Concrete's Carbon Capture and Structural Longevity with Supercomputing

Researchers at the University of Southern California have achieved a major milestone in material science with the development of Allegro-FM, an advanced artificial intelligence model capable of simulating the behavior of over four billion atoms at once. This capability opens the door to designing carbon-neutral and self-healing concrete, a significant advancement for the construction industry.

Traditional molecular simulations are limited to millions of atoms, but Allegro-FM operates on a scale 1,000 times greater. The model enables researchers to test various chemical compositions virtually, predicting how materials might behave under stress, heat, or environmental exposure. This scalability allows for the rapid development of concrete formulations that could capture and store carbon dioxide, turning a historically high-emission material into a potential carbon sink.

Concrete production accounts for roughly 8% of global carbon emissions. Allegro-FM’s ability to model chemical reactions at the atomic level has shown that carbon dioxide released during production can be reabsorbed within the concrete structure itself. The captured CO₂ forms a carbonate layer, improving material strength and durability.

This process could lead to concrete that not only lasts far longer than modern versions, potentially matching the resilience of ancient Roman concrete, but also actively reduces its environmental footprint. The research also suggests that the new material would perform better under extreme heat, providing improved resistance in fire-prone regions. 

The Allegro-FM model demonstrates the potential of AI-assisted materials research by integrating machine learning, physics, and high-performance computing to advance sustainable construction methods. By automating aspects of quantum mechanical calculations, the approach allows researchers to simulate material properties and chemical interactions at large scales with greater efficiency. This capability may support the development of stronger and more environmentally sustainable materials through accelerated data-driven experimentation.