How Self-Driving Labs in Singapore Can Solve Materials Challenges

Ask anyone on the street what advanced technologies they desire today, and they’ll often identify materials-related challenges: clean energy, unbreakable cell phone screens, long-life batteries, biodegradable plastics, rapid transit, and more.

The reason we don’t have all these solutions today, is that it typically takes many years — 20 is a good number — to develop a new material. Wouldn’t it be great if we could innovate materials in a fraction of the time and cost?

A combination of maturing technologies promises to do just that. “Automation” allows us to execute experiments faster and with greater precision. The large amount of data generated is digested with the aid of “machine learning.” And “computer simulations” perform experiments inside a chip instead of a lab, reducing the amount of work we need to do.

Combining all three enables “self-driving laboratories,” automated molecule makers that create and test thousands of new materials a day. Compare this to the traditional way of making materials, which is about one experiment per week, and you can quickly see the potential.

Researchers have recently been building and testing the first self-driving laboratories or SDLs for materials, some of them right here in Singapore. Right now, these SDLs are highly specialised for certain materials (one kind of plastic, metal, or peptide), and are able to access only a limited set of available elements and molecules. As these self-driving labs become more versatile, they promise to access a wider range of matter.

Researchers at SMART or the Singapore-MIT Alliance for Research and Technology, A*STAR, NUS, NTU, and MIT have collaborated to demonstrate the local value proposition of SDLs. In one collaboration, a machine-learning algorithm guided researchers to make 17 new polymers, which spontaneously assemble into nanoparticles at user-defined temperatures — a feature of growing importance in both the biomedical and petrochemical industries. The researchers collected available data in the lab and literature of polymer structure and nanoparticle formation; trained an algorithm to recognise patterns in the data; and then used the algorithm to predict new materials with custom properties. Then in a few weeks, they made 17 new polymers, which could have taken months without the algorithm. In another collaboration, a local SDL produced another kind of nanoparticle with user-defined colours, guided by a machine-learning optimisation algorithm. Nanoparticle formation is a first step toward designing more complex material properties for energy, biomedical, and chemical applications encountered in daily life. As researchers continue to build these capabilities, they hope to custom-design improved performance, cost, manufacturability, sustainability, and more, whilst also exploring the ethics and preventing misuse of this technology.

This project was successful because it brought together two communities: chemists who understand polymers, and data scientists who understand machine-learning algorithms. It was challenging, like two different nationalities coming together and trying to speak the same language. But they managed to solve the problem together, and built trust in this new way of doing science. The enthusiasm for this approach was evident at recent international conferences, including the June 2019 International Conference on Materials for Advanced Technologies in Singapore.

Envision a future when we can build new materials closer to the speed of our imagination. This would enable engineers to more quickly address societal and environmental problems. It would also make possible mass customised manufacturing, whereby materials in your daily life such as your phone and medical devices could be better customised for your needs.

The challenge on the horizon now is to make these SDLs more versatile, able to access a wider range of molecular building blocks. Another challenge is to get different scientific communities working together from different fields, and to change prevailing mindsets. We need to balance preventing technology misuse, whilst allowing innovators to easily test new ideas. Ultimately, all these tasks rely on improved education and access to fundamentals. For example, SMART is collaborating with A*STAR, NUS, NTU, and MIT to create an “Accelerated Materials Development” YouTube channel where anyone with a computer can learn from our example datasets and code, and upskill. The ultimate goal is to inspire more people to work alongside us and improve our human condition.