties of the uniform electron gas,
generalized gradient approximations, hybrid schemes which include a portion
of exact Hartree-Fock exchange, and other more recent schemes which
include a kinetic energy density contribution to Exc.
These efforts have lead to functionals that are currently widely used
by a large variety of scientists in many different disciplines
including chemistry, physics, biology, and multiple areas of engineering.
DFT is currently used extensively both by theoreticians and
experimentalists.
The numbers of scientific papers in the literature using DFT has
grown exponentially and remains unabated.
The accuracy of energy functionals has steadily improved over the years
and showed a substantial leap-forward in the early 1990s. Although many
properties like interatomic distances and energies are fairly
reproduced by the model when compared to experiment, it would be desirable
to further improve the predictions to so called "chemical accuracy" (1 kcal/mol
in energies and 0.001 A in equilibrium distances). Furthermore, there are
still a large number of properties for which the best current functionals
in DFT do not provide qualitatively correct answers. Among others, the general
description of activation
energy barriers, magnetic properties, and van der Waals interactions could
benefit from better energy functionals.
The workshop will focus on addressing the question:
"Is the quest for an exact Exc leading us to the promised land?"
More specifically:
1. Is DFT just one more "Semiempirical Theory" or will it ever deliver
on its "exactness" promise?
2. Is there a systematic way to construct functionals
that will lead to the right answer for the right reason?
3. What are the strengths and weaknesses of commonly used functionals?
4. Do chemical, biological, and materials related applications
impose different constrains on Exc?
5. Is there a limit to the accuracy of so-called "semiempirical" functionals?
6. Is self-interaction error the main culprit of current functional limitations?
7. Are radical departures from the "conventional wisdom" needed
to substantially and systematically improve accuracy?
8. What is the role of empirical parameters in the development of Exc?
Are they truly needed? Do they improve functionals significantly?
Most of these questions do not have precise answers, but an open, candid
debate among leaders in functional development will certainly impact and shape
the future of this important field.
To get more info contact:
Gustavo Scuseria
