Nonlocal operators in divergence form and existence theory for integrable data

1. Problem Statement and Context

The paper addresses two interconnected challenges in the theory of partial differential equations (PDEs):

1. Existence of Weak Solutions for Integrable Data: Classical existence theories for elliptic equations in divergence form typically require the source term (datum) $f$ to belong to $L^p(\Omega)$ with $p > 1$. This allows the use of standard regularity theory and functional analysis techniques. However, when the datum is only in $L^1(\Omega)$, these methods fail because the solution may not possess sufficient regularity to be a weak solution in the standard sense. While existence results for local operators with $L^1$ data exist under specific structural assumptions (e.g., [AB15, AB18]), extending these to the nonlocal setting has been an open problem.
2. The Limit Process ($s \nearrow 1$): There is a need to rigorously connect nonlocal fractional operators to their local counterparts. Specifically, the authors aim to construct a nonlocal operator in divergence form that, as the fractional order $s$ approaches 1, converges to any given classical second-order elliptic operator in divergence form, $-\text{div}(A(x)\nabla u)$.

The paper focuses on the Dirichlet boundary value problem governed by a nonlocal operator $L_K$ with a kernel $K$ and a nonlinear lower-order term $h(u)$, where the source $f$ and coefficient $a$ are merely in $L^1(\Omega)$.

2. Methodology

The authors employ a multi-faceted approach combining variational methods, approximation techniques, and asymptotic analysis:

• Variational Framework for $L^1$ Data:

  • Instead of relying on standard $L^2$-based regularity, the authors utilize a minimization argument for a tailored energy functional.
  • They introduce a truncation technique using the function $G_k(t)$ (a cutoff function) to handle the lack of regularity. By testing the weak formulation with $G_k(u)$, they derive uniform $L^\infty$ bounds and energy estimates that are independent of the approximation parameter.
  • They prove that the solution space is $H^s_0(\Omega) \cap L^\infty(\Omega)$, establishing that even with $L^1$ data, the solution remains bounded.

• Construction of the Nonlocal Operator:

  • The authors define a nonlocal operator $L_{M, \varrho, s}$ via an interaction kernel $K$ modulated by a matrix-valued function $M(x, y)$.
  • The kernel is defined as:
    $$K(x, z) = \frac{c_{n,s} \chi_{[0, \varrho)}(|x-z|)}{|M(z, x-z)(x-z)|^{n+2s}}$$
  • Crucially, $M$ is constructed to satisfy specific symmetry and ellipticity conditions to ensure the operator is well-defined in divergence form.

• Asymptotic Analysis ($s \nearrow 1$):

  • The paper establishes a "Master Equation" approach. They prove that as $s \nearrow 1$, the nonlocal operator $L_{M, \varrho, s}$ converges to a local operator $-\text{div}(A(x)\nabla u)$.
  • The matrix $A(x)$ is derived from $M$ via an integral formula over the unit sphere $S^{n-1}$:
    $$A_{ij}(x) = \frac{n}{\omega_{n-1}} \int_{S^{n-1}} \frac{\psi_i \psi_j}{|M(x, 0)\psi|^{n+2}} d\mathcal{H}^{n-1}_\psi$$
  • Reverse Engineering: A significant methodological contribution is the proof that for any given symmetric, uniformly positive definite matrix $A(x)$, one can explicitly construct a matrix function $M(x, y)$ such that the limit of the nonlocal operator recovers $A(x)$. This involves an algebraic inversion of the integral relation using spectral decomposition.

• Uniform Estimates:

  • The core of the convergence proof relies on deriving uniform estimates for the sequence of nonlocal solutions $u_s$ that do not depend on $s$. These estimates allow the extraction of a convergent subsequence in $H^1_0(\Omega)$ as $s \to 1$.

3. Key Contributions and Results

A. Existence and Uniqueness for $L^1$ Data (Theorem 1.1)

The paper proves the existence and uniqueness of a weak solution $u \in H^s_0(\Omega) \cap L^\infty(\Omega)$ for the problem:
$$\begin{cases} L_K u + a(x)h(u) = f & \text{in } \Omega \\ u = 0 & \text{in } \mathbb{R}^n \setminus \Omega \end{cases}$$
where $f, a \in L^1(\Omega)$, $h$ is continuous, odd, and strictly increasing, and $|f| \leq Q a$ for some $Q$.

• Novelty: This result is new even for the fractional Laplacian (where $K$ is homogeneous) with $L^1$ data.
• Estimates: The solution satisfies $\|u\|_{L^\infty} \leq h^{-1}(Q)$ and a specific energy bound involving the $L^1$ norms of $f$ and $a$.

B. Convergence to the Local Case (Theorem 1.4)

The authors prove that the sequence of nonlocal solutions $u_s$ converges (as $s \nearrow 1$) to the unique weak solution $u_1$ of the local problem:
$$\begin{cases} -\text{div}(A(x)\nabla u) + a(x)h(u) = f & \text{in } \Omega \\ u = 0 & \text{on } \partial\Omega \end{cases}$$
where $A(x)$ is the matrix generated by the limit of the nonlocal kernel. This provides a unified framework where the classical result is recovered as a limit case of the nonlocal theory.

C. Inverse Construction and Alternative Proof (Theorems 1.6 & 1.7)

• Theorem 1.6: Provides an explicit formula to construct the matrix $M$ from a given classical matrix $A$. This "reverse engineering" allows one to start with any classical elliptic operator and find a corresponding nonlocal operator that converges to it.
• Theorem 1.7: Uses this construction to provide an alternative proof of the existence of solutions for classical elliptic equations with $L^1$ data (previously established in [AB21]). The proof constructs the classical solution as the limit of nonlocal solutions, bypassing some traditional local regularity arguments.

4. Significance and Impact

1. Extension to Nonlocal $L^1$ Theory: The paper successfully bridges the gap between local and nonlocal existence theories for integrable data. It demonstrates that the structural assumptions allowing existence in the local case (specifically the divergence form and the domination of $f$ by $a$) are sufficient to guarantee existence in the nonlocal setting, despite the lack of standard regularity.
2. Unified Perspective: It offers a "unified scenario" where classical PDEs are viewed as limits of nonlocal problems. This is significant for numerical analysis and modeling, as it suggests that nonlocal models can approximate local ones with high fidelity, even for rough data.
3. New Proof Technique: By proving the existence of classical solutions via a limit of nonlocal problems, the authors introduce a novel methodological pathway. This "nonlocal-to-local" approach could be applied to other problems where direct local analysis is difficult.
4. Flexibility of the Operator: The ability to tailor the nonlocal operator $M$ to recover any specific local operator $A$ (Theorem 1.6) is a powerful theoretical tool. It allows researchers to study specific local geometries or anisotropies by analyzing their nonlocal counterparts.

5. Conclusion

This paper establishes a robust existence theory for nonlocal elliptic equations with $L^1$ data, a regime where standard regularity theory fails. By developing tailored uniform estimates and a sophisticated asymptotic analysis, the authors not only prove existence for the nonlocal problem but also demonstrate that the classical local theory emerges naturally as a limit. The work provides a constructive link between nonlocal and local operators, offering new tools for both theoretical analysis and the approximation of classical PDEs.