Global UCP For Parabolic Fractional $p$-Laplace Equation With Very Rough Potentials

1. Problem Statement

The paper investigates the Global Unique Continuation Property (UCP) for weak solutions to a specific class of non-linear, non-local parabolic partial differential equations. The equation under study is:

$$\partial_t u + (-\Delta_p)^s u + V(x, t)u(x, t) = 0 \quad \text{in } \Omega \times I$$

Where:

• $(-\Delta_p)^s$ is the fractional p-Laplacian operator (a non-local, non-linear generalization of the fractional Laplacian).
• $V(x, t)$ represents a very rough potential, belonging to the space $L^{p'}_t W^{-s, p'}_x$.
• The domain is $\Omega \subset \mathbb{R}^N$ (open) and $I \subset \mathbb{R}$ (an open time interval).

The Core Question: If a weak solution $u$ vanishes on a non-empty open subset $G \times J$ (where $G \subset \Omega$ and $J \subset I$), does it necessarily vanish everywhere in $\mathbb{R}^N \times J$?

2. Background and Novelty

• Context: Unique continuation for non-local operators is a significant area of research, with applications in inverse problems, stability of Cauchy problems, and nodal set analysis.
• Existing Literature:
  • For the linear fractional Laplacian, UCP was established by Riesz and recently refined by Berger and Schilling using functional analytic methods.
  • For the fractional heat equation (linear, local time derivative), UCP results are scarce and typically rely on delicate Carleman estimates via Caffarelli-Silvestre extensions.
  • For the local p-Laplace equation ($s=1$), UCP remains an open problem for dimensions $N > 2$.
  • For the parabolic fractional p-Laplace equation, no results existed prior to this work, even for zero potentials.
• Novelty: This paper provides the first global UCP result for the parabolic fractional p-Laplace equation with rough potentials. Crucially, the proof eschews the standard techniques of Carleman estimates and extension problems (Caffarelli-Silvestre), offering a more direct, elementary approach.

3. Methodology

The proof relies on a functional analytic approach involving convolution conditions and density arguments, rather than energy estimates or extension techniques. The methodology proceeds in four logical steps:

Step 1: Reduction to a Convolution Condition

The author starts with the weak formulation of the equation. Since the solution $u$ is assumed to be zero on the open set $G \times J$, the local terms in the weak formulation (the time derivative $\partial_t u$ and the potential term $Vu$) vanish when tested against functions supported in $G \times J$.
This leaves only the non-local double integral. By decomposing the integration domain and using the fact that $u=0$ on $G$, the author derives a condition where the non-local interaction between the zero region ($G$) and the non-zero region ($G^c$) must vanish:
$$\int_{\mathbb{R}^N \setminus G} \frac{|u(y, t)|^{p-2}u(y, t)}{|x-y|^{N+sp}} \, dy = 0 \quad \text{for a.e. } x \in G.$$

Step 2: Zeroth Moment Vanishes

By fixing a time $t$ and utilizing the openness of $G$, the author defines a function $\Phi(x)$ representing the integral above. Using the tail condition of the solution space ($u \in L^\infty_{loc}(I; L^{p-1}_{sp}(\mathbb{R}^N))$), it is shown that $\Phi(x)$ is continuous and identically zero on a ball $B_\epsilon(x_0) \subset G$.

Step 3: All Moments Vanish

Because the kernel $|x-y|^{-(N+sp)}$ is smooth for $x$ inside the ball and $y$ outside the ball, the author differentiates $\Phi(x)$ under the integral sign to arbitrary orders. This generates a system of equations showing that the "moments" of the function $f_t(y) = |u(y, t)|^{p-2}u(y, t)$ against specific polynomial-weighted kernels vanish.
Specifically, for multi-indices $\alpha$, integrals of the form:
$$\int_{\mathbb{R}^N \setminus G} f_t(y) y^\alpha |y|^{-\gamma - 2m} \, dy = 0$$
are shown to be zero.

Step 4: Density Argument (Stone-Weierstrass)

The author constructs an algebra $\mathcal{B}$ generated by functions of the form $y^\alpha |y|^{-2|\alpha|}$.

• It is shown that $\mathcal{B}$ separates points and vanishes nowhere on $\mathbb{R}^N \setminus B_\epsilon(0)$.
• By the Stone-Weierstrass theorem, $\mathcal{B}$ is dense in $C_0(\mathbb{R}^N \setminus B_\epsilon(0))$.
• Since the signed measure $d\mu_t = f_t(y)|y|^{-\gamma} dy$ annihilates all elements of $\mathcal{B}$ (due to Step 3), and $\mathcal{B}$ is dense, the measure must be zero by the Riesz Representation Theorem.
• Consequently, $f_t(y) = 0$ almost everywhere, implying $u(y, t) = 0$ almost everywhere in $\mathbb{R}^N$.

4. Key Contributions

1. Global UCP for Non-Linear Non-Local Parabolic Equations: The paper establishes that if a solution vanishes on an open set in space-time, it vanishes globally in space for that time interval.
2. Handling Rough Potentials: The result holds for potentials $V$ in the rough space $L^{p'}_t W^{-s, p'}_x$, which is a significant generalization over previous works requiring smoother coefficients.
3. Elementary Proof Technique: The proof avoids the heavy machinery of Carleman estimates and extension problems, which are standard for fractional operators but often difficult to apply to non-linear or parabolic settings.
4. Purely Non-Local Phenomenon: The result highlights a fundamental difference between local and non-local operators. In local PDEs (like the standard heat equation), vanishing on an open set does not imply vanishing everywhere (due to finite propagation speed). Here, the non-local nature forces global vanishing.

5. Main Theorem (Theorem 1.1)

Statement: Let $u$ be a weak solution to the equation in $\Omega \times I$. If there exists a non-empty open subset $G \subset \Omega$ and a subinterval $J \subset I$ such that $u(x, t) = 0$ for almost every $(x, t) \in G \times J$, then $u(x, t) = 0$ for almost every $(x, t) \in \mathbb{R}^N \times J$.

6. Significance and Implications

• Theoretical Advancement: This resolves a long-standing open problem regarding the unique continuation for the parabolic fractional p-Laplace equation, even in the absence of a potential ($V=0$).
• Methodological Shift: By demonstrating that extension techniques are not strictly necessary for UCP in this context, the paper opens the door for applying similar functional analytic arguments to other non-linear, non-local kinetic equations.
• Robustness: The ability to handle "very rough" potentials suggests the result is stable under significant perturbations, making it applicable to a wider range of physical and mathematical models involving anomalous diffusion with irregular environments.
• Future Directions: The author notes in Remark 3.1 that this technique can be extended to equations with nonlinear time derivatives ($\partial_t \phi(u)$) and nonlocal kinetic equations, suggesting a broad applicability of the derived method.