AI Enables Development of Carbon-Neutral, Long-Lasting Concrete

A team of researchers at the USC Viterbi School of Engineering has developed an AI model called Allegro-FM that could revolutionize concrete production by making it carbon-neutral and significantly more durable. The model, which simulates the behavior of billions of atoms simultaneously, addresses two critical challenges: concrete’s role as a major source of carbon dioxide (CO₂) emissions and its limited lifespan. Concrete production accounts for approximately 8% of global CO₂ emissions, but Allegro-FM’s simulations suggest it’s possible to recapture and store CO₂ within the material itself, effectively turning it into a carbon sink. This breakthrough could help mitigate climate change while also creating structures that last centuries, inspired by the longevity of ancient Roman concrete. The project was sparked by concerns over climate disasters, including the 2023 wildfires in Los Angeles, which prompted professors Aiichiro Nakano and Ken-Ichi Nomura to collaborate on a new approach. Alongside Priya Vashishta and Rajiv Kalia, they focused on CO₂ sequestration—capturing and storing emissions—to address both environmental and structural issues. Allegro-FM’s scalability is a key innovation: unlike traditional molecular simulations limited to thousands or millions of atoms, it can model over 4 billion atoms with 97.5% efficiency on the Aurora supercomputer. This represents a 1,000-fold increase in computational power, enabling the team to predict atomic interactions across 89 chemical elements. By analyzing how CO₂ interacts with concrete’s molecular structure, the model identifies ways to embed the gas into a “carbonate layer,” enhancing the material’s strength and durability. Concrete’s traditional vulnerabilities—such as cracking and degradation over time—have been exacerbated by modern climate extremes, including wildfires. The AI’s ability to simulate complex material behaviors, from mechanical properties to structural interfaces, could lead to breakthroughs in creating fire-resistant, long-lasting concrete. Nakano emphasized that the process would make concrete “carbon-neutral,” while Nomura noted that the carbonate layer could improve its resilience. This aligns with global efforts to reduce emissions, as the construction industry’s reliance on concrete is a major environmental issue. The research, published in The Journal of Physical Chemistry Letters and highlighted as its cover image, leverages machine learning to bypass the need for laborious quantum mechanics calculations. Instead of deriving formulas from scratch, the model uses training data to predict atomic interactions efficiently. This reduces computational demands, allowing supercomputers like Aurora to focus on more advanced tasks. The team plans to expand their work, exploring complex geometries and surfaces to refine the technology further. Industry experts highlight the significance of this development. The ability to simulate entire materials systems at scale could accelerate innovations in sustainable construction, addressing both environmental and infrastructural needs. The professors’ interdisciplinary approach—combining computer science, physics, and materials engineering—demonstrates AI’s potential to transform scientific discovery. While practical implementation remains a challenge, the model’s efficiency and accuracy mark a pivotal step toward decarbonizing construction. The research underscores AI’s role in solving real-world problems, from climate change to aging infrastructure. By bridging computational power with material science, Allegro-FM could redefine how humanity builds and sustains the future.