Found by people not folding paper

The shortest Möbius band that folds flat is √3 times as long as it is wide, and the number falls out of composing three reflections and solving two linear equations.

The same number has a substantial literature in a different subject, where it took a great deal longer to establish and where nobody was folding anything.

Two objects #

The distinction matters and it is easy to lose.

A folded Möbius band is flat everywhere except on finitely many creases. It is made by creasing a strip and joining the ends, it can be pressed onto a table, and it is the object this collection computes.

A smooth Möbius band is curved everywhere and has no creases at all. It is made by bending a strip without folding it, it is a developable surface embedded in space, and it is the object of the differential-geometry literature.

Those are different objects. They are not two descriptions of one thing and the second is not the limit of the first in any sense that has been established here.

Two routes #

The folded number comes from an equation and a fitting condition.

The crease angles have to satisfy an alternating sum. The positions come out of a two-by-two linear solve. And the length is where the solved creases stop fitting between the ends of the strip.

The smooth number comes from a genuinely hard analysis of developable surfaces, and it was open for a long time. It is not an equation with a threshold; it is an inequality about a family of embeddings, and proving it is a different order of work.

So the same number, two subjects, two methods, and one of them is arithmetic while the other is a theorem.

Neither cites the other #

That is the historical observation and it is the essay’s point.

The origami literature contains the folded triangle as a construction — a strip folded three times at sixty degrees, standard, unattributed, the sort of thing that appears in a book of paper models without a source. It does not contain the bound as a result, because nobody had a reason to ask for the minimum.

The differential-geometry literature contains the bound as a theorem and does not discuss folded bands, because a folded band is not a smooth surface and is outside the question being asked.

Two literatures with the same number in them and no traffic between them, which is the shape this collection has found several times before.

What this collection can claim #

Narrowly, and it is worth being explicit because the temptation to claim more is strong.

Computed here: that a three-crease Möbius band folds only when its crease angles satisfy an alternating-sum condition; that the shortest strip admitting the equilateral solution is 1.7320508 widths; that the next-best admissible triple needs 1.9225; and that no square-creased band closes at any count.

Not established here: any relationship between the folded problem and the smooth one. That the two have the same bound is reported. Why they do is not, and the plausible explanation — that a folded band is a degenerate limit of smooth ones — is an assertion rather than an argument.

That distinction is the sort of thing that gets lost when a number turns up in two places, and losing it produces a claim of the form this collection has proved the differential-geometry result, which it has not and could not.

The vocabulary problem, in detail #

The two literatures do not share a word and it is worth listing the pairs, because the mismatch is total.

What origami calls a crease pattern, differential geometry calls nothing — a developable surface has rulings rather than creases, and a crease is a degenerate ruling where the surface is not smooth.

What origami calls flat-foldable, differential geometry has no term for, because the objects it studies are embedded in space and being flat is not a property it asks about.

What differential geometry calls a developable surface, origami calls a sheet of paper, and treats as the ambient assumption rather than as a definition.

What differential geometry calls the strip’s aspect ratio, origami calls the paper proportion, which it usually means as a shape rather than as a bound.

Four terms, four mismatches, and a search for any of them in the other subject’s index returns nothing. That is a fair mechanical explanation for two literatures not meeting, and it is more convincing than the usual one about disciplinary boundaries.

What would settle the relationship #

For anybody wanting the connection established rather than reported, the shape of the argument is clear enough to describe.

A folded band is a limit of smooth ones: take a smooth band, concentrate its curvature into three narrow regions, and let the regions narrow. In the limit the surface is flat except on three lines, which is the folded band.

If that limit is continuous in the aspect ratio — if a smooth band of ratio $r$ exists whenever a folded one does, and vice versa in the limit — the two bounds coincide and the coincidence is explained.

Establishing it means showing that the concentration can be done without lengthening the strip, and that the limit of admissible smooth bands is admissible. Neither is obviously true and neither is obviously false, and both are analysis rather than arithmetic.

Nothing here attempts it. What is worth saying is that the connection is the kind of thing that could be established, so the coincidence is not mysterious — merely unexplained.

Why the folded case was never asked #

The most likely reason nobody computed the folded bound is that nobody needed a bound.

A folder making a Möbius triangle takes a strip that is obviously long enough and folds it. If it is too short, they cut a longer one. The question what is the minimum arises only when somebody is optimising, and nobody optimises a paper curiosity.

The differential-geometry question arose because the smooth case is a natural question about developable surfaces, and what is the shortest is the standard form such questions take.

So the two subjects asked about the same object and only one of them had a reason to ask for a minimum, which is a sufficient explanation for the asymmetry in what each knows.

It also explains why the folded bound is easy and unnoticed rather than hard and unnoticed: it was available to anybody who wanted it and nobody wanted it.

Why the folded case is easy and the smooth one is not #

The asymmetry is worth explaining, since it is unusual for the discrete version of a problem to be so much simpler.

The folded band has finitely many pieces of data: three angles and three positions. The condition on them is a composition of six reflections’ worth of matrix arithmetic, which is linear in the positions and closed-form in the angles.

The smooth band has a function as its unknown: the surface’s shape, subject to being developable and to closing up with a flip. Conditions on a function are an infinite-dimensional problem, and infinite-dimensional problems are hard.

So the discrete version is not an approximation to the smooth one that happens to be easier. It is a different problem with finitely many unknowns, and having a finite problem is why the arithmetic is short.

The fourth instance #

This collection keeps meeting the same historical shape and it is worth naming again.

The same vertex was found four times in four traditions. The cure was named first — a technique published before the problem it solves. A test was imported without its hypothesis from a literature where the sheet is a disc.

Each is a case of two bodies of knowledge overlapping and not noticing, and the cause is always the same: the two describe the same object in vocabularies that do not share a word.

Here the shared object is a strip of paper joined with a half twist. One subject calls the question what is the shortest developable Möbius band and the other calls it how long does the strip have to be, and neither phrase would find the other in an index.

The pattern, across four instances #

Setting the collection’s cases of this shape side by side makes the mechanism visible.

The same vertex, four times. A degree-four vertex’s conditions were found independently in four traditions, each with its own name and notation, none citing the others. Cause: the object is elementary enough that anybody working on folding meets it.

The cure named first. A technique was published before the problem it solves was posed. Cause: the technique was invented for a different purpose and its application was noticed later.

A test imported without its hypothesis. A layer-ordering test travelled from a literature where sheets are discs, and the sentence naming that assumption did not travel with it. Cause: the hypothesis was true of everything in the receiving subject too.

A bound found in another subject. The Möbius band’s √3, established in differential geometry and unknown in folding. Cause: no shared vocabulary and no reason for either to ask the other’s question.

Four causes and they are all different, which is worth noting. There is no single mechanism producing rediscovery; there is a family of ways two bodies of knowledge fail to meet, and knowing which one is operating says where to look for the next.

The band itself, briefly #

Since the object is a hundred and sixty years old, a sentence on where it comes from.

The Möbius band is named for a nineteenth-century mathematician and was described independently at almost the same time by another — which is itself an instance of the pattern this essay is about, and the earliest one available.

It entered folding not at all. It is a standard example in topology, it appears in recreational mathematics, and the paper model made by taping a strip is a familiar demonstration of one-sidedness. None of that is folding: the demonstration is about the surface rather than about creasing it.

So the object arrived in this collection from mathematics rather than from the folding tradition, which is unusual — most of what is here arrived the other way, as a construction folders had been making that turned out to have a theorem in it.

That reversal is worth noticing. The Möbius band is the first object in this collection that folding did not supply, and the questions it raises are consequently ones the folding tradition has no opinion about.

How the collection came to ask #

The provenance of the question here is worth recording, because it is an instance of the thing the essay is about.

The gluing machinery was built to ask what a boundary costs a search: how many free letters a rectangle’s rim supplies, and what removing it does to the cost of finding a lettering. That is a question about tessellations and complexity and it has nothing to do with Möbius bands.

Building it meant writing a rule that says which boundary points are the same point. Gluing with a flip is one extra line of the same rule, and having written it, a sheet with one side existed.

Then the questions came for free: what the parity does, what the closure condition becomes, which crease angles admit a solution, and how short a strip can be.

So the whole of this essay’s subject arrived as a by-product of an unrelated construction. That is the ordinary way a subject reaches a neighbour’s territory — not by looking for it, but by building something for another reason and finding it has more in it than was asked for.

Two subjects, one strip #

The closing observation is about what each subject wanted from the same piece of paper.

Topology wanted the surface: one side, one edge, non-orientable, and the paper model is a way of holding an abstraction.

Differential geometry wanted the embedding: how a developable strip can be bent into that surface, and how short a strip suffices.

Folding wanted the object: a pretty triangle made from a strip, three creases, a demonstration that a one-sided thing can be flattened.

Three subjects, one strip, and the shared number belongs to the second and third. That the first has no number at all is worth noticing too: the topology is entirely indifferent to how long the strip is, and length is exactly what the other two are about.

What the paper model contributes #

The folded triangle has been made by people for a long time and it is worth asking what that tradition knows that the mathematics does not.

It knows the construction: three creases at sixty degrees, alternating, and a half twist. That is complete and it is the recipe.

It knows the feel: the band settles into the triangle with almost no persuasion, which is the sort of thing a person notices and a computation does not report.

It knows that it works at a range of proportions, because anybody making one cuts a strip by eye and it folds. That is the bound’s practical content, discovered empirically and never quantified.

What it does not know is where the boundary is, and the reason is that nobody making one ever cut a strip so short that it failed for the interesting reason. A strip too short by a lot fails obviously; a strip too short by five per cent fails in a way that looks like bad folding.

So the traditional knowledge is correct, complete for its purposes, and silent about exactly the question the computation answers. That is a reasonable division and it is worth respecting rather than treating the tradition as having missed something.

The claim, stated once #

Not claimed: the smooth Möbius band’s bound. That is somebody else’s theorem and it is a substantial one.

Not claimed: a relationship between the folded and the smooth problems. The bounds agree and the agreement is reported, not explained.

Not claimed: the folded triangle. It is a traditional construction and has been made for a long time.

Claimed: the alternating-sum condition on a three-crease band’s angles, the linear solve for its positions, the bound of 1.7320508 widths for the equilateral triple, the ranking of the other admissible triples, and the fact that no band creased square across the strip closes at any count.

All of it elementary, all of it computed, none of it hard, and the reason it appears here rather than anywhere else is that nobody had built a sheet with one side to ask about.

Where else to look #

Given four instances, the question is what else this subject shares with a neighbour and does not know it.

Developable surfaces. The whole of curved-crease folding is about them, and differential geometry has studied them for two centuries. This collection’s curved-crease work touches that literature and does not systematically draw on it.

Layer ordering. Sorting and partial orders are a large subject, and the collection’s layer-ordering machinery is combinatorics that somebody has certainly done in another setting.

Rigid mechanisms. Kinematics is an engineering discipline with a century of results, and rigid-folding is a special case of it.

Three neighbours, each of which has almost certainly answered a question this subject is still asking. That is not a criticism of anybody; it is the ordinary state of two fields with different vocabularies, and it is the reason this collection keeps finding the same shape.

The result, filed properly #

For the record, so the attribution is right.

The folded triangle is a traditional construction with no identifiable origin, appearing in paper-folding books as a curiosity.

The √3 bound for smooth bands belongs to differential geometry, was conjectured in the nineteen-seventies and settled much later, and is not this collection’s result in any form.

The bound for the folded case, and the alternating-sum condition on the angles, are computed here, by a method that is elementary and that nobody appears to have written down because nobody had a reason to ask.

That is a modest claim and it is the accurate one.