🔬 applied physics

Orbital and Spin-Orbit Torque Interplay in Ta/W-based Magnetic Tunnel Junctions with Vertical Non-local Switching

This paper demonstrates that integrating a Ta/W bilayer system into SOT-MTJ devices significantly enhances spin-orbit torque efficiency through orbital Hall contributions, enabling robust perpendicular magnetic anisotropy, high-temperature compatibility, and a novel proof-of-concept for vertical non-local switching to simplify MRAM fabrication.

Original authors: Marco Biagi (Univ. Grenoble Alpes, CEA, CNRS, Grenoble-INP, SPINTEC), Corrado C. M. Capriata (Univ. Grenoble Alpes, CEA, CNRS, Grenoble-INP, SPINTEC), K. Subham Senapati (Univ. Grenoble Alpes, CEA, CN

Published 2026-05-27

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Marco Biagi (Univ. Grenoble Alpes, CEA, CNRS, Grenoble-INP, SPINTEC), Corrado C. M. Capriata (Univ. Grenoble Alpes, CEA, CNRS, Grenoble-INP, SPINTEC), K. Subham Senapati (Univ. Grenoble Alpes, CEA, CNRS, Grenoble-INP, SPINTEC), Ioannis Trikoilis Koll (Univ. Grenoble Alpes, CEA, CNRS, Grenoble-INP, SPINTEC), Corentin Bouchard (Univ. Grenoble Alpes, CEA, CNRS, Grenoble-INP, SPINTEC), Ricardo C. Sousa (Univ. Grenoble Alpes, CEA, CNRS, Grenoble-INP, SPINTEC), Louis Hutin (CEA-Leti Minatec, Grenoble, France), Bernard Viala (CEA-Leti Minatec, Grenoble, France), Kevin Garello (Univ. Grenoble Alpes, CEA, CNRS, Grenoble-INP, SPINTEC)

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 or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to flip a tiny magnetic switch inside a computer chip. This switch is the heart of a new type of memory called MRAM, which is designed to be faster and more energy-efficient than the memory we use today. To flip this switch, you usually need to send a "spin current"—a flow of electrons that carry a specific kind of rotation called "spin."

For a long time, scientists have used heavy metals (like Tungsten) to generate this spin current. However, this process is a bit like trying to push a heavy boulder up a hill: it requires a lot of energy, and the conversion from electricity to "spin" isn't very efficient. The paper you shared proposes a clever new way to do this by using a different kind of physics called orbital physics.

Here is a breakdown of their discovery using simple analogies:

1. The Problem: The "Heavy" Push

In standard devices, scientists use a heavy metal layer to turn electricity into the spin current needed to flip the magnetic switch. Think of this like a water wheel. You pour water (electricity) onto the wheel, and it spins (spin current). But in current technology, the wheel is heavy and the water doesn't turn it very efficiently. You need a huge amount of water just to get the wheel moving.

2. The New Idea: The "Orbital" Shortcut

The researchers discovered that electrons have another property besides spin called orbital motion. Imagine an electron not just spinning like a top, but also orbiting a nucleus like a planet around the sun.

The paper suggests we can use this "orbital" motion to help push the switch.

The Analogy: Imagine you have a conveyor belt (the orbital current) moving very fast. It's carrying boxes (orbital momentum). But the machine you want to power (the magnetic switch) only accepts spinning tops (spin current).
The Solution: You need a "converter" to turn those boxes into spinning tops. The researchers found a way to do this using a sandwich of two metals: Tantalum (Ta) and Tungsten (W).

3. The Magic Sandwich: Tantalum and Tungsten

The team created a stack where:

Tantalum (Ta) acts as the conveyor belt. It generates a massive amount of orbital current (the fast-moving boxes).
Tungsten (W) acts as the converter. It sits on top of the Tantalum and instantly turns that orbital motion into the spin current needed to flip the magnetic switch.

The Result: By adding just a tiny layer of Tungsten on top of Tantalum, they got four times more "push" than they would have gotten from Tantalum alone. It's like adding a small gear to a machine that suddenly makes it four times more powerful.

4. Why This Matters for Computers

The researchers tested this new "sandwich" in actual memory devices (called Magnetic Tunnel Junctions).

Efficiency: The new system is just as good at flipping the switch as the old standard Tungsten systems, but it offers a new path to make it even better in the future.
Durability: The new system can survive high heat (400°C), which is a strict requirement for manufacturing computer chips in factories.
Stronger Magnet: The new setup makes the magnetic switch "stickier" (more stable), meaning it holds its data better.

5. The "Non-Local" Trick: The Invisible Wire

The most creative part of the paper is a "proof-of-concept" for a new way to build these chips.

The Old Way: Usually, the wire that sends the current must be directly underneath the magnetic switch. This is hard to build because you have to be incredibly precise with your tools.
The New Trick: The researchers showed that the "orbital current" can travel through a spacer (a layer of Tantalum) to reach the switch from a distance.
The Analogy: Imagine you are trying to turn on a light switch, but the switch is covered by a thick wall. Usually, you can't do it. But with this new physics, it's as if the "signal" can walk through the wall to reach the switch. This allows them to build "bottom-pinned" switches (where the magnet is on the bottom) much more easily, simplifying the manufacturing process.

Summary

The paper claims that by stacking Tantalum and Tungsten, they can use orbital physics to create a much more efficient "spin current." This acts like a super-charged engine for flipping magnetic switches in computer memory. They proved this works in real devices, survives factory heat, and even allows for a new, simpler way to build these memory chips by letting the current travel through a spacer layer to reach the switch.

Note: The paper focuses entirely on the physics of the materials and the device performance. It does not claim these devices are ready for consumer products yet, nor does it discuss medical or clinical applications. It is a step toward better computer memory, but the work is currently in the research and development phase.

Technical Summary: Orbital and Spin-Orbit Torque Interplay in Ta/W-based Magnetic Tunnel Junctions

Problem Statement
Spin-orbit torque (SOT) is a promising mechanism for next-generation non-volatile magnetic random-access memories (MRAMs), particularly for cache applications, due to its potential for ultra-fast and energy-efficient magnetization switching. However, current SOT-MRAM devices face significant limitations in write efficiency. The charge-to-spin conversion efficiency ( $\xi_{DL}$ ) in existing heavy metal systems typically reaches only $\sim45\%$ , falling short of the projected $\sim80\%$ required to meet the specifications of advanced transistor nodes. While materials like topological insulators offer higher efficiencies, their industrial integration is hindered by thermal budget constraints and compatibility issues with magnetic tunnel junction (MTJ) specifications. Furthermore, there is a need for materials that support robust perpendicular magnetic anisotropy (PMA) and can withstand high-temperature annealing (e.g., 400°C) required for back-end-of-line (BEOL) processing.

Methodology
The authors investigated a Ta/W bilayer system designed to leverage orbital current physics to enhance SOT efficiency. The study proceeded in three main stages:

Material Characterization: Using the harmonic Hall voltage (HHV) method on annealed Hall bar samples, the authors quantified the damping-like SOT efficiency ( $\xi_{DL}$ ) for reference stacks (pure W, pure Ta) and orbital stacks (Ta/W, Ta/Pt, Ta/Ni). Samples were annealed at 300°C and 400°C to assess thermal stability and BEOL compatibility.
Device Fabrication and Benchmarking: Three-terminal SOT-MTJ devices with 80 nm pillar diameters were fabricated. The authors compared a standard W-based reference device against an OtS (Orbital-to-Spin) device utilizing a Ta(10)/W(1.5) stack. Switching performance was characterized via $R$ - $H$ loops and switching current ( $I_{SW}$ ) measurements as a function of pulse length ( $\tau_p$ ) down to 500 ps.
Vertical Non-Local Switching Proof-of-Concept: To explore a simplified fabrication route for "bottom-pinned" MTJs, the authors fabricated devices where the SOT track and the free layer were separated by a thick Ta spacer (acting as a hard mask). This setup tested the feasibility of vertical non-local switching mediated by orbital torques. COMSOL simulations were employed to model current shunting and distinguish between local current effects and long-range orbital diffusion.

Key Contributions and Results

Enhanced Efficiency via Orbital Physics: The Ta/W system demonstrated a significant enhancement in SOT efficiency compared to single-layer references. Specifically, the Ta(20)/W(1.5) stack achieved a saturation efficiency ( $\xi_{sat}$ ) of $-34.7\%$ , representing a four-fold increase over Ta alone and a 3.2-fold increase over a thin W(1.5) reference. This enhancement is attributed to the conversion of orbital currents generated in Ta (via the Orbital Hall Effect) into spin currents by the W layer (via strong spin-orbit coupling).
Robustness and Compatibility: The Ta/W system maintained high efficiency after 400°C annealing, with only a marginal reduction in $\xi_{sat}$ (to $-32.4\%$ ). Crucially, the annealing process significantly enhanced the perpendicular magnetic anisotropy ( $B_k$ ), increasing from $\approx0.5$ T to $\approx1.1$ T, which improves thermal retention for device operation.
Device Performance: In 3-terminal MTJ devices, the Ta/W system exhibited switching symmetry consistent with SOT mechanisms. While the absolute switching current was higher due to the thicker SOT track, normalization by the track thickness and coercivity revealed that the Ta/W system offers comparable or superior current-to-spin conversion efficiency ( $I_{c0}/t_{SOT}$ ) compared to standard W-based systems.
Vertical Non-Local Switching: The study successfully demonstrated vertical non-local switching in SOT-MTJs. By separating the current injection point from the free layer with a Ta spacer, the authors showed that switching could still be achieved. Analysis of the critical switching current ( $I_{c0}$ ) versus spacer thickness indicated that while current shunting is a dominant factor, an additional long-range contribution (likely orbital diffusion) persists, as the critical current in the active zone decreased with increasing spacer thickness in a manner not fully explained by local current reduction alone.

Significance and Claims
The paper claims that the Ta/W system offers a viable strategy to enhance SOT-MRAM efficiency by integrating orbital physics into industrially compatible materials. The key significance lies in:

Efficiency Enhancement: Demonstrating that orbital currents can be effectively converted to spin currents in a Ta/W bilayer, providing a boost in $\xi_{DL}$ that rivals or exceeds standard heavy metal systems.
Industrial Viability: The system meets critical industrial requirements, including compatibility with 400°C annealing (BEOL compatible) and the ability to support robust PMA.
Simplified Fabrication: The demonstration of vertical non-local switching suggests a pathway to simplify the fabrication of bottom-pinned SOT-MTJs. This approach could eliminate complex steps required to remove hard masks without damaging the free layer, potentially improving device yield and scalability.
Future Outlook: The authors conclude that while the current orbital diffusion length in Ta/W is a constraint, the results underscore the potential of orbital transport mechanisms to improve SOT writing performance. They propose that exploring materials with longer orbital diffusion lengths could further optimize this approach for non-local control of MTJs.