Using first principles calculations for modelling energy materials: concepts and challenges

by Dr Vladimir Tripkovic

Wednesday 13 Jun 2018, 14:00 → 15:00 Europe/Zurich

401 (OFLG)

Density Functional Theory (DFT) has become the most widely used computational method for studying bulk and surface properties of extended systems at the nanoscale level. Nowadays, DFT does not only provide information not amenable to experimental determination, but arguably, can drive catalyst design in silico. Nevertheless, one has to be extremely careful when applying DFT, because the method per se and approximations used to estimate material properties are approximative. I will support this statement by two examples. A known artefact of DFT is a self-interaction error (SIE), which tends to delocalize electron density. This is particularly an issue in transition metal (TM) compounds whose electronic structure contains many local features. This problem is commonly tackled by applying a localization correction (Hubbard U term to the localized mostly d and f states that participate in chemical bonding) or adding part of a Hartree-Fock exchange as in hybrid functionals (mixing  term). However, both of these parameters are empirical and have to be fitted to a property of interest, i.e. a value usually taken from experiments. Reaction energetics is extremely sensitive to the choice of the parameters and selecting the right value and property to fit is of paramount importance. I will present a simple empirical method that gives accurate reaction energetics independent of  and U values. Besides the method, it is also important to have good theoretical concepts for assessing material properties from DFT calculations. For instance, a simple descriptor approach based on reaction thermochemistry is often used for deducing the most active catalysts within the field of heterogeneous catalysis and electro-catalysis. This approach is neat in the sense that one only needs to calculate a single parameter; however I will demonstrate on several examples that this holds only to a first approximation. To get a complete picture one has to take explicitly into account the chemical environment, e.g. spectator surface species, the role of solvent, competing reactions and last but not the least reaction kinetics. In the conventional approach many of these are either neglected or roughly approximated. Finally, I will share a few thoughts and ideas about future directions and instances in which I believe theory can provide valuable information to understand and complement experimental results.