A discretization for the nonlinear parabolic evolution equation of fractional order in space

1. Problem Statement

The paper investigates the nonlinear parabolic evolution equation of fractional order in space, given by:
$$\frac{\partial u}{\partial t} = D^\gamma u + f(u)$$
Specifically, the authors focus on the semilinear equation:
$$\frac{\partial u}{\partial t} = -(-\Delta)^\alpha u + u^p$$
where:

• $u(t, x)$ is the unknown function.
• $(-\Delta)^\alpha$ represents the fractional power of the Laplacian (with $\alpha \in (0, 1)$).
• $f(u) = u^p$ is a nonlinear source term ($p > 1$).

Key Phenomenon: The study focuses on finite-time blow-up, where solutions become unbounded in finite time ($t < \infty$). This behavior mimics explosive physical phenomena. The paper aims to analyze the theoretical properties of these solutions and propose a robust numerical discretization scheme to simulate them.

2. Methodology

A. Functional Analytic Definition of Fractional Derivatives

The authors critique traditional definitions (Riemann-Liouville, Caputo, Grünwald-Letnikov) for their difficulties with boundary conditions and the requirement of a "pivot" point, which conflicts with periodic or symmetric domains.

Instead, they adopt a functional analytic approach:

• Let $A$ be a self-adjoint, positive-definite operator in a Hilbert space $X$.
• The fractional power $A^\alpha$ is defined via the integral representation (based on Kato's theory):
  $$A^\alpha u = \frac{\sin(\pi \alpha)}{\pi} \int_0^\infty \lambda^{-\alpha} A(\lambda + A)^{-1} u \, d\lambda$$
• Advantages: This definition naturally incorporates boundary conditions into the domain of the operator $D(A^\alpha)$. It preserves positivity (if the resolvent $(\lambda + A)^{-1}$ preserves positivity, so does $A^\alpha$) and aligns with spectral definitions for periodic functions.

B. Theoretical Analysis

The paper establishes several theoretical results regarding the existence and blow-up behavior:

1. Local Existence: Proved for initial data in the domain $D(A^\alpha)$ under specific Sobolev embedding conditions.
2. Blow-up Criteria:
   • Theorem 2: Adapts Levine's concavity method. If the initial data satisfies a specific energy inequality involving the inner product $(u, A u)$, the solution blows up in finite time.
   • Theorem 4 (Fujita-type Exponent): The authors generalize Fujita's critical exponent theorem to the fractional case. They prove that for the equation on $\mathbb{R}^m$, blow-up occurs for all non-negative solutions if $m(p-1) < 2\alpha$. If $m(p-1) > 2\alpha$, small initial data leads to global existence, while large data leads to blow-up.
3. Positivity: While the maximum principle does not hold for $\alpha < 1$, the positivity preserving property is maintained, which is crucial for the analysis.

C. Numerical Discretization

The authors propose a numerical scheme for the equation on a bounded domain with periodic boundary conditions.

• Spatial Discretization: Uses the Spectral Method. The operator $A^\alpha$ is computed by diagonalizing the Laplacian in the Fourier basis. For a function expanded as $u(x) = \sum \gamma_n \cos(nx)$, the action of $A^\alpha$ is simply multiplication by $n^{2\alpha}$ in the frequency domain.
• Temporal Discretization:
  • Explicit Scheme: $u^{m+1} = (I - \tau A^\alpha)u^m + \tau (u^m)^p$. The time step $\tau$ is adaptive, shrinking as the solution magnitude increases to prevent instability near blow-up.
  • Implicit Scheme: $(I + \tau A^\alpha)u^{m+1} = u^m + \tau (u^m)^p$. This scheme is unconditionally stable and avoids the severe time-step restrictions of the explicit method for large $\alpha$. The inverse operator $(I + \tau A^\alpha)^{-1}$ is efficiently computed in the spectral domain.

3. Key Contributions

1. Definition Framework: The paper advocates for the functional analytic definition of fractional derivatives over traditional integral definitions, specifically highlighting its suitability for handling boundary conditions in numerical simulations without arbitrary pivot points.
2. Generalized Fujita Exponent: The authors rigorously derive the critical exponent condition $m(p-1) = 2\alpha$ for the fractional heat equation, extending classical results from $\alpha=1$ to $0 < \alpha < 1$.
3. Numerical Implementation: They provide a concrete, implementable spectral discretization for the fractional operator $A^\alpha$ and compare explicit vs. implicit time-stepping strategies for blow-up problems.
4. Monotonicity Conjectures: Based on numerical experiments, the authors conjecture that for a fixed initial data, the blow-up time $T_\alpha$ is a monotonically increasing function of $\alpha$. In other words, higher fractional orders (stronger diffusion) delay the blow-up.

4. Numerical Results and Discussion

• Smoothing Effect: Simulations confirm that the fractional term acts as a smoothing force. As $\alpha$ increases, the solution profile becomes smoother, and the peak value at any given time $t$ decreases.
• Blow-up Time:
  • Numerical experiments with initial data $u(0, x) = \cos x + 1$ show that solutions with larger $\alpha$ (e.g., $\alpha=0.7$) blow up later than those with smaller $\alpha$ (e.g., $\alpha=0.5$).
  • This supports Theorem 6: $T_\alpha$ is monotonically increasing with respect to $\alpha$.
• Scheme Comparison:
  • The Explicit scheme requires very small time steps for large $\alpha$ to maintain stability.
  • The Implicit scheme is more robust for large $\alpha$ but may require careful handling for small $\alpha$ where the explicit scheme is naturally stable.
• Non-Monotonicity in Space: The authors note that while the maximum value $u(t, 0)$ decreases as $\alpha$ increases, the inequality $u_\alpha(t, x) < u_\beta(t, x)$ does not necessarily hold for all $x$ simultaneously, indicating complex spatial interactions.

5. Significance

This paper bridges the gap between the abstract functional analysis of fractional operators and practical numerical simulation of blow-up phenomena.

• Theoretical Impact: It clarifies the conditions for finite-time blow-up in fractional diffusion-reaction systems, providing a rigorous extension of the famous Fujita exponent.
• Computational Impact: It offers a reliable spectral method for computing fractional powers of differential operators, which is essential for simulating anomalous diffusion and fractional PDEs where boundary conditions are critical.
• Physical Relevance: By modeling how the fractional order $\alpha$ influences the timing of "explosive" events (blow-up), the work provides insights into physical systems governed by non-local diffusion, such as plasma physics, fluid dynamics, and biological population models.

The authors conclude by posing open questions regarding the spatial profile of the blow-up (whether it remains localized to a point) and the precise behavior of the blow-up profile, suggesting directions for future research.