A database for functional transition metal complex discovery
Abstract
Metal-organic bonding in functional transition metal complexes imparts unique optical, magnetic, and catalytic properties that are challenging to predict a priori, typically requiring experimental characterization or computational screening with quantum mechanical (QM) modeling. In the past decade, computational chemistry-derived high-throughput screening tools (i.e., with QM modeling) have accelerated the discovery of new materials as well as design principles (e.g., structure-property relationships). These same tools also have the promise of providing unified, deterministic, and curated large databases for materials discovery, similar to those available in solid state, crystalline materials or in small molecule organic chemistry. A database approach is beneficial both for discovering and repurposing materials but also for the development and testing of machine learning models. Although the design of transition metalcomplexes as functional materials and catalysts is expected to benefit from this data driven approach, no such database of properties and complexes exist. Unique aspects of open shell transition metal chemistry prevent direct application of existing tools to this region of chemical space. A computational chemistry/machine-learning-driven approach for accelerated transition metal chemical discovery has distinct challenges due to: i) low probability of feasibility computationally or synthetically of the specific complex, ii) questions of the suitability of method (e.g., density functional theory or DFT) for the region of chemical space, iii) the lack of heuristic rules for de novo complex generation, and iv) the large size of the transition metal complexes that slow DFT calculations. A scalable, computationally efficient approach to robustly characterizing both previously synthesized and feasible de novo complexes with quantified method accuracy is needed for data driven discovery in transition metal chemistry. Our long-term goal is to develop computational tools that enable the rapid computational discovery and design of functional transition metal complex materials. The overall objective of this proposal is to develop a readily augmented public database of known and de novo transition metal complexes with robust properties obtained from DFT. The proposed work broadly leverages our experience in developing an open-source software toolkit to enable highthroughput simulation (i.e., with DFT) and machine learning model development for the accelerated discovery of catalysts and materials. This proposal aims to populate a database of transition metal complex properties robustly, mitigating sources of uncertainty and calculation failure during data acquisition, and use it to address outstanding challenges in materials design through four thrusts. 1) Develop infrastructure for transition metal complex database deployment and extension. 2) Design decision logic for autonomous calculation control to maximize promise and fidelity. 3) Augment databases through systematic and generative approaches for new inorganic compound development. 4) Demonstrate databases and algorithms for multi-objective materials design. The expected outcomes of this research include i) a public database and utilities for the design of transition metal complex properties, ii) algorithms for robust database population and new molecule discovery, and iii) new strategies for multi-objective transition metal complex design.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Mar 11, 2020
- Source ID
- N000142012150
Entities
People
- Heather J. Kulik
Organizations
- Massachusetts Institute of Technology
- Office of Naval Research
- United States Navy