Researchers from the University of Toronto’s Faculty of Applied Science & Engineering and Fujitsu have applied quantum-inspired computing to find the promising, previously unexplored chemical family of Ru-Cr-Mn-Sb-O₂ as acidic oxygen evolution reaction catalysts for hydrogen production.

The best catalyst shows a mass activity eight times higher than state-of-the-art RuO₂ and maintains performance for 180 h. A paper on their work appears in the journal Matter.

Choubisa et al.

Scaling up the production of what we call green hydrogen is a priority for researchers around the world because it offers a carbon-free way to store electricity from any source. This work provides proof-of-concept for a new approach to overcoming one of the key remaining challenges, which is the lack of highly active catalyst materials to speed up the critical reactions.

—Ted Sargent, senior author

Nearly all commercial hydrogen is produced from natural gas. The process produces carbon dioxide as a byproduct; if the CO₂ is vented to the atmosphere, the product is known as grey hydrogen, but if the CO₂ is captured and stored, it is called blue hydrogen. Green hydrogen is a carbon-free method that uses an electrolyzer to split water into hydrogen and oxygen gas. The low efficiency of available electrolyzers means that most of the energy in the water-splitting step is wasted as heat, rather than being captured in the hydrogen.

Researchers around the world are striving to find better catalyst materials that can improve this efficiency. Because each potential catalyst material can be made of several different chemical elements, combined in a variety of ways, the number of possible permutations quickly becomes overwhelming.

One way to do it is by human intuition, by researching what materials other groups have made and trying something similar, but that’s pretty slow. Another way is to use a computer model to simulate the chemical properties of all the potential materials we might try, starting from first principles. But in this case, the calculations get really complex, and the computational power needed to run the model becomes enormous.

—Jehad Abed, co-lead author

To find a way through, the team turned to the emerging field of quantum-inspired computing. They made use of the Digital Annealer, a tool that was created as the result of a long-standing collaboration between U of T Engineering and Fujitsu Research. This collaboration has also resulted in the creation of the Fujitsu Co-Creation Research Laboratory at the University of Toronto.

Digital Annealer (DA) is a computer architecture developed to solve large-scale combinatorial optimization problems rapidly using CMOS digital technology. DA is unique in that it uses a digital circuit design inspired by quantum phenomena and can solve problems that are very difficult and time-consuming or even impossible for classical computers to address.

Digital Annealer is inspired by quantum mechanics, but unlike quantum computers, does not require cryogenic temperatures. DA makes use of a method called annealing—named after the annealing process using in metallurgy. During this procedure, metal is heated to a high temperature before the structure stabilizes as it is slowly cooled to a lower energy, more stable state.

Using the analogy of placing blocks in a box, in the classical computational approach, the blocks are placed in sequence. If a solution is not found, the process is restarted and repeated until a solution is found. With the annealing approach, the blocks are placed randomly and the entire system is “shaken.” As the shaking is gradually reduced, the system becomes more stable as the shapes quickly fit together.

DA is designed to solve fully connected quadratic unconstrained binary optimization (QUBO) problems and is implemented on CMOS hardware. The second-generation Digital Annealer expands the scale of problems that can be solved from the 1,024 bits of the first generation, launched in May 2018, to 8,192 bits and an increase in computational precision.

This leads to substantial gains in precision and performance for enhanced problem-solving and new applications, expanding by a factor of one hundred the complexity that the second-generation Digital Annealer can tackle now. Its algorithm is based on simulated annealing, but also takes advantage of massive parallelization enabled by the custom application-specific CMOS hardware.

The Digital Annealer is a hybrid of unique hardware and software designed to be highly efficient at solving combinatorial optimization problems. These problems include finding the most efficient route between multiple locations across a transportation network, or selecting a set of stocks to make up a balanced portfolio. Searching through different combinations of chemical elements to a find a catalyst with desired properties is another example, and it was a perfect challenge for our Digital Annealer to address.

—Hidetoshi Matsumura, senior researcher at Fujitsu Consulting (Canada)

In the paper, the researchers used a technique called cluster expansion to analyze an enormous number of potential catalyst material designs—they estimate the total as a number on the order of hundreds of quadrillions. For perspective, one quadrillion is approximately the number of seconds that would pass by in 32 million years.

Quantum annealers and similar quantum-inspired optimizers have the potential to provide accelerated computation for certain combinatorial optimization challenges. However, they have not been exploited for materials discovery because of the absence of compatible optimization mapping methods. Here, by combining cluster expansion with a quantum-inspired superposition technique, we lever quantum annealers in chemical space exploration for the first time. This approach enables us to accelerate the search of materials with desirable properties 10–50 times faster than genetic algorithms and bayesian optimizations, with a significant improvement in ground state prediction accuracy.

—Choubisa et al.

The results pointed toward a promising family of materials composed of ruthenium, chromium, manganese, antimony and oxygen, which had not been previously explored by other research groups.

The team synthesized several of these compounds and found that the best of them demonstrated a mass activity that was approximately eight times higher than some of the best catalysts currently available.

The new catalyst has other advantages too: it operates well in acidic conditions, which is a requirement of state-of-the-art electrolyzer designs. Currently, these electrolyzers depend on catalysts made largely of iridium, which is a rare element that is costly to obtain. In comparison, ruthenium, the main component of the new catalyst, is more abundant and has a lower market price.

The team aims to optimize further the stability of the new catalyst before it can be tested in an electrolyzer. Still, the latest work serves as a demonstration of the effectiveness of the new approach to searching chemical space.

I think what’s exciting about this project is that it shows how you can solve really complex and important problems by combining expertise from different fields. For a long time, materials scientists have been looking for these more efficient catalysts, and computational scientists have been designing more efficient algorithms, but the two efforts have been disconnected. When we brought them together, we were able to find a promising solution very quickly. I think there are a lot more useful discoveries to be made this way.

—Hitarth Choubisa, co-lead author

Resources

• Hitarth Choubisa, Jehad Abed, Douglas Mendoza, Hidetoshi Matsumura, Masahiko Sugimura, Zhenpeng Yao, Ziyun Wang, Brandon R. Sutherland, Alán Aspuru-Guzik, Edward H. Sargent (2022) “Accelerated chemical space search using a quantum-inspired cluster expansion approach,” Matter doi: 10.1016/j.matt.2022.11.031