Interplay between large materials databases, semi-empirical approaches, neuro-computing and first principle calculations

Analyzing the conditions that make it possible to search for materials science concepts, it is shown that it was the amassing of a critical volume of experimentally-determined data in the literature that permitted an individual with deep insight to perceive an underlying pattern not previously apparent. Extending these facts to a new area of materials design leads to the following four key-points: 1) creation and use of huge, critically evaluated materials databases; 2) computer-aided reduction of elemental parameters and systematic combinations of them to find the relevant feature sets which can link materials properties qualitatively with the chemical species present; 3) refinement and optimisation of qualitatively-obtained results under (2) with the help of neurocomputing leading to more explicit quantitative results; and 4) focusing on predicted, most-promising materials systems with the aim to reduce the experimental work for verification, as well as trying to create a theoretically-based explanation for such quantitative results