Autonomy in materials research: a case study in carbon nanotube growth

By P. Nikolaev, D. Hooper, Frederick Webber, R. Rao, Kevin T. Decker, Michael P. Krein, J. Poleski, Richard R. Barto, B. Maruyama. 2016

345 citations recorded by the source corpus.

Advances in materials are an important contributor to our technological progress, and yet the process of materials discovery and development itself is slow. Our current research process is human-centred, where human researchers design, conduct, analyse and interpret experiments, and then decide what to do next. We have built an Autonomous Research System (ARES)—an autonomous research robot capable of first-of-its-kind closed-loop iterative materials experimentation. ARES exploits advances in autonomous robotics, artificial intelligence, data sciences, and high-throughput and in situ techniques, and is able to design, execute and analyse its own experiments orders of magnitude faster than current research methods. We applied ARES to study the synthesis of single-walled carbon nanotubes, and show that it successfully learned to grow them at targeted growth rates. ARES has broad implications for the future roles of humans and autonomous research robots, and for human-machine partnering. We believe autonomous research robots like ARES constitute a disruptive advance in our ability to understand and develop complex materials at an unprecedented rate. A robot that learns how to optimize carbon nanotube production using feedback algorithms may be a valuable partner in materials discovery. A team of scientists at the Air Force Research Laboratory and Lockheed Martin Advanced Technology Laboratories have built an Autonomous Research System (ARES), a robot guided by artificial intelligence in an iterative learning loop, that is capable of designing, executing and analyzing its own experiments orders of magnitude faster than current research methods. Teaming with human researchers, ARES learned to control the growth of carbon nanotubes by converging on a set of conditions that yielded an objective growth rate. This approach creates a rapid, high-throughput approach that is broadly applicable to other materials research problems and enables researchers to pursue more challenging experimental campaigns.

Subfields: Biomedical Engineering, Electrical and Electronic Engineering, Materials Chemistry

View Paper: https://www.semanticscholar.org/paper/20a8aa0295c43b5f553155b470c643d1708a7eaf

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Biomedical Engineering

In Electrical and Electronic Engineering

In Materials Chemistry

Citations in the corpus

Accelerating the discovery of materials for clean energy in the era of smart automation (2018, background): “Several groups have demonstrated this closed-loop approach in a range of applications, including carbon nanotubes, Bose–Einstein condensates, alloys, substituted functional organic molecules, oil droplets and the search for new chemical reactions 23–26, 57, 282– 284 .”
An autonomous laboratory for the accelerated synthesis of inorganic materials (2023, neutral): “Pioneering efforts have demonstrated autonomy in several aspects of materials research, including robotic and Bayesian-driven optimization of carbon nanotube yield 3,4 , photo-voltaic performance 5 and photocatalysis activity 6 .”
Closed-loop optimization of fast-charging protocols for batteries with machine learning (2020, methodology): “Adaptive OED algorithms have been successfully applied to physical science domains, such as materials science 1,2,12–14 , chemistry 15,16 , biology 17 and drug discovery 18 , as well as to computer science domains, such as hyperparameter optimization for machine learning 19,20 .”
Accelerating Materials Development via Automation, Machine Learning, and High-Performance Computing (2018, methodology): “New tools being developed bymachine-learning specialists invite the possibility of modifying hardware design to take advantage of machine-learning tools, rather than the other way around.(45)”