Combinatorial synthesis and high-throughput characterization have become powerful tools to accelerate the discovery and design of novel materials.Correctly extracting information about the constituent phases and gaini...Combinatorial synthesis and high-throughput characterization have become powerful tools to accelerate the discovery and design of novel materials.Correctly extracting information about the constituent phases and gaining materials insight from high-throughput X-ray diffraction data of combinatorial libraries is a crucial step in establishing the composition–structure–property relationship.Basic information includes the number,identity,and fraction of present phases in all the samples,while advanced information includes the lattice change,texture information,solid solution behavior,etc.Encoding domain-specific knowledge,such as crystallography,X-ray diffraction,thermodynamics,kinetics,and solid-state chemistry,into automated algorithms is crucial for the development of automated phase mapping algorithms.In this study,we present an unsupervised optimization-based solver to tackle the phase mapping challenge in high-throughput X-ray diffraction datasets.Besides leveraging robust fitting abilities of neural-network optimization algorithms,we integrated various material information,including first-principles calculated thermodynamic data,crystallography,X-ray diffraction,and texture into our automated solver.Our approach exhibits robust performance across multiple experimental datasets.We emphasize the importance of correctly integrating material information for automated solvers,contributing to the development of future automated characterization tools.展开更多
基金the funding support from the Research Center for Industries of the Future(RCIF)from Westlake University,and computation resources from the High-Performance Computing Center(HPC)at Westlake UniversityS.G.and C.W.acknowledge the funding support from US AFOSR Multidisciplinary University Research Initiative(MURI)under award FA9550-18-1-0136T.L.was supported by funding from the Toyota Research Institute.The DFT calculations were performed on Quest computing facility at Northwestern University,which is jointly supported by the Office of the Provost,the Office for Research,and Northwestern University Information Technology.
文摘Combinatorial synthesis and high-throughput characterization have become powerful tools to accelerate the discovery and design of novel materials.Correctly extracting information about the constituent phases and gaining materials insight from high-throughput X-ray diffraction data of combinatorial libraries is a crucial step in establishing the composition–structure–property relationship.Basic information includes the number,identity,and fraction of present phases in all the samples,while advanced information includes the lattice change,texture information,solid solution behavior,etc.Encoding domain-specific knowledge,such as crystallography,X-ray diffraction,thermodynamics,kinetics,and solid-state chemistry,into automated algorithms is crucial for the development of automated phase mapping algorithms.In this study,we present an unsupervised optimization-based solver to tackle the phase mapping challenge in high-throughput X-ray diffraction datasets.Besides leveraging robust fitting abilities of neural-network optimization algorithms,we integrated various material information,including first-principles calculated thermodynamic data,crystallography,X-ray diffraction,and texture into our automated solver.Our approach exhibits robust performance across multiple experimental datasets.We emphasize the importance of correctly integrating material information for automated solvers,contributing to the development of future automated characterization tools.
