Intelligent chemistry for autonomous chemistry
Abstract
The ever-growing demand for functional molecules and advanced materials requires to rethink conventional experimentation and disrupt the current approaches to significantly accelerate the discovery process. State-of-the-art discovery techniques are inherently slow, capital intensive and their pace of discovery has arguably reached a plateau. Substantial advances are possible when redesigning traditional experimentation protocols. The Materials Genome Initiative (MGI) and the Mission Innovation Challenge 6 (MI6) propose to supply automated experimentation platforms with high-performance computing and artificial intelligence in a single platform for accelerated materials discovery. Such an integrated platform, referred to as a self-driving laboratory, can execute discovery experiments without human supervision in full autonomy. Artificial intelligence algorithms propose promising candidate materials for computational or experimental evaluation, and improve their recommendations based on the computed or measured properties. Nevertheless, the implementation of a self-driving laboratory poses significant challenges on developing a robust yet versatile control software, which substantially limits their massive deployment. We will develop a strongly modular open-source software package to control self-driving laboratories for unsupervised experimentation. Building upon a previously released prototype, which we named ChemOS, we will implement the required software layers indispensable for autonomous materials discovery. We will focus in the integration of heterogenous platform for molecule/material synthesis and characterization, including computational methods, and supply the proposed control software with various layers of artificial intelligence methods to fully embrace the vision of unsupervised experimentation. Notably, we plan to enable the self-driving laboratory to develop data-driven chemical intuition based on conducted experiments across multiple applications. This enables the self-driving laboratories to identify relevant trends and general concepts, which further accelerate the discovery process. Furthermore, researcher can conceptualize the identified trends to formulate scientific insights from the discovery experiments. The outcome of this project will be an intuitive yet ubiquitous tool to lower the barriers to the deployment of self-driving laboratories. The results from this work will disrupt the established approaches to scientific discovery and liberate researchers from repetitive tasks to instead raise their creativity and productivity.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Apr 25, 2019
- Source ID
- N000141912134
Entities
People
- Alán Aspuru-Guzik
Organizations
- Office of Naval Research
- United States Navy
- University of Toronto