The plant's pattern is not a hard case

A hornbeam leaf leaves its bud already the right shape. It has spent the winter folded into a corrugation — a zigzag along the midrib with a herringbone of secondary folds running off it — and it opens by unfolding rather than by growing into position.

That is a genuinely striking thing for a plant to do, and the striking part is easy to locate wrongly. A crease pattern that folds flat is a rare object; almost every pattern fails; so a plant that arrives at one looks like a plant that has solved something hard.

It has not. Every combinatorial question this collection knows how to ask of a crease pattern comes back easy on the leaf, and the interesting thing about the plant is somewhere else entirely.

What the pattern costs to letter #

Searching the leaf’s pattern for a consistent lettering — an assignment of mountains and valleys satisfying the conditions at every vertex whose implied statements about which panel lies above which do not contradict one another — costs one step per panel.

Nine, twelve, fifteen, eighteen, twenty, twenty-four. No backtracking anywhere, at any geometry, and the same result at six deliberately awkward ones — even columns, a one-sided ramp, a violent taper, taller rows, a steeper zigzag.

That is the same cost as the orthogonal grid a box-pleated design is drawn on and the same cost as a sheet crumpled at random. The leaf is not distinguishable from either by any measure of search difficulty.

Why it is easy, in one sentence #

Every interior vertex of a leaf corrugation has degree four with a strictly smallest sector, and such a vertex has exactly two admissible labellings once one of its creases is known.

That is the whole of it. The conditions at each vertex determine almost everything from almost nothing, so a letter written anywhere propagates outward and settles the sheet, and the search’s only real decisions are one per vertex. A pattern built entirely of such vertices has as many genuine choices as it has vertices and no more, and no arrangement of any search can do better or worse.

The leaf is built entirely of such vertices because a corrugation is: two families of creases crossing, each crossing making a degree-four point. So is a Miura, so is a box-pleated grid, and so is a crumpled sheet, which is why all four cost the same.

There is a way of putting this that makes the plant’s position clearer. The leaf does not choose an assignment from among the possibilities and then discover it folds. The assignment follows from the geometry — a crease that is a ridge is a ridge because of how the tissue grew there, and the vertex conditions are satisfied because the growth that produced the angles also produced the letters.

That is a genuinely different situation from a designer’s. A designer draws a pattern and then has to letter it, and at sixteen divisions ninety-nine hand-drawn letterings in a hundred cannot be folded. A plant never has the two steps apart, so the failure mode does not exist for it.

The right analogy is not a designer working out a pattern. It is a sheet of paper that has been folded and flattened: the letters a crumple was given are consistent because a physical process produced them, and no search was involved on either side.

The vertex is a two-by-two #

The one-sentence explanation deserves its arithmetic, because the phrase strictly smallest sector is carrying a precise factor and it is worth knowing which.

A degree-four vertex admits sixteen letterings. The counting theorem leaves the eight with a three-to-one split. The smallest-sector lemma then requires the two creases flanking the strictly smallest sector to differ, leaving four.

Those four are not a scatter. Writing them out with the smallest sector first, they are $VMMM$, $MVMM$, $MVVV$ and $VMVV$ — two independent binary choices: which of the two flanking creases is the odd one out, and which letter is in the majority. A two-by-two.

Fix any single crease to any letter and exactly one of those two coordinates is settled. Two labellings remain, whichever crease was fixed and whichever letter it was given. That is the essay’s claim, and it holds for all four creases rather than for a convenient one.

Which is where the factor of two lives #

Now take the lemma away, which is what happens when the smallest sector is tied rather than strict.

The counting theorem still leaves eight, and nothing removes half of them. So a tied vertex is a three-by-two rather than a two-by-two, and fixing one crease leaves four labellings instead of two.

So the strictly smallest sector is worth exactly one halving, and the halving is the whole difference between a vertex that propagates and one that does not. With four survivors, a known crease forces the vertex to a single binary choice and the choice forces its neighbours. With eight, a known crease leaves two binary choices, and a search has to make one of them on its own.

That is the mechanism behind the flat cost curve, stated as a number rather than as a property. A corrugation’s vertices are strict everywhere — the taper, the zigzag and the row heights all conspire to leave no two sectors equal — so every vertex is a two-by-two and the propagation never stalls.

It also says what a leaf would have to be like to be hard, more sharply than the later section does. Not a higher degree, necessarily: a corrugation with equal sectors would do it, because ties silence the lemma and a silenced lemma doubles the branching at every vertex at once. A plant whose veins happened to make its sectors equal would have a pattern the search had to work at, and the fact that no such leaf is known is a fact about veins rather than about folding.

The rule table is the Miura’s #

The stronger version of the same point was measured earlier and is worth restating here, because it says the leaf is not merely easy but not distinct.

A corrugation of this kind is generated by a repeating rule — a handful of bits saying which letter each row and column carries. Sweeping every rule for the leaf and for the Miura gives identical tables: sixteen of sixty-four fold in each case, forty-eight are refused by the mountain-valley count alone, and thirty-eight close a four-panel circle. The same sixteen, by number, at six geometries chosen to be as unlike each other as the construction allows.

So the leaf’s combinatorics are the Miura’s combinatorics. What differs between a hornbeam and a solar array is the geometry — the taper, the vein angles, the way the cells change size along the blade — and none of that changes which rules fold.

What the plant’s difficulty actually is #

Four things, and none of them is a search.

Growing the geometry. The leaf’s creases are laid down by differential growth in the primordium, and the pattern’s angles are set by where the veins go. Growth as a metric is the language for that, and the constraint is severe: a sheet that grows unevenly acquires curvature, and a leaf that acquired curvature would not fold flat into a bud at all.

Fitting the bud. The corrugation has to reduce the blade to something that fits inside a bud of a given radius, and the packing is what decides how many folds are needed. That is a real optimisation and the plant is under real pressure on it.

Opening once. The fold is used once. It has to open reliably, in one direction, without a mechanism to pull it — and opening with nothing to pull is where the difficulty of that is set out.

Thickness. A leaf has one, it is not negligible relative to a bud, and it is twice as thick where it is thickest as the corrugation stacks.

Every one of those is a difficulty of geometry, material or growth. None is a difficulty of assignment, and the plant never faces the question that the search answers.

What the plant could not have done #

The negative result is more interesting with its converse beside it, so it is worth asking what a hard biological folding pattern would look like and why none is expected.

It would need vertices whose conditions leave real choices — degree six or more, or sectors that tie — so that the letters were not forced by the geometry. It would need those vertices to be numerous and coupled, so that a local choice had distant consequences. And it would need the correct global assignment to be rare among the locally admissible ones, so that getting it right was not automatic.

Nothing in a growing tissue would produce that, and the reason is mechanical rather than mathematical. A crease in a leaf is a line of differently-grown cells, and the growth that lays it down is local — a cell responds to its neighbours. A pattern whose correct letters depend on distant coordination is a pattern no local growth process could lay down, because there is nothing to carry the coordination.

So the easiness of the leaf’s pattern is not an accident of this particular leaf. It is close to a requirement: a pattern a growth process can produce is a pattern whose letters are locally determined, and a pattern whose letters are locally determined is one a search settles at one step per panel. The measurement and the biology point at the same conclusion from opposite ends.

Why a negative result here is worth having #

It is tempting to leave an easy case unreported. The reason not to is that the alternative is a plausible sentence nobody has checked.

“The leaf solves a hard combinatorial problem” is the sentence, and it is the kind that is repeated because it sounds right and because the underlying fact — flat-foldable patterns are rare — is true. Rarity gets read as difficulty routinely, and the two are unrelated quantities: the share of admissible letterings that agree with themselves falls sharply as a corrugation grows, and the cost of finding one does not move.

So the useful thing to say about the leaf is not that it is impressive. It is which part is impressive, and the measurement is what makes the distinction sayable. A plant that had to search would be an extraordinary claim; a plant whose pattern’s letters are forced by its shape is an ordinary and correct one.

The two counts that do differ #

There is one place where the leaf and the Miura come apart, and it is worth naming so the “identical” above is not read too widely.

They differ in how much they shrink, and by how much in each direction. A Miura’s two in-plane dimensions fall together as it folds; a tapered leaf’s do too, but unequally along the blade, because the cells are not the same size at the tip and at the base. That is a geometric difference and it is exactly the difference a plant is under selection about — the shrinkage is what has to fit the bud.

They also differ in where the thickness accumulates. A uniform corrugation stacks evenly; a tapered one stacks more where the cells are widest, which is the middle of the blade, and that is where a bud is widest too.

So the honest summary is that the leaf and the Miura have identical combinatorics and different geometry, and every difference between them that matters to the plant is on the geometry side. Which is a satisfying place for the difference to be, and it is what makes the corrugation a good thing for a plant to have arrived at: the part that has to be right is the part evolution can adjust continuously.

Which theorem was checked, and how #

Every lettering counted is written back onto the leaf’s pattern and put past two instruments that did not produce it: the four conditions at every interior vertex, read by the pattern’s own reader, and a folded sheet rebuilt from the coordinates and walked for a circle.

The linear cost is asserted rather than observed: every leaf geometry in the ladder is required to cost no more than one step per panel, and the assertion fails on the first that does not. That is stronger than a straight line through four points, since a search could visit exactly n nodes while taking a wrong turn and recovering.

The rule table’s identity with the Miura’s is asserted mask for mask rather than by count — the same sixteen rules, not sixteen rules — because two tables of the same size are not the same table.

What the picture cannot show #

That the leaf here is a model. It is a corrugation with a taper, built to the angles a hornbeam’s veins take, and it is not a leaf: it has no thickness, no midrib of different stiffness, no cells, no turgor and no growth. The organism is not the model sets out the gap, and everything above is a statement about a piecewise-flat sheet whose creases are where a leaf’s are.

The gap matters most for the negative result. Saying the plant faces no combinatorial difficulty is a statement about the model’s combinatorics, and a real leaf’s folding involves a great deal the model omits — the crease is a region of differently-grown tissue rather than a line, the fold has a radius set by the tissue’s thickness, and opening is driven by turgor rather than by geometry. None of that reintroduces a search, which is why the conclusion survives; but the confidence should attach to no assignment problem rather than to nothing difficult.

And four geometries is four. The claim generalises on the mechanism — degree-four vertices with a strictly smallest sector — rather than on the sample, and a leaf whose veins produced a vertex of degree six would need measuring rather than assuming.

What an engineer takes from it #

A great deal of deployable-structure work is justified by pointing at a leaf, and the reading here changes what the pointing is worth.

It does not diminish it. A hornbeam’s corrugation is an excellent engineering solution to the problem the hornbeam has — pack a large flat thing into a small round space, open it once, reliably, with no actuator — and copying it is entirely reasonable.

What it changes is what has been copied. Taking the leaf’s pattern gives its geometry: the taper, the vein angles, the way the cell size changes along the blade, all of which are tuned by selection to the bud-packing problem. It does not give any combinatorial insight, because there is none to give — the same rule table is the Miura’s, and the Miura was arrived at without looking at a plant.

So an engineer copying a leaf should copy the shape and not expect the letters to be the clever part. And an engineer whose deployment problem has a different packing constraint from a bud’s has no reason to prefer a leaf’s proportions at all, since the proportions are the part that was fitted to the bud.

What is still unmeasured about the leaf #

Two questions this collection can ask of a crease pattern have not been asked of the leaf, and both are the expensive ones.

How many folded states it has. The letters agreeing among themselves is a necessary condition; how many distinct stackings of the panels satisfy the two non-crossing rules is a separate count, and on a leaf of twenty-four panels it is beyond what the ordering search will attempt. A plant that had many available stackings and needed a particular one would face a genuine difficulty, and nothing here says whether it does.

Whether it folds rigidly. A leaf’s cells do not stretch much, so the folding is closer to a rigid one than a paper model is, and whether the pattern admits a rigid motion at all is a different question from whether it folds flat. The Miura does; a tapered corrugation need not; and the answer would say something about whether the leaf’s opening is a mechanism or a relaxation.

Both are worth having and neither is cheap. Recording them as unmeasured is the honest alternative to letting the easy result stand for the whole object.

Where the ladder goes next #

The pattern running through all of this has been that a difficulty kept turning out to belong to an instrument rather than to a subject, and that shape has a literature of its own. The cure was named before the disease was caught here records where the heavy-tailed runtime, the restart strategy and the reasoning behind both came from, which is not this subject and never was.