Koopman spectral analysis of skew-product dynamics on Hilbert $C^*$-modules

Dimitrios Giannakis, Yuka Hashimoto, Masahiro Ikeda, Isao Ishikawa, Joanna SławińskaarXiv 2023 · DOI · arXiv

Abstract

We introduce a linear operator on a Hilbert $C^*$-module for analyzing skew-product dynamical systems. The operator is defined by composition and multiplication. We show that it admits a decomposition in the Hilbert $C^*$-module, called eigenoperator decomposition, that generalizes the concept of the eigenvalue decomposition. This decomposition reconstructs the Koopman operator of the system in a manner that represents the continuous spectrum through eigenoperators. In addition, it is related to the notions of cocycle and Oseledets subspaces and it is useful for characterizing coherent structures under skew-product dynamics. We present numerical applications to simple systems on two-dimensional domains.

Introduction

In the study of dynamical systems, Koopman operator theory has emerged as a powerful tool for analyzing nonlinear systems by examining how functions (observables) evolve linearly over time. Traditionally, this analysis relies on finding the operator's eigenvalues and eigenvectors to understand long-term behavior. However, standard Hilbert space methods face significant hurdles when dealing with continuous spectra—common in measure-preserving fluid flows—and skew-product dynamics, where a "base" process drives a "fiber" process. These complexities often lead to noisy spectral information that fails to capture the underlying structure of the flow.

Mathematical Framework

The core innovation of this work is the transition from standard Hilbert spaces to Hilbert $C^*$-modules. In a standard Hilbert space, the inner product of two vectors is a complex number. In a Hilbert $C^*$-module, the inner product takes values in a $C^*$-algebra (specifically, a space of bounded operators). This algebraic shift allows the mathematical framework to "store" more information about the system's internal structure.

Schematic of the mathematical framework connecting skew-product dynamics, Hilbert $C^*$-modules, and eigenoperator decomposition.

As illustrated in the figure above, the authors define a novel "eigenoperator decomposition": $K_T \hat{w} = \hat{w} \cdot \hat{M}$. Here, $\hat{M}$ is an eigenoperator rather than a scalar eigenvalue. This generalization allows the continuous spectrum of the fiber dynamics to be "absorbed" into the eigenoperator, treating a complex continuous spectrum as a single structured object. This framework effectively bridges Koopman operator theory with Multiplicative Ergodic Theory and Oseledets subspaces, providing a unified language for both autonomous and non-autonomous systems.

To implement this numerically, the authors utilize Reproducing Kernel Hilbert Spaces (RKHS) to approximate the infinitesimal generator of the dynamics, enabling the computation of these decompositions from system equations.

Numerical Results and Insights

The method was validated using synthetic numerical experiments on two-dimensional domains:

  • Moving Gaussian Vortex: This experiment demonstrated that as the dimension of the generalized Oseledets subspace increases, the visualization of flow features becomes significantly clearer. The eigenvectors successfully capture the essential geometry of the moving vortex.
  • Idealized Stratospheric Flow: In this time-dependent model of atmospheric transport, the spectrum of the eigenoperator provided much more distinct information about coherent patterns—regions of air that remain together despite turbulence—than the standard Koopman generator, which appeared significantly noisier.

Theoretical proofs confirmed that the eigenoperator decomposition exists and can perfectly reconstruct the original Koopman operator of the skew-product system. This suggests that eigenoperators are directly related to the detection of coherent structures, the "skeletons" of a flow that dictate transport and mixing.

Conclusion and Future Directions

While the Hilbert $C^*$-module framework offers superior spectral resolution and a unified theoretical approach, it is computationally more intensive than traditional methods. Current implementations also rely on knowing the underlying equations of the system. Future research will focus on developing purely data-driven versions of this framework—where only trajectories are known—and exploring potential applications in quantum computation, where the operator-based decomposition may align naturally with quantum algorithms.

Summary updated

Abstract

We introduce a linear operator on a Hilbert $C^*$-module for analyzing skew-product dynamical systems. The operator is defined by composition and multiplication. We show that it admits a decomposition in the Hilbert $C^*$-module, called eigenoperator decomposition, that generalizes the concept of the eigenvalue decomposition. This decomposition reconstructs the Koopman operator of the system in a manner that represents the continuous spectrum through eigenoperators. In addition, it is related to the notions of cocycle and Oseledets subspaces and it is useful for characterizing coherent structures under skew-product dynamics. We present numerical applications to simple systems on two-dimensional domains.
BibTeX
@misc{giannakis2023koopman,
  author = {Dimitrios Giannakis and Yuka Hashimoto and Masahiro Ikeda and Isao Ishikawa and Joanna Sławińska},
  title = {{Koopman spectral analysis of skew-product dynamics on Hilbert $C^*$-modules}},
  year = {2023},
  eprint = {2307.08965},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2307.08965},
}
Figure