Finite Spectral-Band Optimal Control of Acoustic Waves via Subwavelength Point-Like Resonant Actuators

Technical Summary: Finite Spectral-Band Optimal Control of Acoustic Waves via Subwavelength Point-Like Resonant Actuators

Problem Statement
This work addresses the optimal control of acoustic waves generated by local clusters of sub-wavelength resonant bubbles. Unlike traditional formulations where actuators are idealized as static Dirac sources imposed at the model level, this paper treats the actuators as physical resonant systems. The bubbles oscillate near their Minnaert resonance, generating localized monopole fields through multiple scattering. The control objective is to steer the acoustic field toward a desired finite-band trajectory while minimizing a quadratic cost functional. The central challenge is to quantify how the physical resonant dynamics of the bubble clusters—specifically their collective scattering poles—interact with and influence the optimal control design and cost.

Methodology
The authors employ a rigorous asymptotic reduction strategy to bridge the gap between the full acoustic transmission problem and a finite-dimensional control model:

1. Asymptotic Reduction: Starting from the time-domain acoustic transmission problem with high-contrast inclusions, the scattered field is approximated using a retarded Foldy–Lax expansion. The bubbles are grouped into local clusters where intra-cluster distances are small ($O(\epsilon^p)$) compared to inter-cluster distances.
2. Spectral Analysis: The Laplace-domain Foldy–Lax matrix is analyzed to identify principal collective Minnaert poles. For each cluster $\alpha$, a distinguished simple pole $s^\epsilon_\alpha = -\gamma^\epsilon_\alpha + i\omega^\epsilon_\alpha$ is isolated, where $\gamma^\epsilon_\alpha > 0$ represents radiation damping arising from the scattering mechanism, not an ad-hoc assumption.
3. Finite-Band Projection: The dynamics are projected onto a finite acoustic spectral band. This yields a coupled finite-dimensional state-space system (denoted as system M):
   • Acoustic Modes: $\ddot{a} + \Lambda a = C^\epsilon \eta$
   • Collective Actuator States: $\ddot{\eta} + 2\Gamma^\epsilon \dot{\eta} + K^\epsilon \eta = u$
     Here, $a$ represents acoustic modal coefficients, $\eta$ represents the collective microstructural states of the clusters, and $u$ is the control input. The matrices $\Gamma^\epsilon$ and $K^\epsilon$ are diagonal, derived from the real and imaginary parts of the isolated poles, while $C^\epsilon$ encodes the coupling between acoustic modes and cluster channels.
4. Optimal Control Formulation: A linear-quadratic tracking problem is formulated with a cost functional $J_\mu$ that penalizes the deviation from a target trajectory and the control effort, regularized by a parameter $\mu > 0$.

Key Contributions and Results

• Existence and Uniqueness: The paper proves the existence and uniqueness of the optimal control $u^*_\mu$ for the finite-dimensional system. It derives the associated adjoint optimality system (Pontryagin minimum principle), characterizing the optimal pair via a coupled forward-backward system.
• Resonant Source-Lifting Estimates: A primary quantitative contribution is the derivation of "source-lifting" estimates. The authors establish that if a desired effective source profile $\eta_r$ is spectrally concentrated in specific bands $I_\alpha$, the required control input $u_r$ satisfies:
  $$\|u_r\|_{L^2}^2 \leq \sum_{\alpha} \left( \sup_{\nu \in I_\alpha} d_{\alpha,\epsilon}(\nu) \right)^2 \|\eta_r\|_{BT(I_\alpha)}^2$$
  where $d_{\alpha,\epsilon}(\nu) = |(\omega^\epsilon_\alpha)^2 + (\gamma^\epsilon_\alpha)^2 - \nu^2 + 2i\gamma^\epsilon_\alpha \nu|$.
• Quantification of Resonant Gain: At exact frequency matching ($\nu = \omega^\epsilon_\alpha$), the multiplier behaves as $d_{\alpha,\epsilon}(\omega^\epsilon_\alpha) \approx 2\gamma^\epsilon_\alpha \omega^\epsilon_\alpha$. Since $\gamma^\epsilon_\alpha$ is small (scaling with the bubble size $\epsilon$), the control energy required to generate a prescribed effective source is $O(\epsilon^2)$ relative to the squared source output. This quantifies the efficiency gain provided by resonant tuning.
• Exponential Stabilization: The paper demonstrates that the same small radiation damping ($\gamma^\epsilon_\alpha$) that regularizes the control cost also provides intrinsic stabilization. Under an explicit modal coupling condition (requiring that no acoustic eigenmode in the selected band is invisible to the cluster channels), the homogeneous coupled system exhibits exponential decay with a rate proportional to the cluster-induced damping scale ($\delta_\epsilon = O(\gamma^\epsilon)$).

Significance and Claims
The paper claims to provide a rigorous bridge between the physics of resonant microstructured actuators and optimal control theory. Its significance lies in:

1. Explicit Cost-Spectral Connection: It establishes an explicit link between the spectral properties of the resonant actuators (specifically the location of the collective poles) and the resulting control cost. The work shows that resonant tuning reduces actuator effort not by eliminating cost, but by leveraging the small real part of the poles to achieve a finite, favorable resonant gain.
2. Beyond Idealization: By deriving the control model from the underlying physical scattering problem rather than postulating static sources, the framework identifies not only which source profiles can be produced but how efficiently they can be produced through physical resonant dynamics.
3. Unified Mechanism: The analysis reveals that the collective Minnaert poles play a dual role: their imaginary parts define the frequency windows for efficient coupling, while their real parts govern both the finite gain in source-lifting estimates and the intrinsic stabilization of the reduced dynamics.

The authors conclude that this framework offers a systematic route for designing finite-band wave controls where the geometry and material parameters of the clusters are optimized jointly with the control input, rather than treating the resonant structure as a passive correction to an idealized actuator.