Spintronics for AI

Overview

In I2S, we are pushing the boundaries of innovation at the intersection of biomedical applications, neuromorphic computing, spintronics, and hardware security. Our research focuses on developing intelligent biomedical devices that enable precision medicine, personalized treatment plans, and disease diagnosis through spintronics platforms. By combining neuromorphic computing's ability to learn and adapt with spintronics' unique properties for information processing, we are creating novel solutions for secure data storage and transmission. Our work in hardware security ensures the integrity of these advancements, safeguarding against cyber threats and ensuring the confidentiality and authenticity of sensitive medical data. Through our research, we aim to transform industries worldwide by revolutionizing healthcare, advancing artificial intelligence, and shaping the future of innovation.

Spintronics for AI

The significant energy waste inherent in traditional computer chips has hindered the quest for efficient artificial intelligence (AI) processing. Although graphics processing units (GPUs) and application-specific integrated circuits (ASICs) have shown impressive power for deep neural networks, their scalability and performance are limited by growing energy costs and a fundamental bottleneck. To overcome these challenges, innovative hardware approaches are needed that can mimic the brain's efficient computing capabilities. Emerging neuromorphic computing technologies, which simulate biological neural networks, may help address the unsustainable energy demands of current AI systems. As AI applications spread, finding sustainable and scalable solutions to mitigate energy waste concerns is crucial. Our research group is at the forefront of advancing artificial intelligence through innovative spintronic devices and neuromorphic computing architectures. We are dedicated to overcoming traditional computing systems' energy inefficiencies and scalability limitations by exploring cutting-edge physics and novel device concepts.

Our research focuses on harnessing the unique properties of magnetic domain walls and skyrmions in the magnetic tunnel junction (MTJ) to emulate the complex functionalities of neurons and synapses in hardware. Through the development of magnetic multilayer structures and advanced fabrication techniques, we have achieved significant milestones in creating devices that mimic the behavior of neurons, essential for applications in spiking neural networks (SNNs) and convolutional neural networks (CNNs) for pattern recognition and AI tasks.

By studying domain wall and skyrmion micromagnetic dynamics and transport characteristics, we have shown energy-efficient solutions that promise to revolutionize data storage and non-conventional computing architectures for hardware ANN and SNN applications. Our experimental successes include demonstrating the dynamic control and manipulation of domain walls and skyrmions in magnetic thin films and showcasing our commitment to pushing the boundaries of neuromorphic computing.

Domain wall and Skyrmion Spiking Neuron Devices for SNN Applications