ESMI Innovation
ESMI Innovation
The ESMI project at PNNL is pioneering new R&D approaches and developing new technologies to transform the field of materials science and accelerate development of a new generation of battery materials and chemistries. These innovations are included below.
High-throughput Experimentation
ESMI team members are using high-throughput experimentation (HTE) equipment that enables them to accomplish in a day what used to take weeks or months. Part robot, part workstation, part intelligent database, PNNL’s HTE laboratory enables automated combinatorial materials synthesis, high-throughput screening, and optimizations for large-scale data generation that significantly accelerate material discovery and predictive material design through advanced data analytics. These HTE capabilities enable the ESMI team to perform hundreds of experiments per day (compared to only a handful before) with less labor, all while maintaining a bird’s-eye view of results by using physics-informed data sets to improve experimental design and eliminate errors.
PNNL's High-throughput Experimentation Equipment enables ESMI materials scientists to accomplish in a day what used to take weeks or months.
Machine Learning and Database Innovations
ESMI team members are developing physics-informed database structure that integrates data to organize and understand highly complex data. The team is developing a database that maps solubility and other key parameters, from material structure and properties all the way to cell pouch performance. This will contain solubility data for about 10,000 experimental aqueous candidate molecules in a multi-modal database.
Physics-informed Modeling
The team is using multi-scale modeling, combined with physics-informed machine-learning techniques to establish physics-based models for prototype cells and cell stack, the accuracy of which will be incrementally improved through iterative validation using the test data from the corresponding prototype cells. Physics-based models of increasing complexity with scale will be developed first. These models cover the intermediate scales between materials discovery and kW-systems and are essential to facilitate information flow across scales.