The leaf's rules are the Miura's

A leaf that opens by corrugating is not a rectangle. It is broad in the middle and narrow at both ends, and the pattern it packs into has to taper — which it does in the column widths, because the vertex angles are set by the zigzag alone and the widths never enter any of them.

That was derived before it was drawn, and it has a consequence nobody had chased. If the widths are invisible to the conditions, they are invisible to everything downstream of the conditions — including the whole table of which repeating labellings fold.

The table #

A repeating rule for a grid corrugation is six bits: a letter for each row parity, and a letter for each combination of column parity and row parity. Sixty-four rules.

Sixteen fold. They are the rules whose column creases carry different letters above and below a row, with the row letters free — which is the Miura fold’s table exactly, the same sixteen by number, with the same forty-eight failures, the same forty-eight refusals by the counting theorem alone, and the same thirty-eight of those closing a circle of four panels.

Not similar. Identical, item for item — and it is worth saying that this was not the expected answer. The two patterns have different sector angles at every vertex, and the smallest-sector lemma reads sector angles; a table that depended on it at all would have moved.

Six geometries, one table #

The identity survives everything a plant could plausibly vary.

The printed proportions — seven columns tapering from 0.09 to 0.22 and back — give sixteen. Even columns, all seven the same width, give sixteen. A one-sided ramp, from 0.04 to 0.34, gives sixteen. A violent taper, two narrow columns then two wide ones then two narrow again, gives sixteen. Six taller rows instead of four give sixteen. A steeper zigzag, at 0.8 radians instead of 0.42, gives sixteen.

Same masks, same failures, same circles, in all six. The bar chart is the same bar six times, which is a strange sort of figure and is the finding.

Why nothing can move it #

The reason is one paragraph and it is entirely local.

At an interior vertex of a corrugation, four creases meet. Two are the halves of one zigzag row running straight through the point, so they are collinear and under any repeating rule they carry the same letter. The other two are the halves of a column crease, one in the row above and one in the row below.

The counting theorem wants three of one letter and one of the other. With the two row-halves already equal, the two column-halves must differ — and that is the whole condition, at every vertex, in every geometry.

Nothing in that argument mentions a width, a height, a row count or an angle. The angles enter only through the alternating angle condition, which a corrugation satisfies identically whenever the row above and the row below have the same height, and through the smallest-sector lemma, which has nothing to say at a vertex whose sectors come in equal pairs.

So the table is decided by the cyclic arrangement of the creases at a vertex — which two are collinear — and by nothing else. Two patterns with the same arrangement have the same table however differently they are drawn.

The failures are identical too #

It is easy to be impressed by sixteen matching sixteen and to overlook that the other forty-eight match as well, which is the stronger half of the claim.

Every one of the forty-eight rules that does not fold fails the counting theorem and fails nothing else: the alternating angle sum holds at every vertex where the rule breaks, and so does the smallest-sector lemma. That is true at all six geometries and it is true of the Miura’s forty-eight.

Of those forty-eight, thirty-eight go further and produce a labelling whose arcs send four panels round in a circle — a proof that no arrangement of the layers exists anywhere in the pattern, rather than a vertex that will not close. Thirty-eight at every geometry, on the leaf and on the Miura alike.

And the ten that fail without a circle are the same ten: the rules where the columns keep their letter across a row and the row letter agrees with them, which is the closed form read off three bits and agrees with the arc walk on all sixty-four rules of both families.

Three numbers — sixteen, thirty-eight, ten — matching across two patterns and six geometries. A table that matches only in its successes could be a coincidence of counting; one that matches in its failure modes is the same table.

Sixteen is two to the fourth, and that is the argument #

The paragraph above gives the reason in words, and the same reason written as arithmetic says something the words leave implicit — why the answer is a power of two, and what would have to be true for it not to be.

A repeating rule is six bits. The condition at a vertex is that the two column-halves carry different letters, which written over the two-element field is one equation: the two bits sum to one. A corrugation has two kinds of interior vertex, one per row parity, so the filter is two linear equations in six unknowns.

Six unknowns less two independent equations leaves four, and $2^4$ is sixteen.

That is why nothing geometric can move the table. A linear equation over the two-element field has no room in it for a length. Its coefficients are ones and zeros supplied by which creases meet at a vertex, and a width or a height cannot enter a coefficient that is a statement about incidence. The six geometries were the right test to run and the algebra says they could not have failed.

What a non-power of two would have meant #

The reading has a check attached, and the sweep across families already ran it.

If every corrugation’s filter were linear, every family’s count would be a power of two. The Yoshimura’s is twenty-six, which is not — and the reason is where its vertices are: at degree six, the counting theorem admits two mountains or four, while the parity that captures it at degree four admits zero and six as well. Excluding those two extra cases is not a linear condition, and a table shaped by it has no reason to be a power of two.

So the two families’ counts carry their vertex degrees on their faces. Sixteen says the filter was linear and therefore geometry-proof; twenty-six says something non-linear is biting. The leaf and the Miura are both degree-four throughout, so both are sixteen, and the identity of their tables is a consequence rather than a coincidence.

The converse does not hold, and it is worth saying so — a power of two can also arise from a non-linear filter that happens to land on one. What a non-power proves is the presence of a non-linear clause; what a power suggests is its absence.

What this says about the leaf #

The temptation with a biological pattern is to read every feature as a solution to something, and this is a case where a feature turns out not to be a choice at all.

A hornbeam leaf’s corrugation has its ridges and valleys in a definite arrangement, and that arrangement is one of sixteen — or rather one of four, since the row letters are free and the four ways of writing them are the same object seen from either side and shifted by a cell. The plant did not select it against alternatives. Every alternative outside the sixteen is a pattern that does not fold flat, so a bud packing that way would not pack.

What the plant does choose is the geometry: how many rows, how tall, how broad each column, how steep the zigzag. Those are the parameters under selection — they set how tightly the leaf packs, how it deploys, how the veins run — and the measurement above says that none of them reaches the letters.

So the letters are not a biological fact about leaves. They are a fact about corrugations, and the leaf inherits them the same way an engineered Miura does.

The convergence, stated exactly #

This collection has said several times that a leaf’s corrugation and a Miura fold are the same pattern, and the statement has always been a little loose. Now it can be made precise, and the precision matters because it is narrower than the loose version.

What is identical: the table of repeating rules, the sixteen that fold, the forty-eight that do not, which theorem refuses them, and which of the failures close a circle.

What is not: the geometry. A leaf tapers and a Miura does not; a leaf’s rows are set by its growth and a Miura’s by whatever the engineer wanted; the two fold to different shapes and pack to different densities.

So the convergence is at the level of the letters and the conditions on them, and it is complete there. Two patterns arrived at independently — one by a plant, one by an engineer — have not merely similar labelling rules but the same rules, because there is only one set available.

That is a sharper version of what this collection found about the vertex itself: a degree-four vertex with a three-to-one assignment turns up in a buckled cylinder, a Miura, a Resch tessellation and a crumpled sheet, and it is not coincidence or influence but a small solution space. Here the same argument runs one level up, over whole rules rather than single vertices.

What a sweep at six geometries is for #

There is a methodological point here worth separating from the biology, because it is the reason the measurement was run six times rather than once.

A single sweep establishes that this drawing of the leaf has sixteen surviving rules. It says nothing about whether that number is a property of the corrugation or of the particular widths and heights somebody chose — and the widths were chosen to look like a leaf rather than for any principled reason.

Six sweeps at deliberately different geometries turn a fact about a drawing into a fact about a family. The geometries were picked to be awkward rather than representative: a one-sided ramp is not a shape any leaf has, and a violent taper with a factor of twenty between its narrowest and widest columns is a drawing rather than a plant. If the table were sensitive to geometry at all, those are where it would show.

That is the same discipline as feeding a checker the input it must refuse. A measurement that comes out the same on the cases chosen to break it is worth more than one that comes out the same on cases chosen to resemble each other.

What would have made it a choice #

It is worth asking what the leaf would have to be like for its letters to be under selection, because the answer says why this case came out the way it did.

The letters would have to be free — the conditions would have to leave more than one genuinely different option — and the options would have to differ in something a plant cares about.

Neither holds. The sixteen surviving rules are four objects up to symmetry, and all four fold to the same corrugation; there is nothing to select between. And the conditions leave nothing else: forty-eight of the sixty-four rules produce a sheet that does not lie flat, which for a leaf in a bud is not a variant strategy but a failure to pack.

A pattern where this went the other way would need vertices of higher degree, where more labellings survive. The Yoshimura’s degree-six vertices leave twenty-six rules of sixty-four rather than sixteen, and a corrugation built on that arrangement would have genuine alternatives. No leaf packs that way, as far as this collection’s material goes.

What the taper does instead #

The taper is not idle. It does exactly one thing and the thing is geometric.

Tapering across the folds rather than along them — varying the row heights instead of the column widths, which is what anybody drawing a narrowing leaf would try first — puts the alternating angle sums at 186.4° and 173.6°, and the sheet does not close. So the direction of the taper is forced, and getting it wrong produces a pattern that will not fold at any labelling whatever.

That is the sense in which the leaf’s shape is constrained: not in its letters, which are free of it, but in which axis it is allowed to narrow along. The plant has one degree of freedom in its taper and none in its assignment, and both facts come out of the same vertex.

What is a fact about leaves after all #

Stripping the letters out of the biology leaves the question of what remains, and the answer is most of what matters to a plant.

The number of rows decides how much the leaf shortens when packed and how far it has to travel to open. That is under selection and it is measurable: the consistency of a corrugation’s labellings falls as the rows grow, but the packing ratio rises, and a bud has to trade the two.

The direction of the taper is forced, and getting it right is not free — it means the leaf narrows along its folds rather than across them, which constrains where the veins can run.

The zigzag angle decides how flat the packed leaf is and how much of the sheet ends up in the pleats rather than in the panels. Steeper packs tighter and wastes more.

The letters are none of the above. Sixteen rules, four objects, no alternatives, no trade-off, nothing to select on.

That division is worth having explicitly, because a paper about a folding leaf will describe its mountain-valley pattern alongside its proportions as though both were findings about the plant, and only one of them is.

The comparison the engineering makes #

There is a reason this identity is worth stating in a collection that is mostly about paper rather than about plants, and it runs the other way from the usual bio-inspiration story.

The Miura fold was arrived at as engineering: a way to pack a large flat thing into a small volume and open it with one motion, published in 1970 and used on solar arrays. The leaf’s corrugation was arrived at by growth. The usual framing is that the engineer might have looked at the leaf and been inspired.

What the rule table says is that inspiration was not available in the place it would have been looked for. The letters are not a design choice either party made; both patterns have them because there are sixteen rules and forty-eight failures and no third option. What an engineer could have taken from a leaf is the geometry — the taper, the row proportions, the packing ratio — and that is exactly the part the two patterns do not share.

So the convergence is real and it is uninformative in the direction people usually want it to run. Two independent arrivals at the same table is evidence about how small the table is, not about either arriver.

Where the ladder goes next #

The immediate question is whether any real leaf departs from the sixteen. Everything measured here is a construction; a photograph of an unfurling hornbeam or beech leaf, with its ridges and valleys read off, would say whether the pattern in the bud is one of the four objects the table permits. This collection has no instrument that reads a photograph, and building one is a different kind of work from anything here.

The second is the corrugations this collection has not built. A plicate leaf is a grid corrugation; other plants pack by rolling, by conduplicate folding along a single midrib, and by patterns with vertices of degree six. The bud chooses the pattern surveys which, and the rule sweep has been run on exactly one of them.