Semigroup decay for the wave equation with unbounded damping

This paper establishes sharp polynomial time-decay rates for the energy of solutions to the damped wave equation with singular, unbounded damping coefficients by analyzing the resolvent norm at low frequencies, despite the absence of uniform exponential decay caused by zero being in the spectrum of the generator.

The Gist

Imagine you are in a giant, endless hall (the unbounded domain). You shout a word, and the sound waves bounce around. This is your Wave Equation.

Now, imagine the walls of this hall are lined with special, sticky foam. This foam is your Damping. Its job is to eat the energy of the sound waves, turning the shout into silence over time.

In most physics problems, this foam is uniform—it's the same thickness everywhere. If the foam is thick enough, the sound dies out quickly and predictably (exponentially).

But this paper asks a different question: What if the foam gets infinitely thick as you walk toward the horizon? What if, at the very center of the room, the foam is thin, but as you go further out, it becomes a giant, crushing wall of absorption?

This is the problem of Unbounded Damping. The authors (Arnal, Gerhat, Royer, and Siegl) wanted to know: Does the sound still die out? If so, how fast?

The Big Problem: The "Zero" Trap

Usually, if you have a damping system, you expect the energy to vanish completely and quickly. However, the authors discovered a tricky "trap" in this specific setup.

Because the damping gets infinitely strong at the edges but might be weak or zero in the middle, the system has a "ghost" frequency (zero frequency) that refuses to die out immediately. It's like a pendulum that, instead of stopping, just swings very slowly forever. Because of this, you can't get a "uniform" decay where every sound stops at the same speed. Some sounds linger.

The Solution: The "VIP Section"

So, does the sound ever stop? Yes, but only if you start with the right kind of sound.

The authors realized that if you start with a "messy" shout (random noise), it might get stuck in that slow-swinging ghost mode. But, if you start with a "clean" shout (mathematically speaking, initial conditions in a specific subspace called K), the system behaves beautifully.

They proved that for these "clean" shouts, the energy doesn't vanish exponentially (like $e^{-t}$), but it vanishes polynomially (like $1/t$ or $1/t^2$).

The Analogy:

• Exponential Decay: Like a cup of hot coffee in a freezer. It cools down super fast, then slows down, but it's gone quickly.
• Polynomial Decay (This Paper): Like a cup of coffee cooling in a room with a draft. It cools down steadily, but it takes a long time to reach room temperature. It never truly hits zero instantly, but it gets there eventually.

The "Low Frequency" Detective Work

How did they prove this? They used a mathematical tool called Resolvent Analysis.

Think of the wave equation as a complex machine with gears. The "gears" spin at different speeds (frequencies).

1. High Frequencies (Fast Gears): These are the high-pitched sounds. The authors showed that because the damping gets huge at the edges, these fast gears get crushed and stop very quickly. This part was easy.
2. Low Frequencies (Slow Gears): These are the deep, rumbling sounds. This is where the magic happens. The "ghost" frequency (zero) makes the machine wobble. The authors had to do a very detailed, microscopic analysis of how these slow gears behave.

They found that even though the machine wobbles, the "sticky foam" eventually wins. They calculated exactly how long it takes for the wobble to settle down.

The "Diffusive" Surprise

One of the coolest findings is related to Heat.

Usually, waves (like sound) and heat (like a warm spot on a metal rod) behave very differently. Waves bounce; heat just spreads out.
However, the authors found that in this specific "unbounded foam" scenario, the wave equation starts to act like the heat equation after a long time.

The Metaphor:
Imagine dropping a drop of ink in water.

• Short term: The ink spreads out in a wave-like pattern.
• Long term: The ink just diffuses, spreading slowly and evenly like heat.

The paper shows that for this specific type of damping, the wave eventually forgets it was a wave and starts behaving like heat, spreading out and fading away at a predictable, slow rate.

Why Does This Matter?

1. It solves a puzzle: Previous math could only handle this problem in 3D or higher dimensions. This paper solved it for all dimensions (1D, 2D, 3D, etc.).
2. It handles the "messy" stuff: Real-world materials aren't perfect. The damping might be rough or irregular. This paper works even if the damping coefficient is just "locally integrable" (a fancy way of saying it doesn't have to be perfectly smooth, just not infinitely broken).
3. It gives the exact speed: They didn't just say "it fades." They gave the exact recipe: "If you start with condition X, the energy will drop by $1/t^2$."

Summary

The paper is about a wave in a room where the walls get infinitely sticky as you go further out.

• The Bad News: You can't make every wave stop instantly.
• The Good News: If you start with a "good" wave, it will definitely stop.
• The Rate: It stops at a steady, polynomial pace (like $1/t$), not a fast exponential pace.
• The Twist: After a while, the wave stops acting like a wave and starts acting like heat spreading out.

The authors used advanced spectral analysis (looking at the "gears" of the machine) to prove that even with infinite damping at the edges, the system is stable and predictable, provided you start with the right initial conditions.

Technical Summary

1. Problem Statement

The paper investigates the long-time behavior of solutions to the damped wave equation (DWE) on an open domain $\Omega \subseteq \mathbb{R}^d$:
$$\begin{cases} \partial_{tt}u + a(x)\partial_t u = (\Delta - q(x))u, & t > 0, x \in \Omega, \\ u = 0, & t > 0, x \in \partial\Omega, \\ (u(0), \partial_t u(0)) = (f, g). \end{cases}$$
Key Features of the Setting:

• Unbounded Damping: The damping coefficient $a(x)$ is non-negative and may be unbounded at infinity (e.g., $a(x) \sim |x|^\beta$).
• Minimal Regularity: The coefficients $a$ and $q$ are only assumed to be locally integrable ($L^1_{\text{loc}}$), allowing for singularities.
• Failure of Uniform Decay: In many such settings (specifically when $a$ is not dominated by the potential operator $-\Delta + q$), zero belongs to the essential spectrum of the generator. Consequently, uniform (exponential) energy decay is impossible for general initial data in the energy space.
• Goal: The authors aim to establish sharp polynomial decay rates for the energy and solution norms, specifically for initial data belonging to a suitable subspace of the energy space.

2. Methodology

The authors employ a semigroup approach combined with spectral and resolvent analysis.

A. Functional Framework

• Energy Space ($H$): Defined as $H = W \oplus L^2(\Omega)$, where $W$ is the Hilbert completion of $C_c^\infty(\Omega)$ under the norm $\|u\|_W^2 = \|\nabla u\|_{L^2}^2 + \|q^{1/2}u\|_{L^2}^2$.
• Generator ($A$): The DWE is rewritten as an abstract Cauchy problem $\partial_t U = AU$. Due to the unboundedness of $a$, the domain of $A$ is defined carefully using dominant Schur complements (a technique from [27]) to ensure $A$ generates a $C_0$-contraction semigroup.
• Subspace ($K$): Since uniform decay fails, the analysis is restricted to a subspace $K \subset H$ defined by:
  $$K = \{ (f, g) \in (H^1_0 \cap \text{Dom}(q^{1/2})) \times L^2 : af \in L^1_{\text{loc}}, af+g \in W^* \}.$$
  This space corresponds to the range of the generator $A$ (i.e., $K = \text{Ran}(A)$).

B. Spectral Analysis Strategy

The decay rates are derived from bounds on the resolvent $(A - \lambda)^{-1}$ in the complex plane, specifically distinguishing between two regimes:

1. High Frequencies ($\lambda \to \pm i\infty$):
   • Under the assumption that $a(x) \geq a_0 > 0$ (uniformly positive), the authors prove that the resolvent norm is uniformly bounded on the imaginary axis away from zero.
   • This relies on the Geometric Control Condition (GCC) being satisfied by the damping.
2. Low Frequencies ($\lambda \to 0$):
   • This is the core difficulty. Since $0 \in \sigma(A)$, the resolvent has a singularity at zero.
   • The authors derive precise estimates for $\|(A - \lambda)^{-1}\|$ as $\lambda \to 0$.
   • Key Insight: They reduce the non-self-adjoint problem to a self-adjoint spectral problem involving the Schur complement $T_\lambda = -\Delta + q + \lambda a + \lambda^2$.
   • Using Neumann bracketing and asymptotic perturbation theory, they establish coercivity estimates for the associated quadratic forms.

C. From Resolvent to Time Decay

The paper utilizes an abstract theorem (Theorem 5.2) that converts resolvent bounds on the imaginary axis into time-decay estimates for the semigroup $e^{tA}$.

• If $\|(A - \lambda)^{-1}\| \sim |\lambda|^{-\kappa}$ near zero, the solution decays as $t^{-\kappa}$.
• The high-frequency boundedness ensures the decay is determined solely by the low-frequency singularity.

3. Key Contributions and Results

A. Main Theorem (Theorem 1.2)

For initial data $F \in K$, the solution $u(t)$ exhibits the following sharp polynomial decay rates (where $\langle t \rangle = \sqrt{1+t^2}$):

1. Gradient and Potential Energy:
   $$\|\nabla u(t)\|_{L^2} + \|q^{1/2}u(t)\|_{L^2} \leq C \|F\|_K \langle t \rangle^{-1}$$
2. Velocity:
   $$\|\partial_t u(t)\|_{L^2} \leq C \|F\|_K \langle t \rangle^{-3/2}$$
3. Solution and Weighted Damping:
   $$\|a^{1/2}u(t)\|_{L^2} + \|u(t)\|_{L^2} \leq C \|F\|_K \langle t \rangle^{-1/2}$$

Refined Rates for Unbounded Damping:
If $\Omega$ is an exterior domain and $a(x) \geq c|x|^\beta$ for large $|x|$, the decay rates improve depending on $\beta$:

• $\|\partial_t u(t)\|_{L^2} \lesssim \langle t \rangle^{-\frac{3}{2} - \frac{\beta}{2(2+\beta)}}$
• $\|u(t)\|_{L^2} \lesssim \langle t \rangle^{-\frac{1}{2} - \frac{\beta}{2(2+\beta)}}$

B. Sharpness and Optimality

• The authors prove in Section 6 that these rates are sharp for the model case $\Omega = \mathbb{R}^d$, $q=0$, and constant damping $a=1$.
• The optimality is linked to the diffusive phenomenon: for large times, the damped wave equation behaves like the heat equation. The decay rates match the known decay rates for the heat equation in $\mathbb{R}^d$.

C. Generalization of Previous Work

• The results generalize the work of Ikehata and Takeda (2014), who established similar decay rates for $d \geq 3$ using a modified Morawetz multiplier method.
• Advantages of this paper:
  • Works for all dimensions $d \geq 1$ (solving the open problem for $d=1, 2$).
  • Requires minimal regularity ($L^1_{\text{loc}}$) for coefficients, whereas previous methods required higher regularity.
  • Handles general open domains $\Omega$, not just $\mathbb{R}^d$.
  • Uses a spectral/semigroup approach rather than energy multipliers, providing a more robust framework for unbounded coefficients.

D. Non-Uniform Damping and GCC Violation

• The paper discusses cases where $a(x)$ is not uniformly positive.
• If the Geometric Control Condition (GCC) holds (even with unbounded $a$), high-frequency resolvent bounds remain valid.
• In a specific example (Section 7.2) where GCC is violated (a strip with damping vanishing on a line), the resolvent grows at high frequencies ($\|(A-ib)^{-1}\| \sim |b|^{n/(n+1)}$). However, the authors show that this growth is milder than the singularity at zero, so the low-frequency analysis still dictates the polynomial decay rate.

4. Significance

• Theoretical Breakthrough: This work provides the first complete spectral characterization of the damped wave equation with unbounded, singular damping in arbitrary dimensions.
• Resolution of Open Problems: It settles the question of decay rates in low dimensions ($d=1, 2$) where previous multiplier methods failed due to Sobolev embedding limitations.
• Methodological Innovation: The reduction of the non-self-adjoint damped wave operator to a self-adjoint Schur complement problem allows the use of classical perturbation theory and Neumann bracketing, bypassing the need for complex multiplier techniques.
• Physical Insight: The results rigorously confirm the diffusive nature of the damped wave equation at large times, even in the presence of unbounded damping, linking the decay rates directly to the spatial growth of the damping coefficient.

In summary, the paper establishes that for unbounded damping, the decay of energy is polynomial and determined by the behavior of the damping at infinity and the low-frequency spectral properties of the generator, with the rates being optimal and dimension-independent (in the sense of the specific exponents derived).