🔢 mathematics

Uniform large deviation principles for the stochastic heat equation over unbounded sets of initial data

This paper establishes uniform large deviation principles for the stochastic heat equation with multiplicative space-time white noise over highly unbounded sets of initial data, including $L^1$ -bounded subsets, by proving novel well-posedness and regularity results that enable the analysis of exit times in mass-conserving physical systems.

Original authors: Michael Salins, Esmée Theewis

Published 2026-08-21

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Michael Salins, Esmée Theewis

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 world where heat, or perhaps a chemical concentration, spreads through a material not in a smooth, predictable wave, but in a jittery, chaotic dance driven by invisible, random kicks. This is the realm of the stochastic heat equation, a mathematical model used to describe how things diffuse when they are constantly poked by unpredictable noise. In physics and engineering, this model helps us understand everything from the movement of pollutants in the air to the spread of heat in a metal rod. For decades, mathematicians have studied what happens when this random noise is very small. They wanted to know: if the noise is tiny, how likely is it that the system will suddenly behave in a wildly unexpected way, straying far from its normal path? This question is about "rare events," and the tools used to answer them are called large deviation principles. These tools allow scientists to calculate the precise odds of these unlikely jumps, which is crucial for designing safe systems that must not fail, even when faced with extreme, improbable conditions.

The challenge has always been that these calculations work best when the starting conditions of the system are very well-behaved and tightly controlled. In the real world, however, things are rarely so neat. A system might start with a massive amount of mass concentrated in a tiny spot, or a distribution that is rough and uneven. Previous mathematical methods could only guarantee their predictions if the starting conditions were smooth and bounded in a very strict sense. If the starting data was too rough or too large, the old methods broke down, leaving a gap in our understanding of how these systems behave under stress.

In a new study, researchers Michael Salins and Esmée Theewis have bridged this gap. They have developed a way to predict the behavior of these noisy heat systems even when the starting conditions are extremely rough and unbounded. Their work focuses on a specific type of noise that grows as the system itself grows, a scenario that is common in physical reality but mathematically difficult to handle. The team proved that their new method works uniformly across a vast range of starting points, including those that are so large or irregular that they would have been impossible to study before. This means they can now calculate the odds of rare events for systems starting with a bounded total amount of "mass," even if that mass is distributed in a jagged, uneven way that defies simple description.

The significance of this achievement lies in its ability to handle the "messy" reality of physical systems. In many applications, such as the diffusion of molecules, the total amount of substance is conserved, but its distribution can be highly uneven. The researchers showed that their mathematical framework remains robust even when the starting distribution is only bounded in a very weak sense, allowing for initial states that are effectively infinite in the standard measures used by previous theories. They demonstrated that the system's behavior remains predictable in terms of rare events, provided the noise does not grow too quickly relative to the system's size. This is a major step forward because it moves the theory from the realm of idealized, smooth scenarios to the complex, irregular conditions found in nature.

The researchers achieved this by carefully analyzing how the heat equation smooths out rough edges over time. They realized that even if the starting point is very jagged, the heat equation itself acts as a powerful filter, smoothing the data as it evolves. By tracking exactly how this smoothing interacts with the random noise and the starting conditions, they were able to prove that the system's path remains stable enough to apply their new prediction tools. They established that for a wide class of starting data, including those with a finite total mass but infinite peak density, the probability of the system taking a rare, extreme path can still be calculated with high precision. This allows for a much deeper understanding of "exit times," or how long it takes for a system to drift out of a safe zone, which is a critical factor in risk assessment for physical and biological systems.

This work does not just extend the range of numbers that can be plugged into an equation; it fundamentally changes the types of questions that can be asked. By proving that these predictions hold true for unbounded sets of initial data, the authors have opened the door to studying systems where the starting conditions are not just large, but potentially infinite in their local intensity. They have shown that the mathematical machinery required to understand rare events is far more resilient than previously thought. The findings suggest that even in the most chaotic and irregular starting scenarios, the underlying laws governing these systems are strong enough to provide reliable predictions about their most extreme behaviors. This brings the abstract mathematics of large deviations closer to the messy, unpredictable reality of the physical world, offering a new lens through which to view the stability of complex systems.

Technical Summary: Uniform Large Deviation Principles for the Stochastic Heat Equation Over Unbounded Sets of Initial Data

Problem Statement
This paper investigates the small-noise large deviation principles (LDPs) for the stochastic heat equation (SHE) on the one-dimensional torus $D := \mathbb{R}/2\pi\mathbb{Z}$ . The equation is given by:
$\partial_t X^\varepsilon_x = \Delta X^\varepsilon_x + b(X^\varepsilon_x) + \sqrt{\varepsilon}\sigma(X^\varepsilon_x)\dot{W},$
where $W$ is space-time white noise, $\varepsilon > 0$ is a small parameter, and the coefficients $b$ and $\sigma$ satisfy Lipschitz continuity and (sub)linear growth conditions. Specifically, $|\sigma(z)| \leq C(1+|z|^\lambda)$ for $\lambda \in (0, 1]$ .

The primary objective is to establish Uniform Large Deviation Principles (ULDPs) for the family of solutions $\{X^\varepsilon_x\}$ as $\varepsilon \to 0$ . Unlike previous works that typically prove uniformity only over compact or bounded sets of initial data in the state space, this paper aims to prove ULDPs over unbounded sets of initial data. Specifically, the authors seek to establish uniformity over sets of initial data that are bounded in an $L^q(D)$ norm but unbounded in the state space topology (e.g., $L^p(D)$ or $C(D)$ ), where $q < p$ (or $q < \infty$ in the continuous case).

Methodology
The authors employ the weak convergence approach to large deviations, originally developed by Budhiraja, Dupuis, and Maroulas, and extended to non-compact sets by the first author in [39]. The core of the methodology involves:

Controlled Equations: Analyzing the convergence of controlled mild solutions $X^{\vare,u}_x$ to the skeleton equation $X^{0,u}_x$ as $\varepsilon \to 0$ . The convergence must be uniform over initial data $x$ in $L^q$ -bounded sets and control functions $u$ in bounded subsets of $L^2(0, T; L^2(D))$ .
Regularity and Well-Posedness: A significant portion of the technical work involves establishing new well-posedness and regularity results for the SHE with rough initial data ( $x \in L^q(D)$ $x \in L^{q} (D)$ where $q$ $q$ may be less than the state space index $p$ $p$ ).
- The authors utilize the smoothing properties of the heat semigroup $S(t)$ , specifically the estimate $\|S(t)x\|_{L^p} \lesssim t^{-\frac{1}{2}(\frac{1}{q} - \frac{1}{p})} \|x\|_{L^q}$ .
- They decompose the solution into the semigroup term $S(\cdot)x$ and integral terms (deterministic and stochastic convolutions).
- Crucially, they prove that while the solution $X^\varepsilon_x$ itself may not lie in the desired path space (e.g., $C([0, T]; L^p)$ ) if $x \in L^q$ with $q < p$ , the remainder term $X^\varepsilon_x - S(\cdot)x$ possesses the necessary regularity.
Estimates: The proofs rely on:
- Deterministic and Stochastic Convolution Estimates: Using Minkowski's inequality, Young's inequality, and the factorization method to bound the integral terms.
- Gronwall Schemes: Establishing uniform bounds on the difference $\|X^{\vare,u}_x - X^{0,u}_x\|$ to satisfy the sufficient criteria for ULDPs.
- Uniform Moment Bounds: Proving that the skeleton solutions $X^{0,u}_x$ satisfy uniform $L^{\lambda \tilde{p}}$ bounds over $L^q$ -bounded initial data, which is essential for controlling the noise term in the limit.

Key Contributions and Results

ULDP on $C([0, T]; C(D))$ (Theorem 0.2):
- The authors prove a ULDP in the uniform topology $C([0, T]; C(D))$ uniformly over initial data bounded in $L^q(D)$ .
- Condition: $2\lambda < q$ .
- Significance: If $\lambda < 1/2$ , this allows for $q=1$ . This establishes a ULDP over $L^1$ -bounded subsets of initial data, which is highly unbounded in the $C(D)$ topology. This is a novel result enabling the study of systems with mass conservation.
ULDP on $C([0, T]; L^p(D))$ (Theorem 0.4):
- The authors prove a ULDP in the $L^p$ -path space uniformly over initial data bounded in $L^q(D)$ .
- Condition: $\frac{2\lambda p}{p+2} < q$ .
- Significance: This condition allows for $q < p$ $q < p$ , meaning the set of initial data is unbounded in the state space $L^p(D)$ $L^{p} (D)$ .
  - If $\lambda \leq 1/2$ , the result holds for $q=1$ for all $p \geq 2$ .
  - If $\lambda \in (1/2, 1)$ , the result holds for $q=1$ for a specific range of $p$ (specifically $p < \frac{2}{2\lambda - 1}$ ).
Novel Well-Posedness (Theorems 2.5 and 2.7):
- The paper establishes existence and uniqueness of mild solutions for initial data in $L^q(D)$ where $q$ can be strictly less than the state space index ( $p$ or $\infty$ ).
- The results show that the solution splits into a singular semigroup part $S(\cdot)x$ and a regular part belonging to $C([0, T]; L^p(D))$ (or $C(D)$ ).
Extended ULDPs (Section 5):
- Corollary 5.1: Establishes ULDPs for the remainder term $X^\varepsilon_x - S(\cdot)x$ in $C([0, T]; L^p(D))$ for initial data purely in $L^q(D)$ .
- Corollary 5.3: Extends these results to weaker topologies, specifically $L^r(0, T; L^p(D))$ , allowing for initial data in $L^q(D)$ without requiring $x \in L^p(D)$ .
- Corollary 5.4: Confirms ULDPs in $C([0, T]; L^r(D))$ for $r \in [1, 2]$ over $L^1$ -bounded initial data when $\lambda < 1$ .

Significance and Claims
The paper claims that its primary contribution is extending the scope of uniform large deviation principles to unbounded sets of initial data in infinite-dimensional spaces. Previous literature (e.g., [5, 9, 40]) largely restricted uniformity to compact or bounded sets in the state space norm.

Physical Relevance: The ability to handle $L^1$ -bounded initial data (mass conservation) is highlighted as a major improvement. In physical systems where solutions represent mass densities, the total mass is the $L^1$ norm. Previous results could only guarantee uniformity over $L^\infty$ -bounded densities. This work enables the rigorous study of exit time problems (Freidlin–Wentzell theory) for systems with conserved mass, where the initial mass is bounded but the density profile can be arbitrarily large in $L^\infty$ .
Methodological Advancement: The work demonstrates that the variational principle underlying the weak convergence approach can be successfully applied to non-compact, unbounded sets provided the controlled equations satisfy stronger convergence modes. The authors explicitly state that their focus is on establishing this uniformity with respect to unbounded sets, rather than generalizing to multi-dimensional systems or non-Lipschitz drifts, which they leave for future work.

In summary, the paper provides a rigorous framework for analyzing rare events in stochastic heat equations with multiplicative noise, specifically addressing scenarios where the initial conditions are only controlled in a weaker norm (like $L^1$ ) than the topology of the solution space, a regime previously inaccessible to uniform large deviation analysis.

Technical Summary: Uniform Large Deviation Principles for the Stochastic Heat Equation Over Unbounded Sets of Initial Data

More like this