🔬 mesoscale physics

Cooperative STT and SOT switching in perpendicular magnetic tunnel junctions: Role of the pulse-end magnetization state

This paper proposes and analyzes a field-free switching scheme in perpendicular magnetic tunnel junctions where cooperative spin-orbit torque and spin-transfer torque drive ultrafast, reliable magnetization reversal by ensuring the pulse-end state is sufficiently close to the target direction to prevent back-switching, thereby establishing the pulse-end magnetization state as the decisive factor for switching performance.

Original authors: Zeyi Zhu, Ding Ma, Kai Zhang, Mingwei Wu, Yi Huang, Jiarui Niu, Shuai Zhang, Mengdi Yin, Ka-San Mak, Lina Chen, Bo Liu, Tiejun Zhou, Yongbing Xu, Ronghua Liu

Published 2026-10-09

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Zeyi Zhu, Ding Ma, Kai Zhang, Mingwei Wu, Yi Huang, Jiarui Niu, Shuai Zhang, Mengdi Yin, Ka-San Mak, Lina Chen, Bo Liu, Tiejun Zhou, Yongbing Xu, Ronghua Liu

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

In the world of modern computing, the race to build faster, more efficient memory is a constant struggle. Engineers are constantly looking for ways to store data without electricity, creating memory that remembers its state even when the power is cut. One promising technology is called magnetic random-access memory, which stores information by pointing tiny magnetic arrows in different directions. To write new data, these arrows must be flipped quickly and reliably. A leading method for doing this uses a phenomenon called spin-orbit torque, where an electric current flowing through a heavy metal layer pushes on the magnetic layer to flip it. However, this method has a stubborn flaw: without an external magnetic field to help guide the way, the magnetic arrow often wobbles near the middle of its flip and then snaps back to its original position, causing the write operation to fail. This instability creates a narrow window where writing is possible, limiting the speed and reliability of the device.

Researchers at Nanjing University and their collaborators have proposed a solution to this problem by combining two different forces to guide the magnetic arrow. Instead of relying on a single push, they use a dual-engine approach. One force, generated by the spin-orbit torque, acts like a strong shove that knocks the magnetic arrow off its stable vertical position and sends it swinging wildly. The second force, known as spin-transfer torque, acts more like a steady hand that guides the arrow toward its new destination. By carefully balancing these two forces, the team discovered they could flip the magnetic state without needing any external magnetic field. Their work, detailed in a series of computer simulations, reveals that the key to success is not just how hard you push, but exactly where the arrow is when the push stops.

The researchers modeled a tiny magnetic disk, just 15 nanometers wide, sitting on top of a heavy metal channel. In their simulation, they applied two currents simultaneously: one flowing through the metal channel to generate the first force, and another flowing directly through the magnetic disk to generate the second. They found that the first force is excellent at getting the magnet moving, creating a large, energetic swing that takes it away from its starting point. However, if this were the only force, the magnet might swing too far or not far enough, and when the current stops, it could easily fall back to where it started. The second force solves this by constantly pulling the magnet toward the opposite side as it swings. This ensures that by the time the current pulse ends, the magnet is already leaning heavily toward its new home, making it much harder for it to slip back.

A critical discovery in this study is that the timing of the current pulse is just as important as its strength. The simulations showed that if the pulse is too short, the magnet might be swinging vigorously but hasn't yet crossed the invisible line that separates the old state from the new one. When the power cuts off in this situation, the magnet relaxes back to its original position, and the write fails. Conversely, if the pulse lasts just long enough for the magnet to cross that line and settle into the new state's territory, the switch is successful. The team found that this "crossing point" is not a simple straight line; it is a complex boundary where the two forces interact in a non-linear way. Increasing one force does not simply add to the other; instead, they work together in a specific balance to overcome the magnet's natural resistance to change.

The study also explored how the material properties of the magnet itself affect this process. They found that if the magnet is very "sticky," meaning it resists changes in motion, a stronger current is needed to get it moving. However, once the magnet is moving, having more of this stickiness actually helps it settle down faster, reducing the time it takes to complete the write operation. Similarly, making the magnet more resistant to being tilted away from its vertical axis requires a stronger push, but once it is flipped, that same resistance helps it snap quickly into its new position. The researchers used a mathematical model to predict these behaviors, and their predictions matched the detailed computer simulations very closely, confirming that their understanding of the physics is sound.

Perhaps the most significant finding is that the speed of the memory device depends heavily on the state of the magnet at the exact moment the current stops. If the current is stopped while the magnet is still far from its target, it has to travel the rest of the way on its own, which takes time and risks failure. But if the current is tuned so that the magnet is already close to its target when the pulse ends, it arrives there almost instantly. This means that simply increasing the power of the current is not always the best way to make the memory faster. Instead, finding the right mix of the two forces is crucial. The simulations showed that by adjusting the balance between the two currents, the researchers could steer the magnet to a position that minimizes the time needed to finish the flip.

This work suggests a clear path forward for designing better memory chips. By using a combination of two different torque mechanisms, engineers can create devices that switch faster and more reliably without needing bulky external magnets. The research highlights that the final outcome of a switching event is determined not just by the peak force applied, but by the precise trajectory the magnet follows and where it ends up when the energy is removed. This insight shifts the focus from simply making the currents stronger to carefully engineering the pulse duration and the balance of forces to ensure the magnet lands exactly where it needs to be. The findings provide a practical guide for building the next generation of ultrafast, non-volatile memory that could one day power everything from smartphones to advanced computer processors.

Technical Summary: Cooperative STT and SOT Switching in Perpendicular Magnetic Tunnel Junctions

Problem Statement
Spin–orbit torque MRAM (SOT-MRAM) is a leading candidate for next-generation nonvolatile memory due to its high speed, endurance, and architectural compatibility. However, conventional field-free SOT switching in perpendicular magnetic anisotropy (PMA) systems faces a fundamental reliability challenge. In standard geometries, the magnetization often remains near the in-plane region at the termination of the current pulse. Consequently, the final state is highly sensitive to post-pulse relaxation dynamics. The presence of an effective in-plane field can induce oscillatory precession that drives the magnetization back across the film plane, resulting in "back-switching" where the device returns to its initial state. This phenomenon severely constrains the effective write-current window and compromises write reliability, particularly at high current densities.

Methodology
The authors propose and investigate a field-free writing scheme that synergistically combines perpendicular spin-transfer torque (STT) and in-plane spin-orbit torque (SOT) in a three-terminal perpendicular magnetic tunnel junction (p-MTJ). The study employs two primary approaches:

Micromagnetic Simulations: Using OOMMF, the authors solve the Landau–Lifshitz–Gilbert (LLG) equation including both STT and SOT terms. The simulated device consists of a 15 nm diameter CoFeB free layer (1 nm thick) with specific material parameters ( $M_s = 1.3 \times 10^6$ A/m, $\alpha = 0.1$ , $K_u = 1.35 \times 10^6$ J/m $^3$ ). The fixed layer and heavy-metal underlayer are modeled implicitly via effective spin-torque terms.
Analytical Macrospin Modeling: An analytical model is derived based on the stability analysis of the current-induced equilibrium state. This model linearizes the LLG equation to determine the critical conditions for dynamic anti-damping instability and static instability, providing a theoretical framework for the switching boundaries.

Key Contributions and Results
The study reveals a cooperative mechanism where STT and SOT play distinct, non-equivalent dynamical roles:

Distinct Dynamical Roles: The SOT component (driven by in-plane current) primarily provides transverse excitation, driving the magnetization out of its initial potential well and into large-angle precession near the in-plane region. Conversely, the perpendicular STT component biases the dynamical trajectory toward the target reversed state ( $-z$ ), assisting in final-state selection and stabilizing the outcome.
Nonlinear Switching Boundary: The switching boundary in the $J_{STT}$ – $J_{SOT}$ parameter space is distinctly nonlinear. This indicates that the torques cannot be described by a simple linear superposition. Deterministic reversal occurs only if the combined pulse drives the magnetization across the dynamical separatrix before the pulse terminates.
Pulse-End State Dependence: A critical finding is that the magnetization state at the moment the pulse ends is the decisive factor for switching speed and reliability. Successful switching requires the magnetization to enter the reversed-state attraction basin before the pulse termination. If the pulse ends while the magnetization is still near the in-plane region, post-pulse relaxation may lead to failure or back-switching.
Parameter Dependencies:
- Pulse Duration: There is a sharp transition in switching success based on pulse duration. A marginal increase in duration (e.g., 0.01 ns) can allow the magnetization to cross the separatrix, converting a failure into a success.
- Material Parameters: Increasing the Gilbert damping constant ( $\alpha$ ) or the anisotropy constant ( $K_u$ ) raises the critical current thresholds. However, within the switching regime, higher $\alpha$ and $K_u$ can actually reduce the total switching time by accelerating the decay of residual oscillations and increasing the precessional frequency toward the equilibrium state.
- STT-SOT Balance: While increasing SOT drives precession, increasing STT more effectively biases the trajectory toward the final state. Optimizing the STT-to-SOT ratio allows for faster switching by ensuring the magnetization is closer to the target equilibrium at pulse termination, thereby minimizing post-pulse relaxation time.

Significance
The paper establishes that in coupled STT–SOT systems, the pulse-end magnetization state, rather than the instantaneous torque amplitude, is the primary determinant of switching speed and reliability. The proposed cooperative mechanism offers a physical framework for achieving robust, field-free switching in perpendicular magnetic heterostructures. By demonstrating that precise control of the pulse-end trajectory—specifically by balancing STT and SOT contributions—can minimize terminal relaxation time, the work provides concrete guidelines for designing optimized STT-SOT-assisted MRAM devices with enhanced write margins, accelerated switching speeds, and improved operational reliability. The analytical model developed further offers a unified theoretical description of the critical conditions for this cooperative switching, distinguishing between dynamic and static instability branches.

More like this