Speeding Up Scientific Discovery

In the chemical universe, a vast number of molecules and materials remain undiscovered. In this enormous and unmapped space, a self-driving lab (SDL) is like a rocket ship that uses AI as its navigational GPS to boldly chart the unknown. The combined technologies empower researchers to explore this frontier efficiently and arrive at scientific discoveries light-years faster than ever before.

Bringing together the expertise of human scientists with robotics, AI and lab automation, SDLs expedite the rate at which we can solve challenges in healthcare, energy, the environment and more. Unlike traditional manual experimentation, these labs can operate continuously and perform high-throughput experiments that run thousands of small, simultaneous tests at breakneck pace.

Using machine learning, SDLs also intelligently plan and perform follow-up experiments until they find the right “recipe” for new molecules or materials. Compared to conventional chemical and materials science, SDLs accelerate discovery up to 100 times faster.

“The biggest challenge is time to discovery,” says Milad Abolhasani, the Alcoa Professor and a University Faculty Scholar in chemical and biomolecular engineering. With Abolhasani at the helm, NC State University is a leader in SDL technology.

“SDLs allow us to search the chemical landscape faster and more intelligently. They can find a promising route, help explain why it works and then keep improving the process,” Abolhasani says.

“This is the type of human-AI-robot collaboration we need to accelerate discovery of next-generation materials and molecules.”

Racing Towards Solutions

Today’s global challenges — from renewable energy to sustainability to healthcare — require us to find solutions faster than ever before.

Discovering new materials and molecules has historically required ample time. In a conventional lab, a human scientist drives everything from designing a hypothesis, to running experiments, to analyzing the results — and then deciding what to do next. While tried-and-true, this process can take as long as decades when the number of undiscovered chemical and molecular recipes is so vast. It can also require large investments in research and development before a promising solution emerges.

“The societal cost of slow discovery is real,” Abolhasani says. “There may be better materials for clean energy, better catalysts for chemical manufacturing or better molecules for health applications, but we need tools that can find them on a much shorter timeline.”

SDLs are designed to remove that bottleneck. We can find solutions for these problems within weeks rather than years, and with hundreds of thousands of dollars in investment instead of hundreds of millions.

One example is Rainbow, a multi-robot SDL that has found the best-in-class quantum dots — tiny photoactive compounds that emit light when excited by electricity or photons. Already commercially viable, quantum dots are used in TVs and other LED devices. In solar cells, they can be used to harvest light and potentially open the door to next-generation quantum information technology.

Optimizing next-generation quantum dots is challenging because their properties depend on many variables. Historically, the process required significant space and multiple years for large-scale synthesis.

Rainbow takes up less space and can conduct and analyze up to 1,000 experiments per day without human intervention, dramatically accelerating the search for a synthetic route of high-performing quantum dots. Within a day of running experiments, Rainbow identified quantum dots considered the best-in-class. 

“That’s one day of autonomous experimentation using these self-driving labs versus about seven years of academic literature using human-centered experimentation,” says Abolhasani. 

Another example of an SDL is Fast-Cat, which transforms catalysis. Catalysts are chemical tools that help reactions happen faster and more efficiently. They’re essential for making fuels, plastics, pharmaceuticals, agricultural chemicals and other products.

Fast-Cat is designed to accelerate the discovery and optimization of catalytic reactions for chemical research and manufacturing. These reactions can be difficult and time-consuming to study because they often involve many interacting variables. 

Some catalytic reactions also require challenging operating conditions, including flammable or toxic gases, elevated temperatures, or high pressure. Automated robotic systems can help make these studies more consistent, data-rich and safer.

Fast-Cat uses AI and robotic experimentation to explore catalyst systems and reaction conditions far more efficiently than conventional testing. In mere days, the platform can generate the type of information that would otherwise require months of manual experimentation.

“In catalysis, we are often trying to optimize several goals at the same time,” Abolhasani says. “We want reactions to be faster, more selective, more efficient and more practical for industry. 

“Self-driving catalysis labs give us a way to navigate those trade-offs much more effectively.”

Improving Exponentially

SDLs do not simply run experiments faster; they also learn from every experiment. Their AI builds a model of the system, predicts which experiments are most useful and then sends those instructions back to the robotic platform.

“We are not just automating trial and error,” says Abolhasani. “We are creating laboratories that can learn, adapt and improve as they go.”

In a traditional workflow, scientists use their expertise and intuition to decide which experiments to run next. With SDLs, AI helps make that decision by analyzing all available data, predicting future outcomes and selecting the best informed next step. The result is a closed loop between experimentation and learning where the system runs autonomously as a self-correcting feedback cycle.

Each new experiment sharpens the model; each improved model makes the next experiment more targeted. 

“We use that model to predict and rank future experiments, then select the one with the highest probability of finding the material or molecule we’re looking for,” Abolhasani says. Over time, the system becomes better at navigating complex chemical and materials spaces.

Another SDL called PoLARIS is like Rainbow and Fast-Cat in speed and how it harnesses AI to constantly fine-tune its experiments, learn from them and choose which ones to carry out next.

PoLARIS specializes in finding nanoplatelets — ultra-thin semiconductor nanomaterials with potential applications in solar-energy technologies and other optoelectronic devices. 

The challenge with these nanomaterials has been that they usually contain lead, raising concerns about toxicity and long-term use. 

PoLARIS searched through billions of possible synthesis recipes, ran 120 experiments and identified brighter, lead-free nanoplatelets — in just 12 hours. While doing that, it analyzed the properties of the materials it produced and fine-tuned its parameters for each new round of experiments.

What’s more, PoLARIS also built a map of how chemistry, composition and synthesis conditions affect performance, helping researchers understand which recipes worked best and why.

“The beauty of PoLARIS is that it is both a GPS for materials discovery and a miniature materials factory,” Abolhasani says. “It can help us discover promising materials faster, use a lower amount of chemicals during discovery and optimization — and understand the chemistry behind the results.”

Leading the Future of Discovery

Far from the sci-fi scenarios of robots taking over, SDLs do not replace scientists. Instead, they change what scientists can focus on.

“SDLs are about giving scientists more powerful tools. Abolhasani says. “Human researchers define the goal, set the constraints and interpret the meaning of the results. The AI and robots help us get there faster.”

Abolhasani believes that SDLs can also expand who can participate in discovery.

“By creating autonomous and remotely accessible laboratories, we can make powerful discovery platforms available to more researchers, students, startups and institutions,” Abolhasani says.

“Ultimately, SDLs can change not only what we discover, but who gets to participate in discovery.”

SDLs can change not only what we discover, but who gets to participate in discovery.

In that quest, NC State has become a major hub for Abolhasani’s vision and currently houses the largest operational SDL ecosystem of any U.S. academic institution.

The Integrative Sciences Initiative (ISI) will further expand that potential. One of NC State’s newest scientific endeavors, ISI will harness molecular innovation, leveraging robotics, machine learning and artificial intelligence as a catalyst for discovery. Currently under construction, Woodson Hall will house ISI and become a dedicated hub for SDLs at NC State.

“Without NC State leadership’s support, I don’t think this vision would have actually been accomplished in the last decade,” says Abolhasani, who will be ISI’s director of accelerated technologies. 

And momentum continues to grow. The U.S. Department of Energy recently awarded the College of Engineering $2.9 million to develop an SDL to make fuel and chemical manufacturing processes faster and more cost-efficient. 

Like a rocket ship, SDLs enable us to venture boldly into the unknown. A powerful AI-guided engine, they are built to chart the unexplored reaches of the chemical universe and carrying scientists, students — and society at large — toward molecular and materials discovery.