Most of the "first principles" simulations we do are with a theory known as
density-functional theory (DFT). This is a very powerful theory. In principle it
is exact but in practice it relies on an approximation for how electrons interact
with each other. We are tackling the issue of the accuracy of DFT through extensive
series of studies of small gas phase complexes [1-5], and water-solid
interactions [6]. These benchmark studies with techniques such as Møller
Plesset perturbation theory, coupled cluster, or quantum Monte Carlo often
come with extreme computational burdens. However, these benchmarks are essential
to establish the accuracy of more traditional methods such as DFT, and help to ensure
that the numbers we produce stand the test of time and experiment. Some specific examples of this work includes:
Water clusters: The four structures below are some of the low energy arrangements of a gas phase water hexamer. Which one has the lowest
energy has been hotly debated for decades. Our simulations have helped to resolve this debate [2].
Chemical accuracy for adsorption: "Chemical accuracy" - an accuracy of 1 kcal/mol or ~43 meV - is a much-discussed and highly sought after precision
in the simulation of chemical processes and materials. It is the level of accuracy that must be achieved in order to make quantitative predictions of
reaction rates. There are established techniques for obtaining this accuracy when dealing with the full range of interactions that govern molecular systems,
as afforded through the application of explicitly correlated quantum chemistry methods with very large basis sets. However, when it comes to moderately
complex problems in condensed matter, such as the calculation of the adsorption energy of a molecule on a surface, such techniques have not been used to
demonstrate that chemical accuracy can be achieved. Given the central importance of the molecule-solid interaction to disciplines as diverse as catalysis,
electrochemistry, and semiconductor processing, it remains a major challenge to theory to demonstrate that chemical accuracy in this area can be obtained or
at least approached. We have been working on trying to obtain chemical accuracy in adsorption energies of molecules at solid surfaces. One system where we
have almost achieved this is water on common salt (NaCl): our best estimate of the binding energy for a water monomer on salt is about half an electron volt
(ca. 50 kj/mol) [6]. Another system we are examining at present is water on graphene, where the binding energy is much smaller.
Dispersion forces from density functional theory: London dispersion interactions are ubiquitous in nature contributing to the binding of biomolecules
such as DNA, molecular crystals, and molecules on surfaces. The accurate description of dispersion, which often occurs in conjunction with hydrogen bonds,
is a major challenge for many electronic structure theories. Density functional theory (DFT), the most widely used electronic structure theory, often
doesn't meet this challenge. Indeed, it is well established that popular generalized gradient approximation (GGA) or hybrid exchange-correlation functionals
are inadequate for the description of dispersion interactions. Many schemes have been developed that allow dispersion to be accounted for within
DFT in a more or less approximate manner. One of the most promising and rigorous methods is the nonlocal van der Waals density functional (vdW-DF) of
Langreth and Lundqvist and co-workers [M. Dion et al., Phys. Rev. Lett. 92, 246401 (2004)]. In our recent work [5] we have shown how the accuracy of vdW-DF
can be dramatically improved both for dispersion and hydrogen bonded complexes through the judicious selection of its underlying exchange functional.
New and published exchange functionals were identified that deliver much better than chemical accuracy from vdW-DF for the S22 benchmark set of weakly
interacting dimers and for water clusters. Improved performance for the adsorption of water on salt was also obtained.
Differences in interaction energies for vdW-DF (Eint[DFT]) with various exchange functionals from accurate reference data. Our "optB88" exchange functional yields much smaller errors than other choices of exchange functional.
References:
[1] B. Santra, A. Michaelides, and M. Scheffler, J. Chem. Phys. 127, 184104 (2007). [pdf]
[2] B. Santra, A. Michaelides, M. Fuchs, A. Tkatchenko, C. Filippi, and M. Scheffler, J. Chem. Phys. 129, 194111 (2008). [pdf]
[3] B. Santra, A. Michaelides, and M. Scheffler, J. Chem. Phys. 131, 124509 (2009). [pdf]
[4] J. Ma, D. Alfè, A. Michaelides, and E. Wang, J. Chem. Phys. 130, 154303 (2009). [pdf]
[5] J. Klimeš, David R. Bowler and Angelos Michaelides, J. Phys.: Condensed Matter 22, 074203 (2010). [pdf]
[6] B. Li, A. Michaelides, and M. Scheffler, Surf. Sci. 602, L135 (2008). [pdf]
|
+44 (0)20 7679 2000 - Copyright © 1999-2005 UCL
