🔬 mesoscale physics

Dynamical control of random telegraph noise in magnetic tunnel junctions

This paper theoretically demonstrates that mode-selective parametric excitation and a proposed nonlinear cooling mechanism can dynamically control random telegraph noise in magnetic tunnel junctions, offering a pathway to accelerate switching speeds for probabilistic computing applications.

Original authors: Mehrdad Elyasi, Shun Kanai, Hideo Ohno, Shunsuke Fukami, Gerrit E. W. Bauer

Published 2026-06-16

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Mehrdad Elyasi, Shun Kanai, Hideo Ohno, Shunsuke Fukami, Gerrit E. W. Bauer

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 a tiny magnetic switch inside a computer chip, so small it's measured in nanometers. This switch, called a Magnetic Tunnel Junction (MTJ), is the heart of a new type of "probabilistic bit" (or p-bit) used for advanced computing. Unlike a standard light switch that is either strictly ON or OFF, this p-bit is like a coin spinning on a table. It constantly flips back and forth between two states (Heads or Tails) randomly. This random flipping is called Random Telegraph Noise (RTN).

The speed at which this coin flips is crucial. If it flips too slowly, the computer is sluggish. If it flips just right, the computer can solve complex problems very fast.

The paper you provided is a theoretical study that figures out how to control the speed of this flipping coin using invisible "microwaves" (like the ones in your kitchen, but tuned to specific magnetic frequencies). Here is the breakdown of their findings using simple analogies:

1. The Problem: The "Brake" on the Coin

Usually, scientists thought the coin flipped based on how hot the environment was (thermal energy). However, the authors discovered that inside these tiny magnets, there are invisible ripples called spin waves (or magnons).

Think of the main magnet as a giant spinning top (the macrospin). The spin waves are like tiny, chaotic ripples on the surface of that top. The paper argues that these ripples act like a brake. When the ripples get too active, they slow down the giant top, making the coin flip much slower than expected.

2. The Solution: Tuning the Ripples with Microwaves

The authors propose using microwaves to actively control these ripples, acting like a conductor directing an orchestra. They found two main ways to do this:

A. Speeding Up the Flip (Heating the Main Spin)

If you aim the microwave energy directly at the main spinning top (the "Kittel mode"), you can make it flip faster.

The Analogy: Imagine pushing a child on a swing. If you push at just the right rhythm (parametric excitation), the swing goes higher and faster.
The Catch: While this makes the "attempt" to flip more frequent, it also lowers the "hill" (energy barrier) the coin has to climb to flip. The paper shows that the lowering of the hill wins out, so the coin actually flips faster overall.

B. Slowing Down the Flip (Exciting the Ripples)

If you aim the microwaves to create those chaotic ripples (spin waves with specific shapes) instead of pushing the main top directly, the coin flips slower.

The Analogy: Imagine the spinning top is trying to stand up, but you are shaking the floor underneath it (creating ripples). The top gets confused and wobbles, taking longer to decide which way to fall.
The Result: The paper confirms that exciting these specific ripples acts as a brake, slowing down the switching speed.

3. The New Trick: "Cooling" the Ripples

The most creative part of the paper is a proposed method to make the coin flip even faster by "cooling" the ripples.

The Analogy: Imagine the chaotic ripples on the top are like a crowd of people running around wildly, bumping into the top and slowing it down. The authors propose using a specific high-frequency microwave to act like a vacuum cleaner or a siphon.
How it works: They suggest using a high-frequency "pump" to suck the energy out of the chaotic ripples and transfer it elsewhere. By "cooling" these ripples (making them less active), the brake is released.
The Outcome: This "nonlinear cooling" could theoretically make the coin flip ten times faster than it does naturally, without needing to change the physical size of the chip.

4. The Experimental Plan

The paper doesn't just guess; they ran complex computer simulations to prove these ideas work.

They modeled a tiny magnetic disk (20 nanometers wide).
They simulated hitting it with different microwave frequencies.
The Results:
- Hitting the main spin directly = Faster flipping.
- Hitting the ripples directly = Slower flipping.
- Using the "cooling" trick = Much faster flipping.

Summary

In short, this paper is a recipe book for controlling the speed of tiny magnetic switches. It tells us that:

Ripples (spin waves) usually slow things down.
Microwaves can be used to either push the main spin (speeding it up) or create more ripples (slowing it down).
A new "cooling" technique using microwaves could remove the ripples entirely, making these switches incredibly fast.

The authors conclude that by mastering these microwave "knobs," we can build faster, more efficient probabilistic computers, provided we can build the experiments to test these specific microwave frequencies.

Technical Summary: Dynamic Control of Random Telegraph Noise in Magnetic Tunnel Junctions

Problem Statement
Random telegraph noise (RTN) arising from stochastic magnetization switching in magnetic tunnel junctions (MTJs) is a critical mechanism for probabilistic computing (p-bits). While the N´eel-Arrhenius law describes the average switching time based on thermal activation over an energy barrier, recent experimental observations indicate that the attempt frequency ( $\tau_0^{-1}$ ) in small magnetic particles is slower than predicted by rigid macrospin dynamics. The authors identify that nonlinear interactions between the macrospin and spin waves (magnons) with finite wave numbers act as a "brake" on RTN, reducing the switching frequency. Current passive methods to engineer switching dynamics (e.g., tuning magnetic fields or anisotropy) are insufficient for actively manipulating these magnon fluctuations to optimize operation speeds.

Methodology
The paper employs a combination of analytical theory, numerical simulations, and micromagnetic modeling to investigate the dynamic control of RTN:

Analytical Modeling: The authors utilize the Fokker-Planck equation and Kramers' escape theory to derive switching rates. They model the system using a macrospin Hamiltonian coupled to magnon modes, incorporating parametric excitation via microwave drives. Nonlinearities, specifically self-Kerr interactions and Suhl (four-magnon) interactions, are included to address the divergence of linear theories at high drive powers.
Stochastic LLG Simulations: To validate the analytical approximations, the authors solve the stochastic Landau-Lifshitz-Gilbert (LLG) equation for a 20 nm diameter CoFeB disc. This allows for the assessment of switching rates ( $f_s$ ) and autocorrelation functions under various parametric drive conditions.
Micromagnetic Calculations: A spatially resolved micromagnetic code is used to simulate a 2D lattice of magnetic elements. This approach captures the interplay between exchange and dipolar interactions, verifying the single-domain assumption and testing the model's limits regarding domain wall nucleation and non-uniform spin wave excitations.
Cooling Mechanism Modeling: A theoretical model for "in-situ" mode-selective cooling is developed using a four-magnon interaction Hamiltonian, where a high-frequency magnon mode is driven to cool a target low-frequency mode ( $\pm K$ ) via anti-Stokes-like scattering.

Key Contributions and Results

Parametric Excitation of the Kittel Mode (Macrospin):
The authors demonstrate that parametrically exciting the uniform Kittel mode (macrospin) using microwaves at twice its resonance frequency ( $2\omega_0$ ) can increase the RTN switching frequency. While the drive increases the attempt time ( $\tau_0$ ) due to squeezing of fluctuations, it simultaneously lowers the effective energy barrier ( $E_B$ ) via the shift of minima relative to the saddle point. The reduction in the barrier dominates, leading to a net acceleration of the switching rate ( $f_s$ ). Numerical results confirm that $f_s$ increases monotonically with drive power until coherent ringing obscures the stochastic switching.
Excitation of Finite-Momentum Spin Waves:
Conversely, parametric excitation of spin waves with finite wave vectors ( $\pm K$ ) suppresses the RTN frequency. The interaction between the macrospin and these spin waves (mediated by Suhl interactions) increases the effective damping of the macrospin correlation, thereby increasing the attempt time and reducing the switching rate. The relationship is non-monotonic: at low powers, $f_s$ decreases as spin wave population grows; at very high powers, the excitation of the Kittel mode itself may eventually compete, though the model suggests a breakdown of the single-pair approximation in the strongly driven regime.
Mode-Selective Nonlinear Cooling:
A novel mechanism is proposed to accelerate switching by actively "cooling" the specific spin wave modes that hinder RTN. By parametrically driving a high-frequency magnon mode ( $\vec{k}$ ) via a four-magnon interaction term, the population of the target low-frequency modes ( $\pm K$ ) can be reduced below their thermal equilibrium values. Theoretical calculations indicate this nonlinear cooling reduces the magnon occupation $\langle n_{\pm K} \rangle$ , thereby removing the "brake" on the macrospin and potentially increasing the RTN frequency by an order of magnitude.
Experimental Pathways:
The paper outlines concrete experimental proposals to realize these effects. It suggests using microwave striplines or voltage-controlled magnetic anisotropy (VCMA) via AC gate voltages to drive the necessary parametric excitations. Specifically, modulating the hard-axis anisotropy is proposed for in-plane magnetized samples, while modulating the easy-axis anisotropy is suggested for perpendicular samples.

Significance and Claims
The authors claim to provide a theoretical framework and numerical validation for actively controlling the speed of probabilistic bits by manipulating the interplay between macrospins and magnons. The work highlights that the "braking" effect of spin waves is a fundamental limitation in small MTJs that can be overcome not just by passive design, but by active, mode-selective manipulation.

Acceleration: Parametric excitation of the macrospin and the proposed nonlinear cooling of spin waves are presented as viable strategies to maximize the operation speed of MTJ-based p-bits.
Suppression: The ability to suppress switching via spin wave excitation offers a mechanism to tune the stochasticity of p-bits.
Limitations: The authors modestly note that their single-domain model is appropriate for MTJs with radii smaller than the domain wall width (~30 nm for CoFeB). They acknowledge that for larger devices, domain wall nucleation and motion will become the dominant reversal mechanism, requiring future work to address these complications.

The paper concludes that these active control methods, utilizing existing microwave or voltage technologies, offer a pathway to engineer the switching dynamics of magnetic devices for next-generation probabilistic computing.