Combinatorial Discovery of Heterogeneous Catalysts Utilizing Emission Spectroscopy and Advanced Machine Learning

In this SBIR Phase I project, AccuStrata, Inc. will work with researchers at the University of South Carolina to create a high-throughput flame spray pyrolysis (FSP) system for the rapid, combinatorial discovery and optimization of heterogeneous catalysts. The proposed system will comprise three key features: (1) Optimization and operation of the high-throughput FSP system that is uniquely capable of creating Pd-CeO2-MnOx solid solution catalysts; (2) Integration of an in-situ laser induced breakdown spectroscopy system capable of monitoring the particle synthesis in real time; and (3) Development of an advanced machine learning algorithm that will utilize process parameters (flow rates, burner geometry, temperatures, precursor concentrations, etc.), in-situ laser induced breakdown spectroscopy measurements and post-synthesis characterization datato discover critical signatures within the emission spectra that can help narrow material search space and speed up materials discovery. The proposed system will integrate these features to provide a holistic, commercialize solution for combinatorial discovery of heterogeneous catalysts. A system with these unique capabilities will be of great interest to laboratories both at the university and industry levels. The SBIR proposal team will validate the approach by developing stable solid solution catalysts for natural gas combustion engines. Flame spray pyrolysis can be used to create catalysts from a wide array of materials. In addition, nanoparticle synthesis through FSP allows for precise control over crystallite size, crystalline phase, degree of aggregation and agglomeration, surface area and porosity, which makes it an ideal technique for heterogeneous catalysis discovery. While the technique provides incredible flexibility, complete characterization of the nanoparticles quality post-synthesis is often a slow process that hinders the discovery process. Laser induced breakdown spectroscopy is a processing in-situ technology for monitoring FSP but is an especially difficult characterization method due to the various emission lines originating from the fuel, precursors and by-products. The challenge of correlating the emission spectra to nanoparticle properties may be resolved using advance machine learning algorithms that can correlate the spectral response to the resulting nanoparticle properties as well as the processing parameters. Once the algorithm is trained, it can be used with real-time emission data as a prescreening so that only the most “promising” candidates (as determined by the algorithm) will be flagged for further study. The goal of this SBIR phase I work will be to provide a proof-of-concept for the metrology and algorithms approach. Once the teams have validated the approach, a completely integrated system will be developed and commercialized in a subsequent phase II.