Fig. 1: Closed-loop autonomous materials exploration and optimization (CAMEO).
From: On-the-fly closed-loop materials discovery via Bayesian active learning
The autonomous cycle begins with loading data from databases including composition data for the materials on the composition spread and computed materials data from the AFLOW.org41 density functional theory database. The collected data is then used to begin analysis of the data using physics-informed Bayesian machine learning. This process extends knowledge of structure and functional property from materials with data to those without, predicting their estimated structure and functional property, along with prediction uncertainty. Physics-informed active learning is then used to identify the most informative next material to study to achieve user-defined objectives. For this work, active learning can select the next sample to characterize through autonomous control of the high-throughput X-ray diffraction system at a synchrotron beamline or it can (optionally) request specific input from the human-in-the-loop. Future implementations will include autonomous materials synthesis and simulation. The data collected from measurements and from human input are added to the database and used for the next autonomous loop. For more information, see “Methods”.
