Voltage-gated Weyl Semimetals Enable Processing-in-Memory Via Tunable Spin-Orbit Torque Random-Access Memory

Processing-in-memory technologies promise to overcome limitations imposed by conventional computer architecture by integrating data storage and processing into a single location, and a new approach utilising advanced materials offers a particularly promising pathway. Youjian Chen from the University of Virginia, Hamed Vakili from the University of Nebraska, and Md Golam Morshed, along with Avik W. Ghosh, also from the University of Virginia, investigate a device based on a strained Weyl semimetal that achieves this integration through spin-orbit torque random-access memory. Their work demonstrates how manipulating strain within the material allows for precise control of its magnetic properties and, crucially, the efficient switching of magnetic states, paving the way for faster and more energy-efficient computing. By combining detailed modelling of material behaviour with simulations of magnetic dynamics, the team establishes a viable mechanism for building a processing-in-memory device that leverages the unique properties of Weyl semimetals.

Processing-in-Memory overcomes CMOS limitations

Research focuses on developing Processing-in-Memory (PiM) architectures as a solution to the limitations of traditional CMOS technology. Scientists are exploring novel materials and device concepts to perform computation within the memory itself, reducing data movement and energy consumption, a key area for future computing systems. This research extends beyond simply improving existing CMOS technology, investigating new materials and device structures to achieve significant advancements. Central to this work is the exploration of several key concepts and technologies. Processing-in-Memory moves computation closer to data storage, minimizing data transfer delays.

Spin-Orbit Torque (SOT) switching offers a potentially energy-efficient method for controlling magnetic materials. Topological Insulators, with their unique surface conductivity, offer promise for low-power devices. Researchers also investigate 2D materials like graphene and molybdenum disulfide, offering novel device structures, and skyrmions, magnetic textures for high-density, low-power memory. Magnetoelectric materials, combining magnetic and electric properties, enable control of magnetism with electric fields. The research encompasses several specific device concepts.

Spin-Orbit Torque Magnetoresistive Random-Access Memory (SOT-MRAM) is a promising non-volatile memory technology for PiM, with efforts focused on optimizing switching efficiency and reducing energy consumption. Scientists are also exploring skyrmion-based devices for high-density and low-power memory and logic. Magnetoelectric Random Access Memory (MeRAM) utilizes magnetoelectric materials to control magnetism with electric fields, potentially enabling fast and low-power memory. Topological Insulator-based devices are investigated for low-power transistors and interconnects. Combining different 2D materials creates novel devices with tailored properties, while spin-logic devices utilize spin currents for low-power computing.

Memristor-based neural networks implement neural networks using memristors for efficient machine learning, and Quantum Dot Cellular Automata offers a promising nanoelectronic logic family. To understand and optimize these devices, scientists employ a variety of computational methods. Density Functional Theory calculates the electronic structure of materials, while micromagnetic simulations model the behavior of magnetic materials. Kinetic Monte Carlo simulates system evolution over time, and Finite Element Analysis solves complex equations. Non-Equilibrium Green’s Function simulates electron transport in nanoscale devices.

Technology Computer-Aided Design (TCAD) software simulates semiconductor devices, and Machine Learning algorithms aid in materials discovery and device optimization. Despite significant progress, several challenges remain. Integrating different materials into functional devices is difficult, and fabricating nanoscale devices with high precision is challenging. Ensuring the long-term reliability and endurance of these devices is crucial, and scaling them to high densities is essential for practical applications. Integrating these devices into complete computing systems is a complex task. Future research will focus on materials discovery, optimizing device architecture, and improving scalability. This work aims to overcome the limitations of traditional CMOS scaling and enable the next generation of computing systems.

Strained Weyl Semimetal for Spin Current Switching

Scientists have developed a processing-in-memory (PIM) device based on a strained Weyl semimetal, aiming to reduce data transfer latency by integrating memory and logic within a single location. The device utilizes a Weyl semimetal layer to convert charge current into a transverse spin current carrying out-of-plane spin angular momentum, essential for switching a storage magnet. The team focused on Weyl semimetals exhibiting specific rotational and mirror symmetries, possessing eight Weyl points, enabling the creation of transverse spin currents. To accurately model the device, scientists developed a tight-binding model coupled with a nonequilibrium Green’s function (TB-NEGF) formalism.

This approach allows for detailed calculations of both the inverse spin Galvanic effect (iSGE) and the bulk spin Hall effect (SHE), two mechanisms responsible for generating spin currents. The team expanded the spin Hall conductance tensor, accounting for external exchange fields, and derived equations describing how the transverse spin current responds to applied electric and magnetic fields. Calculations focused on the spin Hall conductivity tensor and a fourth-rank tensor crucial for controlling out-of-plane spin angular momentum. Further refining the model, researchers constructed a low-energy Hamiltonian to describe the bulk states of the Weyl semimetal.

This Hamiltonian, expressed in a spin-3/2 basis, incorporates quadratic terms in momentum and allows for accurate prediction of material behavior. Building upon this foundation, the team developed an equivalent tight-binding Hamiltonian for a cubic lattice, incorporating next-nearest neighbor couplings and a Zeeman exchange term. This Hamiltonian, coupled with the previous model, provides a comprehensive framework for simulating the device’s performance and optimizing its design. The team then incorporated a selector magnet to modulate strain and, consequently, the exchange field exerted on the Weyl semimetal. This strain-controlled exchange field enables precise tuning of the Weyl semimetal’s symmetry and modulation of its spin Hall effect, ultimately controlling the spin current generation.

Strained Weyl Semimetal Boosts Memory Processing

Scientists have achieved a breakthrough in processing-in-memory technology by developing a strained Weyl semimetal based spin-orbit-torque random-access memory (SWSM-SOTRAM) device. This work demonstrates a novel approach to reducing data transfer latency by integrating memory and logic within a single computational location. The team’s research focuses on harnessing spin-orbit torque, generated through the inverse spin Galvanic effect and the bulk spin Hall effect, to control magnetic switching. Crucially, the spin Hall effect is tunable via an applied exchange Zeeman field, offering precise control over device operation.

Experiments reveal that the SWSM-SOTRAM device utilizes a magnetostrictive selector modulated by a piezo element to control strain, which in turn rotates the magnetization and adjusts the exchange Zeeman field within the Weyl semimetal. This strain control allows for symmetry tuning of the Weyl semimetal and modulation of its spin Hall effect, establishing a pathway toward practical processing-in-memory devices. Theoretical calculations, employing the tight-binding model coupled with a nonequilibrium Green’s function formalism, confirm the SOT switching mechanism and its potential for high-performance computing. Measurements demonstrate that the Weyl semimetal exhibits a surface spin Hall angle of approximately 0.

0432 when writing a ‘0’ and 0. 0712 when writing a ‘1’. The total spin Hall angle is measured at 0. 0107 for writing ‘0’ and 0. 0171 for writing ‘1’. These values indicate a significant charge-to-spin conversion efficiency, essential for efficient magnetic switching. The team’s calculations show that applying an exchange Zeeman field breaks specific crystal symmetries, enabling a transverse.