Six teams in five labs from around the world collaborated using self-driving labs to accelerate materials discovery resulting in least 21 top-performing OSL gain candidates being discovered in a two-month optimization campaign
The Acceleration Consortium’s (AC’s) mission is to build a global network that will work together to build and use self-driving labs to accelerate materials discovery. As new research published in the journal Science describes, an AC research team, led by Dr. Felix Strieth-Kalthoff and Dr. Han Hao, demonstrated the power of global collaboration as six teams in five labs across three continents worked together to discover new molecules for organic solid-state lasers (OSLs) through a delocalized, asynchronous approach using self-driving labs that leverage AI and automation.
“This project started back in 2018 when I was thinking about what could be done with self-driving labs,” said Dr. Alán Aspuru-Guzik, Director, Acceleration Consortium. “I was thinking about fields that were nascent- fields that had very few example materials in the literature. I was asking myself, ‘where does the starting line begin when it comes to discovering materials faster?’ And that’s the cool thing about this paper; as I was designing what to do about this question, I spoke to Martin Burke, and he told me if I wanted a challenge, I should pick OSLs.”
OSLs are lasers that use an organic gain medium – the substance that will amplify the light – that is organic rather than an inorganic single crystal which usually relies on rare elements and is hard to fabricate. The first indication of electrically pumped organic lasing was reported in 2019, which made the timing perfect for Dr. Aspuru-Guzik.
The Research
The project began in earnest when the call came out from the Defense Advanced Research Projects Agency, or what it is more commonly known as, DARPA, for the Accelerated Molecular Discovery (AMD) program for teams to compete to develop “new, systematic approaches that increase the pace of discovery and optimization of high-performance molecules.” It also stipulated that the research should involve a closed loop system that would “design, discover, validate and optimize new molecules” in an iterative and active learning way. Dr. Aspuru-Guzik and the AC team knew OSLs and self-driving labs were a fit for this challenge.
A self-driving lab, or SDL, uses advanced technologies such as artificial intelligence and robotic synthesis to streamline the process of identifying novel materials – in this case, materials with exceptional lasing properties. Up until now, SDLs have usually been confined to one physical lab in one geographic location; in this research, the team employed the concept of distributed experimentation, where tasks are delocalized among different research sites, to achieve the joint goal. For this research, labs from Toronto and Vancouver in Canada, Glasgow in Scotland, Illinois in the USA, and Fukuoka in Japan were involved.
“The reality is that you’re not always going to be able get every researcher and piece of equipment you need to tackle large problems all in the same facility,” said Dr. Felix Strieth-Kalthoff, incoming Assistant Prof. of Digital Chemistry at Bergische Universität Wuppertal. “Even though we’re building a world-leading self-driving lab hub in Toronto, there are excellent labs around the world that we can connect with to create new materials- and that’s what we did here.”
By using this method, each lab was able to contribute its unique expertise and resources- which ultimately played a key role in the success of this project. This decentralized workflow, managed by a cloud-based platform, not only enhanced efficiency but also allowed for the rapid replication of experimental findings, ultimately democratizing the discovery process.
“The way the AI works is that it looks at the data and picks the best molecules to synthesize and then test,” said Dr. Strieth-Kalthoff. “But you don’t want the lab in Toronto and the lab in Glasgow synthesizing the same, or nearly the same molecule at the same time- that would be a waste of resources. So, by using the cloud, the AI could make sure each lab was making and testing different molecules ensuring a wide variety of data was fed back up into the cloud. That helped us get results even faster.”
Ultimately, this collaborative model not only maximized the collective knowledge and resources of the research teams but also set a precedent for future delocalized discovery campaigns in the field of materials science and self-driving labs.
“This is a great example how building infrastructure that can be made compatible between the labs, can lead to faster more impactful discoveries using digital chemistry.” - Dr. Lee Cronin, Regius Chair of Chemistry in the School of Chemistry at the University of Glasgow and AC scientific leadership team member.
Why OSLs matter
Just as impressive as the way the OSLs were discovered, are the OSLs themselves.
Researchers in this study focused on the automated design, synthesis, and test of gain materials for OSLs, with a particular emphasis on optimizing emission properties and performance in thin-film devices. By segmenting the candidate space into a building block framework, the researchers were able to expedite the assembly of OSL gain candidates, following a "synthesis-to-function" paradigm. This strategic approach bypassed traditional synthesis bottlenecks by providing a standardized and modular approach to creating complex molecular structures and accelerated the discovery of functional materials.
“The synthesis bottleneck has thus far precluded the substantial power of AI-guided closed-loop discovery to be realized in small molecule science. We were excited to see how a team from many different scientific backgrounds was able to collectively harness automated modular chemistry with TIDA boronates to shatter this barrier and thereby discover a top-in-class molecular function. Perhaps most importantly, I think this paper may provide a general playbook for democratizing molecular innovation, in the area of OSL research, and far beyond.” - Dr. Martin Burke, professor of Chemistry at University of Illinois Urbana-Champaign.
In the previous few decades, only 10-20 new materials have been tested as candidates for new OSLs.
“OSLs are very difficult to make, so we didn’t set ourselves an easy challenge to test our SDL,” said Dr. Han Hao, Staff Scientist at the Acceleration Consortium. “There are also over 150,000 possible experiments we could have run to find new OSL materials. The AI helped us prioritize them so that we could use the available experimental resources as efficiently as possible by predicting what the next best experiment was.”
After this experimental workflow was running, during the two-month optimization campaign, the team synthesized and tested over one thousand potential candidates. Ultimately, they were able to discover 21 top performing materials as OSL gain candidates.
“For decades, scientists have been experimenting and testing a diverse range of materials and techniques to enhance the performance of organic light-emitting diodes. My team at Center for Organic Photonics and Electronics Research (OPERA), Kyushu University, specializes in the evolution of Organic Light-Emitting Diodes to Organic Semiconductor Laser Diodes. However, global collaboration with AC partners enabled us to work in unison with other researchers than in a silo to create breakthroughs at a more rapid pace.” - Dr. Chihaya Adachi, Professor, OPERA, Kyushu University.
The significance of organic solid-state lasers lies in their potential to revolutionize various industries, from telecommunications to medical devices. These lasers offer a unique combination of flexibility, efficiency, and color tunability, making them ideal for a wide range of applications. By harnessing the power of advanced technologies and collaborative networks, these results are paving the way for the development of next-generation OSL devices with enhanced performance and functionality.
“There is so much potential for OSLs,” said Dr. Hao. “The change from the lasers we are all familiar with today to OSLs is like the change from LEDs to organic LEDs, or OLEDs. OLEDs are the screens you see wrapped around buildings these days and are a huge step from the little blinking lights LEDs used to be. It’s the same with OSLs- they let the user go from one static point of light to having much more sophisticated control. This could have applications in areas like healthcare, telecommunications and data storage. We have only scratched the surface of what they can do.”
Looking to the future
“This is an amazing example of how many teams with many disciplines can collaborate across both time and space. I think this is a perfect example for how globally a new wave of AI enabled research is going to be accomplished.” - Dr. Jason Hein, Associate Professor in the Department of Chemistry at the University of British Columbia and Acceleration Consortium Scientific Leadership Team member.
The success of this delocalized, asynchronous, closed-loop discovery model is just the first of many global collaborations in the AC’s future.
“What this paper shows is that a closed-loop approach can be delocalized, researchers can go all the way down from the molecular state down to devices and you can accelerate the discovery of materials that are very early in the process of commercialization,” said Dr. Aspuru-Guzik. “The team designed an experiment that went all the way down from molecule to device- with the final devices being made in Japan. They were scaled up in Vancouver and then transferred to Japan for characterization.”
This work is a first look at the promise SDLs bring to the accelerated materials discovery process.
“My goal for this SDL was to reduce the barrier to commercialization from 25 years down to two and a half years. We didn’t quite hit that mark in the OSL project – we’re at six years now, however, the commercialization process is underway. We’re not quite at 10 times faster yet, but that’s the whole point of the AC; it lets us do a learning curve and improve the process over time. The more SDLs we make, the time it takes to develop an SDL will be reduced. And as we get better, we will get faster and faster. You’re going to see that with the other projects we fund over time as well.”
“While there is a significant amount of data in the literature, the reporting bias, limited reproducibility, and machine readability makes it challenging to extract the right information that is essential for experiment planning with AI. SDLs, by their nature, generate high quality, reproducible, digitized data that is more efficient than all other methods, which has moved us into the era of AI-powered chemistry discovery.” – Dr Bartosz Grzybowski, Ulsan National Institute of Science and Technology
It is this learning curve that Dr. Aspuru-Guzik says will contribute to Canada remaining the world leader in Accelerated Discovery.
“We originally designed the AC’s lab to discover this laser, so really, this experiment is where it all began in Toronto,” said Dr. Aspuru-Guzik. “Just last month, we hosted Deputy Prime Minister Freeland to show her the power of SDLs in Canada and I was able to show her the organic laser molecules. It really was a full circle moment and underscores the importance of AI in Canada.”
The AC is just getting started. In 2023 it was awarded a $200-million grant from the Canada First Research Excellence Fund (CFREF) – the largest any Canadian university as ever received.
“When Deputy Prime Minister Freeland visited, while I was touring her around the lab, I was thinking, ‘wow, just beyond the wall there is a big hole in the ground where our new lab is going to be. That’s where we’re building six new self-driving lab spaces. It’s all happening here, right now.’ And so, I am proud that we’re working to make Toronto the world capital of self-driving labs. You watch, in the near future there will be many more buildings with SDLs- public, private and government in our city. And that is really exciting.”
Read the full paper in Science.
