Materials learn from chaos, not just order
In laboratories and in life, the materials around us are quietly learning — not from perfect repetition, but from the unpredictable pressures of ordinary existence. Researchers at the Tata Institute of Fundamental Research, working with European collaborators, have demonstrated that amorphous solids like glass can encode precise mechanical memories even when subjected to random, chaotic deformation — overturning the long-held assumption that orderly, periodic stress is required for a material to 'remember.' The discovery suggests that the capacity to learn from disorder is not a flaw in our models, but a fundamental feature of matter itself.
- For decades, the science of material memory rested on a tidy assumption: only perfectly repeated, rhythmic stress could teach a disordered solid to remember — but that assumption has now cracked.
- Using computer simulations, researchers trained amorphous materials with irregular, unpredictable deformations and found the materials retained an exact memory of the amplitude at which they were trained — a result that surprised even those who suspected it might be possible.
- The shoe analogy cuts to the heart of the disruption: real-world materials are never trained by metronomes, yet they adapt anyway, and science is only now catching up to what our feet have always known.
- There is a hard boundary to this memory — push a material past its yielding point and the record dissolves, permanent deformation erasing everything the material had quietly learned.
- The findings, published in the New Journal of Physics, reorient the field toward conditions that actually exist in the world, where chaos is not the exception but the medium through which all learning happens.
Glass and other amorphous solids — materials whose molecules arrange themselves in no repeating pattern — have long been known to carry a kind of mechanical memory. The way they respond to a new force depends on what forces shaped them before. But for decades, scientists believed this memory required perfectly regular, predictable training: the same stress, applied again and again in controlled cycles. A team at the Tata Institute of Fundamental Research in Hyderabad, collaborating with researchers in France and Germany, has now shown that this assumption was too narrow.
The researchers asked a deceptively simple question: does a material need orderly repetition to learn, or can it learn from the messiness of real experience? To find out, they used computer simulations to subject a disordered material to random, irregular deformations — unpredictable in sequence, but always kept within a fixed amplitude limit. After many such cycles, they tested the material with a single, carefully controlled readout deformation at various amplitudes.
The results were precise and unexpected. The material's particles returned to their trained configuration only when the readout amplitude matched the original training amplitude exactly. At every other amplitude, displacement remained measurable. The material had retained an exact memory of its training — not despite the randomness, but through it.
That memory, however, is not unconditional. It holds only so long as deformation stays below the material's yielding point — the threshold at which it begins to flow or break permanently. Cross that boundary, and the memory is gone, overwritten by irreversible change.
Published in the New Journal of Physics, the findings suggest that the capacity to encode mechanical memory is more robust and more universal than previously understood. Materials do not need laboratory-perfect conditions to learn. Like shoes that gradually conform to the shape of a foot through weeks of unpredictable movement, disordered solids can absorb the chaos of real-world stress and still hold something precise within it.
Glass and other amorphous materials—solids whose molecules are jumbled together with no repeating pattern—possess an unexpected ability. They can remember how they have been pushed and pulled. The way they respond to a new force depends on what forces came before. For decades, scientists have understood this memory formation only under the most controlled conditions: regular, predictable deformations applied over and over, like a metronome. But a team of researchers at the Tata Institute of Fundamental Research in Hyderabad, working with collaborators in France and Germany, has now shown something more surprising. These disordered materials can encode precise memories even when the deformations applied to them are completely random and chaotic.
The question that drove the research is deceptively simple. Do materials really need perfectly repeated stress to learn, or can they learn from the messiness of real life? Consider a new pair of shoes. They feel stiff at first, uncomfortable. But after weeks of wearing them—walking fast sometimes, slowly other times, twisting and turning unpredictably—they gradually mold to the shape of your feet. The movements are anything but regular. Yet the shoes adapt anyway. If shoes can learn from randomness, why couldn't other materials?
To test this, the researchers used computer simulations to study how a disordered material behaves when subjected to random deformation. Rather than pushing and pulling it in a neat, repeating cycle, they applied irregular, unpredictable stress—but always within a fixed limit. Think of it as randomly shoving something back and forth while ensuring it never strays beyond a certain boundary. They trained the material this way over many cycles, then performed what they called a readout: they applied a single, carefully controlled deformation and measured how the material responded.
The results were striking. After training the material with random deformations at a specific amplitude—say, 5 percent strain over 100 cycles—the researchers tested it with readout deformations of various amplitudes: 1 percent, 2 percent, 3 percent, and so on. They measured the mean squared displacement, essentially how much the particles had shifted relative to one another. The displacement dropped to zero only when the readout amplitude matched the original training amplitude of 5 percent. The material had returned to its trained state. This happened only at that precise amplitude. The material, in other words, had retained an exact memory of the deformation it had experienced during training, despite the randomness of how that deformation was applied.
But the memory has boundaries. The researchers discovered that this encoding works only when the training deformation stays below the material's yielding point—the threshold beyond which it begins to break or flow permanently. Cross that line, and the memory dissolves. The material loses its record of what came before.
These findings, published in the New Journal of Physics, shift how scientists think about memory in disordered solids. The conventional view held that materials needed perfectly regular, predictable training to learn. The new work suggests otherwise. Even under chaotic, irregular stress—the kind that dominates in the real world—these materials can still encode and retain precise mechanical memories. It is a step toward understanding not just laboratory glass, but the behavior of materials as they actually exist in the world, where randomness is not an exception but the rule.
Notable Quotes
New shoes often feel stiff and uncomfortable at first, but after continuous use they gradually adjust to the shape of the feet, even though the movements are irregular and unpredictable— Roni Chatterjee, Tata Institute of Fundamental Research
The Hearth Conversation Another angle on the story
So the material remembers the amplitude of the random deformation, but only that amplitude. Why does it matter that the deformation was random rather than regular?
Because in the real world, nothing is perfectly regular. Shoes don't experience the same footfall twice. Bridges don't vibrate in neat sine waves. If materials could only learn from perfectly controlled stress, the science would be elegant but useless. This shows they learn anyway.
But the material still needs to stay below the yielding point. Doesn't that limit how useful this is?
It does set a boundary. But that boundary exists in real materials too. A shoe can adapt to your foot, but if you jump off a building in it, the leather tears. The memory survives within the regime where the material survives.
How do the particles actually encode this memory? What changes inside the material?
That's the deeper question the study doesn't fully answer. The particles rearrange themselves into a configuration that corresponds to that training amplitude. When you apply the same amplitude again, they slip back into that configuration. When you apply a different amplitude, they settle into a different state. It's like a landscape with valleys—each amplitude has its own valley.
Could this be used for anything practical?
Potentially. If you could train a material to respond in a specific way to a specific stress, you might use it as a sensor or a filter. But that's speculation. For now, the value is in understanding that materials are smarter than we thought—they learn from chaos, not just order.