Progress and prospects for accelerating materials science with automated and autonomous workflows

Abstract

Integrating automation with artificial intelligence will enable scientists to spend more time identifying important problems and communicating critical insights, accelerating discovery and development of materials for emerging and future technologies.

Benchmarking the acceleration of materials discovery by sequential learning

“…That is, poor choices for ML model and/or acquisition function for a given experiment budget or research object can lead to substantially worse performance than random sample selection, a critical lesson that illustrates the importance of comprehensive workow design in the context of specic research objectives. 2 We also note that these accelerations are with respect to random sample selection, which is not a commonly applied experiment design strategy. Development of more illustrative baselines is an important area for future research, although we note that comparing to traditional human-rational catalyst selection will also motivate further development of the acceleration metrics, as the relative costs of the experiment design mechanisms should be considered.…”

“…Accelerating materials discovery is of utmost importance for realization of several emergent technologies, particularly to combat climate change through the adoption of zero or negative emission technologies such as hydrogen driven cars and other means of clean chemical energy generation, storage and utilization. One method of accelerating materials research is through integration of automated experiments [1][2][3][4] that are guided by arti-cial intelligence (AI). 5,6 Specically, AI sampling strategies 7,8 hold great promise for resource-constrained activities such as materials research due to their potential to minimize the number of experiments necessary for achieving a desired objective.…”

Benchmarking the Acceleration of Materials Discovery by Sequential Learning

“…That is, poor choices for ML model and/or acquisition function for a given experiment budget or research object can lead to substantially worse performance than random sample selection, a critical lesson that illustrates the importance of comprehensive workflow design in the context of specific research objectives. 2 The design of an appropriate SL algorithm must be performed in the context of the research task at hand, which is consistent with general best practices in design of experiments. The any, all and model active learning metrics of the present work are designed to span a range of common research goals from the most applied to the more fundamental.…”

“…Accelerating materials discovery is of utmost importance for realization of several emergent technologies, particularly to combat climate change through the adoption of zero or negative emission technologies such as hydrogen driven cars and other means of clean chemical energy generation, storage and utilization. One method of accelerating materials research is through integration of automated experiments [1][2][3][4] that are guided by artificial intelligence (AI) 5,6 . Specifically, AI sampling strategies 7,8 hold great promise for resource-constrained activities such as materials research due to their potential to minimize the number of experiments necessary for achieving a desired objective.…”

Scanning electrochemical microscopy and its potential for studying biofilms and antimicrobial coatings

“…Screening approaches for photocatalyst and electrocatalyst have been demonstrated via SECM, although automated workflows and advanced robotics as demonstrated in i.e. materials science [140] have not yet been demonstrated for SECM.…”