Existence and uniqueness of spectral submanifolds beyond hyperbolicity, nonresonance, locality, and weak forcing

Spectral submanifolds (SSMs) are the unique smoothest invariant manifolds tangent to spectral subspaces of the linearization at a fixed point or recurrent motion of a dissipative nonlinear system, and they form the mathematical foundation of exact nonlinear model reduction. Haller and Ponsioen (2016) proved their existence, uniqueness in a suitable smoothness class, and persistence, but only under four restrictive hypotheses: the linearized spectrum must be asymptotically stable (all eigenvalues in the open left half-plane), low-order resonances between the chosen spectral subspace and its spectral complement must be excluded, the constructed manifolds are purely local objects, and in the non-autonomous (forced) case the exact nonresonance condition is only known to be $\mathcal{O}(\epsilon)$ -close to the linear criterion at zero forcing. The open problem is to complete this existence-uniqueness theory: cover spectra with zero real parts, characterize existence, smoothness, and uniqueness when resonances are present, determine the exact non-autonomous nonresonance condition at finite forcing amplitude, and construct globally valid, fold-free parametrizations. A decade of intense follow-up work has resolved the computational side of the program but only partially advanced each theoretical component.

Background

Linear vibration theory decomposes small-amplitude motion into independent normal modes, each a linear invariant subspace of phase space spanned by eigenvectors of the linearized dynamics. For nonlinear systems this decomposition fails, and two classic generalizations exist. Rosenberg defined a nonlinear normal mode (NNM) of a conservative mechanical system as a synchronous periodic oscillation in which all generalized coordinates reach their extrema simultaneously. Shaw and Pierre proposed instead to regard a mode as an invariant manifold tangent to a modal subspace of the linearization at an equilibrium: restricting the full nonlinear dynamics to such a manifold yields an exact low-dimensional reduced model. The second view is the one that supports model reduction, but it faces a fundamental ambiguity: invariant manifolds tangent to a given eigenspace are generally non-unique (already in a linear system there are infinitely many), and the formal Taylor expansions used to approximate them can diverge, so no member of the family is distinguished. Classical invariant-manifold theory gives only partial guidance: Fenichel's theory of normally hyperbolic invariant manifolds (invariant manifolds whose normal contraction dominates their tangential contraction, which guarantees persistence under perturbation) provides existence and robustness but no distinguished smoothest member and often low smoothness, while the parametrization method of Cabre, Fontich and de la Llave constructs invariant manifolds as embeddings solving an invariance equation, obtaining uniqueness in a high smoothness class under nonresonance assumptions.

Haller and Ponsioen unified this picture for dissipative oscillatory systems. They define an NNM as any set of eternally recurrent motions with finitely many frequencies - a fixed point, a periodic orbit, or the closure of a quasiperiodic orbit - and a spectral submanifold (SSM) as the smoothest invariant manifold tangent to a spectral subspace $E$ of the linearization along an NNM. (A spectral subspace is a direct sum of real eigenspaces of the linearized operator $A$ ; each real eigenspace is spanned by the real and imaginary parts of the generalized eigenvectors belonging to one eigenvalue.) For the autonomous system $\dot{x}=Ax+f(x)$ with $f(x)=\mathcal{O}(|x|^2)$ and $A$ asymptotically stable - all eigenvalues satisfy $\mathrm{Re}\,\lambda_j<0$ , so the fixed point attracts nearby trajectories - their main theorem states: under the low-order nonresonance conditions $\langle m,\lambda\rangle_E\neq\lambda_l$ for all eigenvalues $\lambda_l$ of $A$ outside the spectrum of $A$ restricted to $E$ and all integer vectors $m$ with $2\le|m|\le\sigma(E)$ , there exists a class- $C^r$ SSM tangent to $E$ , unique among all $C^{\sigma(E)+1}$ invariant manifolds with that tangency. Here $\sigma(E)$ , the relative spectral quotient, is the integer part of the ratio between the weakest decay rate outside $E$ and the strongest decay rate inside $E$ ; it can be large even for weakly damped systems, which is why damping is essential to the conclusions. Depending on where $E$ sits in the spectrum one speaks of fast, intermediate, and slow SSMs; slow SSMs, built over the slowest-decaying modes, are the most useful for model reduction but have the most restrictive uniqueness. Under periodic or quasiperiodic forcing of amplitude $\epsilon$ , an analogous theorem gives time-dependent SSMs attached to a unique nearby NNM, with an absolute spectral quotient $\Sigma(E)$ replacing $\sigma(E)$ and nonresonance conditions involving only the real parts of the eigenvalues.

This theory became the foundation of a productive program: automated SSM computation in the open-source package SSMTool, analytic prediction of forced-response curves, of backbone curves (the amplitude-frequency loci of the unforced periodic responses), and of isolated resonance curves (isolas - detached response branches that numerical continuation from zero amplitude generally misses), analysis of internally resonant forced systems, and data-driven extraction of SSMs from measured trajectories. The existence-uniqueness theorems themselves, however, rest on four restrictions that the 2016 paper explicitly flags or exemplifies without removing. First, hyperbolicity: eigenvalues must lie strictly in the open left half-plane (the purely repelling case follows by time reversal, but eigenvalues with zero real parts are excluded altogether, so conservative, gyroscopic (systems with velocity-dependent forces of Coriolis type, such as rotating shafts), and bifurcating systems are not covered). Second, nonresonance: when the low-order nonresonance conditions fail, the paper's simple polynomial examples show that either no invariant manifold of the required smoothness exists, or infinitely many analytic invariant manifolds coexist with no smoothest member - and the general boundary between these outcomes is unknown. Third, locality: every construction (graph-based or parametrization-based) is local to a neighborhood of the NNM, and graph representations break down at folds, where the SSM ceases to be a single-valued graph over $E$ . Fourth, in the non-autonomous case the exact nonresonance condition ensuring convergence at finite forcing amplitude is not known - only that it is $\mathcal{O}(\epsilon)$ -close to the linear ( $\epsilon=0$ ) criterion, because the spectrum of the infinite-dimensional transfer operator appearing in the proof is computable only at $\epsilon=0$ . For context on the zero-damping regime: a Lyapunov subcenter manifold (LSM) is the analytic manifold of periodic orbits tangent to a two-dimensional elliptic (purely imaginary) eigenspace of a fixed point of a conservative system with a conserved quantity, guaranteed by Lyapunov's subcenter theorem; it is the natural zero-damping relative of an SSM, and how LSMs relate to SSMs under dissipation is itself only partially understood.

Problem Statement

For finite-dimensional autonomous nonlinear systems $\dot{x}=Ax+f(x)$ with $f(x)=\mathcal{O}(|x|^2)$ , a fixed point at the origin, and $A$ hyperbolic and asymptotically stable, and for their periodically or quasiperiodically forced counterparts, the theorems of Haller and Ponsioen decide the existence and the uniqueness class of the spectral submanifold over any spectral subspace $E$ by explicitly checkable spectral conditions. The open problem is to complete this existence-uniqueness theory by removing its remaining restrictions, so that the same questions are settled for general spectra and forcing regimes:

Non-hyperbolic spectra. Establish existence, smoothness, and uniqueness-class results for SSMs when the spectrum of $A$ contains eigenvalues with zero real part - conservative and gyroscopic systems, equilibria at or near Hopf bifurcations (where a complex-conjugate eigenvalue pair of $A$ crosses the imaginary axis and a periodic orbit is born) and other local bifurcations, and mixed attracting/center/repelling spectra - deciding in particular when a distinguished smoothest nonlinear continuation of $E$ exists and when it provably does not.
Resonant spectra. When the low-order nonresonance conditions between $E$ and its spectral complement fail, give a complete characterization: necessary and sufficient conditions for the existence of an invariant manifold tangent to $E$ , the exact smoothness class attainable, and the uniqueness class - separating in general the regimes of non-existence, of finite-smoothness existence, and of infinite non-uniqueness.
Exact non-autonomous nonresonance. In the forced setting, determine the exact nonresonance condition at finite forcing amplitude $\epsilon$ under which the time-dependent SSM over $E$ exists and its Taylor-Fourier parametrization converges; currently the condition is only known to be $\mathcal{O}(\epsilon)$ -close to the linear criterion obtained at $\epsilon=0$ .
Global validity. Construct parametrizations of SSMs that remain valid beyond a single local patch over $E$ - in particular through folds, where the SSM ceases to be a graph over $E$ - with an explicitly characterized domain of validity.

These four components are the hypotheses and the domain of validity of one and the same existence-uniqueness theorem, so they form a single objective: a completed theory in which, for any spectral configuration and forcing regime, the existence question, the attainable smoothness, the uniqueness class, and the domain of validity of the SSM parametrization are all decided by explicitly checkable conditions. Settling only one of these components, or only a proper subset of them, leaves the theory incomplete and does not resolve the objective; the four are jointly required, and the objective is met only when all four are settled.

References

George Haller and Sten Ponsioen, 'Nonlinear normal modes and spectral submanifolds: existence, uniqueness and use in model reduction', Nonlinear Dynamics 86 (2016) 1493-1534. DOI: 10.1007/s11071-016-2974-z, https://doi.org/10.1007/s11071-016-2974-z. arXiv:1602.00560, https://arxiv.org/abs/1602.00560.
R. M. Rosenberg, 'On nonlinear vibrations of systems with many degrees of freedom', Advances in Applied Mechanics (1966) 155-242. DOI: 10.1016/s0065-2156(08)70008-5, https://doi.org/10.1016/s0065-2156(08)70008-5.
S. W. Shaw and C. Pierre, 'Normal modes for non-linear vibratory systems', Journal of Sound and Vibration 164 (1993) 85-124. DOI: 10.1006/jsvi.1993.1198, https://doi.org/10.1006/jsvi.1993.1198.
X. Cabre, E. Fontich and R. de la Llave, 'The parameterization method for invariant manifolds I: manifolds associated to non-resonant subspaces', Indiana University Mathematics Journal 52 (2003) 283-328. DOI: 10.1512/iumj.2003.52.2245, https://doi.org/10.1512/iumj.2003.52.2245.
N. Fenichel, 'Persistence and smoothness of invariant manifolds for flows', Indiana University Mathematics Journal 21 (1971) 193-226. DOI: 10.1512/iumj.1972.21.21017, https://doi.org/10.1512/iumj.1972.21.21017.
Rafael de la Llave and Florian Kogelbauer, 'Global persistence of Lyapunov subcenter manifolds as spectral submanifolds under dissipative perturbations', SIAM Journal on Applied Dynamical Systems 18 (2019) 2099-2142. DOI: 10.1137/18M1210344, https://doi.org/10.1137/18M1210344. (arXiv:1809.00924, https://arxiv.org/abs/1809.00924.)
Sten Ponsioen, Tiemo Pedergnana and George Haller, 'Automated computation of autonomous spectral submanifolds for nonlinear modal analysis', Journal of Sound and Vibration 420 (2018) 269-295. DOI: 10.1016/j.jsv.2018.01.048, https://doi.org/10.1016/j.jsv.2018.01.048. (arXiv:1709.00886, https://arxiv.org/abs/1709.00886.)
Sten Ponsioen, Tiemo Pedergnana and George Haller, 'Analytic prediction of isolated forced response curves from spectral submanifolds', Nonlinear Dynamics 98 (2019) 2755-2773. DOI: 10.1007/s11071-019-05023-4, https://doi.org/10.1007/s11071-019-05023-4. (arXiv:1812.06664, https://arxiv.org/abs/1812.06664.)
Sten Ponsioen, Shobhit Jain and George Haller, 'Model reduction to spectral submanifolds and forced-response calculation in high-dimensional mechanical systems', Journal of Sound and Vibration 488 (2020) 115640. DOI: 10.1016/j.jsv.2020.115640, https://doi.org/10.1016/j.jsv.2020.115640.
Shobhit Jain and George Haller, 'How to compute invariant manifolds and their reduced dynamics in high-dimensional finite-element models', Nonlinear Dynamics 107 (2021) 1417-1450. DOI: 10.1007/s11071-021-06957-4, https://doi.org/10.1007/s11071-021-06957-4. (arXiv:2103.10264, https://arxiv.org/abs/2103.10264.)
Thomas Breunung and George Haller, 'Explicit backbone curves from spectral submanifolds of forced-damped nonlinear mechanical systems', Proceedings of the Royal Society A 474 (2018) 20180083. DOI: 10.1098/rspa.2018.0083, https://doi.org/10.1098/rspa.2018.0083.
Mingwu Li, Shobhit Jain and George Haller, 'Nonlinear analysis of forced mechanical systems with internal resonance using spectral submanifolds, Part I: Periodic response and forced response curve', Nonlinear Dynamics 110 (2022) 1005-1043. DOI: 10.1007/s11071-022-07714-x, https://doi.org/10.1007/s11071-022-07714-x. (arXiv:2106.05162, https://arxiv.org/abs/2106.05162.)
Mingwu Li and George Haller, 'Nonlinear analysis of forced mechanical systems with internal resonance using spectral submanifolds, Part II: Bifurcation and quasi-periodic response', Nonlinear Dynamics 110 (2022) 1045-1080. DOI: 10.1007/s11071-022-07476-6, https://doi.org/10.1007/s11071-022-07476-6. (arXiv:2108.08152, https://arxiv.org/abs/2108.08152.)
George Haller, Balint Kaszas, Aihui Liu and Joar Axas, 'Nonlinear model reduction to fractional and mixed-mode spectral submanifolds', Chaos 33 (2023) 063138. DOI: 10.1063/5.0143936, https://doi.org/10.1063/5.0143936. (arXiv:2305.15796, https://arxiv.org/abs/2305.15796.)
George Haller and Roshan S. Kaundinya, 'Nonlinear model reduction to temporally aperiodic spectral submanifolds', Chaos 34 (2024) 043152. DOI: 10.1063/5.0187080, https://doi.org/10.1063/5.0187080. (arXiv:2404.05355, https://arxiv.org/abs/2404.05355.)
Thomas Thurnher, George Haller and Shobhit Jain, 'Nonautonomous spectral submanifolds for model reduction of nonlinear mechanical systems under parametric resonance', Chaos 34 (2024) 073127. DOI: 10.1063/5.0168431, https://doi.org/10.1063/5.0168431. (arXiv:2307.10240, https://arxiv.org/abs/2307.10240.)
Zhenwei Xu, Roshan S. Kaundinya, Shobhit Jain and George Haller, 'Nonlinear model reduction to random spectral submanifolds in random vibrations', arXiv:2407.03677 (2024), https://arxiv.org/abs/2407.03677.
James King, Balint Kaszas, Gergely Buza, William Jussiau and George Haller, 'Parametric spectral submanifolds across Hopf bifurcations with applications to fluid dynamics', arXiv:2603.19079 (2026), https://arxiv.org/abs/2603.19079. (Preprint submitted to Chaos.)
Robert Szalai, 'Conservative limit of two-dimensional spectral submanifolds', arXiv:1805.07837 (2018, withdrawn by the author: 'The proof has an error and it requires a different approach'), https://arxiv.org/abs/1805.07837.
Gergely Buza, 'Spectral submanifolds of the Navier-Stokes equations', SIAM Journal on Applied Dynamical Systems 23 (2024) 1052-1089. DOI: 10.1137/23M154858x, https://doi.org/10.1137/23M154858x. (arXiv:2301.07898, https://arxiv.org/abs/2301.07898.)
Gergely Buza and George Haller, 'Existence of spectral submanifolds in time delay systems', arXiv:2512.22062 (2025), https://arxiv.org/abs/2512.22062.
Gergely Buza and George Haller, 'Spectral submanifold reduction for PDEs describing nonlinear continuum vibrations', arXiv:2607.10675 (2026), https://arxiv.org/abs/2607.10675.
M. Krack and J. Gross, 'Discussion of S. Ponsioen, S, Jain and G. Haller: "Model reduction to spectral submanifolds and forced-response calculation in high-dimensional mechanical systems, Journal of Sound and Vibration 488, 2020, pages 1-23"', Journal of Sound and Vibration 597 (2025) 118808. DOI: 10.1016/j.jsv.2024.118808, https://doi.org/10.1016/j.jsv.2024.118808.
Sten Ponsioen, Shobhit Jain and George Haller, 'Response to "Discussion of S. Ponsioen, S. Jain and G. Haller: Model reduction to spectral submanifolds and forced-response calculation in high-dimensional mechanical systems, Journal of Sound and Vibration 488, 2020, pages 1-23"', Journal of Sound and Vibration 597 (2025) 118791. DOI: 10.1016/j.jsv.2024.118791, https://doi.org/10.1016/j.jsv.2024.118791.