Computer Sciences and Information Technology
Title : | Ferrimagnet based artificial synaptic device for neuromorphic computing |
Area of research : | Computer Sciences and Information Technology |
Principal Investigator : | Dr. Chandrasekhar Murapaka, Indian Institute Of Technology (IIT) Hyderabad, Telangana |
Timeline Start Year : | 2023 |
Timeline End Year : | 2026 |
Contact info : | mchandrasekhar@msme.iith.ac.in |
Equipments : | Magneto-Optic Kerr Effect (MOKE) microscopy
Vibration isolation table |
Details
Executive Summary : | Neuromorphic computing by mimicking the human brain using artificial neural networks (ANN) has several advantages over conventional Von-Neumann architecture such as parallel processing, adaptability to complex and varying inputs, and energy efficiency. The Von-Neumann architecture is built with CMOS transistors works based on a digital computing system with binary bits. However, the human brain has two key components neurons and synapses. In mimicking the brain, ANNs are realized with neurons that are connected with synapses. The synapses store the information in an analog manner and contribute for the computation by adjusting synaptic weights according to the signals from the pre-and post-neurons. The human brain has 4 orders higher number of synapses as compared to neurons. So it is of paramount interest to realize artificial synaptic devices using emergent technologies beyond CMOS. Spin-based technology is attractive due to its non-volatility and plasticity. The traditional spintronic memory devices represent two binary states based on their relative orientation. However, an essential property of synapse is its analog tunable synaptic weight to perform complex tasks. Though ferromagnetic materials are highly established for various spintronic applications, they have fundamental limitations with scalability and speed. Currently, antiferromagnetic materials are attracting great interest to overcome these limitations. The absence of stray field brings in high scalability and exchange-driven THz precession dynamics enable faster device operation. However, the absence of a net magnetic moment makes it challenging to read two different magnetic states. Ferrimagnetic materials are a class of magnetic materials with antiferromagnetically coupled non-equivalent sublattices, resulting in a small but finite magnetization. Ferrimagnetic materials hold great promise in combining the advantages of both ferromagnet and antiferromagnetic materials. The inequivalence of magnetic moment in sublattices allows ferrimagnets to have a finite Zeeman coupling and spin polarization. In this work, we propose to prepare rare-earth – transition-metal (RE-TM) based ferrimagnetic thin films using co-sputtering technique. The effect of the relative composition of RE-TM on the magnetization reversal and domain wall motion will be investigated by Kerr microscopy. Current induced spin-orbit torque dynamics will be studied in Hall cross structures fabricated from ferrimagnetic/heavy metal heterostructures. The goal is to achieve multistate switching synaptic behavior with high speed and low power consumption. Direct observation using Kerr imaging sheds light on the domain wall nucleation and motion to tune the spin-orbit torque induced magnetization switching in ferrimagnetic materials. The realization of the synaptic devices will open new avenues for ferrimagnetic materials to be employed in ANNs for neuromorphic computing applications. |
Total Budget (INR): | 64,83,845 |
Organizations involved