Introduction
This proposal aims to establish an electrical characterization facility to support the experimental validation of a ferroelectric spiking neural network (FSNN) based on Ferroelectric-FETs (Fe-FETs) devices for neuromorphic computing applications. Our team is currently designing normally-on ferroelectric FinFET/MOSFET devices using TCAD tools, with the necessary licenses already in place at the institute. Further, we are collaborating with NTHU Taiwan and have opportunity to get samples fabricated, compatible with the most-advanced technology node. However, our institute currently lacks the required characterization tools.
The proposed electrical characterization facility will address this gap and play a pivotal role in characterizing the synaptic crossbar arrays made from these devices. This will enable us to develop a Process Design Kit (PDK) and further investigate the circuit- and system-level performance of FSNNs, with simulations informed by data obtained through electrical characterization.
The characterization facility will consist of two key components:
- A mini electrical probe station equipped with 16 micro-manipulators to probe an 8×8 cross-bar array, and
- An electrical source and measure unit capable of acquiring data from an 8×8 switch matrix.
Estimated Timeline for the Project
12 Months
Expected Impact
Synaptic devices are a fundamental element in neuromorphic systems, enabling intelligent, low-power operations and compact chip designs. These devices make neuromorphic systems more efficient and scalable, providing an ideal foundation for next-generation computing systems. Fe-FETs have garnered significant attention due to their potential as synaptic devices, especially with the discovery of impressive ferroelectric properties in a CMOS compatible zirconium-doped hafnium oxide.
The successful integration of Fe-FETs into synaptic devices can advance artificial neural networks and bring us closer to realizing neuromorphic computing systems. Although current synapse array technology has made significant progress, there is still considerable potential for improvement, particularly in terms of performance, scalability, and power efficiency.
The impact of the project will be far-reaching, both in terms of technological advancements and institutional growth. The facility will play a crucial role in the validation of FSNNs, supporting the development of a PDK and enabling high-performance simulations. It will also allow the institute to contribute to national and global efforts in neuromorphic engineering, artificial intelligence, and high-performance computing, ultimately advancing the field and improving the capabilities of neuromorphic systems.
