[News] Chinese Research Team Developed Ultra-High-Speed Flash Memory

Fudan University has achieved a key breakthrough in the field of integrated circuits. The research team led by Zhou Peng and Liu Chunsen has, by constructing a quasi-2D Poisson model, theoretically predicted a phenomenon called “super-injection,” breaking through the existing theoretical limits of memory speed and developing the “PoX” picosecond flash memory device. Its read/write speed can be boosted to sub-1-nanosecond levels. With a speed of 400 picoseconds — equivalent to 2.5 billion operations per second — this marks the fastest semiconductor charge memory technology in the world to date.

Amid the AI wave, high-speed storage of big data is of great importance. How to break through the speed limits of information storage has long been one of the most fundamental challenges in the integrated circuits field and remains a critical bottleneck restricting AI computing power. Achieving high-speed storage for big data means the memory must perform excellently across speed, energy consumption, and capacity — becoming a true “hexagonal warrior.”

As the basic storage unit of flash memory, the floating-gate transistor consists of a source, drain, and gate. When electrons flow from the source through the channel toward the drain, pressing the gate “switch” to capture the electrons into the floating-gate storage layer, thus achieving information storage.

“In the past, the strategy to speed up flash memory was to allow electrons to ‘warm up’ and accelerate along the channel before pressing the switch when they gained enough energy,” Liu explained vividly. However, under traditional theoretical mechanisms, the “run-up” distance for electrons was long and acceleration was slow. The unique electric field distribution in semiconductors also imposed a theoretical limit on electron acceleration, preventing the storage speed of flash memory from breaking through the injection extreme point.

Starting from the fundamental theoretical mechanisms of memory devices, the team proposed a brand-new approach: by combining the Dirac energy band structure and ballistic transport properties of two-dimensional materials, they modulated the Gaussian length of the 2D channel to achieve super-injection of channel charges into the floating-gate storage layer. Under the super-injection mechanism, electrons can reach high speed directly without a “run-up,” and injection can continue without being limited by a critical injection point.

By constructing a quasi-2D Poisson model, the team successfully predicted the super-injection phenomenon theoretically. Based on this, they developed a picosecond flash memory device that achieved a read/write speed breakthrough below 1 nanosecond (400 picoseconds), enabling up to 2.5 billion operations per second. Its performance surpasses the world’s fastest volatile memory (SRAM) technologies at the same technology node

This marks the fastest semiconductor charge storage technology in the world so far, achieving almost the same storage and computing speed. Once large-scale integration is completed, it could potentially revolutionize existing memory architectures. With this technology, future personal computers may no longer need separate memory and storage layers, eliminating the need for hierarchical storage and allowing local deployment of large AI models.

As the foundational pillar of the intelligent era, expanding the speed limits of storage technology could trigger exponential innovations across application scenarios and become one of the key drivers for China’s leadership in AI, cloud computing, communication engineering, and other related fields. Next, the team plans to scale up integration to tens of megabits within 3 to 5 years, at which point the technology could be licensed to enterprises for industrialization.