Limit Cycles and Poincare-Bendixson

A loop that nothing else can reach

So far in this rung every long-term destination has been a resting point. Linearization told you whether a nearby trajectory spirals in to an equilibrium or runs away from it, and Lyapunov functions proved stability even where the Jacobian sat on the borderline. But a great many real systems never settle to a point at all. A heartbeat, a firing neuron, a clock circuit, a chemical reaction that flashes between colours — these run in a sustained, self-renewing oscillation. In the phase plane that steady rhythm shows up not as a still point but as a closed loop the system traces over and over.

Not every loop is special. A frictionless pendulum is filled with closed orbits, one through every starting amplitude — a whole nested family, none of them isolated, each just a neighbour of the next. A limit cycle is the rarer, sturdier kind: an *isolated* closed orbit, a single loop with no other closed orbit immediately beside it. Because it stands alone, nearby trajectories cannot also be loops — they have nowhere to go but to spiral *toward* the cycle (a stable limit cycle) or *away* from it (an unstable one). That isolation is the whole point: a stable limit cycle is an attractor with a definite shape and a definite period, and the system locks onto it no matter where nearby it starts.

The trapping-region idea

How could you ever prove a loop exists when you cannot solve for x(t)? Here is the beautiful, almost topological argument behind the Poincare-Bendixson theorem. Suppose you can find a region of the plane — picture an annulus, a ring-shaped band — that the flow can enter but never leave: on its whole boundary the velocity vector field points strictly inward. Once a trajectory steps into such a trapping region, it is caught there for all future time. The question becomes: with infinite time but only finite room, where can it possibly end up?

In the plane the answer is sharply limited. A trapped trajectory cannot wander forever without repeating, because two-dimensionality leaves it no room to cross itself or escape — the trajectory would fence itself in. If the trapping region contains no equilibrium for the trajectory to fall into, the only remaining destiny is to approach a closed orbit. The wandering path must wind onto a loop. That loop is a limit cycle, and you have proved its existence without ever writing down a solution.

1. Build a trapping region: a closed, bounded ring in the phase plane that the flow enters but cannot leave, with the vector field pointing inward all along its border.
2. Make sure the ring encloses no equilibrium — usually by punching a small hole around the rest point so it lies outside the band.
3. Invoke Poincare-Bendixson: a trajectory trapped in a planar region with no equilibrium must approach a closed orbit.
4. Conclude that at least one limit cycle lives inside the ring — its existence is now proven, even though its exact shape stays unknown.

What the theorem promises — and what it doesn't

Be honest about the fine print, because every clause earns its keep. First, the plane is essential. Poincare-Bendixson is a strictly two-dimensional theorem: it rests on the fact that a curve in the plane separates inside from outside, so a trajectory cannot tangle without crossing itself. Add a third dimension and that fence vanishes — a path can loop over and under itself indefinitely without ever closing. That extra freedom is exactly where chaos becomes possible: the Lorenz system in three dimensions wanders forever on a strange attractor, never settling to a point and never closing into a cycle, something a planar autonomous system is simply forbidden to do.

Second, the system must be autonomous and smooth — no explicit time-dependent forcing, or the no-self-crossing argument breaks. Third, the theorem proves a closed orbit *exists*; it does not hand you its shape, its period, or even tell you the cycle is stable rather than unstable, nor how many cycles the ring holds. And it never gives you a formula. Like everything qualitative in this subject, it answers *whether* and *roughly where*, trading the unsolvable demand for an exact x(t) for a true statement about long-term behaviour. To pin down the loop's actual shape and period you still fall back on the numerical methods from earlier rungs.

Ruling loops out: Bendixson's criterion

There is a companion result that does the opposite job, and it is often easier to apply. Suppose you want to know there is *no* limit cycle in some region. The Bendixson negative criterion looks at the divergence of the field, df/dx + dg/dy. If that quantity keeps a single sign — always positive, or always negative — throughout a simply-connected region (one with no holes), then no closed orbit can fit inside it. The reasoning is a clean application of the divergence theorem: a closed orbit would enclose an area whose net inflow-minus-outflow must cancel to zero around the loop, which a never-zero divergence makes impossible.

  system:   x' = f(x, y) ,   y' = g(x, y)

  divergence of the field:   D(x, y) = df/dx + dg/dy

  Bendixson:  if D keeps one sign (and is never 0) on a
              hole-free region R  =>  NO closed orbit lies in R

Van der Pol: a loop you can almost see forming

The cleanest concrete example — the one the final guide in this rung studies in full — is the van der Pol oscillator, x'' - mu(1 - x^2) x' + x = 0, born from early vacuum-tube radio circuits. Read the damping term mu(1 - x^2): when the swing is small, x^2 < 1, so the bracket is positive and the term *pumps energy in*, pushing the oscillation to grow. When the swing is large, x^2 > 1, the bracket flips negative, the term *drains energy*, and the oscillation is reined back. Small motions are amplified, large motions are damped — and the system gets squeezed from both sides onto a single self-sustaining loop in between.

That squeezing is precisely a trapping region in action: trajectories from the inside grow outward, trajectories from the outside shrink inward, so the flow is funnelled into a ring that holds no stable rest point. Poincare-Bendixson then guarantees a limit cycle inside — and here it happens to be a unique, stable one. Every trajectory, whatever its starting amplitude, winds onto the *same* loop with the *same* eventual period. That is the signature of a true van der Pol oscillation, and the reason such circuits make dependable clocks: the rhythm is a property of the system itself, not of how you started it.

Step back and see how this rung now hangs together. Linearization and the Jacobian read off the *local* fate at each equilibrium; Hartman-Grobman said when that local reading is trustworthy; Lyapunov functions proved stability where the linear picture went silent. Poincare-Bendixson is the last piece — the *global* statement that, in the plane, once trajectories are trapped away from any rest point, an oscillation must form. Together they let you sketch a faithful phase portrait of a system you can never solve, equilibria and loops alike. The next and final guide takes the pendulum and van der Pol and walks every one of these tools through them in full.