On the limits of ab initio and DFT methods for bioinorganic iron models

Iron is one of the most common metal centers in bioinorganic chemistry. The many processes in which iron plays a part are governed by its electronic structure. Particularly, the different accessible oxidation and spin states can display distinct physical and chemical properties. In this regard, computational chemistry provides an invaluable tool for the study of the electronic structure of the different iron centers. However, not all the theoretical methods are equally precise or accurate. Furthermore, it is not always easy to discern which answer is the correct one whenever different methods are in conflict. The most often used methodology for this type of studies is density functional theory, but its accuracy for the description of transition metal centers is not always consistent, especially when studying spin-state energetics. Ab initio (or wavefunction) methods can sometimes be more accurate, but they tend to be much more computationally expensive, and only applicable to small or medium-sized systems.In this thesis we will explore the accuracy of different wavefunction and density functional methods. The systems that we have studied are related to iron sites of biological importance, such as non-heme high-valent iron centers or iron-sulfur clusters. We have found limitations in some of the wavefunction methods, and further study is needed to understand the reasons behind these limitations and in order to improve the methods. We also show how composite multireference methods that have a balanced treatment of electron correlation provide a description in best agreement with experimental evidence. With respect to the density functional methods we have developed metrics to assess their sensitivity to small changes in the density. These metrics show that some biologically important iron-sulfur clusters, such as the iron molybdenum cofactor of nitrogenase, are extremely sensitive to small changes in the density, which stresses the need for better, consistently accurate, density functionals.

Publication year:2021

Accessibility:Open