GENESIS: A Next Generation Synthesis Center

Project Objectives: GENESIS combines emerging experimental and computational approaches to construct a framework capable of systematizing materials synthesis, and a platform that will allow machine assisted and, eventually, machine initiated synthesis of functional materials.

Project Description:The 'cook-and-look' technique remains the mainstay of materials research and development. Researchers seal chemical reactants in a vessel, 'cook' to initiate a reaction and after some time 'look' at the recovered products to determine if they are in a form required to be useful. The formidable task of repeated synthesis-recovery-characterization as a strategy for materials discovery can be accelerated using combinations of in situ techniques and data science.

Paths taken by a reaction are opaque to researchers without the use of in situ techniques, which allow us to 'look inside' reactors using penetrating beams at DOE User Facilities. During reactions transient species form, grow, and transform to other species. These processes are critical to the final product and all are invisible without the ability to follow them in situ in real time. Discovery of these pathways alone is insufficient. The science of synthesis lies in not only mapping the reaction pathway but also in understanding at the atomic level the underlying operational mechanisms that occur all the way along the pathway, determining at the speed of the reaction which atoms are doing what. That requires computational methods that identify what phases are forming when. In order to speed the development of transformational materials the challenge of tracking the evolution of phases along the reaction pathways must be met.

While computational methods for the prediction of atomic arrangements likely to lead to desirable properties have been developed, little headway has been made in the complementary prediction of reaction pathways. Theory struggles to describe the non-equilibrium diffusional processes governing real syntheses due to the incredibly complex, dynamically-evolving, defect-driven, and multi-dimensional parameter space associated with the synthesis mechanisms of real materials. Can we develop a data-driven approach to design synthesis pathways ab initio? Can we build a framework that moves us forward from think-cook-look-repeat strategies to machine-suggested synthesis pathways, in a manner analogous to now-mature computational structure- and properties-prediction methods?

The tremendous amounts of minable data on reaction pathways produced by in situ experiments, not only by GENESIS but by the community of scientist worldwide, can be extracted using natural language searches. In natural language and artificial intelligence (AI) processing applications that are increasingly popular in industry, features computed from data are linked to a label which describes actual or desired output. GENESIS will capture all synthesis parameters (which will become features), including temperature history, process gas composition and flow rates, concentrations of all liquid species, and the composition, into structured-data databases. The structure, defects, and morphological parameters (size, crystallinity) of solid phases are the outcomes (labels) that also will be captured.

Potential Impact of the Project: GENESIS builds the tools and understanding to discover new materials more rapidly. In the end, to be useful a material must be made. Computational approaches can now identify a library of potential materials but synthesis and physical realization of these new predicted materials remains a critical limit. Control over synthesis pathways and products is an essential requirement for materials design, without which concepts developed in computers cannot be brought into reality. The GENESIS approach eliminates this bottleneck by exploiting advanced real time diagnostics coupled with data science tools, thereby accelerating the synthesis of new materials by 100-1000 times. The long-term goal is for adaptive control - the ability to 'steer' reactions in real time using AI coupled to in situ as a guide.