Vitamin-powered batteries: how ‘self-driving labs’ are using AI to develop remarkable new materials

Aug 8, 2023

It can take many years for a new substance to get from a successful lab experiment to a solution ready to roll out across the world. But AI can be used to sidestep this bottleneck

Emily Hopkins, in mask and gloves, operates a materials acceleration platform, a large steel case wired to many computers. — *Labs can now use AI to develop radical new materials, saving time and money. Photograph: Johnny Guatto/University of Toronto*

This article was originally published on theguardian.com as part of the University of Toronto and Guardian Labs What’s Next? Ask Toronto campaign.

Led by Prof Alán Aspuru-Guzik, the University of Toronto’s Acceleration Consortium has an ambitious plan: to dramatically reduce the time it takes to go from a scientific inquiry to a ready-to-use application by employing the latest approaches in automation. The Acceleration Consortium is an Institutional Strategic Initiative at the University of Toronto that designs and builds self-driving labs, an emerging technology that uses artificial intelligence (AI) and robotics to drastically alter the timeline and costs of developing advanced materials. A discovery process that would take an average of 20 years and $100m, for example, can be streamlined to as little as one year and $1m.

This is achieved in part due to the ability of the self-driving lab to invert the usual scientific discovery process. Instead of researchers spending untold hours performing tedious trial and error experiments, self-driving lab technologies enable scientists to pre-define the desired properties of a material, leaving the lab to then work autonomously – using computational modelling to predict what molecular combinations will best suit a particular application. A robotic lab then uses these predictions to autonomously synthesise and test for the desired properties. This data is then fed back into the AI system, so that it can learn from the results to generate a new, better slate of candidates. After rounds of predictions, syntheses, and tests, a winner emerges.

A number of research initiatives at the University of Toronto are employing this “self-driving” lab approach to push forward much-needed technologies aimed at solving issues in the present, as well as getting ahead of the issues of the future. Among those, two projects working to produce a range of advanced materials promise gamechanging benefits for industry and consumer alike …

Built-to-last

“Corrosion is pervasive, and the mitigation and remediation of it costs Canadian taxpayers $38bn dollars a year,” says Prof Jason Hattrick-Simpers, who leads the Built-to-Last project, using AI to find innovative combinations for corrosion-resistant alloys. “The human cost [of corrosion] is higher still: lead problems in Flint, Michigan were related to corrosion of lead pipes there, and it is almost a certainty that corrosion contributed to the Surfside Condominium collapse [in Miami].”

The project seeks out novel “high-entropy materials”, meaning alloys made of many elements. To put this in context, consider that bronze, the first alloy ever created, is composed of just two elements: copper and tin. Given the sheer number of possible combinations of metals and elements there are potentially billions of undiscovered alloys. Testing these combinations one by one to sort those with properties that would scale up to implementation, and those that wouldn’t, would be impossible. The hours in human labour required simply do not exist. But that is not the case once the process is augmented by artificial intelligence (AI).

Corrosion is pervasive, and it costs Canadian taxpayers $38bn dollars a year. I’m interested in materials for the green economy.

I’m interested in materials for the green economy,” says Hattrick-Simpers. “One potential application for corrosion-resistant alloys is more robust electrical contacts: our phones, laptops, and even vehicles need to be recharged every night, but the electrical connections aren’t exactly kept in a pristine environment or treated gently,” he says.

“Yet we want to know that at the end of the night, when we plug the EV we just covered in salt or mud into the connector we left in our humid garage, our battery will charge at the same efficiency as when we purchased it. Unfortunately, even the process of plugging the car in degrades the electrical connections. Imagine what the electrical connections for a wind turbine must go through, worse still one that is near saltwater.”

Corrosion is pervasive, and it costs Canadian taxpayers $38bn dollars a year. I’m interested in materials for the green economy.

The Matter Lab

Researchers at the University of Toronto’s Matter Lab have collaborated with the Acceleration Consortium to hasten and deepen studies into new organic materials to store energy. Prof Yang Cao, a researcher from the Matter Lab, is interested in the creation of new battery materials, moving away from using conventional metal ions such as lithium and towards organic compounds derived from vitamins.

Such batteries could be cheaper and more environmentally friendly than those currently available. “We hope to find molecules that can store electricity generated during the day on solar farms to be used at night. This technology can help stabilise the energy grid and significantly reduce our reliance on fossil fuels,” says Cao, who leads the project.

Self-driving labs can run 24/7 and make decisions on the fly, liberating researchers from repetitive tasks and allowing them to focus on creating new molecules and materials.

Much like the hunt for new alloys, finding the right molecules is a herculean human task made possible thanks to the auxiliary power of AI. “Self-driving labs help eliminate human bias or error and make results more reliable,” says Cao. “They can also run 24/7 and make decisions on the fly, liberating researchers from repetitive tasks and allowing them to focus on creating new molecules and materials. And, thanks to self-driving labs, we can quickly validate historical results so that better predictions can be made for future molecules.”

Self-driving labs can run 24/7 and make decisions on the fly, liberating researchers from repetitive tasks and allowing them to focus on creating new molecules and materials.

Elsewhere at the Matter Lab, Dr Han Hao is researching novel organic semiconducting materials, which lie somewhere on the electrical conductivity scale between metals and materials that insulate electricity, such as glass. Organic semiconducting materials are crucial for developing better lasers, with applications in a whole host of technologies, from phone screens and solar panels, to wearable devices.

“By using self-driving labs, we were able to explore more than 200 laser molecule candidates within one month,” says Hao, noting that, without the support of self-driving labs, a predecessor in the field published fewer than 10 molecule candidates in five years.

“We were also able to improve the success rate of synthesis of complicated organic laser molecules by 150% or higher.”

A magnet for top talent

Alongside all of this advanced research, a virtuous circle is emerging: the more exciting the technological developments, the more talent and funding the University of Toronto’s Acceleration Consortium attracts, says Aspuru-Guzik. “One of the things that is really cool is that we’re growing really fast,” he says. “You can talk about the science, but basically we’re building a movement here.”

Meet the extraordinary community that’s pushing the boundaries of what’s possible. utoronto.ca/news

By Alfredo Carpineti