🔢 mathematics

Strong convergence of finite element schemes for the stochastic Landau--Lifshitz--Bloch equation

This paper establishes strong convergence with explicit rates for semi-implicit and implicit fully discrete finite element schemes approximating the stochastic Landau--Lifshitz--Bloch equation in one and two dimensions, utilizing localized error estimates and new exponential moment bounds that also yield stability and uniqueness results.

Original authors: Agus L. Soenjaya

Published 2026-02-23

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Agus L. Soenjaya

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

The Big Picture: Magnetism in a Hot, Chaotic World

Imagine a tiny, invisible army of compass needles (magnetic particles) packed tightly together inside a hard drive. When the computer is cool, these needles all point in the same direction, creating a strong, stable magnetic signal that stores your data.

But what happens when the hard drive gets hot?

Thermal Jitters: The heat makes the needles shake and vibrate wildly.
Random Noise: Imagine a chaotic wind blowing through the needles, pushing them in random directions.
The Goal: We want to predict exactly how this army of needles will move over time so we can build better, faster hard drives (like the "HAMR" technology mentioned in the paper).

The Stochastic Landau–Lifshitz–Bloch (sLLB) Equation is the master rulebook that describes this chaotic dance. It's a very complicated math formula that mixes smooth physics (how magnets naturally want to align) with random chaos (the heat and noise).

The Problem: The Formula is Too Hard to Solve

The problem is that this rulebook is so complex that no one can solve it with a simple pencil and paper. It's like trying to predict the exact path of every single raindrop in a hurricane.

So, scientists use computers to approximate the answer. They break the continuous space (the hard drive) into tiny Lego blocks (a "mesh") and simulate the movement step-by-step. This is called a Finite Element Scheme.

However, previous methods had two big flaws:

They were only "probably" right: They could say, "99% of the time, our answer is close," but they couldn't guarantee it for every possible scenario.
They were slow to converge: To get a more accurate answer, you had to make the Lego blocks incredibly tiny, which takes forever to compute.

The Solution: A New, Smarter Way to Simulate

The author of this paper, Agus Soenjaya, has developed new, smarter ways to run these computer simulations. Think of it as upgrading from a shaky, hand-drawn map to a high-definition GPS.

Here are the three main upgrades he introduced:

1. The "Safety Net" (Exponential Moment Bounds)

In the old methods, if the random wind (noise) got too strong, the simulation could go wild and the math would break.

The Analogy: Imagine trying to balance a broom on your finger. If the wind is too strong, it falls.
The Innovation: The author proved a new mathematical "safety net." He showed that even with the random wind, the system has a natural tendency to stay within certain limits. He proved that the "energy" of the system doesn't explode to infinity, even in the worst-case scenarios. This safety net allows the computer to trust the simulation even when things get chaotic.

2. The "Two-Speed" Strategy (Semi-Implicit vs. Implicit)

The paper tests two different ways to calculate the next step in the simulation:

The Semi-Implicit Method (The Fast Runner): This is like taking a quick guess based on where you were a moment ago. It's fast and works well for 2D surfaces (like a square hard drive), but it's a bit less precise.
The Implicit Method (The Careful Climber): This is like looking at where you will be before you take the step. It's harder to calculate (like solving a puzzle before moving), but it is much more stable.
The Result: The author showed that for 1D problems (like a thin wire), this "Careful Climber" method is optimal. It gives the most accurate answer possible for the amount of computing power used.

3. The "Strong Convergence" Guarantee

This is the most technical part, but here is the simple version:

Old Way: "If you run this simulation 1,000 times, the average result will be close to the truth."
New Way: "If you run this simulation, the result will be close to the truth every single time, and we can tell you exactly how close it is."
The Metaphor: Imagine throwing darts at a board.
- Old method: "On average, you hit the bullseye."
- New method: "Every single dart you throw will land within 1 millimeter of the bullseye."

Why Does This Matter?

Better Hard Drives: By understanding exactly how magnets behave at high temperatures, engineers can design hard drives that store more data and write it faster without melting or losing data.
Mathematical Certainty: The author didn't just guess; he proved that these computer methods are mathematically sound. He showed that as you make the "Lego blocks" smaller, the answer gets closer to the truth at a predictable speed.
Stability: He proved that if you start with two slightly different magnetic states, they will eventually settle into the same behavior (unless the noise is too crazy). This means the system is reliable.

Summary in a Nutshell

The paper is about building a better, more reliable GPS for the chaotic world of hot magnets. The author proved that his new computer algorithms don't just work "most of the time," but work strongly and reliably every time, giving engineers the confidence to build the next generation of super-fast storage devices. He did this by creating a mathematical "safety net" that keeps the chaotic equations from running wild.

1. Problem Statement

The paper addresses the numerical approximation of the Stochastic Landau–Lifshitz–Bloch (sLLB) equation, which models the dynamics of magnetization in ferromagnets at temperatures above the Curie temperature. In this regime, thermal fluctuations are significant, necessitating a stochastic model.

The governing equation (1.1) is a vector-valued quasilinear stochastic partial differential equation (SPDE) defined on a bounded domain $D \subset \mathbb{R}^d$ ( $d=1, 2$ ):
$du = \left( \kappa_1 \Delta u + \gamma u \times \Delta u - \kappa_2(1 + \mu |u|^2)u \right) dt + (\kappa_1 g + \gamma u \times g) \circ dW$
where $u$ is the magnetization vector, $W$ is a Wiener process, and the noise term is interpreted in the Stratonovich sense.

Key Challenges:

Nonlinearity: The equation contains non-globally Lipschitz and non-monotone nonlinearities (specifically the cross product and the cubic term), which prevent the direct application of standard convergence theories for monotone SPDEs.
Regularity: For $d=2$ , higher-order regularity results for the exact solution are not known, requiring a regularized version of the equation (sLLBar) for numerical analysis.
Convergence: Previous works (e.g., [14]) established convergence in probability with suboptimal rates and strong convergence only up to a discrete stopping time. The goal is to establish strong convergence in $L^2(\Omega)$ over the entire sample space with explicit, sharper rates.

2. Methodology

The author employs a combination of probabilistic analysis, functional analysis, and finite element methods (FEM).

A. Analytical Preliminaries

Existence and Uniqueness: The paper relies on known results for the existence of unique pathwise strong solutions for $d \le 2$ .
Exponential Moment Bounds (New Contribution): A critical component of the analysis is the derivation of new exponential moment bounds for the exact solution (Theorem 3.5). These bounds control the growth of the solution in $L^2$ $L^{2}$ and $H^1$ $H^{1}$ norms.
- These bounds are used to prove mean-square exponential stability and uniqueness of the invariant measure for $d=1$ under small noise assumptions.
- They are essential for deriving strong convergence rates without relying on stopping times.

B. Numerical Schemes

The paper analyzes three specific finite element schemes:

Semi-implicit Scheme for $d=1$ (C0-conforming):
- Uses linear Lagrange elements ( $P_1$ ).
- Treats the Laplacian and linear terms implicitly, while the cross-product nonlinearity is treated semi-implicitly (using the previous time step).
- Analyzed in Section 4.
Semi-implicit Scheme for $d=2$ (C1-conforming):
- Applied to a regularized version of the sLLB equation (sLLBar) to handle the lack of $H^3$ regularity in 2D.
- Uses $C^1$ -conforming elements (e.g., Hsieh–Clough–Tocher).
- Analyzed in Section 5.
New Implicit Scheme for $d=1$ :
- A fully implicit scheme where the nonlinear cross-product term is also treated implicitly.
- This scheme offers better stability properties, allowing for error estimates in the stronger $H^1$ norm.
- Analyzed in Section 6.

C. Error Analysis Technique

Localization: The analysis utilizes a localization technique (inspired by [7, 4]) to define a "good" set $\Omega_{\kappa, m}$ where the solution remains bounded.
Decomposition: The error is decomposed into a projection error (deterministic) and a discrete error (stochastic).
Global Convergence: By combining the exponential moment bounds of the exact solution with the localized error estimates, the author extends the convergence from a subset of the probability space to the entire space $L^2(\Omega)$ .

3. Key Contributions and Results

A. Theoretical Advances

Exponential Moment Bounds: Theorem 3.5 establishes new bounds of the form $E[\exp(\alpha \|u\|^2)] < \infty$ . This is a novel result for the sLLB equation and implies mean-square exponential stability for $d=1$ under small noise.
Strong Convergence in $L^2(\Omega)$ :
- The paper proves strong convergence for the semi-implicit schemes (previously only known up to stopping times).
- Rates for $d=1$ (Semi-implicit): $O(h + k^{1/2-\delta})$ in probability and strong $L^2(\Omega)$ .
- Rates for $d=2$ (Regularized): $O(h^2 + k^{1/2-\delta})$ in probability and strong $L^2(\Omega)$ .
Optimal Rates for New Implicit Scheme ( $d=1$ ):
- The proposed implicit scheme achieves an optimal convergence rate in the $H^1$ norm: $O(h + k^{1/2-\delta})$ in probability.
- Strong convergence is established with a rate of $O(|\log(h^2 + k^{1/2-\delta})|^{-r})$ .

B. Comparison with Previous Work

Sharpening Rates: The convergence rates in probability are improved compared to [14] (e.g., from $O(h^{1/2})$ to $O(h)$ for $d=1$ ).
Removing Stopping Times: Unlike [14], the strong convergence results hold over the full probability space, not just up to a discrete stopping time.

4. Numerical Experiments

Section 7 validates the theoretical findings using the FEniCS framework:

Simulation 1 ( $d=1$ , Implicit Scheme): Confirms the theoretical convergence orders. The $H^1$ error converges with order 1 in space and 1/2 in time, matching Corollary 6.5. The $L^2$ error shows even higher convergence rates.
Simulation 2 ( $d=2$ , Semi-implicit Scheme): Although the rigorous error analysis for $d=2$ relies on regularization, numerical experiments suggest the method exhibits the expected convergence orders.
Simulation 3 (Stability): Verifies the exponential stability result (Proposition 3.6). For small noise, the difference between two solutions with different initial data decays exponentially. For large noise, this contraction is lost, confirming the "small noise" condition is necessary.

5. Significance

Rigorous Foundation: This work provides the first rigorous proof of strong convergence with explicit rates for finite element schemes applied to the sLLB equation, overcoming the difficulties posed by non-monotone nonlinearities.
Practical Implications: The results support the use of these numerical schemes for simulating Heat-Assisted Magnetic Recording (HAMR) and other high-temperature micromagnetic applications where thermal noise is critical.
Theoretical Tools: The derivation of exponential moment bounds for the sLLB equation is a significant theoretical contribution that may be applicable to other stochastic micromagnetic models.
Algorithmic Improvement: The introduction of the implicit scheme for $d=1$ offers a path to optimal convergence rates in stronger norms, which is crucial for accurately capturing gradient-dependent physical quantities.

In summary, the paper bridges the gap between the theoretical existence of solutions and the practical reliability of numerical approximations for stochastic micromagnetics, providing sharper error estimates and stronger convergence guarantees than previously available.