Computational Materials Design

Considering for example organic electronics, the possible design space of small molecular crystals alone easily ranges in the tens of millions of compounds. This awards unparalleled flexibility of possible materials- or electronic properties where crystals can be tailored to fit any purpose. On the other hand, the huge variability is also the main factor hampering computational materials design, as the potential search spaces are enormous and not easily amenable to standard theoretical methods. We therefore work on the identification of good descriptors for e.g. the conductivity of organic solids and methods to determine them cheaply but accurately. With these we can then in a screening scheme not only identify good candidates for optimised materials but also generate vast amounts of data for further knowledge-based exploration of design space and a systematic analysis of large chemical databases pointing towards promising design strategies. Indeed, in our work we strive to extract general design criteria through a combination of advanced, relative visualisation techniques as well as mathematically sound statistical testing. Visualisation of chemical space networks thereby allows us to identify regions of chemical design space where promising examples have not yet undergone systematic improvements e.g. through synthesis of systematically modified molecules and crystals. In this regards, using our rigorously tested design rules, such as e.g. certain combinations of molecular building blocks, could be a way to fully explore design space regions so far under-utilized in experiments.