Francisco Martin-Martinez1,Isaac Vidal-Daza1,2,Anna Bachs-Herrera1
Swansea University1,University of Granada2
Francisco Martin-Martinez1,Isaac Vidal-Daza1,2,Anna Bachs-Herrera1
Swansea University1,University of Granada2
Nature has been traditionally mimicked in the design of biomaterials for her capacity to achieve performance, but not that often for her ability to degrade at the end of life. To design materials for degradation as well as performance, molecular degradation needs to be better modelled. Degradation is a complex phenomenon largely dependent on the molecular system, the environmental conditions, and the timescales being considered. A common factor in the degradation process is molecular reactivity, which can be predicted using conceptual density functional theory (DFT). Conceptual DFT reactivity descriptors either quantify the global tendency of a molecule to engage in chemical reactions or the areas of the molecule that would undergo such engagement, but these global and local reactivity descriptors are usually detached. Atomic reactivity is not usually quantified in relation to the global molecular reactivity, and it limits our ability to accurately predict degradation. In fact, quantifying the reactivity of any individual atom within the global reactivity of a molecule, is not a trivial task, and it requires the conjunction of different theories in the framework of quantum chemistry. In this work, we combine a topological analysis of the electron density with a non-arbitrary partition of the molecular space into atomic domains to define topology-based atom-condensed reactivity indexes to predict molecular degradation. High-throughput calculations of these <i>ab initio</i> atom-condensed reactivity descriptors enabled the analysis and identification of existing patterns among data, which provides a better understanding of molecular reactivity, and the definition of new indexes that describe the local reactivity of individual atoms, in the global reactivity of target molecules. The data generated can be used to train machine learning (ML) models that predict molecular and biomaterials degradation.