Autonomous efficient experiment design for materials discovery with Bayesian model averaging

Abstract

The accelerated exploration of the materials space in order to identify configurations with optimal properties is an ongoing challenge. Current paradigms are typically centered around the idea of performing this exploration through high-throughput experimentation/computation. Such approaches, however, do not account for-the always present-constraints in resources available. Recently, this problem has been addressed by framing materials discovery as an optimal experiment design. This work augments earlier efforts by putting forward a framework that efficiently explores the materials design space not only accounting for resource constraints but also incorporating the notion of model uncertainty. The resulting approach combines Bayesian Model Averaging within Bayesian Optimization in order to realize a system capable of autonomously and adaptively learning not only the most promising regions in the materials space but also the models that most efficiently guide such exploration. The framework is demonstrated by efficiently exploring the MAX ternary carbide/nitride space through Density Functional Theory (DFT) calculations.

Machine-learned impurity level prediction for semiconductors: the example of Cd-based chalcogenides

“…Today, machine learning (ML) has become an integral component of materials design 24 . Researchers have extracted models and design rules from materials data to drive the accelerated discovery of NiTi alloys for thermal hysteresis 25 , design of polymer dielectrics for improved energy storage in capacitors 26,27 , synthesis of new classes of compounds 28,29 , identification of new and improved catalysts 30,31 , and the design of experiments in a smart and 'adaptive' fashion 32 . ML-based design of materials usually begins with the generation of sufficient data for candidate materials in terms of a property P, and the conversion of all materials in the chemical space into a unique numerical representation X, referred to as descriptors, feature vectors, or fingerprints.…”

Rocketsled: a software library for optimizing high-throughput computational searches

“…A previous study identified stable structures for stage-I and stage-II lithium-graphite intercalation compounds using 4%-6% of the calculations required to explore the entire search space, a combinatorial problem containing more than 16.7 million total possibilities [51]. Others have found adaptive design can accelerate searches-particularly those where exhaustive computation would be problematic-for a variety of applications including novel crystalline interfaces [56], elastic properties [52], high-pressure Mg-silicate phases [47], ultra-low thermal conductivity structures [53], inorganic/organic molecular interfaces [54], layered materials [50], stable carbide/nitrides [57], piezoelectrics [34], Poisson-Schrödinger simulations of LEDs [48], and stable crystal structures [55] for Y 2 Co 17 . In this study, we demonstrated two additional proofs-of-concept searching for superhard materials and photocatalysts.…”

“…The growing toolkit for running high-throughput calculations [39][40][41][42], the potentially immense search space (e.g. often more than 10 10 viable compounds [43]), and the growing number of studies using adaptive design in the materials domain to guide both simulations [34,[44][45][46][47][48][49][50][51][52][53][54][55][56][57][58][59] as well as experiments [44,[60][61][62][63] makes computational materials design an exciting field to explore with Rocketsled. In the two following case studies, we demonstrate the applicability of Rocketsled to adaptive design for materials discovery.…”

Multi-objective Bayesian materials discovery: Application on the discovery of precipitation strengthened NiTi shape memory alloys through micromechanical modeling

“…Very recently, global optimization frameworks--including gradient based [9], [10], direct search methods [11], [12] and Bayesian optimization approaches [13]- [19] have emerged to guide the efficient exploration of the material space [4], [20]- [29]. Unlike high throughput-based approaches to materials discovery, Bayesian Optimization (BO) enables the (global) optimization of materials while minimizing the number of evaluations of the materials space.…”