Materials innovations enable new technological capabilities and drive major societal advancements but typically require long and costly development cycles. Materials Genomics aims at realizing the transition to a new paradigm in materials development from a traditional “trial and error” mode to a “rationally designed experiments” mode. The key element of this highly promising approach is the availability of materials data which can be searched and analyzed in order to understand structure – property relationships and to select new candidate materials for further investigation.
In this presentation I will discuss two examples which are currently under investigation.
Covalent Organic Frameworks (COFs) have gained a lot of interest during the last years because of their potential application in several fields. The properties of the synthesized materials depend on the characteristics of the corresponding organic building blocks, which leads to nearly endless combinatorial possibilities. This complexity poses a formidable challenge for theory and simulations in order to guide the selection of precursor molecules. I will discuss our simulation efforts to predict electro-mechanical properties of two-dimensional COFs and their relation to the properties of the respective molecular building-blocks [1,2]. The results are based on more than 500 structures, most of which have been already experimentally synthesized.
Compared to the visual and auditory senses, which have long since successfully found their way into artificial intelligence applications, the sense of smell has been comparatively poorly understood. In order to change that, data analysis via artificial intelligence plays a crucial role: on the one hand, it is key for the intelligent discovery and prediction of complex chemical interactions of smelling substances. On the other hand, artificial intelligence is crucial for mimicking human perception of smells. In order to find relations between the structure of molecules and their perception, we use the concept of the physiochemical odor space which is built from 4094 molecular descriptors of 1389 odor molecules . In a second study, we use liquid-phase-exfoliated functionalized graphene sensors to discriminate odor molecules based on their sensor fingerprint. Relating the fingerprints to the physicochemical space will allow further optimization of the sensor material towards digital olfaction.