Finite element approximations of the stochastic Benjamin-Bona-Mahony equation with multiplicative noise

This paper establishes the existence, uniqueness, and stability of solutions to the stochastic Benjamin-Bona-Mahony equation with multiplicative noise and derives optimal strong error estimates for a fully discrete finite element approximation under bounded noise, as well as sub-optimal convergence rates in probability for general noise via a localization technique.

The Gist

Imagine you are watching a long, rolling wave in the ocean. In a perfect, calm world, you could predict exactly how that wave will move using a set of rules called the Benjamin-Bona-Mahony (BBM) equation. It's like a recipe for how water waves travel, bounce, and interact with each other.

But the real world isn't calm. The wind gusts, the rain falls, and tiny, unpredictable ripples disturb the water. These are random noises. When we add this randomness to our wave recipe, we get the Stochastic BBM Equation. It's like trying to predict the wave's path while someone is constantly throwing pebbles into the water at random times.

This paper is about building a computer simulation to solve this messy, unpredictable wave problem. Here is the story of how the authors did it, broken down into simple concepts:

1. The Problem: A Wave in a Storm

The authors are trying to model a wave that is being pushed around by random forces (multiplicative noise).

• The Challenge: The "noise" isn't just a constant drizzle; it depends on how big the wave is. If the wave gets huge, the random kicks get stronger. This makes the math incredibly difficult because the rules change depending on the wave's size.
• The Goal: They want to create a computer program that can calculate the wave's shape at any future time, even with all this chaos, and prove that the computer's answer is actually close to the "true" answer.

2. The Tools: Grids and Time Steps

To solve this on a computer, you can't just write down the answer; you have to break the problem into tiny pieces.

• Spatial Grid (The Mesh): Imagine laying a fine fishing net over the ocean. The computer calculates the wave height at every knot in the net. This is called the Finite Element Method. The finer the net (more knots), the more accurate the picture.
• Time Steps: The computer doesn't see time as a smooth flow; it sees it as a series of snapshots. It calculates the wave at 1:00, then 1:01, then 1:02. This is the Euler-Maruyama scheme.

3. The Secret Weapon: A "Damping" Brake

One of the biggest hurdles in this math is that the wave could theoretically grow infinitely large, causing the computer to crash.

• The Analogy: Imagine the wave is a car speeding down a hill. Without brakes, it might go too fast to control. The authors added a damping term (a friction brake) to their mathematical model.
• Why it helps: This brake doesn't change the physics of the wave significantly; it just keeps the energy from exploding. It allows them to prove that the wave stays "tame" enough for the math to work. They showed that even if you remove the brake later, the math still holds up.

4. Two Scenarios: The "Safe" Wave and the "Wild" Wave

The authors tested their computer program under two different conditions:

• Scenario A: The Bounded Noise (The Safe Wave)
  Imagine the random pebbles thrown at the wave are always small and predictable. They never get bigger than a certain size.

  • The Result: The computer simulation is highly accurate. The error (the difference between the computer and reality) shrinks perfectly as you make the net finer and the time steps smaller. It's like a perfect recipe.

• Scenario B: The General Noise (The Wild Wave)
  Imagine the pebbles can be any size. Sometimes they are tiny, but occasionally, a giant boulder might hit the wave. The math gets messy because the "brake" might not be enough to stop a giant boulder.

  • The Result: The authors used a clever trick called Localization. Instead of trying to guarantee the computer is right every single time, they proved it is right most of the time (with very high probability).
  • The Analogy: It's like saying, "If you drive this car, 99% of the time you will arrive safely. The 1% chance of a crash is so tiny we can ignore it for practical purposes." The math shows the computer works well, but the error rate is slightly higher than in the "Safe" scenario.

5. The Proof: "It Works!"

After doing all the heavy math (which involves complex inequalities and probability theory), they ran numerical experiments.

• They created a virtual ocean and threw random pebbles at it.
• They compared their computer's prediction against a "super-accurate" reference solution (like a high-resolution satellite photo).
• The Outcome: The computer's predictions matched the theory perfectly. The errors decreased exactly as they predicted when they made the simulation finer.

Summary

In short, this paper is a guidebook for how to simulate chaotic, random waves on a computer.

1. They built a mathematical safety net (stability estimates) to keep the wave from going crazy.
2. They created a digital fishing net (finite elements) to catch the wave's shape.
3. They proved that whether the random noise is gentle or wild, their computer method gives a reliable answer.
4. They showed that for the wild noise, the method works almost always, which is good enough for real-world engineering and science.

It's a triumph of turning a chaotic, unpredictable natural phenomenon into something a computer can reliably calculate.

Technical Summary

Here is a detailed technical summary of the paper "Finite Element Approximations of the Stochastic Benjamin-Bona-Mahony Equation with Multiplicative Noise" by Nguyen, Thieu, and Vo.

1. Problem Statement

The paper addresses the numerical analysis of the Stochastic Benjamin-Bona-Mahony (BBM) equation driven by multiplicative noise in two spatial dimensions. The equation is given by:
$$du - d(\Delta u) + [\nu u + \text{div}(F(u))] dt = G(u) dW(t)$$
where:

• $u$ represents the wave amplitude.
• $\nu \geq 0$ is a linear damping coefficient (introduced specifically to aid stability analysis).
• $F(u) = \langle u + \frac{1}{2}u^2, u + \frac{1}{2}u^2 \rangle^T$ is the nonlinear transport term.
• $G(u)$ is a diffusion operator depending on the solution $u$, introducing multiplicative noise via a real-valued Wiener process $W(t)$.
• The domain $D$ is a 2D square with periodic boundary conditions.

Key Challenges:

• Non-Lipschitz Nonlinearity: The convective term is non-Lipschitz, preventing the direct application of standard SPDE techniques relying on global Lipschitz or monotonicity assumptions.
• Multiplicative Noise: The noise intensity depends on the solution, leading to potential amplification of stochastic perturbations in regions of large amplitude.
• Dispersive Structure: The BBM operator complicates the derivation of classical energy estimates compared to parabolic equations.
• Lack of Literature: While deterministic BBM and stochastic BBM with additive noise have been studied, rigorous strong convergence analysis for fully discrete schemes with multiplicative noise was previously unavailable.

2. Methodology

A. Analytical Framework (Well-Posedness)

The authors first establish the theoretical foundation for the continuous problem:

1. Variational Framework: They define a suitable variational setting in Sobolev spaces ($H^1, H^2$) with periodic boundary conditions.
2. Existence and Uniqueness: Using a regularization approach (adding a higher-order dissipation term $\epsilon \Delta^2 u$), they construct martingale solutions and prove path-wise uniqueness, establishing the existence of a unique strong solution.
3. Stability Estimates:
   • High-Moment Bounds: Derivation of $L^{2n}$ bounds for the solution in $H^1$ and $H^2$ norms.
   • Exponential Stability: A crucial result showing that if the noise is bounded and damping $\nu > 0$, the solution satisfies exponential moment bounds ($E[\exp(\gamma \|u\|_{H^1}^2)] < \infty$). This is essential for proving optimal error rates.
   • Time Regularity: Proof of Hölder continuity in time for the solution.

B. Numerical Scheme

The paper proposes a Fully Discrete Finite Element Method:

• Spatial Discretization: Conforming finite elements using piecewise linear polynomials ($P_1$) on a uniform triangular mesh ($V_h$).
• Temporal Discretization: The Implicit Euler-Maruyama scheme. The scheme is fully implicit in the linear and nonlinear terms to ensure stability.
• Algorithm: At each time step, a nonlinear system is solved (using a fixed-point iterative solver rather than Newton's method for better performance in stochastic sampling).

C. Error Analysis Strategy

The convergence analysis is split into two cases based on the properties of the noise coefficient $G(u)$:

1. Case 1: Bounded Multiplicative Noise

   • Assumption: $G(u)$ is bounded (e.g., $G(u) = \sin(u)$).
   • Technique: The authors combine the exponential stability of both the continuous and discrete solutions with a discrete stochastic Gronwall inequality.
   • Result: This allows for the derivation of optimal strong error estimates in full expectation ($L^p_\omega L^\infty_t H^1_x$) for moments $p < 4$.

2. Case 2: General Multiplicative Noise (Unbounded)

   • Assumption: $G(u)$ may be unbounded (e.g., $G(u) = u$).
   • Technique: Since exponential moment bounds cannot be guaranteed globally, the authors employ a localization technique. They define high-probability events $\Omega_{\rho, n}$ where the solution and numerical approximation remain bounded by a threshold $\rho$.
   • Result: This leads to sub-optimal convergence rates in probability. The error is bounded on the set $\Omega_{\rho, n}$, and the probability of the complement set vanishes as the time step $k \to 0$.

3. Key Contributions

1. First Rigorous Well-Posedness for Original Stochastic BBM: The paper provides the first systematic proof of existence and uniqueness for the original stochastic BBM equation (without artificial regularization terms like $\Delta u$ added for analytical convenience) with multiplicative noise in 2D.
2. Optimal Strong Convergence for Bounded Noise: The authors derive optimal strong error estimates of order $O(k^{1/2} + h)$ in the $L^p_\omega L^\infty_t H^1_x$ norm for bounded noise, utilizing a novel combination of exponential stability and stochastic Gronwall inequalities.
3. Convergence in Probability for General Noise: For unbounded noise, they establish convergence in probability with sub-optimal rates, overcoming the lack of global exponential stability via localization arguments.
4. Discrete Exponential Stability: A significant theoretical contribution is the proof of discrete exponential stability for the fully implicit Euler-Maruyama scheme, mirroring the continuous case.

4. Main Results

• Theoretical Error Bounds:
  • Bounded Noise: $\left( E \left[ \max_{1 \le n \le M} \|u(t_n) - u_h^n\|_{H^1}^{2q} \right] \right)^{1/2q} \le C(k^{1/2} + h)$.
  • General Noise: $\left( \max_{1 \le n \le M} E \left[ \mathbb{1}_{\Omega_{\rho,n}} \|u(t_n) - u_h^n\|_{H^1}^2 \right] \right)^{1/2} \le C \sqrt{\ln(1/k)} (k^{1/2} + h)$.
• Numerical Experiments:
  • Simulations were conducted on a unit square with $J=400$ Monte Carlo paths.
  • Two noise types were tested: $G(u) = \frac{1}{4}u$ (unbounded) and $G(u) = \frac{1}{10}\sin(1+u)$ (bounded).
  • The results confirmed the theoretical convergence rate of approximately order 1 in space and time (specifically $O(h)$ when $k=h^2$), validating the theoretical predictions.

5. Significance

• Theoretical Advancement: This work bridges a critical gap in the numerical analysis of dispersive stochastic PDEs. It moves beyond additive noise and simplified models to handle the physically relevant, mathematically challenging case of multiplicative noise in the BBM equation.
• Methodological Innovation: The successful application of localization techniques combined with discrete stochastic Gronwall inequalities offers a robust framework that can be extended to other nonlinear dispersive SPDEs (e.g., stochastic KdV, Kuramoto-Sivashinsky) where global Lipschitz conditions fail.
• Practical Utility: The proposed fully implicit finite element scheme is shown to be stable and convergent, providing a reliable tool for simulating wave phenomena subject to random environmental fluctuations where noise intensity scales with the wave amplitude.

In summary, this paper establishes a rigorous mathematical and numerical framework for the stochastic BBM equation with multiplicative noise, proving well-posedness, deriving optimal and sub-optimal error estimates, and validating these results through numerical experiments.