Twenty-two chapters here compute a return map. None has ever asked what its itinerary is. The answer turns out to be a restriction rather than a discovery, and it holds for every member of the family at once.
A return map is what you build when you want to replace a flow by a map: pick a section, follow each point until it comes back, and study the map instead of the differential equation. Strogatz introduces it at §8.7, p. 281. This corpus builds them constantly — WP-122 produced one in closed form — and has never once asked the question the construction exists to answer.
Counted at HEAD with tools/corpus_count.py, entity-aware, with this page and its
script classified out:
| pattern | files | chapters |
|---|---|---|
| limit cycle | 120 | 110 |
| return map | 34 | 22 |
| Poincaré section / map | 15 | 12 |
| symbolic dynamics | 1 | 1 |
| topological entropy | 1 | 1 |
| horseshoe | 0 | 0 |
| shift map | 0 | 0 |
| Sharkovskii | 0 | 0 |
| logistic map | 0 | 0 |
| lap number | 0 | 0 |
| Smale | 2 | 2 |
| Levinson | 6 | 4 |
| Cartwright | 0 | 0 |
The same shape ch-van-der-pol found: the phenomenon is everywhere and the object that reads it is nowhere. What follows is the answer, and it is worth saying at the top that it is negative — and that the negative is sharper than the positive would have been, because it holds for the whole family and not for one system.
Stephen Smale was born in Flint, Michigan, in 1930, took his doctorate at Michigan in 1957
under Raoul Bott, and had the Fields Medal by 1966. The Bott is worth a
pause: his advisor's name is the one attached to Bott periodicity, which is rung 28
— the floor WP-82 says this series
never built. Measured here, the bare pattern /bott/ matches 789
tracked files and every single one of them is the word bottom;
“Bott periodicity” reaches exactly one chapter, and
ch-conley put it there earlier today.
“Raoul Bott” appeared in no file at all before this page. The
gallery's two newest chapters are a student of Moser's and a student of Bott's, and both of
those teachers are gaps.
In 1960 Smale was in Rio de Janeiro, and he had conjectured something close to the opposite of what he is now famous for: that structurally stable systems were, in essence, tame. Norman Levinson wrote to him and pointed him at papers by Mary Cartwright and J. E. Littlewood on a second-order equation from radio engineering — the forced van der Pol oscillator. Levinson's own 1949 paper had found, in that equation, infinitely many periodic solutions where Smale's conjecture said there should be finitely many. Smale did not argue with the analysis. In his own account he says he recast it: “So I put what I did in a very geometric context so I could understand it.” The geometric context was a rectangle stretched, folded, and laid back across itself — the horseshoe.
For ṙ = r(1−r²), θ̇ = 1, one turn is exactly t = 2π, so the return map to any ray is the time-2π flow map. Block [1] gives it in closed form and checks it against RK4 at eight radii, worst disagreement 8×10−15:
P′ is a positive quantity divided by a positive quantity. It is never zero and never negative, at any radius — checked on a grid of 2001 points in block [2]. So P is strictly increasing, and a strictly increasing map has nothing to fold.
The lap number ℓ(g) counts maximal monotone intervals; Misiurewicz and Szlenk give the topological entropy of a piecewise-monotone interval map as lim (1/n) log ℓ(gn). Block [3] computes it for P, and prints something it would have been easy to leave out:
The n ≥ 6 rows are not evidence. P contracts by e−4π per turn, so by the sixth iterate the whole of (0, 3] has landed on the single double-precision number nearest 1.0, and a lap count of 1 is then a true statement about a constant function rather than about this one. Four rows carry information and three are an artefact of the arithmetic. The script says so and asserts only on the four. The reason the entropy is zero is the next section, not this table.
The same instrument on a map that does fold, so that a reading of 0 can be believed: the logistic map at μ = 4 gives ℓ(Ln) = 2n exactly, and (1/n) log ℓ = log 2 = 0.693147180559945 at every n tested. “Logistic map” is another zero in the table above — the most famous single cell of Strogatz's Figure 1.3.1, and this page is the first to write it down here.
Let ṙ = f(r), θ̇ = ω > 0 on the punctured plane, with f smooth. The return map P to a ray is the time-T flow map of a scalar autonomous ODE, T = 2π/ω. Differentiating the flow in its initial condition gives v′ = f′(r(s)) v, v(0) = 1, so
dP/dr0 = exp( ∫0T f′(r(s)) ds )
which is an exponential, hence strictly positive, for every f and every T. So P is a strictly increasing homeomorphism, ℓ(Pn) = 1 for all n, and h(P) = 0. Orbits of a scalar autonomous equation cannot cross, and a fold is a crossing. ∎
Block [5] checks the identity on four fields, two of them deliberately not monotone — Example 7.1.1's, WP-120's frozen field at a = 5.4366, sin(3r) − 0.3, and r − r⁵ — and sweeps T, which turns out to be the methodological part:
At T = 0.2 the identity verifies against a central difference to about 10−8. By T = 2π the derivative for WP-120's field is e−110, the two perturbed flows land on the same double, and the difference quotient reports nonsense — only one of thirty sample points is even comparable. The integral still returns the number. A derivative that small is computed, not measured, and a script that had only differenced would have reported a fold where there is none.
This is now the third property of the corpus's closure family traced to one source, the radial speed being a function of r alone:
| restriction | consequence | where |
|---|---|---|
| circle-preserving | every admissible closed orbit is a circle, so van der Pol's is in no closure | ch-van-der-pol |
| uniqueness route tied to it | the Dulac function g = 1/r³ does not survive the move off circles | WP-120 |
| entropy zero | the return map is monotone, for every f; no horseshoe is available | here |
The corpus's operator chain is G = U ∘ F ∘ K ∘ C, and F is the Fold. The published return map cannot fold, and by Part V no member of the family can. Those two statements are not in contradiction — F acts on the operator chain, not on a return map — but nothing in the corpus has ever said which of the two senses of “fold” it means where, and a reader is entitled to ask. Stated as a question, not as a defect.
Poincaré–Bendixson forbids chaos in the plane outright, so Part V's restriction is not the binding one for a planar system; what it adds is which feature is responsible, which is the part that transfers. Escaping needs three dimensions, and the corpus has them — the contact manifold of Volume II. But there the obstruction is the one ch-conley found from the other side: on Γ, ż ≡ 1. The height climbs at unit rate forever, no orbit returns to a section, and a return map to that section does not exist to have an itinerary. The two chapters meet on the same fact from opposite directions — one looking for a compact invariant set and one looking for recurrence, both stopped by the same identically-equal-to-one.
Which is also the honest place to say what would change it. Smale's horseshoe came from a forced oscillator: a periodic drive, making the system non-autonomous, which is exactly the hypothesis Part V's one-line proof needs and does not have. The corpus's modulation e−z is not periodic; it is monotone. A periodic modulation in place of a monotone one is the smallest change to this corpus's own equations that could produce what Levinson found, and it has not been tried here.
| Smale 1998 | S. Smale, “Finding a Horseshoe on the Beaches of Rio”, The Mathematical Intelligencer 20, no. 1. |
| Smale 1967 | S. Smale, “Differentiable dynamical systems”, Bulletin of the American Mathematical Society 73. |
| Levinson 1949 | N. Levinson, “A second order differential equation with singular solutions”, Annals of Mathematics 50. |
| Cartwright–Littlewood | M. L. Cartwright and J. E. Littlewood, on non-linear differential equations of the second order, 1945. For the history and its reception see J. Guckenheimer, “The legacy of the Cartwright–Littlewood collaboration”, J. London Math. Soc., 2026. |
| Misiurewicz–Szlenk | The lap-number formula for the entropy of a piecewise-monotone interval map, used in block [3] and block [4]. |
| Strogatz | S. H. Strogatz, Nonlinear Dynamics and Chaos, 2nd ed. §8.7 pp. 281–282 (Poincaré maps); §10.5; Figure 1.3.1 p. 10. |
| in-corpus | WP-122 · WP-120 · ch-conley · ch-van-der-pol · WP-82 |
| verification | book7/ch-smale-verify.py — seven blocks, standard library only, entity-aware counts via tools/corpus_count.py. Every number on this page is printed by it. |
No priority is claimed. Smale's horseshoe, the Misiurewicz–Szlenk lap-number formula, and the monotonicity of scalar autonomous flow maps are all classical and are used as such. What is new to this corpus is that its own published return map has been put through them, and that the resulting restriction was stated for the family rather than for one system.