In the past decade, the demand for computing power has skyrocketed with the emergence of advanced artificial intelligence (AI) technologies. This has led to a need for energy-efficient hardware designs that can handle the increasing computational requirements, improve processing speed, and enable AI training directly on devices.
According to Shan Wang, the Leland T. Edwards Professor in the School of Engineering at Stanford University, we have transitioned from the internet era to the AI era. The goal now is to enable AI at the edge, allowing local training on devices like computers, smartphones, or smartwatches for applications such as heart attack detection or speech recognition. To achieve this, a fast and non-volatile memory is essential.
Recently, Wang and his colleagues made an exciting discovery, which they published in the journal Nature Materials. They identified a promising material called manganese palladium three, which possesses the necessary properties to facilitate a new type of working memory. This memory stores data based on electron spin directions and is known as spin orbit torque magnetoresistive random access memory (SOT-MRAM).
SOT-MRAM has the potential to store data more efficiently and rapidly compared to current methods that rely on electric charge. Unlike traditional memory systems, SOT-MRAM does not require a continuous power input to maintain stored data.
Wang believes that their research lays the groundwork for future energy-efficient storage elements, marking a significant breakthrough in this field. The findings offer a fundamental building block towards the commercialization of advanced memory technologies.
Harnessing electron spin
SOT-MRAM, the memory technology that utilizes electron spin, relies on a fascinating property of electrons. To understand spin, imagine an electron as a spinning basketball balanced on a skilled athlete’s finger. As electrons are charged particles, their rotation generates a tiny magnet with polarization along its axis, similar to the line extending from the athlete’s finger. When the spin direction of the electron switches, the magnet’s north and south poles also switch. This magnetic dipole moment, represented by the up or down direction of magnetism, can be used to encode the binary data of computer memory.
In SOT-MRAM, a current passing through the SOT layer induces specific spin directions in electrons. The movement and spin directions of these electrons create a torque that can switch the spin directions and associated magnetic dipole moments of electrons in an adjacent magnetic material. By employing suitable materials, it becomes possible to store magnetic data by simply changing the direction of the electrical current in the SOT layer.
However, finding the right materials for SOT-MRAM is a challenging task. The hardware design of SOT-MRAM allows for denser data storage when electron spin directions are oriented up or down in the z-direction. Unfortunately, most materials polarize electron spins in the y-direction when the current flows in the x-direction.
Fen Xue, a postdoctoral researcher in Wang’s lab, explains that conventional materials typically generate spin in the y-direction. This means that an external magnetic field would be required to achieve switching in the z-direction, which consumes more energy and space. To minimize energy consumption and achieve higher memory density, the goal is to achieve switching without the need for an external magnetic field.
The researchers discovered that manganese palladium three possesses the desired properties. This material can generate spins in any orientation because its internal structure lacks the crystal symmetry that would enforce a specific electron orientation. By utilizing manganese palladium three, the researchers successfully demonstrated magnetization switching in both the y- and z-directions without relying on an external magnetic field. While not specifically showcased in the paper, magnetization in the x-direction can also be switched without the need for an external magnetic field.
The ability to have three different spin directions with the same input current is a significant advancement. Mahendra DC, the first author of the paper, highlights that depending on the application, magnetization can be controlled in any desired direction.
The success of this research can be attributed to the collaborative efforts of multiple institutions and disciplines. Evgeny Tsymbal’s lab at the University of Nebraska conducted calculations to predict the unexpected spin directions and movement, while Julie Borchers’s lab at the National Institute of Standards and Technology led the measurement and modeling efforts to uncover the intricate microstructures within manganese palladium three. The achievement truly exemplifies the collaborative nature of scientific advancement.
Manufacturing possibilities
Apart from its unique structure, manganese palladium three possesses additional properties that make it highly suitable for SOT-MRAM applications. One notable advantage is its ability to withstand and maintain its properties during the post-annealing process, which is crucial for electronics.
“The post-annealing process requires electronics to be exposed to temperatures of 400 degrees Celsius for 30 minutes,” explains DC. “This poses a challenge for new materials in these devices, but manganese palladium three can handle it.”
Furthermore, the deposition of the manganese palladium three layer utilizes magnetron sputtering, a technique already employed in other aspects of memory-storage hardware. This means that no new tools or techniques are required for the integration of this material.
“We don’t need specialized conditions or a textured substrate to deposit this material,” says Xue. “It seamlessly fits into current manufacturing techniques.”
As a result, manganese palladium three not only exhibits novel properties that can address the increasing demands of computing but also enables smooth integration with existing manufacturing processes. The researchers are already working on prototypes of SOT-MRAM using manganese palladium three, aiming to incorporate them into real devices.
“We are currently facing limitations with current technology,” notes DC. “Therefore, we need to explore alternative options.”
The utilization of manganese palladium three in SOT-MRAM presents a promising avenue for overcoming existing barriers and pushing the boundaries of memory technology.
Source: Stanford University