Poincaré maps and Floquet multipliers
Reduce periodic-orbit stability to a transverse return map, distinguish multipliers from exponents, and account for the neutral flow direction.
On this page
The question for today
What you’ll understand
- Define a local return map on a transverse section through a periodic orbit.
- Connect the monodromy spectrum with transverse orbital stability.
- Compute a radial Floquet multiplier from a period and a decay exponent.
A periodic orbit never attracts every nearby point to the same clock position: a small phase shift remains a phase shift. What can attract is the orbit as a geometric set. A transverse section removes that neutral direction and turns repeated revolutions into a discrete map near a fixed point. Floquet theory analyzes the linearized perturbation over one period. Together, these tools distinguish attraction toward a cycle from motion around it and make stability measurable one return at a time.
Take a section where the flow crosses
Let a smooth autonomous vector field have a nonconstant periodic solution gamma of minimal period T. At a point on it, the velocity is nonzero. Choose a codimension-one local section whose normal is not perpendicular to that velocity. Transversality lets the implicit function theorem determine a smooth nearby return time and return map for points sufficiently close to the periodic point. This is a local construction near a known returning orbit, not a claim that every trajectory in the whole state space returns.
The periodic orbit intersects this chosen local section at a fixed point of the return map. If all derivative eigenvalues there have modulus strictly less than one, that fixed point attracts and the orbit is locally orbitally asymptotically stable. A multiplier outside the unit circle gives instability. Unit-modulus multipliers require nonlinear analysis, just as zero-real-part eigenvalues do for equilibria. A chosen section can have multiple global intersections per revolution; define the intended local return and orientation before inferring a period from an intersection count.
Propagate the variational equation over one period
Along gamma, infinitesimal perturbations satisfy a linear equation with periodic coefficient matrix A(t), the Jacobian evaluated on the orbit. Let X solve X dot = A(t)X with X(0) equal to the identity. The matrix X(T) is the monodromy matrix; its eigenvalues are Floquet multipliers. Their moduli measure growth or decay per revolution. Floquet exponents describe growth per unit time and exponentiate over T to the multipliers, with logarithm branches relevant for complex values.
An autonomous periodic orbit always has a unit multiplier in the tangent flow direction, because a time-shifted trajectory remains a solution. The transverse return map removes that direction; its spectrum equals the remaining multipliers, counting multiplicity. For an attracting hyperbolic cycle the tangent unit multiplier is simple and every other multiplier lies strictly inside the unit circle. Requiring every full monodromy multiplier to have modulus less than one would incorrectly exclude every autonomous cycle, including stable ones.
Separate phase from transverse decay
For the polar model r dot = mu r−r cubed and theta dot = omega, with positive mu and omega, the cycle radius is square root of mu. Linear radial perturbations have exponent minus twice mu. A section at a fixed angle returns after T=2pi divided by omega, so its scalar multiplier is exp of minus twice mu times T. It is positive and less than one. A phase perturbation along the cycle, however, neither grows nor decays in the constant-frequency model.
An oriented planar-flow return map has positive scalar derivative: nearby ordered trajectories cannot cross and reverse that order during one return. Higher-dimensional sections can have more varied multiplier spectra, including negative multipliers associated with flip phenomena. Estimating return derivatives from a coarse event detector can mix transverse error with crossing-time error. Integrate the variational equation or refine the return event when quantitative stability matters. Finite-time contraction seen in a few returns also does not replace checking that the cycle and section satisfy the local theorem's assumptions.
Work it through
Convert a radial exponent to a return multiplier
For r dot = (1/4)r−r³ and theta dot = 2, find the cycle radius, minimal period, transverse multiplier, and full autonomous neutral multiplier.
The nonzero stationary radius is one half. Its radial linear derivative is 1/4−3/4=−1/2, giving radial exponent minus one half per unit time.
Angular speed two gives minimal period T=π. One return therefore multiplies a sufficiently small radial error to first order by exp(−π/2), approximately 0.20788.
The phase direction contributes the full monodromy multiplier one. The section eliminates it, leaving only the contracting scalar multiplier. Thus the orbit attracts as a set even though an initial phase shift need not disappear.
The cycle has radius 1/2 and transverse multiplier about 0.208 per revolution. The unavoidable tangent multiplier one does not contradict orbital attraction.
The living notebook
Explore the model
Interactive lab · optimization
Read contraction from multiplier magnitude
Pause and reset, then set both controls to step size η = 0.6 and initial x₀ = 1. Predict multiplier 1−η = 0.4 and successive iterates 1, 0.4, 0.16. For a fresh comparison, pause and reset, then set η = 1.6 and x₀ = 1; predict multiplier −0.6 and alternating decay. This exact scalar map illustrates the unit-circle stability test. It is not the lesson's return map; in particular its negative-multiplier branch is not an oriented planar-flow return, whose scalar multiplier is positive.
- Position x0
- 2
- Objective x²/2
- 2
- Iteration behavior
- Stable step
Gradient descent on f(x) = x²/2: xₙ = (1 − η)ⁿx₀. Playback advances two discrete iterations per second. Values outside |x| ≤ 4 are explicitly reported. Parameter changes pause playback and recompute the model at the current elapsed time. Reset restores both parameters and returns time to zero. Playback starts paused. Adjust either parameter, then choose Play or Step.
Look forFor discrete return dynamics, multiplier magnitude controls linear attraction; sign also records orientation and must respect the underlying flow.
Pause & reason
Check your understanding
Your turn
Try it yourself
For r dot = r−r³ and theta dot = 1, calculate the cycle period and its transverse multiplier. If a small initial radial error is 0.01, give the linear prediction after one return and state its limitation.
Take a moment to make a prediction or work it out on paper. Open a hint when you need one.
Give me a hint
Differentiate the radial field at radius one and exponentiate over one full revolution.
Compare with the solution
The cycle radius is one, period is 2π, and radial exponent is 1−3=−2. The transverse multiplier is exp(−4π), approximately 0.00000348734. Linearization predicts an error about 0.0000000348734 after one return from an initial error 0.01. This is the first-order perturbation prediction, not an exact finite-error formula for the nonlinear radial equation. The full autonomous monodromy also has a neutral multiplier one for phase, while the section retains only the radial contraction.
Keep the idea
Take it with you
- A transverse return map reduces periodic-orbit stability by one neutral flow direction.
- Floquet multipliers measure one-period growth; exponents measure growth per unit time.
- Orbital attraction concerns transverse contraction and permits persistent phase shifts.
Read further
Original sources and deeper explanations to follow at your own pace.
Optional progress, saved in this browser when available. Revisit any lesson at any time.
Lesson not marked complete.