How AI helped Japanese researchers solve a decades-old hydrogel design problem

Designing a hydrogel that is simultaneously strong, highly adhesive, flexible and capable of self-healing has long challenged materials scientists, because improving one property has typically come at the cost of another. Researchers at Osaka University and collaborating institutions in Japan say they have now solved that problem with the help of artificial intelligence.

The team used machine learning, data mining and high-throughput laboratory experiments to optimise hydrogel composition, allowing an AI model to analyse large datasets and predict which molecular structures would produce the best overall performance — rather than manually testing thousands of chemical combinations. The predicted hydrogels were then synthesised and experimentally validated.

The study, published in Nature as ‘Data-driven de novo design of super-adhesive hydrogels,’ states that the team has established “an AI-driven materials discovery framework for multifunctional hydrogels.” The resulting material achieved underwater adhesive strength exceeding 1 MPa alongside high mechanical toughness, elasticity and rapid self-healing — a combination conventional hydrogels have rarely managed, particularly underwater.

Researchers see potential applications across medical adhesives, wearable electronics, soft robotics and underwater engineering, positioning the discovery as an example of how AI is beginning to accelerate materials science research that once took years of manual experimentation.