High Performance TiO2 Ferroelectric Field Effect Transistors

This invention has been demonstrated on silicon substrate, indicating its CMOS compatible capability. The devices are made with atomic layer deposition method, which features film thickness uniformity and results in high yield, an indication of large scale capability and commercialization applications.

Background

Traditional computers separate memory and processing causing energy- and time-intensive data transfer. Inspired by the human brain neuromorphic systems have memory and computation co-located in networks of neurons and synapses. This brain-like approach enables massive parallelism, adaptability, and ultra-low power operation. As a result, neuromorphic computing holds great promise for tasks that require real-time learning, pattern recognition, and sensory data processing—such as autonomous driving, robotics, and AI applications.

 

Current existing neuromorphic computing systems use conventional complementary metal-oxide semiconductor (CMOS) technology that lack scalability and energy efficiency. Alternative devices have been proposed for neuromorphic computing including Ferroelectric field-effect transistors (FeFETs) which have emerged as a promising platform for neuromorphic computing due to their ability to integrate nonvolatile memory functionality with energy-efficient logic within a simple transistor structure. Although many hurdles still exist for the practical deployment.

Technology

To overcome these hurdles, UD Researchers have developed a TiO₂ thin-film transistor platform and have demonstrated the first TiO₂ FeFETs with significantly enhanced performance using HfZrO as the ferroelectric material. The devices exhibit a memory window exceeding 5 V and an on/off ratio greater than 107, offering distinct advantages for neuromorphic computing applications, including disturbance immunity, multilevel states, and low switching energy. This invention has been demonstrated on silicon substrate, indicating its CMOS compatible capability. The devices are made with atomic layer deposition method, which features film thickness uniformity and results in high yield, an indication of large scale capability and commercialization applications.

Benefits

  • Superior Signal Control: Features a robust memory window exceeding 5 V and an industry-leading on/off ratio (>107).

  • Optimized for pattern recognition and real-time learning: Delivers multi-level states and exceptional disturbance immunity, critical for precise pattern recognition and real-time learning.

  • Ultra-Low Power: Significantly reduces switching energy, enabling sustainable operation for intensive workloads.

  • Seamless Integration: Fully CMOS-compatible and fabricated using atomic layer deposition, ensuring high yield and effortless scaling for commercial manufacturing.

 

Applications

  • Extensive circuit applications such as autonomous driving, robotics, and AI applications