Koopman Nonlinear Non-Hermitian Skin Effect
Shu Hamanaka · 2026
Read Koopman Nonlinear Non-Hermitian Skin Effect by Shu Hamanaka (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-03636.txt.
Introduction
Shu Hamanaka, writing alone, rethinks a boundary effect in non-Hermitian physics for systems that are nonlinear. In linear non-Hermitian systems, the skin effect piles states at the boundary, but nonlinear systems have no clean eigenstates to pile. Hamanaka proposes describing the effect through the Koopman framework, which lifts dynamics into a space of observables where functions, not states, carry the behavior.
Using a mildly nonlinear version of the Hatano-Nelson model, the study shows dominant Koopman functions localizing sharply on higher-order observables, unlike the linear case. That lifted piling controls how strongly the system reacts to boundary changes, giving a clear fingerprint of the nonlinear skin effect. The setup gives physicists a faithful language for boundary phenomena in driven, nonlinear devices.
Catalog record
| CREATOR | Shu Hamanaka |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 30 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-03636
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Koopman Nonlinear Non-Hermitian Skin Effect Shu Hamanaka Abstract Non-Hermitian skin effects are conventionally manifested as boundary localization of eigenstates in linear systems. In nonlinear settings, however, where eigenstates are no longer well defined, it becomes unclear how skin effects should be faithfully characterized. Here, we propose a Koopman-based characterization of nonlinear skin effects, in which localization is defined in terms of Koopman eigenfunctions in a lifted observable space, rather than physical states. Using a minimal nonlinear extension of the Hatano-Nelson model, we show that dominant Koopman eigenfunctions localize sharply on higher-order observables, in stark contrast to linear skin effects confined to linear observables. This lifted-space localization governs the sensitivity to boundary amplitude perturbations, providing a distinct dynamical signature of the nonlinear skin effect. Our results establish the Koopman framework as a natural setting in which skin effects unique to nonlinear non-Hermitian systems can be identified. Linearity plays a fundamental role in physics, as it enables a spectral decomposition of the Hamiltonian in terms of eigenvalues and eigenstates. This property allows complex time evolution to be reduced to a superposition of elementary modes, providing a powerful organizing principle across a wide range of physical systems 37 . The non-Hermitian skin effect 46 is among the most prominent examples where a mode-based description plays a central role: nonreciprocity causes eigenstates to accumulate at system boundaries. In linear non-Hermitian systems, the skin effect has been formulated in terms of non-Bloch band theory 47 ; 42 and point-gap topological characterizations 53 ; 33 . Here, the central quantities are again eigenvalues and eigenstates, or their generalizations such as pseudospectra 34 . Once nonlinearity is introduced, however, the notions of eigenstates and eigenvalues lose their direct meaning from the outset. As a result, it becomes unclear how the skin effect should manifest itself in nonlinear systems, in particular whether boundary accumulation of stationary states alone provides an adequate diagnostic of the underlying nonlinear dynamics. Previous approaches have mainly focused on stationary solutions, reducing the problem to a nonlinear spatial dynamical system at fixed frequency 52 ; 51 ; 15 . While successful in certain contexts, such approaches may probe a restricted subset of the full nonlinear dynamics. Consequently, the underlying nature of the nonlinear skin effect remains largely elusive. In this Letter, we propose a Koopman-based characterization of nonlinear non-Hermitian skin effects, in which localization is diagnosed at the level of observables rather than physical states. Within this framework, we identify a nonlinear skin effect manifested as localization of dominant Koopman eigenfunctions in a lifted observable space (Fig. 1 ). As an illustrative example, we study a minimal nonlinear extension of the Hatano-Nelson model and demonstrate that the dominant Koopman eigenfunctions localize sharply on higher-order observables in stark contrast to the linear case, where the dominant Koopman eigenfunctions are confined to linear observables. We further introduce a dynamical probe based on the sensitivity to boundary amplitude perturbations, directly reflecting the localization property of Koopman eigenfunctions in the lifted observable space. Our results establish a Koopman-based framework for identifying and probing skin effects unique to nonlinear non-Hermitian systems. Figure 1: Illustration of the Koopman nonlinear non-Hermitian skin effect. (a) The nonlinear Schrödinger dynamics is recast as a linear evolution in an appropriate observable space within the Koopman framework. (b) The physical state X(t) accumulates at the boundary due to nonreciprocal hopping, giving rise to the skin effect. (c) Unlike the linear skin effect, where localization occurs on the linear observable sector, the Koopman nonlinear skin effect is characterized by the localization of observables on higher-order observable bases. Koopman analysis .— Consider an n -component state vector X=(x_{1},\ldots,x_{n})^{\mathsf{T}}\in\mathbb{C}^{n} evolving under a nonlinear equation \frac{d}{dt}X=\mathbf{T}(X), (1) where \mathbf{T}:\mathbb{C}^{n}\to\mathbb{C}^{n} is a generally nonlinear vector field. Let \mathbf{S}^{t}:\mathbb{C}^{n}\to\mathbb{C}^{n} denote the flow map generated by Eq. ( 1 ), defined by \mathbf{S}^{t}(X(0))=X(t) . Equivalently, \mathbf{S}^{t} satisfies \partial_{t}\mathbf{S}^{t}(X)=\mathbf{T}(\mathbf{S}^{t}(X)) . Koopman analysis shifts the focus from the state-space trajectory X(t) to the evolution of observables , i.e., scalar-valued functions f:\mathbb{C}^{n}\to\mathbb{C} belonging to a chosen function space \mathcal{F} . Given an observable f\in\mathcal{F} , its time evolution along the dynamics is the composition f(X(t))=f(\mathbf{S}^{t}(X_{0})) . The Koopman operator \mathcal{K}^{t}:\mathcal{F}\to\mathcal{F} is defined by 19 ; 4 ; 28 [\mathcal{K}^{t}f](X)=f(\mathbf{S}^{t}(X)), (2) i.e., \mathcal{K}^{t} maps an observable to another observable by composition with the flow. Crucially, although the underlying dynamics ( 1 ) is nonlinear, \mathcal{K}^{t} acts linearly on \mathcal{F} . The spectral properties of \mathcal{K}^{t} provide a linear representation of nonlinear dynamics. Koopman eigenfunctions \varphi_{j}\in\mathcal{F} satisfy \mathcal{K}^{t}\varphi_{j}=e^{\lambda_{j}t}\varphi_{j}, (3) where \lambda_{j} are the corresponding (continuous-time) Koopman eigenvalues. Each eigenfunction \varphi_{j} represents a dynamical mode whose temporal evolution is governed by e^{\lambda_{j}t} , allowing the long-time behavior of nonlinear dynamics to be organized in terms of dominant modes with the largest real parts of \lambda_{j} . In the linear limit \dot{X}=\mathbf{L}X , a linear observable \varphi(X)=w^{\dagger}X with w^{\dagger} being a left eigenvector of \mathbf{L} is a Koopman eigenfunction. Indeed, \displaystyle\mathcal{K}^{t}\varphi(X) \displaystyle=\varphi(e^{\mathbf{L}t}X)=w^{\dagger}e^{\mathbf{L}t}X=e^{\lambda t}\varphi(X), (4) where w^{\dagger}\mathbf{L}=\lambda w^{\dagger} . This establishes a direct correspondence between the Koopman framework and the conventional spectral description in linear systems. In practice, we simulate the continuous-time dynamics ( 1 ) and sample the flow map as X_{k+1}=\mathbf{S}^{\Delta t}(X_{k}) with a fixed time step \Delta t . Applying extended dynamic mode decomposition (EDMD) 45 ; 20 ; 1 to snapshot pairs of observables then yields a finite-dimensional approximation of the Koopman operator \mathcal{K}^{\Delta t} associated with this sampling time. In this Letter, we exploit the structure of Koopman eigenfunctions in a lifted observable space to reveal a new type of non-Hermitian skin effect unique to nonlinear systems. Model .— To identify a nonlinear skin effect within the Koopman framework, we study a nonlinear extension of the Hatano-Nelson model described by \text{i}\frac{dx_{n}}{dt}=T_{\rm R}(X)\,x_{n-1}+T_{\rm L}(X)\,x_{n+1}, (5) where T_{\rm R}(X) and T_{\rm L}(X) denote state-dependent nonreciprocal hopping amplitudes T_{\rm R}(X)\!=\!t_{\rm R}\!\left[\!(1-\epsilon)\!+\!\epsilon\,\frac{x_{n-1}}{x_{n}}\right],\,T_{\rm L}(X)\!=\!t_{\rm L}\!\left[\!(1-\epsilon)\!+\!\epsilon\,\frac{x_{n+1}}{x_{n}}\right] (6) with \epsilon\in[0,1] controlling the strength of the nonlinearity. In the linear limit \epsilon=0 , the model reduces to the conventional Hatano-Nelson Hamiltonian 11 ; 12 . For t_{\rm R}>t_{\rm L} , all eigenstates localize exponentially at the right boundary, giving rise to the linear skin effect, which has been extensively studied in a wide range of contexts 3 ; 2 ; 36 ; 21 ; 22 ; 53 ; 33 ; 35 ; 23 ; 54 ; 32 ; 16 ; 49 ; 17 ; 40 ; 5 ; 9 ; 29 ; 39 ; 26 ; 38 ; 31 ; 10 ; 41 ; 30 ; 55 ; 18 ; 7 ; 50 ; 52 ; 51 ; 6 ; 57 ; 27 ; 25 ; 48 . This phenomenon has been experimentally observed in diverse platforms, including ultracold atoms 24 ; 8 ; 56 , electric circuits 13 ; 14 , and photonic lattices 44 . Beyond the linear regime, nonlinear extensions of the skin effect based on stationary nonlinear modes have also been demonstrated experimentally in photonic systems 43 . In the nonlinear regime \epsilon>0 , the hopping amplitudes acquire an explicit dependence on the instantaneous state X , rendering the dynamics nonlinear and spatially inhomogeneous. Importantly, this nonlinearity does not introduce onsite terms but instead acts as a feedback mechanism that modifies the effective hopping processes in a configuration-dependent manner.