Global Analytical Solution of the Identical Kuramoto Model for N=3 via Koopman Eigenfunctions

arXiv:2605.07360 · nlin.AO, math.DS, nlin.SI · Submitted 2026-05-08 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Nonlinear Sciences. The day in nonlinear science: chaos, patterns, networks and adaptation.

Kai: We're looking at the paper: "Global Analytical Solution of the Identical Kuramoto Model for N=3 via Koopman Eigenfunctions".

Mira: Here's the gist of the paper. Global analytical solution for the identical Kuramoto model with N=3 via Koopman eigenfunctions provides an explicit algebraic reconstruction of nonlinear phase dynamics from Koopman eigenfunctions.

Kai: This solution is obtained by constructing Koopman eigenfunctions that relate the phases to time and reducing the phase dynamics to time-dependent quartic equations.

Mira: Model Formulation and Method The study focuses on the all-to-all coupled Kuramoto model with identical oscillators for N=3, where the dynamics are governed by ˙θj = ω + K/N X N k=1 sin(θk − θj), j = 1, 2,..., N (1).

Kai: The approach involves constructing Koopman eigenfunctions that provide three independent relations between the phases and time for the Kuramoto model with identical oscillators for N=3.

Title and authors: Kai: So we’re looking at this paper today called "Global Analytical Solution of the Identical Kuramoto Model for N=three via Koopman Eigenfunctions <ref:2605.07360#pg1,Global Analytical Solution of the Identical Kuramoto Model for N=3 via>." It sounds like a lot of machinery, right?

Mira: Yeah, it does. The title tells you immediately that they've found a complete solution for the phase trajectories of the Kuramoto model when you have three identical oscillators, and they did it using these things called Koopman eigenfunctions.

Lev: And what this implies is that they can get exact solutions for something that was previously very hard to solve beyond the two oscillator case, which is where you have a separable differential equation <ref:2605.07360#pg2>.

Kai: Exactly. It means we can finally track how the phases evolve for three oscillators from any starting point, not just a few special initial conditions <ref:2605.07360#pg1>.

Mira: The authors are tackling this by constructing these specific eigenfunctions that link the phases to time, which then lets them reduce the complex dynamics down to solving some time-dependent quartic equations <ref:2605.07360#pg1>.

Lev: From an engineering standpoint, if you were building a system based on this model, having a closed-form solution for the trajectories would mean you could predict the exact state of the system at any future time without needing to run complex numerical simulations for every scenario <ref:2605.07360#pg2>.

Kai: That’s what I’m thinking. It moves us from just seeing a picture of how synchronization happens to actually knowing the exact path it takes through that state space.

The paper's summary: Mira: So, the core of this work is this explicit algebraic reconstruction of nonlinear phase dynamics from those Koopman eigenfunctions <ref:2605.07360#pg1>. They are using these eigenfunctions to find relations between the phases and time.

Kai: It’s like they find a special set of mathematical "coordinates" that simplify the nonlinear movement into something manageable, specifically a time-dependent quartic equation <ref:2605.07360#pg1>.

Lev: From what I can tell, this construction relies on building these eigenfunctions based on growth rates of observables related to phase differences and differences of phase differences <ref:2605.07360#pg3>.

Mira: Right. They combine those growth rates, and they linearize things down by reducing them to a constant value, which turns out to be negative K, yielding eigenfunctions psi one, psi two, and psi three with that eigenvalue <ref:2605.07360#pg1>.

Kai: So instead of tracking the phases directly in a complicated differential equation, they use these eigenfunctions as tools to reduce the problem down to solving those fourth degree polynomials <ref:2605.07360#pg1>.

Lev: And then they use those solutions to get back to the actual phase differences, which they define using these new variables X j equal e raised to i delta j, and then taking the logarithm of those roots <ref:2605.07360#pg1>.

The paper's improvements: Kai: Now for what’s interesting is how they characterize the dynamics in a different way. They call this phase-amplitude reduction <ref:2605.07360#pg1>.

Mira: They break the dynamics down into three functions: the mean phase eigenfunction, psi ph, which evolves with a constant angular velocity, and an amplitude variable, psi amp.

Lev: And then there's this invariant component, psi inv. The paper says that they can define a ratio of two eigenfunctions as another Koopman eigenfunction satisfying K times that ratio is zero <ref:2605.07360#pg1>.

Kai: So they’ve found a way to describe the dynamics linearly in terms of these three functions, where the dynamics take this specific linear form, d over time of psi ph psi amp psi inv equals some matrix times those three functions <ref:2605.07360#pg1>.

Mira: That’s significant because it shows that for the nontrivial part of the dynamics, even though it started nonlinear, you can characterize it using this linear evolution structure <ref:2605.07360#pg1>.

Lev: For real hardware, running on something physical would mean we need to figure out how to measure these three components—phase, amplitude, and invariant—accurately in a time-dependent way <ref:2605.07360#pg2>.

Conclusion: Kai: So to wrap up this paper on the "Global Analytical Solution of the Identical Kuramoto Model for N=three via Koopman Eigenfunctions," they’ve managed to provide an explicit algebraic reconstruction of nonlinear phase dynamics by solving time-dependent quartic equations derived from their Koopman eigenfunctions <ref:2605.07360#pg1>.

Mira: They also showed a way to characterize the dynamics using a phase-amplitude reduction involving the mean phase, amplitude, and invariant components <ref:2605.07360#pg1>.

Lev: And they suggest that by looking at the energy landscape of this N=three Kuramoto model as a quartic polynomial in terms of these eigenfunctions, you can see how it relates to the two real sheets of an XY energy landscape <ref:2605.07360#pg1>.

Kai: The implication is that for systems like this, we don't just get numerical approximations; we get a concrete algebraic path to understanding the full trajectory from any starting point <ref:2605.07360#pg1>.

Mira: It also hints at some deeper structure, because they found ways to identify phase-locked states, like splay states, by looking for singularities where the denominators of all three Koopman eigenfunctions vanish <ref:2605.07360#pg1>.

Lev: For larger systems, like N=four or N=five the paper suggests that while they can find some nontrivial Koopman eigenfunctions with eigenvalue negative K, they haven't managed to get that explicit global reconstruction for those larger numbers yet <ref:2605.07360#pg2>.

Kai: So the challenge moving forward is figuring out how to scale these methods up from N=three to, say, N=four or higher while keeping that explicit solution structure <ref:2605.07360#pg2>.

Mira: And it opens up a lot of avenues for understanding collective behavior in coupled oscillators because we can now map the dynamics onto this linear phase-amplitude reduction framework <ref:2605.07360#pg1>.

Lev: It’s a solid paper, but the caveat is that if you want to run this on real hardware, you still have to deal with all those measurement challenges for the amplitude and invariant components <ref:2605.07360#pg2>.

Tokyo University of Science

nlin.AO, math.DS, nlin.SI

Submitted: 2026-05-08

Updated: 2026-11-07

Comments: 36 pages, 5 figures

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 76/100

The gist: Global analytical solution for the identical Kuramoto model with N=3 via Koopman eigenfunctions provides an explicit algebraic reconstruction of nonlinear phase dynamics from Koopman eigenfunctions.

Key concepts

Koopman Eigenfunctions
These functions describe how the phases of the Kuramoto oscillators evolve over time. The paper constructs specific eigenfunctions that relate these phases to time, allowing complex nonlinear dynamics to be analyzed through simpler linear evolution in a transformed space.
Time-Dependent Quartic Equations
The core mathematical tool used is reducing the phase problem into solving four-degree polynomial equations that change with time. Solving these equations algebraically allows for the exact reconstruction of the original phase trajectories of the oscillators.
Phase-Amplitude Reduction Dynamics
The system simplifies by characterizing dynamics using three variables: a mean phase eigenfunction, an amplitude variable, and an invariant component. This reduction transforms the complex Kuramoto model into a linear differential equation in these new variables.
Energy Landscape Interpretation
The model's energy landscape is mapped onto the Koopman eigenfunctions. The structure of the resulting quartic equations reveals that the dynamics correspond to two real sheets, mirroring the two-sheet structure found in classical XY models.

Terminology

Summary

Global analytical solution for the identical Kuramoto model with N=3 via Koopman eigenfunctions provides an explicit algebraic reconstruction of nonlinear phase dynamics from Koopman eigenfunctions. p. 0

Where the paper says it

Global Analytical Solution of the Identical Kuramoto Model for N = 3 via Koopman Eigenfunctions Keisuke Taga Department of Physics and Astronomy, Tokyo University of Science, Yamazaki 2641, Noda, 278-8510, Chiba, Japan Abstract The Kuramoto model is a paradigmatic model of collective synchronization in coupled oscillator systems. Although its mathematical properties have been extensively investigated, exact phase trajectories from arbitrary initial conditions have been available only for the simplest case, N = 2.

This solution is obtained by constructing Koopman eigenfunctions that relate the phases to time and reducing the phase dynamics to time-dependent quartic equations. p. 0
Where the paper says it

In this study, we provide a global analytical solution for the phase trajectories of the all-to-all coupled Kuramoto model with identical oscillators for N = 3. This solution is obtained by constructing Koopman eigenfunctions that relate the phases to time and reducing the phase dynamics to time-dependent quartic equations.

Model Formulation and Method

The study focuses on the all-to-all coupled Kuramoto model with identical oscillators for N=3, where the dynamics are governed by ˙θj = ω + K/N X N k=1 sin(θk − θj), j = 1, 2,..., N (1). The approach involves constructing Koopman eigenfunctions that provide three independent relations between the phases and time for the Kuramoto model with identical oscillators for N=3. p. 1

Where the paper says it

The present work has two main components. First, we construct Koopman eigenfunctions that provide three independent relations between the phases and time for the Kuramoto model with identical oscillators for N = 3.

These eigenfunctions are derived using a theorem based on growth rates, which yields six observables related to phase differences and differences of phase differences. p. 4
Where the paper says it

Therefore, the theorem yields growth rates for two families of observables, sin(∆j/2) and sin((∆j − ∆k)/2). Accordingly, we consider the following six observables: three related to phase differences and three from differences of phase differences.

By combining the corresponding growth rates, linear combinations reduce to the constant −K, yielding Koopman eigenfunctions ψ1,ψ2,ψ3 in Eq. p. 5
Where the paper says it

By combining the corresponding growth rates, we find linear combinations that reduce to the constant −K, yielding the following Koopman eigenfunctions (see Appendix B): ψ1 = sin3 ∆1 2 sin ∆1−∆2 2 sin ∆3−∆1 2 , ψ2 = sin3 ∆2 2 sin ∆2−∆3 2 sin ∆1−∆2 2 , ψ3 = sin3 ∆3 2 sin ∆3−∆1 2 sin ∆2−∆3 2 . (18) That is, Kψj = −Kψj (j = 1, 2, 3).

(18) with eigenvalue −K. p. 2
Where the paper says it

(2) Although the vector field F is generally nonlinear, the Koopman generator is linear, so nonlinear dynamics can be analyzed from a linear perspective. An observable ψλ and a scalar λ ∈ C satisfying dψλ dt = Kψλ = λψλ (3) are called a Koopman eigenfunction and the corresponding Koopman eigenvalue, respectively.

Trajectory Reconstruction

The exact phase trajectories are recovered by reducing the inverse problem to solving time-dependent quartic equations. The algebraic analysis introduces new variables Xj = e i∆j, leading to polynomials (P1) and (P2). p. 6

Where the paper says it

For the algebraic analysis, we introduce new variables Xj = e i∆j . By substituting them into Eq.

These are reduced to one-variable quartic polynomials Rj(Xj) = a 4X 4j + a 3X 3j + a 2X 2j + a 1Xj + a 0, (22), whose coefficients depend on the Koopman eigenfunctions. p. 6
Where the paper says it

Since Xj = 0 is not on the unit circle and the factor Xj = 1 corresponds to a fixed point of the Xj dynamics, we separate these factors off. After removing these factors, we obtain the one-variable quartic polynomials Rj (Xj ) = aj,4X 4 j + aj,3X 3 j + aj,2X 2 j + aj,1Xj + aj,0, (22) whose coefficients aj,k depend on the Koopman eigenfunctions.

After removing trivial factors, the reconstruction is reduced to solving R1(X1) = 0 and R2(X2) = 0. p. 6
Where the paper says it

The explicit coefficients are given in Appendix C. Thus, after removing the trivial factors, the reconstruction problem is reduced to solving the quartic equations R1(X1) = 0 and R2(X2) = 0.

The phase differences are then recovered from the selected unit-circle roots as ∆j(t) = −i log Xj(t), where the branches of the logarithm are chosen so that ∆j (0) agrees with the given initial condition. p. 7
Where the paper says it

The trajectory associated with the given initial condition is obtained by choosing the unit-circle root that satisfies the original equations (P1) and (P2) and agrees with the initial condition at t = 0. The phase differences are then recovered from the selected unit-circle roots as ∆1(t) = −i log X1(t), ∆2(t) = −i log X2(t), (23) where the branches of the logarithm are chosen so that ∆j (0) agrees with the given initial condition.

The original phases are then recovered using θ1 = Θ + 2∆1 + ∆2 cubed, θ2 = Θ + ∆2 − ∆1 cubed, and θ3 = Θ + −∆1 − 2∆2 cubed. p. 7
Where the paper says it

The phase differences are then recovered from the selected unit-circle roots as ∆1(t) = −i log X1(t), ∆2(t) = −i log X2(t), (23) where the branches of the logarithm are chosen so that ∆j (0) agrees with the given initial condition. Thus, the phase dynamics of the oscillators θ1, θ2 and θ3 are recovered as follows.

Phase-Amplitude Reduction

The Koopman eigenfunctions characterize the dynamics in terms of phase, amplitude, and invariant components. The mean phase evolves with constant angular velocity as Θ(t) = ωt + Θ(0), represented by the eigenfunction ψph = e iΘ (28). p. 3

Where the paper says it

(8) To derive exact trajectories, we seek three independent relations connecting the three phases to time. First, a trivial Koopman eigenfunction follows from the fact that the mean phase of the Kuramoto model (8) evolves with constant velocity: Θ =˙ ω, Θ := 1 3 X 3 j=1 θj , (9) so that Θ(t) = ωt + 1 3 X 3 j=1 θj (0).

For the nontrivial part of the dynamics, eigenfunctions ψ1, ψ2 and ψ3 evolve with eigenvalue −K. p. 5
Where the paper says it

These eigenfunctions capture the nontrivial exponentially evolving part of the dynamics. Indeed, ψ1 +ψ2 +ψ3 = 0 holds (see Appendix A.

The ratio of two eigenfunctions, for instance ψinv:= ψ1/ψ2, is also a Koopman eigenfunction satisfying Kψinv = 0. p. 5
Where the paper says it

2), and only two of them are independent; for example, ψ1 and ψ2. Together with ψph, these eigenfunctions define nonlinear coordinates for the identical Kuramoto model with N = 3, in which the dynamics is reduced to the linear system d dt   ψph ψ1 ψ2   =   iω 0 0 0 −K 0 0 0 −K     ψph ψ1 ψ2   .

This defines a nontrivial invariant. p. 4
Where the paper says it

In the present case N = 3, however, no Watanabe–Strogatz invariant is available. To construct the nontrivial Koopman eigenfunctions, we apply the construction based on growth rates.

The dynamics can be characterized by three functions: the phase eigenfunction ψph, the amplitude variable ψamp, and the invariant ψinv. p. 5
Where the paper says it

(19) yields ψph(t) = e iωt ψph(0), ψ1(t) = e −Kt ψ1(0), ψ2(t) = e −Kt ψ2(0). (20) By taking suitable combinations of these eigenfunctions, the dynamics can be viewed in terms of phase, amplitude, and invariant components, as discussed in Sec.

In these variables, the dynamics takes the linear form d/dt (ψph ψamp ψinv) = (iω 0 0; 0-K 0; 0 0 0) (ψph ψamp ψinv). p. 5
Where the paper says it

2), and only two of them are independent; for example, ψ1 and ψ2. Together with ψph, these eigenfunctions define nonlinear coordinates for the identical Kuramoto model with N = 3, in which the dynamics is reduced to the linear system d dt   ψph ψ1 ψ2   =   iω 0 0 0 −K 0 0 0 −K     ψph ψ1 ψ2   .

Energy Landscape Interpretation

The identical all-to-all Kuramoto model for N=3 can be viewed as the gradient flow of a three-site classical XY model. The Hamiltonian is H = −J X 3 j=1 cos(θj+1 − θj), (26). p. 0

Where the paper says it

One of the most fundamental models of synchronization is the Kuramoto model [2–4]. ˙θj = ωj + K N X N k=1 sin(θk − θj ), j = 1, 2, .

The energy landscape can be represented in terms of the Koopman eigenfunctions. p. 8
Where the paper says it

However, as discussed in Sec. 2, the difference between the two unitcircle branches may be characterized in terms of the two energy sheets of the XY model.

In the generic case, the discriminant shows that the quartic equation has two real roots, which are labeled as the lower- and higher-energy branches. p. 7
Where the paper says it

By the discriminant analysis in Appendix C, the quartic polynomial has two roots on the unit circle in the generic case. These roots give a finite set of candidate algebraic branches.

This two-sheet structure appears to be related to the two unit-circle branches in Fig. p. 8
Where the paper says it

As a typical example, we show all four candidate branches for a representative initial condition on the complex plane in Fig. Unit-circle branches We first discuss the two unit-circle branches.

2: one represented by (a-1) and (b-1), and the other by (a-2) and (b-2). p. 7
Where the paper says it

Although the roots can be written explicitly, radical expressions involve branch cuts, and a single radical representation may not represent the selected algebraic root over the whole time interval. When such a branch cut is encountered, the same algebraic root is represented by another one of the four radical representations of the quartic equation, chosen so as to be continuous with the previously used representation.

Extension to Larger Systems

For N=4, the same growth-rate construction based on phase differences still yields a nontrivial Koopman eigenfunction with eigenvalue −K. This suggests an integrable structure beyond N=3. p. 1

Where the paper says it

For small-N Kuramoto systems, several studies have clarified the stability, bifurcation, and asymptotic states by analytical and numerical methods [9– 17]. However, obtaining exact phase trajectories from arbitrary initial conditions remains difficult beyond the simplest case N = 2, where the dynamics reduces to a separable differential equation.

However, these relations do not yet yield an explicit reconstruction of the full phase trajectory in the sense achieved here for N=3. p. 0
Where the paper says it

The algebraic branch corresponding to the initial condition is then selected to recover the corresponding phase trajectory. This gives an explicit algebraic reconstruction of the nonlinear phase dynamics from Koopman eigenfunctions.

For N ≥ 5, the present phase-difference construction provides Watanabe–Strogatz invariants but appears to yield no additional Koopman eigenfunctions. p. 4
Where the paper says it

In the present case N = 3, however, no Watanabe–Strogatz invariant is available. To construct the nontrivial Koopman eigenfunctions, we apply the construction based on growth rates.

Whether similarly explicit global analytical reconstructions can be obtained for larger systems remains an open question. p. 0
Where the paper says it

In this study, we provide a global analytical solution for the phase trajectories of the all-to-all coupled Kuramoto model with identical oscillators for N = 3. This solution is obtained by constructing Koopman eigenfunctions that relate the phases to time and reducing the phase dynamics to time-dependent quartic equations.

The Watanabe–Strogatz transformation provides a three-dimensional reduction of the essential dynamics of identical oscillator systems, and N=3 is the smallest nonredundant realization of this reduction. p. 1
Where the paper says it

However, obtaining exact phase trajectories from arbitrary initial conditions remains difficult beyond the simplest case N = 2, where the dynamics reduces to a separable differential equation. For identical-frequency systems, Watanabe and Strogatz showed that the dynamics can be reduced to a three-dimensional system by the Watanabe– Strogatz transformation [18–21].

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Global Phase Trajectory Reconstruction for N=3 Kuramoto Model: The improved system can now provide an explicit algebraic reconstruction of nonlinear phase dynamics from Koopman eigenfunctions by solving time-dependent quartic equations derived from the growth rates of specific observables, as described in Section 3.

  2. Phase-Amplitude Reduction Dynamics: The system can characterize the dynamics using three functions—the mean phase eigenfunction psiph, the amplitude variable psiamp, and the invariant psiinv—allowing for a linear form of dynamics d(t)(psiph psiamp psiinv) = M(t)(psiph psiamp psiinv), as shown in Equation (33).

  3. Energy Landscape Mapping: The improved system can represent the energy landscape of the N=3 Kuramoto model as a quartic polynomial PH(H) in terms of Koopman eigenfunctions, where the two real roots correspond to the two real sheets of the XY energy landscape, as detailed in Appendix E.

  4. Splay State Identification: The system can identify phase-locked states like splay states by analyzing singularities where the denominators of all three Koopman eigenfunctions vanish, leading to fixed points in the phase-difference dynamics, as discussed in Appendix D.

  5. Integrability Suggestions for N=4: For N=4, the construction yields a nontrivial Koopman eigenfunction with eigenvalue -K and suggests an integrable structure, although it does not yet provide an explicit global reconstruction comparable to the N=3 case.

Abstract

The Kuramoto model is a paradigmatic model of collective synchronization in coupled oscillator systems. Although its mathematical properties have been extensively investigated, explicit phase trajectories from arbitrary initial conditions have been available only for the simplest case, N = 2. In this study, we provide a global analytical solution for the phase trajectories of the all-to-all coupled Kuramoto model with three identical oscillators (N = 3). We construct Koopman eigenfunctions from products of observables whose growth rates, weighted by the exponents, sum to a constant, and obtain a nontrivial constant of motion from the ratio of two such eigenfunctions. Using their explicit exponential time dependence, we reduce trajectory reconstruction to time-dependent quartic equations and recover the phase trajectories. The same construction also yields Koopman eigenfunctions for the Kuramoto-Sakaguchi model with arbitrary phase lag and for the N = 4 all-to-all and ring-coupled systems. These results provide a concrete example in which Koopman eigenfunctions are used not only to characterize but also to explicitly solve a widely studied nonlinear model.

Sources

Related papers