Project 3

Non-Volatile Memory Devices for Next-Generation Neuromorphic AI Computing

(Prof. S. Mcnamara, ECE - Desing and research of exciting applications in microsystems (DREAM) LAB

Computers use many types of memory, including cache, main memory, and storage memory.  In this work, we are developing a new type of computer memory that will integrate main memory with storage memory.  This will enable tremendous improvements in computer processing speed, which is critical for applications such as AI.

The technology is based on MEMS technology.  The bit is stored by actuating a beam into one of two stable positions.  The beam position can be read capacitively, and the beam position can be written to using electrostatics.  Most importantly, the power consumption will be orders of magnitude smaller than in flash memory and the write times will be orders of magnitude faster. 

The figure below shows an SEM image of a memory bit consisting of three electrodes in a stack.  This MEMS device was fabricated in the Micro/Nano Technology Cleanroom at the University of Louisville.

The REU student will learn about computer memory, cleanroom processing, and how to design and test devices.  AI will be utilized to help create illustrations that are suitable for communicating semiconductor and MEMS concepts to a variety of audiences.  This REU program will position the student for a job at a semiconductor company or pursuing a graduate degree related to semiconductors.