A reconfigurable photosensitive split-floating-gate memory for neuromorphic computing and nonlinear activation

Area of Science:

• * Materials Science and Engineering
• * Neuromorphic Computing
• * Artificial Intelligence Hardware

Background:

• * The expansion of artificial intelligence (AI) and the Internet of Things (IoT) necessitates compact hardware capable of integrated sensing, computing, and nonlinear processing.
• * Current neuromorphic systems often suffer from limited functionality and rely on heterogeneous integration, hindering scalability and efficiency.

Purpose of the Study:

• * To introduce a novel multi-modal split-floating-gate memory device.
• * To demonstrate monolithic integration of in-sensor computing, in-memory computing, and multiple nonlinear activation functions within a single device.
• * To establish a hardware foundation for scalable, energy-efficient intelligent systems.

Main Methods:

• * Development of a high-speed, reconfigurable multi-modal split-floating-gate memory architecture.
• * Programming of charges in spatially separated floating gates to control photoresponsivity and conductance.
• * Electrical reconfiguration of rectification to emulate ReLU and Sigmoid activation functions.
• * Implementation of a fully hardware-based sensor-processor system using memory arrays for learning tasks.

Main Results:

• * The device monolithically integrates sensing, in-memory computing, and multiple nonlinear activation functions (ReLU, Sigmoid).
• * Non-volatile analog control of photoresponsivity and conductance is achieved through charge programming.
• * A complete sensor-processor system successfully performed unsupervised and supervised learning tasks in hardware.
• * Demonstrated high-speed and reconfigurable operation.

Conclusions:

• * The developed split-floating-gate memory offers a compact and energy-efficient solution for intelligent systems.
• * Monolithic integration of diverse functionalities enhances scalability and efficiency compared to existing neuromorphic approaches.
• * This technology provides a reconfigurable hardware foundation for next-generation AI and IoT applications.