Sensing-storage-computing Integration of Multiterminal ReS2 Neuromorphic Transistors with Sequence-dependent Responses for Dual-mode Authentication Systems

Modern computers are incredibly fast at crunching numbers, but they struggle with the fluid, time-sensitive tasks that living brains handle with ease. In a standard computer, the part that senses the world, the part that remembers it, and the part that thinks about it are physically separated. Information must travel back and forth between these distinct units, creating a bottleneck that slows things down and wastes energy. To build machines that can react to the world in real time—like a self-driving car navigating a sudden change in traffic or a security system that recognizes a specific pattern of behavior—engineers need a new kind of device. They need something that can sense, store, and compute all at once, mimicking the way biological neurons integrate information. The challenge has been creating a single electronic component that can not only remember the past but also understand the order in which events happen, a crucial skill for recognizing complex sequences.

A team of researchers at Xiamen University has developed a new type of electronic transistor that brings this vision closer to reality. By combining a special two-dimensional material with a ferroelectric ceramic, they created a device that can sense light and electricity, store that information, and process it into a decision, all within a single unit. The core of their invention is a thin layer of a material called ReS2, which is sensitive to light, placed on top of a ceramic substrate known as PZT. The PZT is unique because it can hold an electric polarization in two different directions: one pointing straight up and down, and another running sideways along the surface. When the researchers apply voltage or shine light on the device, they can manipulate these internal electric fields to trap or release electrons, effectively changing how easily electricity flows through the material. This allows the device to act like a memory cell that can be rewritten instantly, but with a twist: its behavior depends entirely on the sequence of signals it receives.

The researchers demonstrated that this device can distinguish between different orders of events, a capability they call sequence-dependent response. In a standard memory chip, the final state depends only on the total number of signals sent. In this new transistor, the order matters just as much as the number. For instance, if the device receives a flash of light followed by an electrical pulse, it reacts differently than if it receives the electrical pulse first and the light second. This is because the internal electric fields shift and settle in different configurations depending on the timing and type of the input. The team showed that they could use this property to create a system that remembers a specific sequence of actions, much like a password that must be entered in the exact right order. They tested this by building a security system that required a user to trigger a specific series of light and electrical signals to unlock a gate. If the sequence was wrong, even by a single step, the device would not open, and the system would reset itself, erasing the history of the attempt.

Beyond simple locking mechanisms, the device proved capable of more complex cognitive tasks. The researchers trained an artificial network using these transistors to recognize human faces. They fed the system images of faces, converting the visual data into patterns of light and electrical pulses. Because the device could store the strength of its connections and adjust them based on the order of inputs, the system learned to identify specific individuals with high accuracy. In their tests, the system correctly identified faces from a database of thirty-five different people, achieving a success rate of nearly ninety-four percent when using a combination of light and electrical signals. This performance was superior to using electrical signals alone, suggesting that the ability to process light and electricity together gives the device a richer way to learn. The device also mimicked biological behaviors such as light adaptation, where it could automatically adjust its sensitivity to bright or dim environments, and associative learning, where it learned to link a neutral signal with a meaningful one after repeated exposure.

The physical construction of the device relies on the unique properties of the materials used. The ReS2 layer is extremely thin, only a few atoms thick, which allows it to react quickly to light and electricity. The PZT substrate underneath acts as a dynamic controller, shifting its internal electric alignment to trap electrons in specific spots. When the researchers applied a voltage across the device, they could switch it between a state where it holds onto information permanently and a state where the information fades away quickly, depending on the strength and duration of the signal. This flexibility allows the device to handle both short-term tasks, like reacting to a sudden flash of light, and long-term tasks, like remembering a password. The team also found that the device could perform logical operations, acting as a switch that only turns on when two specific conditions are met simultaneously, effectively functioning as a tiny computer processor within a single component.

This work suggests a path forward for building more efficient and intelligent electronic systems. By integrating sensing, memory, and computing into a single layer of material, the researchers have reduced the need for complex wiring and multiple components. The device's ability to understand the sequence of events means it can be used for applications that require real-time decision-making, such as advanced security systems or adaptive robotics. The researchers confirmed that the device is stable and reliable, maintaining its performance even after many cycles of use and under varying environmental conditions. While the technology is still in the early stages of development, the successful demonstration of dual-mode authentication and facial recognition shows that these materials can support the complex, time-sensitive processing required for the next generation of artificial intelligence. The study provides a concrete blueprint for creating electronic systems that do not just calculate, but truly perceive and remember the world around them.