meander — counting closed loops across a line

A meander is a single closed loop crossing a horizontal line at 2n points, never touching itself. Build one from two halves: a non-crossing arch system above the line and another below. Superpose them — sometimes they fuse into one loop (a meander), sometimes they fall apart into several (a meandric system). Click two points on the same row to draw an arch; the rest reflows to stay non-crossing.

order · n (2n crossings)

4

editing

presets · this row

Each row is a perfect matching of the 2n points by non-crossing arcs — there are C_n (a Catalan number) of them per row. The loop alternates: an arc above, an arc below, above, below. It's a meander exactly when that walk visits all 2n points in one cycle.

How many meanders of order n?

There are C_n non-crossing arch systems per row, so C_n² ways to stack a top and a bottom. The meandric number M_n counts just the stacks that form a single loop. The table below is computed live in your browser by brute force — generating every arch system, superposing every pair, and counting loops.

The sequence — 1, 2, 8, 42, 262, 1828, 13820, 110954, … (OEIS A005315) — has no known closed form. There is no formula, no generating function in closed form, no efficient way to compute M_n beyond direct enumeration. It is one of the cleanest unsolved counting problems in combinatorics: a schoolchild can state it, and nobody can solve it.

Notice how fast the meanders thin out against the C_n² pairs: most ways of stacking two arch systems fall into pieces. The growth tab asks how the survivors multiply.

M_n grows like C · R²ⁿ / n^α. The ratio M_n+1/M_n climbs toward R² ≈ 12.26 — slowly. The exponent α is the strange part.

connectivity R3.5018…

R²12.2626…

exponent α(29+√145)/12

α ≈3.4201…

That α is not combinatorics. Di Francesco, Golinelli and Guitter (1997) derived it from a c = −4 conformal field theory coupled to two-dimensional quantum gravity — a model of a fluctuating random surface, gravitationally dressed à la KPZ/Liouville. A child can count the loops; the rate they multiply at is read off a theory of random geometry. R and α are numerical / field-theory predictions, not theorems.

Meanders are the same object as folded strips of stamps. Take the configuration from the explorer, lay its 2n crossings out as a strip, and fold: the arches above and below the line become the creases of a flat-folded stack seen edge-on. Drag the slider to fold and unfold.

fold

0.50

Folding a strip of stamps flat and reading the folded edges end-on gives exactly an arch system on each side. The connectivity of the paper forces a single thread — a semi-meander; gluing both ends closes it into a meander. The stamp-folding counts (OEIS A000136, A000682) are meander relatives, and every bit as formula-less.

What a meander is

Draw an infinite horizontal line. Now draw a single closed curve — a loop — that crosses the line transversally at 2n points and never touches itself anywhere else. That's a closed meander of order n. Count the essentially different ones and you get the meandric numbers M_n: 1, 2, 8, 42, 262, 1828, 13820, 110954, 933458, …

The name is literal: the curve wanders back and forth across the line like the river Maíandros (modern Büyük Menderes) in Anatolia, whose looping course gave Greek its word for a winding path — and gave us the Greek-key border pattern.

The arch-system picture

Cut the plane along the line. Above it, the loop becomes a set of non-crossing arcs pairing up the 2n crossing points; below it, another. Each half is a non-crossing perfect matching of 2n points — and those are counted by the Catalan number C_n = (2n choose n)/(n+1). A meander is a pair (top matching, bottom matching) whose superposition is a single loop. So meanders live inside C_n² stacked pairs, and the entire difficulty is the word single: deciding when two independently-easy objects happen to interlock into one connected curve.

The pieces are Catalan-simple. The connectivity is the abyss.

Why it's hard — and the quantum-gravity exponent

Both the top and bottom matchings have a clean product formula (Catalan numbers). Their pairing does not. There is no known closed form for M_n, no nice generating function, no polynomial-time formula. The best exact tool is what the count tab does: enumerate.

What is known is the asymptotics. M_n ~ C·R²ⁿ/n^α with a connectivity constant R ≈ 3.5018 and a configuration exponent

α = (29 + √145) / 12 ≈ 3.4201.

Di Francesco, Golinelli and Guitter obtained α in 1997 by mapping meanders to a gas of self-avoiding loops on a random surface — a c = −4 conformal field theory coupled to 2D quantum gravity. The irrational exponent with a √145 in it is the gravitational dressing (KPZ) of a conformal weight. It is genuinely startling that a question about doodling a loop over a line should have its growth rate dictated by a theory of fluctuating spacetime.

The determinant that does have a formula

Here is the sting. Arrange all C_n arch systems and form the meander matrix: entry (p, q) is q^{(number of loops when arch systems p and q are
superposed)}. This is the Gram matrix of the Temperley–Lieb algebra's natural bilinear form. Di Francesco proved its determinant factors completely into a product of Chebyshev-type (cyclotomic-flavoured) polynomials in q — an exact, closed, beautiful product. Same objects, opposite epistemics: the number of single-loop meanders is hopeless, yet a natural determinant over the very same arch systems is one of the prettiest closed forms in the subject.

Meanders as factorizations

Label the crossings 1…2n along the line. The top matching and bottom matching are each fixed-point-free involutions of those labels. Following the loop is exactly composing the two involutions, and the configuration is a meander iff that composition is a single 2n-cycle. So counting meanders is counting factorizations of a long cycle into two non-crossing involutions — a question that lands squarely in the combinatorics of the symmetric group, map enumeration, and Hurwitz theory.

Folding stamps

Meanders are in bijection with ways to fold a strip of stamps flat. View a flat-folded strip edge-on: consecutive stamps are joined by creases that arch alternately over and under the stack — an arch system above and below a line. The single connected strip forces a single thread, a semi-meander; closing both ends gives a meander. The stamp-folding sequences (OEIS A000136, A000682) are cousins of M_n and just as resistant to a formula. The fold tab animates the correspondence.

History

1912 — Henri Poincaré, working on the qualitative theory of differential equations and curves on surfaces, runs into the configurations now called meanders.
1960s–70s — Touchard and Lunnon study the equivalent stamp-folding problem; the sequences enter the early tables of combinatorics.
1980s — Arnold and the Russian school connect meanders to the topology of plane curves, Gauss codes, and singularity theory.
1997 — Di Francesco, Golinelli & Guitter give the matrix-model / quantum-gravity analysis: the connectivity constant R and the exponent α = (29+√145)/12. Di Francesco proves the meander determinant product formula.
today — M_n is known only by enumeration into the low double-digit orders. The exact constant R, a closed form, and even whether M_n is "P-recursive" remain open.

Why this sits in the geometry pack

The pack collects extremal and combinatorial problems with a small-integer heart and a long, open climb — the lonely runner, Heilbronn's triangle. Meanders are the purest of the lot: the statement fits on a napkin, the answer is a sequence of integers, and the growth rate had to be borrowed from quantum gravity because no one can derive it by counting. It also wires this pack to a different one — Viazovska's modular forms and meanders' conformal field theory are both moments where physics-grade special structure crashes a discrete problem.

What's in this site

explorer — stack two non-crossing arch systems; the loops are coloured by connected component, and a badge tells you whether you've made a meander (1 loop) or a meandric system (several). Edit either row by clicking two points; presets and a find a meander search are provided.
count — brute-force M_n in your browser and watch it diverge from C_n².
growth — the ratio M_n+1/M_n creeping toward R² ≈ 12.26, and the quantum-gravity exponent.
fold — the stamp-folding bijection, animated on the current configuration.
docs — this page.

sources

Di Francesco, P., Golinelli, O. & Guitter, E. (1997). Meanders and the Temperley–Lieb algebra. Communications in Mathematical Physics 186, 1–59.
Di Francesco, P. (1997). Meander determinants. Communications in Mathematical Physics 191, 543–583.
Lando, S. K. & Zvonkin, A. K. (1993). Plane and projective meanders. Theoretical Computer Science 117, 227–241.
Lunnon, W. F. (1968). A map-folding problem. Mathematics of Computation 22, 193–199.
OEIS A005315 (closed meanders), A005316 (open meanders), A000682 (semi-meanders / stamp foldings).
Wikipedia — Meander (mathematics)