Maximum principles and spectral analysis for the superposition of operators of fractional order

1. Problem Statement and Motivation

The paper investigates a class of non-local linear operators formed by the continuous superposition of fractional Laplacians of varying orders, modulated by a signed finite Borel measure.

• The Operator: The primary object of study is the operator $L_\mu$ defined as:
  $$L_\mu u := \int_{[0,1]} (-\Delta)^s u \, d\mu(s)$$
  where $\mu = \mu^+ - \mu^-$ is a signed measure on $[0,1]$, and $(-\Delta)^s$ is the fractional Laplacian.
• Motivation: These operators arise naturally in modeling anomalous diffusion in population dynamics. Different sub-populations may exhibit distinct diffusion patterns (e.g., Gaussian vs. Lévy flights with different exponents). The superposition allows for a unified description of these mixed behaviors. Furthermore, allowing the measure $\mu$ to have negative components (signed measures) enables the modeling of "reverse parabolic" phenomena, such as population concentration, which cannot be captured by purely positive diffusion operators.
• Core Challenges:
  1. Maximum Principles: Determining under what conditions the operator satisfies maximum principles, especially when the measure contains negative parts (which typically violate standard maximum principles).
  2. Spectral Analysis: Establishing the existence, properties, and completeness of the eigenvalue spectrum for the associated Dirichlet problem, particularly when the operator involves a "wrong sign" (negative contributions).

2. Mathematical Setting and Methodology

Functional Framework

The authors work within a Hilbert space setting tailored to the measure $\mu$.

• Space $X(\mathbb{R}^N)$: Defined as the completion of $C_c^\infty(\mathbb{R}^N)$ with respect to the norm:
  $$\|u\|_X := \left( \int_{[0,1]} [u]_s^2 \, d\mu^+(s) \right)^{1/2}$$
  where $[u]_s$ is the Gagliardo semi-norm (or $L^2$ norm for $s=0$, gradient norm for $s=1$).
• Space $X(\Omega)$: The subspace of functions in $X(\mathbb{R}^N)$ vanishing outside a bounded domain $\Omega$.
• Measure Constraints: To handle the signed measure, the authors impose structural conditions (1.5)–(1.7):
  • The "positive" part $\mu^+$ must dominate at higher fractional orders (specifically on $[s, 1]$).
  • The "negative" part $\mu^-$ is restricted to lower orders $[0, s)$.
  • A smallness condition is imposed: $\mu^-([0, s)) \leq \gamma \mu^+([s, 1])$ for a sufficiently small $\gamma > 0$. This ensures the negative part can be "reabsorbed" by the positive part, preserving coercivity.

Methodology

• Variational Methods: The spectral analysis relies on the minimization of a Rayleigh quotient functional $I(u)$ associated with the operator.
• Inductive Construction: Eigenvalues are constructed recursively using a minimax principle (Courant-Fischer type) over subspaces orthogonal to previously found eigenfunctions.
• Compactness Arguments: The proof of the existence of an infinite sequence of eigenvalues utilizes the compact embedding of $X(\Omega)$ into $L^2(\Omega)$ and the boundedness of minimizing sequences.
• Counterexamples: To address the failure of maximum principles, the authors construct explicit counterexamples using specific test functions and asymptotic analysis of the fractional Laplacian.

3. Key Contributions and Results

A. Maximum Principles

The paper establishes a sharp dichotomy regarding the validity of maximum principles:

1. Positive Measures ($\mu^- \equiv 0$):

   • Weak Maximum Principle (Theorem 1.1): If $L_\mu^+ u \geq 0$ in $\Omega$ and $u \geq 0$ outside $\Omega$, then $u \geq 0$ in $\Omega$.
   • Strong Maximum Principle (Theorem 1.2): If $u$ attains a non-positive minimum inside $\Omega$, then $u$ is identically zero (assuming $\mu^+$ is not trivial).
   • These results hold for both weak solutions (in $X(\Omega)$) and classical solutions (in $C(\mathbb{R}^N) \cap C^2(\Omega)$).

2. Failure with Signed Measures:

   • The authors prove that if the measure $\mu$ has negative contributions (even if small), the maximum principle fails.
   • Counterexample (Theorem 1.4/Appendix A): They explicitly construct a function $u$ for the operator $-\Delta - \alpha(-\Delta)^s$ (where $\mu = \delta_1 - \alpha\delta_s$) such that $L_\mu u \geq 0$ in $\Omega$ and $u \geq 0$ outside, yet $u < 0$ inside $\Omega$. This demonstrates that the "reabsorbing" property of the measure is insufficient to save the maximum principle if the sign of the leading operator is mixed.

B. Spectral Analysis

The paper completes the spectral analysis of the Dirichlet eigenvalue problem $L_\mu u = \lambda u$ in $\Omega$, $u=0$ outside $\Omega$:

1. Existence of Eigenvalues: Under the smallness condition on $\gamma$, there exists a sequence of eigenvalues $0 < \lambda_1 \leq \lambda_2 \leq \dots$ such that $\lim_{k \to \infty} \lambda_k = +\infty$.
2. Variational Characterization: The eigenvalues are characterized by a recursive minimization formula involving the difference between the positive and negative energy contributions:
   $$\lambda_{k+1} = \min_{u \in E_{k+1} \setminus \{0\}} \frac{\int_{[0,1]} [u]_s^2 d\mu^+(s) - \int_{[0,s)} [u]_s^2 d\mu^-(s)}{\|u\|_{L^2(\Omega)}^2}$$
3. Basis Properties:
   • The corresponding eigenfunctions $\{e_k\}$ form an orthonormal basis for $L^2(\Omega)$.
   • They form an orthogonal basis for the energy space $X(\Omega)$.
4. Finite Multiplicity: Each eigenvalue has finite multiplicity. If an eigenvalue is repeated, the associated eigenfunctions span the corresponding eigenspace.

4. Significance and Impact

• Theoretical Advancement: This work extends the spectral theory of fractional operators to the complex case of continuous superpositions with signed measures. Previous literature largely focused on single fractional operators or positive combinations.
• Resolution of Open Questions: It clarifies the precise conditions under which maximum principles hold for mixed-order operators, proving that the presence of any negative component in the measure (even if dominated by positive parts) destroys the maximum principle. This is a crucial insight for modeling physical systems where "reverse" diffusion or concentration effects are present.
• Mathematical Rigor: The paper provides a robust functional analytic framework (defining the appropriate Hilbert space $X(\Omega)$ and equivalent norms) that handles the interplay between different fractional orders and signed measures.
• Applications: The results are directly applicable to mathematical biology and physics where population dynamics or diffusion processes involve multiple scales and competing mechanisms (diffusion vs. aggregation).

In summary, the paper successfully bridges the gap between the well-understood theory of single fractional Laplacians and the more complex reality of mixed-order, signed-measure operators, providing definitive answers on the validity of maximum principles and a complete spectral characterization.