Secondary Periodic Orbits

• Secondary periodic orbits are additional periodic motions arising from bifurcations and resonance with primary structures in various dynamical systems.
• They manifest in fields like celestial mechanics, fluid dynamics, nonlinear vibrations, and billiards, where they guide transitions to chaos or complex motion.
• Methodologies involve semi-numerical techniques, spectral submanifold reduction, and Floquet analysis to accurately detect and continue these secondary branches.

Secondary periodic orbits are periodic motions that arise beyond the simplest closed trajectories of a dynamical system and are attached to a primary structure by resonance, bifurcation, continuation, symmetry, or hierarchical modulation. Across recent and classical arXiv literature, the expression does not denote a single universally fixed object. In nonlinear vibration it can denote the periodic orbit of a reduced-order model whose lift is a two-frequency torus created in a secondary Hopf bifurcation; in fluid mechanics it can denote periodic states born from secondary bifurcations of coherent equilibria; in celestial mechanics it can denote continued or bifurcated orbit families; in billiards it aligns with additional non-Birkhoff branches beyond the ordered Birkhoff polygons; and in relativistic timing problems it can refer to a secondary orbital clock superposed on a primary recurrence (Liang et al., 2024, Kreilos et al., 2012, Belbruno et al., 2018, Oelen et al., 4 Apr 2025, Arcodia et al., 10 Apr 2026).

1. Terminological range and dynamical role

The literature shows that “secondary periodic orbit” is field-dependent rather than universal. In the forced $1\!:\!2$ internally resonant mechanical systems of "Bifurcation analysis of quasi-periodic orbits of mechanical systems with 1:2 internal resonance via spectral submanifolds" (Liang et al., 2024), a secondary Hopf bifurcation is a bifurcation of a periodic orbit of the forced system, not of an equilibrium, and the resulting object in the full system is a 2D invariant torus. In that setting, the paper’s “secondary periodic orbit” language refers to the periodic orbit of the $4$D reduced-order model whose lift is quasi-periodic motion in the full system. In "Periodic orbits near onset of chaos in plane Couette flow" (Kreilos et al., 2012), secondary periodic orbits are periodic states born from secondary bifurcations of the upper-branch coherent equilibrium. In "A Family of Periodic Orbits in the Three-Dimensional Lunar Problem" (Belbruno et al., 2018), the term is used in the broad celestial-mechanics sense of a family continued from a simple limit family rather than a primary Keplerian one. By contrast, the galactic-potential and irregular-body papers do not formally define the term, but they study resonant or bifurcated periodic-orbit families that play the same structural role (Caranicolas et al., 2012, Jiang et al., 2016).

Context	Meaning of “secondary”	Representative paper
Forced internal resonance	Torus-producing orbit born from a periodic orbit via secondary Hopf	(Liang et al., 2024)
Transitional shear flow	Periodic states from secondary bifurcations of coherent equilibria	(Kreilos et al., 2012)
Celestial mechanics	Continued family beyond a primary/simple limit family	(Belbruno et al., 2018)
Irregular-body dynamics	Additional branches/families generated during continuation and bifurcation	(Jiang et al., 2016)
Symmetric billiards	Non-Birkhoff periodic branches beyond ordered Birkhoff orbits	(Oelen et al., 4 Apr 2025)
Strong-field timing	Secondary orbital clock or super-period modulating a primary recurrence	(Arcodia et al., 10 Apr 2026)

A recurrent feature is that these objects are identified not merely by closure in time, but by their relation to another invariant structure: a resonance, an equilibrium branch, a parent periodic orbit, a symmetry class, or an underlying orbital clock. This suggests that “secondary” is best understood relationally rather than absolutely.

2. Resonance-supported families in smooth Hamiltonian potentials

In "A semi-numerical method for periodic orbits in a bisymmetrical potential" (Caranicolas et al., 2012), the underlying model is the two-dimensional harmonic oscillator plus a Plummer term,

$$V(x,y)=\frac{1}{2}\left(\omega_1^2 x^2+\omega_2^2 y^2\right)-\frac{\mu}{\sqrt{x^2+y^2+\alpha^2}},$$

with

$\mu=0.001,\qquad \alpha=0.25,$

and resonances

$$\omega_1:\omega_2 = 1:1,\quad 2:1,\quad 2:3,\quad 4:3.$$

The paper studies periodic orbits intersecting the $x$-axis perpendicularly and periodic orbits going through the origin. It does not explicitly classify them as primary or secondary, but it explicitly notes that these resonant families are exactly the kinds of orbit families that would commonly be regarded as secondary or family orbits associated with a resonance.

The method is semi-numerical in the strict sense that it combines empirical formulas, harmonic-oscillator intuition, and calibration against direct numerical integration. The full equations are integrated with a Bulirsh–Stöer routine in double precision, with energy conserved to about the $12$th significant figure. The aim is not exact closed-form integration except in the symmetric $1\!:\!1$ case, but accurate formulas for starting positions and periods.

For periodic orbits starting perpendicularly from the $x$-axis, the $1\!:\!1$ resonance is analytically tractable because axial symmetry yields a circular periodic orbit with exact period

$4$0

The $4$1 case is less accurate in position than in period: the semi-numerical starting points differ from direct numerical integration by up to about $4$2, whereas the periods differ by less than $4$3. In the $4$4 and $4$5 resonances, the agreement is described as very good. For origin-crossing families, exact straight-line periodic solutions exist in the symmetric $4$6 case; good semi-numerical estimates are obtained for $4$7 and $4$8; and the authors do not find periodic orbits through the origin in the $4$9 case, even when energy is increased well beyond the tabulated range.

The significance of this study is methodological as well as dynamical. It presents secondary-like resonant families through their geometric character—perpendicular intersection with the $$V(x,y)=\frac{1}{2}\left(\omega_1^2 x^2+\omega_2^2 y^2\right)-\frac{\mu}{\sqrt{x^2+y^2+\alpha^2}},$$0-axis or passage through the origin—and through the resonance ratio constraining the initial data by the energy integral. In this formulation, secondary periodic orbits are not introduced as an abstract class; they are the concrete resonance-supported backbone orbits of the model.

3. Secondary bifurcation, invariant tori, and routes to chaos

In forced internally resonant mechanical systems, secondary periodic orbits appear through a precise codimension-one mechanism. The full system considered in (Liang et al., 2024) is

$$V(x,y)=\frac{1}{2}\left(\omega_1^2 x^2+\omega_2^2 y^2\right)-\frac{\mu}{\sqrt{x^2+y^2+\alpha^2}},$$1

with $$V(x,y)=\frac{1}{2}\left(\omega_1^2 x^2+\omega_2^2 y^2\right)-\frac{\mu}{\sqrt{x^2+y^2+\alpha^2}},$$2 internal resonance

$$V(x,y)=\frac{1}{2}\left(\omega_1^2 x^2+\omega_2^2 y^2\right)-\frac{\mu}{\sqrt{x^2+y^2+\alpha^2}},$$3

and external resonance

$$V(x,y)=\frac{1}{2}\left(\omega_1^2 x^2+\omega_2^2 y^2\right)-\frac{\mu}{\sqrt{x^2+y^2+\alpha^2}},$$4

Reduction on a spectral submanifold $$V(x,y)=\frac{1}{2}\left(\omega_1^2 x^2+\omega_2^2 y^2\right)-\frac{\mu}{\sqrt{x^2+y^2+\alpha^2}},$$5 yields a $$V(x,y)=\frac{1}{2}\left(\omega_1^2 x^2+\omega_2^2 y^2\right)-\frac{\mu}{\sqrt{x^2+y^2+\alpha^2}},$$6D reduced-order model. The paper emphasizes a structural equivalence: a fixed point of the reduced vector field corresponds to a periodic orbit of the full system, whereas a limit cycle of the reduced model corresponds to a 2D invariant torus of the full system. Thus, the secondary periodic orbit is periodic only in the reduced system; in the full system it is quasi-periodic.

Within this framework, the forced response curve of periodic solutions is typically M-shaped and undergoes secondary Hopf bifurcations, also called Neimark–Sacker bifurcations. When the forcing amplitude $$V(x,y)=\frac{1}{2}\left(\omega_1^2 x^2+\omega_2^2 y^2\right)-\frac{\mu}{\sqrt{x^2+y^2+\alpha^2}},$$7 exceeds a critical value, the periodic response loses stability and a family of quasi-periodic responses appears. The paper continues these torus branches by continuing limit cycles of the reduced model and detects local bifurcations—period-doubling and saddle-node bifurcations of quasi-periodic orbits—as well as global events including homoclinic bifurcations, isolas, simple bifurcations, and cusp bifurcations. In the benchmark two-coupled-oscillator example, the quasi-periodic branch terminates in a homoclinic bifurcation at $$V(x,y)=\frac{1}{2}\left(\omega_1^2 x^2+\omega_2^2 y^2\right)-\frac{\mu}{\sqrt{x^2+y^2+\alpha^2}},$$8. In the shallow curved beam, the authors report 20 PD and 20 SN points on the FRC-QO at $$V(x,y)=\frac{1}{2}\left(\omega_1^2 x^2+\omega_2^2 y^2\right)-\frac{\mu}{\sqrt{x^2+y^2+\alpha^2}},$$9. In the high-dimensional shallow shell example, the finite-element model has more than $\mu=0.001,\qquad \alpha=0.25,$0 DOFs and is reduced to a $\mu=0.001,\qquad \alpha=0.25,$1D ROM.

A different but related picture appears in plane Couette flow (Kreilos et al., 2012). There, the primary coherent structure is the Nagata–Busse–Clever state created in a saddle-node at

$\mu=0.001,\qquad \alpha=0.25,$2

The upper branch, initially stable in the symmetry-reduced subspace, loses stability in a supercritical Hopf bifurcation at

$\mu=0.001,\qquad \alpha=0.25,$3

with limit-cycle amplitude scaling like

$\mu=0.001,\qquad \alpha=0.25,$4

That first Hopf-born cycle is the first secondary periodic orbit of the flow. It then undergoes period doubling at

$\mu=0.001,\qquad \alpha=0.25,$5

followed by an incomplete doubling cascade, a stable period-3 window from

$\mu=0.001,\qquad \alpha=0.25,$6

and a boundary crisis at

$\mu=0.001,\qquad \alpha=0.25,$7

The paper identifies seven periodic orbits of symbolic period at most five,

$\mu=0.001,\qquad \alpha=0.25,$8

and states that all of these periodic orbits have exactly one unstable eigenvalue. Their symbolic organization is “almost the same” as the universal Metropolis–Stein–Stein sequence for unimodal maps.

Taken together, these studies show that secondary periodic orbits are often the dynamically relevant descendants of an initial periodic response. They can be born from torus bifurcations, period doublings, or saddle-node events, and they can organize transitions from regular motion to chaotic or transiently chaotic dynamics.

4. Continuation, symmetry, and family structure in celestial mechanics

In spatial three-body dynamics, secondary periodic orbits often appear as continued families from singular or simplified limit problems. The spatial lunar problem studied in (Belbruno et al., 2018) starts from the Hamiltonian of the three-dimensional circular restricted three-body problem in a rotating frame and then applies the standard lunar scaling

$\mu=0.001,\qquad \alpha=0.25,$9

At $$\omega_1:\omega_2 = 1:1,\quad 2:1,\quad 2:3,\quad 4:3.$$0, corresponding to Hill’s lunar problem, the $$\omega_1:\omega_2 = 1:1,\quad 2:1,\quad 2:3,\quad 4:3.$$1-axis is invariant, and one obtains a family of consecutive collision orbits on that axis. These orbits start at collision with the small primary, move out along the positive $$\omega_1:\omega_2 = 1:1,\quad 2:1,\quad 2:3,\quad 4:3.$$2-axis, return to collision, and repeat. Their turning point $$\omega_1:\omega_2 = 1:1,\quad 2:1,\quad 2:3,\quad 4:3.$$3 satisfies

$$\omega_1:\omega_2 = 1:1,\quad 2:1,\quad 2:3,\quad 4:3.$$4

and the period is

$$\omega_1:\omega_2 = 1:1,\quad 2:1,\quad 2:3,\quad 4:3.$$5

with the key estimate

$$\omega_1:\omega_2 = 1:1,\quad 2:1,\quad 2:3,\quad 4:3.$$6

The main theorem asserts that, for $$\omega_1:\omega_2 = 1:1,\quad 2:1,\quad 2:3,\quad 4:3.$$7 sufficiently small and for all but two energies where a non-degeneracy condition fails, there exists a unique periodic orbit

$$\omega_1:\omega_2 = 1:1,\quad 2:1,\quad 2:3,\quad 4:3.$$8

continuing the collision orbit $$\omega_1:\omega_2 = 1:1,\quad 2:1,\quad 2:3,\quad 4:3.$$9, with $x$0 as $x$1. The family is denoted $x$2 and is symmetric with respect to the $x$3-plane. This is a canonical secondary family: it is not a primary Keplerian orbit, but a persistent periodic orbit obtained by continuation from a singular limit family.

The same paper also gives numerical information on the family’s stability and geometry. In the rescaled lunar problem, the polar orbit is elliptic for

$x$4

and complex hyperbolic for

$x$5

It undergoes a period-doubling bifurcation, two simple degeneracies, and a period-halving bifurcation. The reported intervals are

$x$6

for period doubling and halving,

$x$7

for simple degeneracies, and

$x$8

for eigenvalue collision.

The oblate-secondary variant of the spatial Hill problem in (Xu, 2020) provides a different existence result. After passing to Poincaré-Delaunay variables

$x$9

$12$0

and removing the short-period terms by two successive Lie-transform averagings, the paper proves the existence of a class of doubly-symmetric, near-circular periodic solutions around the oblate primary. The system is reversible with respect to two anti-symplectic involutions $12$1 and $12$2, and any orbit intersecting the corresponding Lagrangian planes $12$3 and $12$4 in the prescribed way is periodic with period $12$5. The paper explicitly states that these are spatial analogues of the doubly symmetric secondary periodic orbits known in the spherical Hill problem. It also notes that stability and global bifurcations remain open.

These celestial-mechanics examples make clear that secondary periodic orbits are frequently family-level objects. Their central features are persistence under perturbation, symmetry constraints, and the capacity to bridge singular limit dynamics and the full problem.

5. Relativistic geodesics and hierarchical orbital clocks

In strong-field gravitation, secondary periodic-orbit structure appears both geometrically and observationally. For timelike geodesics around a Schwarzschild-like black hole immersed in a King-type dark matter halo (Sharipov et al., 13 Nov 2025), the motion is studied in a static spherically symmetric spacetime with effective potential

$12$6

and the periodic orbits are organized by the zoom-whirl triplet $12$7. The rational classification parameter is

$12$8

The authors emphasize that every periodic orbit can be labeled uniquely by $12$9. The specific classes displayed are

$1\!:\!1$0

In the paper’s interpretation, the “secondary” periodic orbits are the higher members of this hierarchy beyond the simplest class, such as

$1\!:\!1$1

Increasing the halo parameters $1\!:\!1$2 and $1\!:\!1$3 moves both $1\!:\!1$4 and $1\!:\!1$5 outward, increases $1\!:\!1$6 and $1\!:\!1$7, and decreases $1\!:\!1$8. For periodic orbit classes, the halo causes the required energies to decrease slightly and the required angular momenta to increase. For example, at $1\!:\!1$9, the energy of $x$0 changes from

$x$1

to

$x$2

while $x$3 changes from

$x$4

A complementary use of secondary orbital periodicity appears in the QPE timing analysis of eRO-QPE2 (Arcodia et al., 10 Apr 2026). There the observed eruption times are modeled through an $x$5 analysis with a base period and super-periodic modulations. The recurrence inferred from the XMM1–4 baseline is

$x$6

The best-fit deterministic timing model adds two sinusoidal super-periods. The shorter modulation is measured at

$x$7

with amplitude

$x$8

and the longer modulation settles near

$x$9

with amplitude

$1\!:\!1$0

The short modulation is interpreted as apsidal precession of an eccentric orbit, with

$1\!:\!1$1

The long modulation is inconsistent with EMRI nodal precession; disk precession is allowed only in a limited region of parameter space; and the paper identifies a stable hierarchical triple solution with

$1\!:\!1$2

The authors also report

$1\!:\!1$3

at $1\!:\!1$4, disfavoring high-eccentricity white dwarfs and some IMBH secondaries via GW decay, and they state that the correlated odd/even $1\!:\!1$5 disfavors both disk crossings per orbit being observed.

In this relativistic setting, secondary periodic orbits need not be separate branches in phase space. They can instead be secondary clocks: additional periodicities encoding precession or outer-orbit structure on top of a primary recurrence.

6. Symmetry, topology, and non-primary orbit generation

Around irregular-shaped minor bodies, the language of secondary periodic orbits is most naturally translated into branch generation during continuation. In (Jiang et al., 2016), the dynamics is formulated in the rotating body-fixed frame with gravitational potential obtained from the polyhedron model and effective potential

$1\!:\!1$6

while the Jacobi integral is

$1\!:\!1$7

The paper proves a conserved topological-degree quantity that restricts the number of periodic orbits on a fixed energy surface. Its consequences are that the number of non-degenerate periodic orbits varies in pairs during continuation and is always odd on the fixed energy surface. The numerical procedure fixes a Poincaré section, randomly searches for periodic orbits, uses a tolerance of $1\!:\!1$8, selects roughly $1\!:\!1$9 candidates from about $4$00 computed orbits, computes the monodromy matrix $4$01, and classifies the Floquet multipliers.

The bifurcation structure is encoded by multiplier collisions. The paper states that a period-doubling bifurcation occurs when two multipliers cross $4$02, a tangent bifurcation when two cross $4$03, a Neimark–Sacker bifurcation when two collide on the unit circle away from $4$04, and a real saddle bifurcation when two collide on the real axis away from $4$05. It also stresses an important caveat: collision alone does not necessarily imply a true bifurcation. Around comet 1P/Halley, the authors identify a stable approximately 1:1 resonant family with period ratios in $4$06 undergoing a pseudo Neimark–Sacker bifurcation, where the collision persists without genuine branch splitting. Around asteroid 216 Kleopatra, a single continued family passes through

$4$07

thereby exhibiting two real saddle bifurcations and one period-doubling bifurcation, with orbit periods varying from 2.4257 h to 2.3855 h.

In symmetric convex billiards (Oelen et al., 4 Apr 2025), the corresponding non-primary objects are non-Birkhoff periodic orbits. A periodic billiard sequence is Birkhoff if cyclic order is preserved under time shifts; a non-Birkhoff periodic orbit fails this ordering property. For rational rotation number $4$08 with $4$09, a Birkhoff orbit has minimal period $4$10, whereas a non-Birkhoff orbit with the same rotation number may have minimal period $4$11. The principal existence criterion states that if

$4$12

then the billiard has a non-Birkhoff periodic orbit with minimal period $4$13, winding number $4$14, rotation number $4$15, and spatiotemporal symmetry group $4$16. The paper further states that the lifts of the non-Birkhoff orbit and the reference Birkhoff orbit cross exactly $4$17 times per period $4$18. If

$4$19

then there are infinitely many non-Birkhoff periodic orbits of rotation number $4$20 with arbitrarily long minimal periods. In the $4$21-symmetric case, the condition

$4$22

yields infinitely many non-Birkhoff periodic orbits of rotation number $4$23, with arbitrarily long minimal periods, of types I, II, and V. The paper also proves that any open neighborhood of the circular billiard in the analytic topology contains a billiard with infinitely many non-Birkhoff periodic orbits of any rational rotation number $4$24.

Across irregular-body gravity and symmetric billiards, secondary periodic orbits therefore emerge as topologically constrained and symmetry-selected branches. This suggests a unifying viewpoint: secondary periodic orbits are not defined by a single local normal form, but by their status as additional organized closures beyond a primary ordered family, with their birth and persistence controlled by Floquet topology, continuation, or symmetry-restricted variational principles.