Yale Team Builds a Chip That Remembers Past Flows Using Only Moving Fluid
Tiny flexible fibers in microchannels flip between two stable states, so a network of them keeps a memory of what has flowed through, with no electronics at all.
Researchers at Yale University have shown that a network of flowing fluid can store information about its own past, the way a computer memory chip does. The work was led by graduate student Abhineet Singh Rajput in the lab of Professor Amir Pahlavan in Yale's Department of Mechanical Engineering. It was published in Physical Review X under the title "Fluidic Hysterons and Memory in Flow Networks," and Phys.org covered it on October 14 from a Yale press release.
The team built a small chip with several microchannels, each about half the depth of a human hair. Inside each channel they placed a flexible microfiber that partly blocks the path of the fluid. Pressure from the flow pushes the fiber between two stable positions, and each position changes how easily fluid passes through that channel. The two states can stand for a 0 and a 1, like a bit in computer memory.
The key is that a fiber's current state depends on the flows it has experienced. In physics this behavior, where the present state depends on history, is called hysteresis, and the elements that show it are known as hysterons. Because the fibers switch passively as the imposed flow rises and falls, the chip records its own flow history with no electronics, sensors or active valves.
The researchers also coupled the fibers so they influence one another. That turns memory into a property of the whole network rather than of any single fiber. Rajput put it this way: "While conventional microfluidics rely on active elements, this system uses microfibers to passively store information." He said the interacting fibers can produce "complex, systemwide memory."
The most immediate goal is soft robotics in places where electronics fail, such as environments with strong electromagnetic interference. In such a design, flow rates and pressures would take the place of electrical current, the microfibers would act as valves, and the channels would act as wires. Pahlavan said the aim is to "design a fully fluidic controller that can be integrated into a soft robot" and serve as the robot's brain.
The team also sees a biological angle. They suggest the work could shed light on memory and learning in fungi, plant roots and slime mold, organisms with no neurons that still adapt to what they have experienced. Slime mold, a single-celled organism, is famous for finding the shortest path through a maze.
The demonstration is of memory, not computation. The news summary does not report how many fibers were tested or how fast they switch, and whether such networks can be trained to carry out useful tasks is a longer-term question. The soft-robot controller is a stated goal, not a built device.
