The same vertex, found four times

Four things that have nothing to do with each other: the lattice a crushed cylinder falls into, the Miura fold, Ron Resch’s tessellations, and the creases left in a sheet of paper that has been screwed into a ball and opened out again.

All four are built from the same object. A vertex of degree four, with sectors that satisfy the alternating condition, carrying three creases of one letter and one of the other. Different angles, different arrangements, different purposes — the same local structure.

The count is the explanation #

Start with the arithmetic, because it does most of the work.

A degree-four vertex has four creases and each is a mountain or a valley: sixteen assignments. Maekawa’s condition requires the counts to differ by exactly two, which means three and one — so the twelve assignments with a two-two split are gone immediately, leaving eight.

Of those eight, the big-little-big lemma removes more, depending on the sectors, and Kawasaki’s condition is a constraint on the angles rather than the letters so it decides whether the vertex folds at all.

Sixteen down to a handful. That is a very small space, and a small space is one that independent searches converge on.

Rarity is the general fact #

The degree-four case is the sharpest instance of something this site measures across sizes.

The share of all mountain-and-valley assignments that fold flat falls fast as a pattern grows, while the raw count of valid ones rises. Flat-foldability is a rare property, and the rarity is what makes it structural: the set of valid patterns is not a random subset of all patterns but a highly organised one, described by conditions that hold at every vertex.

So a designer, a buckling shell and a crumpling hand are all sampling the same small target, and they hit it for the same reason: everything else is not a folding.

Doing the count properly #

The sixteen-to-a-handful sketch above deserves to be run exactly, since the site’s habit is to compute rather than to gesture.

Fix the four sectors. Maekawa leaves the eight assignments with a three-to-one split. The big-little-big lemma then says that the two creases flanking a strictly smallest sector must differ, which kills half of the remaining ones whenever such a sector exists — and for a generic vertex it does.

That leaves four. Four flat-foldable assignments of a generic degree-four vertex, out of sixteen: a quarter of the space, and the surviving quarter is completely determined by which sector is smallest. There is no choice left to make.

This site computes those four twice, by routes sharing no code — once by enumerating letters against the local conditions, and once by solving the vertex as a closed spherical linkage in three dimensions — and requires the two to agree on the identity of all four rather than merely on the count.

Four routes to one vertex #

It is worth walking the four cases, because the routes are genuinely unrelated even though the destination is not.

The buckled cylinder arrives by minimising elastic energy under a no-stretch constraint. Nothing chooses the assignment; the deformation determines which creases go out and which go in.

The Miura fold arrives by design, from a requirement: a surface that packs flat, deploys with one motion, and contracts in both directions. The three-to-one assignment is what delivers the single degree of freedom.

The Resch tessellation arrives from an artist’s exploration of what a repeating unit can do, with no mechanics and no requirement, working by folding paper and looking.

The crumpled sheet arrives by nothing at all. It folded, so a folding exists, and the vertices where creases meet are constrained by the same conditions.

What the four assignments are #

Naming them makes the smallness concrete.

Each of the four is: pick which of the four creases is the odd one out. That is the whole parameterisation. The odd crease is the one whose letter differs from the other three, and once it is chosen the assignment is fixed.

And the choice is not free either — the odd crease has to be one of the two bounding the smallest sector. Writing the four out with the smallest sector first, they are the two with a single valley on one of those creases and the two obtained by exchanging every letter, so the flip pairs them off.

So a generic degree-four vertex has, up to turning the sheet over, two flat-foldable assignments: put the odd crease on one side of the smallest sector or on the other. A space with two things in it.

That is why four independent searches agree. There was nothing to disagree about.

Why four and not five #

The degree matters and it is not arbitrary, which is worth establishing because otherwise the convergence looks like a coincidence about the number four.

A degree-two vertex is a straight crease and carries no information. A degree-three vertex cannot satisfy Maekawa, because three letters cannot split with a difference of two — three and zero differs by three, two and one by one. So the smallest flat-foldable interior vertex is degree four.

Degree six and above exist and fold, and they have more freedom: more assignments, more sector arrangements, more ways to be a valid vertex. That freedom is exactly why they are less convergent — a larger solution space is one that independent searches do not agree on.

So four is the smallest, and the smallest is the most constrained, and the most constrained is where everybody ends up.

One degree of freedom is not automatic #

The convergence has a second layer that the count alone does not explain.

A degree-four vertex folds with one degree of freedom: fix one fold angle and every other is determined. That is a property of the vertex as a spherical linkage, and it is the property that makes a tessellation of such vertices behave as a single mechanism rather than as a floppy sheet.

For a designer wanting a deployable, that is the whole requirement. For a buckling shell it falls out because the deformation is a one-parameter family. For an artist it is what makes the model satisfying to open and close. Three different reasons to want the same thing.

The count at degree six, done rather than predicted #

The prediction below can be turned into arithmetic, and the arithmetic gives the factor rather than the direction.

At a vertex of degree $2k$ Maekawa admits the assignments with $k+1$ of one letter or $k-1$, and the smallest-sector lemma then removes half of what is left for each sector that is strictly smaller than both its neighbours. A generic vertex has $k-1$ such sectors — Kawasaki forbids the $k$-th, since $k$ alternating minima would make one alternating sum strictly less than the other — so the surviving count is

$$2^{k} = 2^{d/2}$$

Four at degree four, eight at degree six, sixteen at degree eight, which is the count this site publishes and which comes out of the conditions rather than being fitted to them.

The share of the whole space is therefore $2^{-d/2}$: 25 per cent, 12.5 per cent, 6.25 per cent. The share halves with every two degrees — so a degree-six vertex has more folding assignments in absolute terms and is a rarer target in relative ones, and the conditions bite harder rather than less as the degree rises.

Which gives the convergence factor #

What does grow is the absolute count, and that is the quantity independent searches meet.

Up to turning the sheet over, the count is $2^{d/2-1}$: two objects at degree four, four at degree six, eight at degree eight. Two searches landing on the same degree-four vertex have an even chance of agreeing before anything else is considered; at degree six it is one in four, and at degree eight one in eight.

So the prediction is not merely that degree six should show less convergence. It is that it should show twice less, and degree eight four times less, from the counting alone and before any argument about sector angles or arrangements is made.

Those numbers are small enough to be sobering. A field’s worth of independent rediscovery at degree four, against a space twice the size at degree six, is not a difference between forced and free — it is a difference between two and four, both of which are tiny. The reason the higher-degree patterns are attributable is therefore not that their vertex is unconstrained; it is that four is enough for people to diverge and two is not.

The convergence argument survives the arithmetic and shrinks under it, which is the right outcome for a claim that was made qualitatively and is now measurable. The vertex is forced in the sense that matters and the forcing is a factor of four rather than a wall.

The same argument, at the next size up #

The prediction the account makes is testable at degree six, and the test is worth stating even though this site has not run it.

If convergence is driven by the size of the solution space, then degree-six vertices — with sixty-four assignments, more sector freedom, and a much larger valid set — should show less convergence. Independent designers working with degree-six vertices should arrive at genuinely different things.

That is roughly what the record shows. The famous repeating patterns are overwhelmingly degree-four; the higher-degree ones are individual, attributable, and not independently rediscovered. Nobody is arguing about who found the Resch pattern.

It is a soft test and it points the right way, which is about as much as a historical prediction in this field can offer.

What this does to attribution #

Now the historical consequence, and it is the reason this essay sits in this field.

In a subject whose solution space is this small, independent rediscovery is the expected outcome rather than the surprise. Asking who found the degree-four three-to-one vertex first is close to asking who first noticed that a triangle’s angles sum to a straight angle: the object is forced by the constraints, and anybody who works on the constraints meets it.

That reframes the attribution problem in a useful way. The reason this field has so many near-simultaneous discoveries is not only that it lacked a journal. It is that the space of things to discover is small enough that several people working independently will cover it.

A test the argument could fail #

The claim is not unfalsifiable and it is worth saying what would sink it.

If the convergence were about influence rather than about constraint, the four cases would share more than the vertex. They would share sector angles, or repeating arrangements, or a characteristic proportion, because copying transmits specifics and constraint transmits only structure.

They do not. The Miura’s sectors are set by its slant parameter and are continuously tunable; the Yoshimura’s are set by the cylinder’s aspect ratio; Resch’s are chosen for appearance; a crumple’s are arbitrary. What is shared is precisely and only the part the conditions force.

What the sectors are free to be #

It is worth being exact about which parts of a degree-four vertex are forced and which are not, because the essay’s whole claim depends on the split.

Forced: the number of creases carrying each letter, the identity of the odd one out given the sectors, and the fact of one degree of freedom. All of that follows from the conditions and nothing chooses it.

Free: the sector angles themselves, subject only to Kawasaki’s alternating condition — which leaves a two-parameter family for a degree-four vertex, since the four sectors sum to 360° and satisfy one further linear relation. Two free parameters is a great deal of room.

So the vertex is forced in its combinatorics and free in its geometry, and every visible difference between the Miura, the Yoshimura and Resch’s work lives in the free part. The convergence is on the skeleton; the variety is on everything hung from it.

The trap this creates #

There is a specific error that a small solution space invites, and this site has already walked into a version of it.

When a condition is satisfied identically — for every member of a family, by an algebraic identity rather than by choice — a designer can believe they have selected something when they have selected nothing. The twist vertex is exactly this: its sectors are A, A, 180° − A, 180° − A for any interior angle A, so both alternating sums are a straight angle whatever the polygon, and Kawasaki gives no information at all about which polygons work.

So convergence has a shadow. The same constraints that force independent searches to agree can also stop a constraint from discriminating, and a designer checking a condition that always holds learns nothing while feeling careful.

Where the space is large, the history looks different #

The contrast case is worth drawing out, because it shows the argument is about this subject’s structure rather than about subjects in general.

Designing a base by circle packing is a search over a space that is not small at all: the flap lengths are continuous, the packings are many, and the optimisation is hard. And the history of that part of the subject looks completely different — results are attributable, methods have single authors, and there is very little independent rediscovery.

Same subject, same people, same decades. The difference is the size of the space, and the historical texture follows it.

What is genuinely different between the four #

Having argued that the vertex is common, the differences deserve their due, because they are where all the content is.

What makes the Miura the Miura is not its vertex but its global arrangement: rows offset so that the whole sheet has one degree of freedom rather than many, which is a statement about how the vertices connect. What makes Resch’s tessellations his is the choice of repeating unit and the way the sheet gathers. What makes the Yoshimura what it is, is the cylinder.

The vertex is the alphabet. The patterns are the words, and nobody claims two authors are related because they both used the letter e.

A small space is a good place to start a subject #

There is a constructive reading of all this, and it is probably the more useful one.

A field whose fundamental object is forced is a field where the foundations are not in dispute. Nobody argues about what a flat-foldable degree-four vertex is; there is one answer, several people found it, and the disagreement is entirely about who to name it after. Compare a subject whose basic objects are matters of choice, where the definitions are contested and half the literature is about which definition to prefer.

So the convergence that makes attribution messy is the same property that makes the mathematics clean. The vertex conditions are exact, cheap, and universally agreed, and that is unusual and valuable.

The messy attribution is a small price, and this field pays it in full.

The idealisation, named #

The four cases are treated above as though each has a single well-defined vertex type, and two of them do not.

A crumpled sheet has vertices of many degrees, most of them not flat-foldable in any clean sense because the sheet has stretched and torn slightly and the creases are not exact. The claim about crumpling is a claim about the well-formed vertices in it, not about all of them. A buckled cylinder is likewise idealised: the mode is regular and a real can is not.

So “the same vertex four times” is exactly true of the two designed cases and approximately true of the two found ones, and the approximation is doing work.

One more consequence of that split. Because the variety lives in the sectors and the sectors are continuous, two designs can be arbitrarily close without either being derived from the other — there is no discrete step between them that a claim of influence could point at. That is a genuine difficulty for attribution in this subject and it has no clean resolution.

Where this goes next #

This closes the rediscovery ladder. Below it, found before it was designed is the single clearest case, and sideways one vertex repeated is what happens when the vertex is taken as a unit and tiled deliberately.

The generalisation worth carrying away: a field’s rate of independent rediscovery is a measurement of how constrained its objects are, and it can be read that way rather than as a failure of scholarship. Folding rediscovers things constantly. That is what it looks like from inside a subject whose fundamental objects are forced by four conditions at a point.