Training and memory in a randomly driven fractal gel

Materials that look solid but are actually messy on the inside, like a crumpled piece of paper or a jar of tightly packed sand, often behave in ways that seem to defy simple logic. When you squeeze or stretch these disordered substances, their tiny internal parts shift and rearrange. Usually, scientists expect this rearrangement to be chaotic and permanent, leaving the material in a new, confused state after every push. However, researchers have recently discovered that some of these messy materials can "learn." If you push them back and forth with a specific rhythm and strength, they can train themselves to return to their original shape every time you stop, effectively remembering the exact force you used. This phenomenon, known as training, has been observed in systems like packed grains and crumpled sheets, but it was unclear whether it could happen in soft, jelly-like substances called gels, or if it could occur when the pushing force was random and unpredictable rather than a steady, repeating rhythm.

A team of physicists set out to answer these questions by studying a fractal gel, a substance made of tiny silica particles mixed into mineral oil to form a delicate, sponge-like network. To watch what happened inside this invisible world, they used a powerful technique called X-ray photon correlation spectroscopy. Imagine shining a very focused beam of X-rays through the gel and watching how the light scatters off the moving particles. By analyzing the patterns of this scattered light, the researchers could see the exact arrangement of the particles and track how they moved when the gel was squeezed. They placed the gel between two plates and used a precise machine to slide the bottom plate back and forth, applying shear strain, which is a type of force that tries to slide layers of the material past one another.

The researchers performed two different types of experiments to see how the gel would react. In the first, they applied a predictable, rhythmic push and pull, stretching the gel to a specific limit and then returning it to zero, over and over again. In the second, they used a random approach, making the gel move in a chaotic pattern that wandered up and down but never exceeded the same maximum limits as the first experiment. They found that both methods worked. Whether the gel was pushed in a steady rhythm or a random walk, the internal structure eventually learned to become reversible. This means that after a certain number of cycles, the particles would return to the exact same positions they started in at the end of each stretch, even though the path they took to get there might have been different. The gel had effectively trained itself to remember the boundaries of the force applied to it.

However, the way the gel learned depended on how it was pushed. When the force was steady and rhythmic, the training happened quickly and smoothly. The gel became reversible cycle by cycle, getting better at returning to its starting point with each repetition. When the force was random, the process was much slower and more erratic. The gel's ability to return to its original state would improve, then sometimes dip, before improving again, but the overall trend was still toward greater reversibility. This showed that while the randomness made the learning process messier, it did not prevent the gel from learning. The researchers also measured how far the particles moved irreversibly during these cycles. They found that as the gel trained, the distance the particles moved permanently decreased, eventually becoming so small that the material appeared to bounce back perfectly.

To prove that the gel was truly remembering the specific strength of the force used during training, the team used a "readout" test. They took gels that had been trained with different amounts of force and then subjected them to a new series of stretches that gradually increased in strength. For gels that had not been trained, the correlation between their starting and ending positions dropped immediately as soon as the force became strong enough to move the particles. But for the trained gels, something interesting happened. As long as the new force stayed below the level they had been trained on, the gel remained remarkably stable and reversible. The moment the force reached the exact limit of their training, the stability collapsed, and the correlation dropped sharply. This sharp drop acted like a signature, confirming that the gel had stored a memory of the specific amplitude of the strain it had experienced.

The study revealed that this memory effect is not limited to orderly, rhythmic pushing; it can also be induced by random, chaotic forces, provided those forces stay within certain bounds. While the gel's internal structure did not show obvious signs of changing its overall network or forming new connections, the way it responded to stress suggested that the particles were reorganizing locally to accommodate the force. The researchers noted that for very small forces, the gel was already elastic and reversible, showing no need for training. But once the force reached a certain threshold, the training process became necessary to restore reversibility. This suggests that the gel's ability to remember is a dynamic process that emerges only when the material is pushed hard enough to cause internal shifts, yet not so hard that it breaks down completely. The findings open up new questions about how disordered materials retain information about their mechanical history, hinting that even the most chaotic systems might possess a hidden capacity to learn and adapt.