Data flow modeling, data mining and QSAR in high-throughput discovery of functional nanomaterials

Abstract

Metal oxide nanoparticles are promising materials in applications for fuel cells, gas sensors and fine chemical catalysis. Their functionality depends excessively on composition, structure as well as synthesis and processing conditions. Continuous hydrothermal flow synthesis (CHFS) reactors are an effective technology to make nanoceramics. In order to increase sample throughput of CHFS, a manual high-throughput continuous hydrothermal (HiTCH) flow synthesis process capable of formulating scores of samples per day was developed. More recently, a fully automated nanoceramics synthesis platform called RAMSI (rapid automated synthesis instrument) based on the HiTCH synthesis technology was developed. When large numbers of nanoceramics are made and formulated into appropriate libraries, automated analytical instruments can be used to allow collection of a large amount of useful data. This paper describes the information flow management system of RAMSI (as well as CHFS) and the data mining system for supporting discovery, QSAR (quantitative structure–activity relationship) modeling and DoE (design of experiments). Case studies demonstrating the use of the high-throughput data mining system are presented. These include clustering of Raman spectra, interpretation of X-ray diffraction (XRD) measurements, and QSAR model building linking XRD data and photocatalytic properties. A genetic algorithm method for DoE is also presented that can guide the experiments to search optimal XRD patterns.