Speaker
Description
The development of new hardware components is essential for supporting Artificial Intelligence (AI) computational tasks. Neuromorphic computing is witnessing a shift through the integration of advanced spintronics devices to replace CMOS technology. One example of these devices is the multilevel magnetic tunnel junctions (M2TJs) showing very interesting features to be employed on large neural architectures. My project aims to fabricate multilevel magnetic tunnel junctions (M2TJs) with enhanced levels of state control, specifically targeting 8 and 16 levels (3 and 4 bits) switchable with spin-orbit torque (SOT).
The central innovation lies in the development of M2TJs with one of the magnetic layers replaced by a Synthetic Layered Magnetic Multilayer Structure (SLMMS). This structure facilitates the realization of multiple stable magnetic states, crucial for achieving higher bit depth in memory cells. The focus on SOT as a switching mechanism is driven by its scalability and energy efficiency, especially critical as we downscale the physical dimensions of the M2TJs. A critical aspect of this research is achieving a resistance of less than 5 ohms in the electrical contact lines and ensuring that the voltage breakdown of individual elements exceeds 0.8V.
Furthermore, my project will explore the integration of these M2TJs into a new crossbar architecture. The crossbar design will efficiently manage the multistate cells while maintaining the scalability and density necessary for high-performance computing applications.
