🔬 mesoscale physics

Towards Ultra Scalability of Non-Volatile Magnetic Tunnel Junctions with a 3D Storage Layer

This paper presents a strategy to overcome the scalability limits of sub-20 nm Spin Transfer Torque Magnetic Random Access Memory by utilizing a 3D storage layer with enhanced shape anisotropy to maintain thermal stability at sub-10 nm nodes while achieving faster switching speeds at reduced voltages.

Original authors: Nuno Caçoilo, Shunsuke Fukami, Olivier Fruchart, Ioan-Lucian Prejbeanu

Published 2026-08-24✓ Author reviewed ⓘ

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Nuno Caçoilo, Shunsuke Fukami, Olivier Fruchart, Ioan-Lucian Prejbeanu

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the world of modern computing, data is the lifeblood of everything we do, from the smartphone in our pocket to the massive servers running the internet. Yet, the way we store this information faces a fundamental physical limit. Current memory technologies are like a pyramid: at the bottom, we have vast, cheap storage that is slow to access, while at the top, we have incredibly fast memory that is expensive and small. Bridging the gap between these two extremes requires a new kind of memory that is both fast and dense, but also capable of holding its data without power. One of the most promising candidates for this role is a device called a magnetic tunnel junction. Imagine a tiny sandwich made of two magnetic layers separated by a thin insulating barrier. One layer is fixed, acting as a reference, while the other is free to flip its magnetic direction. By flipping this free layer, the device can store a one or a zero. The state is read by measuring how easily electricity flows through the sandwich; it flows easily when the magnetic layers point in the same direction and with difficulty when they point in opposite directions. The challenge arises when engineers try to shrink these devices to make them smaller and denser. As the magnetic layer gets thinner and smaller, it becomes unstable, like a pencil balanced on its tip, prone to losing its data due to random thermal jiggling.

To solve this problem, a team of researchers led by Nuno Caçoilo has explored a different way to build these memory cells. Instead of keeping the magnetic layer flat and thin, they proposed making it tall and narrow, like a microscopic pillar. This change in shape, known as increasing the aspect ratio, fundamentally alters the physics of the device. In a flat layer, the shape of the material naturally wants the magnetism to lie flat, which fights against the ability to store data vertically. However, by stretching the layer vertically, the shape itself begins to help, encouraging the magnetism to stand up straight. This geometric trick allows the device to remain stable even when shrunk down to diameters smaller than 20 nanometers, a size where traditional flat designs fail. The researchers used advanced computer simulations to understand how the magnetism flips in these tall pillars. They discovered that while the tall shape provides the necessary stability, it introduces a new complication: the magnetism does not flip all at once like a single solid block. Instead, the flip starts at the bottom and travels up the pillar like a wave, or a domain wall, which takes more time and energy to complete.

The team then moved from simulation to the laboratory to see if they could actually build these structures. They developed a precise manufacturing process to create these ultra-small pillars, using techniques to etch the magnetic materials into shapes that are much taller than they are wide. When they tested these devices, they found that the tall pillars did indeed hold their data stably at high temperatures, a crucial requirement for devices used in cars or industrial machinery where heat is a constant factor. The stability of these tall pillars relies on the bulk properties of the magnetic material, which are less sensitive to heat than the thin surface layers used in conventional designs. However, the experiments also confirmed the simulation's warning: flipping the magnetism in these tall pillars requires higher voltages and takes longer than in thinner devices. This creates a trade-off; the tall shape solves the stability problem but makes the writing process slower and more energy-intensive.

To overcome this speed and energy penalty, the researchers looked for a way to keep the stability benefits without needing such extreme heights. They found that by splitting the magnetic storage layer into multiple thin layers separated by insulating barriers, they could multiply the stabilizing effect. This multilayered approach acts like having several thin, stable layers working together, rather than one tall, sluggish one. By stacking these layers, they could achieve the same high stability with a much shorter overall height. This reduction in height meant the magnetism could flip much faster and with less voltage, bringing the performance closer to what is needed for high-speed computing. The study suggests that while a single tall pillar is a good starting point for understanding stability, the future of ultra-dense memory likely lies in these complex, multilayered structures. These designs offer a path forward for creating memory that is not only incredibly small and stable but also fast enough to replace the current generation of computer memory, potentially leading to computers that are both more powerful and more energy-efficient. The work highlights that the path to the future of computing is not just about making things smaller, but about rethinking the very shape and structure of the materials that hold our digital world.

Technical Summary: Towards Ultra Scalability of Non-Volatile Magnetic Tunnel Junctions with a 3D Storage Layer

Problem Statement
Perpendicular Spin Transfer Torque Magnetic Random Access Memory (p-STT-MRAM) is a leading candidate for non-volatile memory, yet its scalability is compromised at technological nodes below 20 nm. In conventional flat Magnetic Tunnel Junctions (MTJs), the storage layer relies on interfacial anisotropy (e.g., at the CoFeB/MgO interface) to maintain perpendicular magnetization. As the lateral dimensions shrink, the thermal stability factor ( $\Delta$ ) decreases quadratically with the junction area, threatening data retention. While increasing the storage layer thickness can enhance stability through volume, it typically reduces the aspect ratio, causing the shape anisotropy to favor an in-plane orientation, which destabilizes the perpendicular state. Conversely, simply increasing the aspect ratio to utilize shape anisotropy introduces new challenges: non-uniform magnetization reversal (domain wall nucleation and propagation) leads to higher switching voltages and longer switching times, and the increased magnetic volume generates significant stray fields that cause inter-cell crosstalk in dense arrays.

Methodology
The paper employs a combination of micromagnetic simulations, analytical modeling, and experimental fabrication to investigate the viability of 3D storage layers with high vertical aspect ratios.

Simulations: The authors utilize micromagnetic simulations based on the Landau-Lifshitz-Gilbert-Slonczewski (LLGS) equation, moving beyond the macrospin approximation to account for spatial variations in magnetization. They model the decay of spin-transfer torque (STT) and surface anisotropy away from the tunnel barrier interface. Minimum Energy Path (MEP) calculations are used to determine switching mechanisms and energy barriers.
Analytical Modeling: The study derives relationships between the effective anisotropy field ( $\mu_0 H_{eff}^K$ ), aspect ratio, and switching dynamics. It analyzes the scaling of thermal stability with temperature, distinguishing between interfacial anisotropy (scaling with $M_s^n, n>3$ ) and shape anisotropy (scaling with $M_s^2$ ).
Fabrication and Characterization: Ultra-small MTJs with high aspect ratios (up to 60 nm thickness) were fabricated using electron-beam lithography, reactive ion etching (RIE), and ion-beam etching (IBE) with specific angles to minimize sidewall redeposition. Devices were characterized via resistance-field hysteresis loops, angular switching field measurements (Stoner-Wohlfarth astroids), and STT-driven switching probability tests at various temperatures.

Key Contributions and Results

Reversal Dynamics in High Aspect-Ratio Layers:
Simulations and experiments reveal that as the storage layer thickness increases, the reversal mechanism transitions from coherent macrospin rotation to non-uniform domain wall nucleation and propagation. While high aspect ratios (e.g., thickness > 15 nm) provide thermal stability at sub-10 nm diameters, they necessitate higher switching voltages and exhibit longer switching times due to the "buckling" reversal mode. The critical switching voltage increases with thickness, potentially exceeding reliable operating limits.
Temperature Resilience:
A significant finding is the superior temperature tolerance of shape-anisotropy-dominated storage layers. Experimental data on 60 nm thick NiFe pillars shows that perpendicular magnetization is maintained up to at least 250°C, at which point the measured magnetic induction only reduces slightly. The thermal stability factor decays with a power law exponent $n \approx 2$ , consistent with bulk shape anisotropy ( $\propto M_s^2$ ), whereas interfacial anisotropy in conventional MTJs decays much faster ( $n > 3$ ). This makes high aspect-ratio MTJs particularly suitable for high-temperature applications (automotive, industrial, and solder reflow).
Stray Field and Inter-Cell Crosstalk:
The study identifies a critical limitation for dense arrays: the long-range nature of the stray field from high aspect-ratio pillars. Unlike flat MTJs where stray fields decay rapidly, the stray field from thick pillars extends significantly, causing substantial variations in the thermal stability factor ( $\Delta$ ) of neighboring cells even at pitches of 60 nm. This crosstalk can lead to asymmetric switching currents and retention failures.
Multilayered Storage Layer Solution:
To address the trade-off between stability, switching speed, and crosstalk, the paper proposes and characterizes multilayered storage layers containing multiple CoFeB/MgO interfaces (e.g., "HEXA" structures with six interfaces).
- Mechanism: By introducing multiple interfaces, the effective anisotropy field ( $\mu_0 H_{eff}^K$ ) is significantly increased without relying solely on extreme geometric aspect ratios.
- Performance: These structures achieve high thermal stability (e.g., 80 $k_B T$ ) at reduced thicknesses and diameters. The increased $\mu_0 H_{eff}^K$ reduces the characteristic switching time ( $\tau_D$ ) and critical voltage ( $V_{c0}$ ), enabling switching times in the tens of nanoseconds (comparable to DRAM) at lower voltages.
- Crosstalk Mitigation: The higher anisotropy field makes the device less sensitive to external stray fields, and the reduced magnetic volume lowers the generated dipolar field, making these structures more viable for ultra-dense packing (projected density ~100 Gbit/cm²).

Significance and Claims
The paper argues that while a simple 3D storage layer with a high vertical aspect ratio offers a pathway to sub-20 nm scalability and high-temperature operation, it is not a standalone solution due to non-uniform reversal dynamics and inter-cell crosstalk. The authors claim that the most viable route for ultra-scalable, high-density STT-MRAM is the combination of moderate aspect ratios with multilayered storage layers. This approach leverages enhanced interfacial anisotropy to maintain stability at small volumes, thereby restoring coherent-like switching speeds and reducing sensitivity to stray fields. The work suggests that with reliable fabrication of these complex multilayer structures, STT-MRAM can compete with DRAM in terms of speed and density while retaining non-volatility, particularly for applications requiring high thermal resilience.

More like this