Fourier restriction norm method adapted to controlled paths: stochastic wave equations

Technical Summary: Fourier Restriction Norm Method Adapted to Controlled Paths: Stochastic Wave Equations

Problem Statement
The paper addresses the pathwise well-posedness of the stochastic nonlinear wave equation (SNLW) with multiplicative noise on the torus $\mathbb{T}^d$:
$$\partial_t^2 u + (1-\Delta)u \pm u^k = u\Phi\zeta, \quad (u, \partial_t u)|_{t=0} = (\phi_0, \phi_1),$$
where $\zeta$ represents a fractional or white-in-time noise. While the Itô solution theory (random field solution theory) for this equation was established in the 1980s, the question of pathwise well-posedness—constructing solutions for almost every realization of the noise using pathwise methods rather than $L^2(\Omega)$ limits—has remained a challenging open problem for over forty years, particularly for white-in-time noise. The primary difficulty lies in the temporal roughness of the noise, which prevents the standard interpretation of the stochastic convolution $\Psi(u)$ as a classical integral, and the incompatibility of standard dispersive PDE tools (like Strichartz estimates in $X^{s,b}$ spaces) with the requirements of rough path theory.

Methodology
The authors develop a unified framework that bridges two distinct mathematical theories:

1. Fourier Restriction Norm Method: Specifically, the adaptation of the method to $U^p$ and $V^p$ spaces (introduced by Koch and Tataru), which are well-suited for handling the dispersive nature of the wave equation and critical regularity problems.
2. Rough Path Theory: Specifically, the theory of controlled paths and Young/rough integration via the sewing lemma (Gubinelli), which provides a pathwise interpretation of integrals driven by irregular signals.

Key technical components include:

• Interaction Representation: The authors transform the equation into an interaction representation $\vec{u}(t) = S(-t)\vec{u}(t)$ to separate the linear wave propagator from the nonlinearity and noise, facilitating the application of rough path techniques.
• Random Tensor Estimates: A crucial innovation is the derivation of random tensor estimates (Lemma 2.12) for multiple stochastic integrals with respect to fractional Brownian motions. These estimates are used to establish the almost sure mapping properties of the random drivers (the noise terms viewed as operators).
• Decomposition Strategy: In the "rough case" (white-in-time noise, $\beta=1/2$), where standard controlled path approaches fail at high dimensions and critical regularities, the authors introduce a decomposition $\vec{u} = \vec{v} + \vec{w}$. Here, $\vec{v}$ captures the rough temporal behavior (solving a rough differential equation) but possesses higher spatial regularity, while $\vec{w}$ solves a deterministic PDE with rougher spatial regularity but smoother temporal behavior. This allows the use of Strichartz estimates on the nonlinear term.

Key Contributions and Results
The paper establishes pathwise local well-posedness in optimal regularity ranges for both fractional-in-time and white-in-time noise scenarios:

1. Young Case ($\frac{1}{2} < \beta < 1$):

   • Theorem 1.4 establishes pathwise local well-posedness for fractional-in-time noise.
   • The solution is constructed as a Young integral. The authors prove that the random driver satisfies the necessary Hölder regularity bounds almost surely, allowing the application of the sewing lemma within the $U^p$-$V^p$ framework.

2. Rough Case ($\beta = 1/2$, White-in-Time Noise):

   • Theorem 1.5 establishes pathwise local well-posedness for white-in-time noise, resolving the forty-year open problem.
   • The solution is constructed as a rough integral. The authors prove that the pair of first and second-order drivers forms a rough path almost surely.
   • Optimality: The results cover the case of "almost space-time white noise" in one dimension (where the noise is slightly smoother than space-time white noise), which is claimed to be optimal within the framework of one-parameter rough paths.
   • Critical Regularity: The results hold for initial data in $H^s$ where $s$ is at or above the scaling-critical regularity $s_{crit}$ (with specific exceptions for certain $(d,k)$ pairs where $s > s_{crit}$ is required), matching the known well-posedness thresholds for the deterministic nonlinear wave equation.

Significance and Claims
The paper claims to provide the first pathwise local well-posedness result for SNLW with multiplicative white-in-time noise. Its significance is framed as follows:

• Unification: It successfully unifies the Fourier restriction norm method (essential for dispersive PDEs) with rough path theory (essential for singular stochastic integrals), overcoming the incompatibility of standard $X^{s,b}$ spaces with controlled path constructions.
• Sharpness: The spatial regularity conditions are essentially sharp, aligning with the known ill-posedness results for the deterministic equation.
• Consistency: The pathwise solutions constructed agree with the classical Itô solutions (proven in Appendix A), validating the pathwise approach as a rigorous alternative to stochastic calculus limits.
• Robustness: The framework is presented as general, capable of being adapted to other stochastic dispersive PDEs with multiplicative noises, though the authors note that extending these results to the stochastic nonlinear Schrödinger equation (NLS) with white-in-time noise presents additional difficulties due to the lack of smoothing from the linear propagator in that context.

The authors explicitly state that their approach allows for the almost sure convergence of various approximation schemes (mollification, piecewise linear approximations), a key advantage of pathwise theories over Itô theories. They do not claim global well-posedness results for the general case, noting that global existence in the energy-critical regime for the defocusing case can be derived via energy conservation arguments adapted from existing literature, but energy-supercritical cases remain open.