The surface has to be supplied

How much surface fits in a body has a curve in it that rises, turns over and falls to nothing. The rise is folding buying area; the turn is the sheet’s own thickness beginning to fill the box; the fall is the box full of sheet with no room left for anything to happen at — the same pile a thick panel runs into from the engineering side.

The curve is right and it is answering a question no organ asks. An organ whose job is to have area does not need the area to fit; it needs the area to work, and a surface works by having something delivered to it. A gut lining absorbs from a lumen and passes into a capillary, and nothing grown has a seam to run the supply through. A gill exchanges with water on one side and blood on the other, and a leaf packs by corrugating with nothing to service at all. A cortex is not exchanging anything, but its sheet is threaded with the vessels that keep it alive.

So there are two things in the box, and only one of them has been charged.

Charging the channel #

The model needs one addition and it is a small one, which is what makes the result surprising.

A box of side S. A corrugation of k folds, each panel running the full side. Without supply, each fold consumes t of the box’s depth to hold its own sheet, so the depth left for the panels to occupy is S − k·t, and the area held is k·S·(S − k·t).

Now require that each unit of surface be serviced, and take the service to be a channel of depth δ lying alongside. Every fold now consumes t + δ instead of t. The depth left is S − k(t + δ), and the area is k·S·(S − k(t + δ)).

That is the whole change. It has no new mechanism in it, no transport model and no new parameter beyond δ, and it is worth being explicit that the form has not changed at all: it is the same parabola with a different coefficient. The corrugation itself is unchanged — the uniform one is still the only member at its ratio.

Which is why the correction is not small #

A parabola k·S·(D − k·c) peaks at k = D⁄2c and reaches S·D²⁄4c. Both of those are reciprocal in c, and c has gone from t to t + δ.

The optimum fold count moves by t ⁄ (t + δ). The peak area moves by the same factor. Those are the two statements and neither has a small parameter in it.

At δ = t — a channel as deep as the sheet is thick, which is modest — the optimum halves and the area halves. At δ = 3t the optimum is a quarter of the unserviced one and so is the area. There is no regime in which supply is a correction to be neglected, because it enters the coefficient rather than being added to the answer.

The figure computes the peak by searching over whole fold counts and checks the result against S ⁄ 2(t + δ), which is what makes the claim testable rather than restated: an arithmetic slip in how the channel is charged would move the numerical peak away from the closed form and the figure would refuse.

A thin sheet with a channel is a thick sheet #

The two quantities enter the arithmetic as t + δ and nowhere separately. That is not a modelling convenience; it is the finding, and it has a consequence that is worth stating carefully.

A sheet of thickness t serviced at depth δ has exactly the same optimum and exactly the same peak area as a sheet of thickness t + δ with no supply at all. Not similarly, not approximately: identically, at every fold count, at every box size.

So the quantity that an anatomist can measure — how much surface an organ holds, and how finely folded it is — cannot distinguish the two. An organ with a thin lining and a deep capillary bed and an organ with a thick lining and no supply produce the same numbers. Everything that distinguishes them is at a scale below the fold.

That is an unusual kind of underdetermination and it deserves the name. It is not measurement error and it is not a limitation of the technique. Two structurally different arrangements produce identical values of every quantity the model computes, which means the model cannot be used to argue from area back to architecture. The layer order of a crease pattern is the same shape of problem: the visible quantity leaves a set of possibilities, and deciding among them needs something the visible quantity does not contain.

What the supply is competing with, which is itself #

There is a second effect and the model above does not have it, so it is worth naming as a limit rather than smuggling in.

A channel that services a unit of surface is not free-standing. It has to be reached in turn, and in a real organ the supply is a branching network whose finest vessels lie against the surface and whose trunks are large. So δ is not a constant: the depth needed to service a unit of surface depends on how much surface is being serviced, because the trunk vessels carrying to all of it are in the box too.

That makes δ an increasing function of k rather than a parameter, which turns the parabola into something steeper and moves the optimum earlier again. The direction is unambiguous and the magnitude is not, so the model here takes δ fixed and states the answer as a bound: the optimum is no later than S ⁄ 2(t + δ), and the peak area no larger than the corresponding value.

That is the honest form of the result. Charging the finest channel is a lower bound on what supply costs, so a correction computed from it is a lower bound on the correction.

What the correction does to the reading of an organ #

The arithmetic is short and its consequences for how a folded organ is described are not, so it is worth working one of them through.

A gut lining is usually described by two numbers: how much its folding multiplies the area of a smooth tube, and how fine the folding is. Both are measurable and both appear in the model here — the first is the packing ratio and the second is the fold count.

Take a lining reported as multiplying its area a certain number of times and ask whether that lining is near its own optimum. Without the supply term the question is answered by comparing the fold count with S ⁄ 2t, which needs the box and the sheet’s thickness and nothing else. With the supply term the same comparison needs t + δ, and δ is a quantity nobody reports because it is not part of the lining.

So a structure that looks under-folded against the thickness of its own epithelium may be exactly at its optimum against the thickness of the epithelium plus the bed beneath it. The two verdicts differ by a factor of four at a plausible δ, and the difference is entirely in a parameter that belongs to a different tissue.

That is the practical form of this rung and it is a warning rather than a result. An optimality claim about a folded surface is a claim about a coefficient that has two tissues in it, and computing it from one of them is not a first approximation — it is a bound in the wrong direction.

Why the two charges are the same charge #

It would be reasonable to expect the sheet’s thickness and its supply to enter differently. The sheet is the thing being folded; the channel is something beside it. One is intrinsic and the other is an accessory, and accessories usually appear as small additive terms.

They enter identically here, and the reason is worth isolating because it explains why the correction is large.

The box has a depth. Every fold lays its panel across that depth, and whatever else has to lie across the depth alongside that panel is spent from the same account. The model does not distinguish between “paper” and “not paper”; it asks only what depth a fold consumes, and a fold consumes whatever is stacked with it.

So anything per-fold competes on identical terms with the sheet. A channel does. So does a layer of mucus, a basement membrane, a sheet of muscle, and the clearance a fold needs to open and close. Every one of them enters the coefficient, and the coefficient is the thing the optimum and the peak are both reciprocal in.

Which suggests the right way to state the result is not about supply at all. The sheet’s thickness in the ladder’s first rung was standing in for everything charged per fold, and calling it the thickness of the sheet was a modelling choice that happened to name the largest term at the time. Once anything else is charged per fold, the parameter is a sum and the first rung’s t was its first term.

The one thing that does not move #

Everything in this rung moves the optimum earlier and the peak lower. One quantity does not move at all, and it is the same one the convergence ladder found at its own optimum.

At k = D⁄2c the depth consumed is k·c = D⁄2, which is half the box. At the best fold count, half the available depth has gone to sheet and supply together, whatever t is, whatever δ is and whatever the box is.

That is the third time this collection has met the same arithmetic, and the three instances are worth putting beside one another because they are in three different fields. A hinge with a radius takes surface per fold and the optimum is where half the sheet is hinge. A sheet with a thickness takes depth per fold and the optimum is where half the box is sheet. A surface with a supply takes depth per fold and the optimum is where half the box is sheet and channel.

The pattern is not about paper. A benefit proportional to a count, against a cost proportional to the same count taken out of a fixed budget, optimises where half the budget has gone. Recognising it is worth more than any of the three results, because the next place it turns up will not be labelled — and the reading it licenses is a strong one: a structure at its own optimum has spent half of whatever it was competing for, so measuring the share spent says how close to optimal a structure is without knowing any of the constants.

Which way size cuts, and it is not the usual way #

The coefficient is the subject of this rung and the box’s size has been held fixed throughout. Letting it move produces a consequence that runs against the standard intuition about surfaces and volumes, and it is worth working out because the intuition is usually right.

The optimum is S ⁄ 2c and the peak area is S·S² ⁄ 4c, where c is whatever each fold consumes. Both are increasing in S, and the second is increasing faster than the box’s own extent. So the area a box holds at its own optimum grows as the cube of its side while the box grows as the square of it — in this two-dimensional cross-section — and the ratio of area to box grows linearly with size.

That says a larger organ, folded optimally, holds more surface per unit of itself rather than less. Which is the opposite of the familiar statement that a growing body loses surface relative to volume, and the two are not in conflict: the familiar statement is about a body that keeps its shape, and this one is about a body that refolds.

The mechanism is easy to see once it is separated out. A fold costs a fixed depth c whatever the box is, so a bigger box has room for proportionally more folds; and every extra fold adds a panel of the box’s own length. Both terms grow, and their product grows faster than either.

So folding is a technology whose returns increase with size, and the constraint on a small organ is not that it needs less area but that it has less room to buy area in. A structure a tenth of the size has a tenth of the optimal fold count and a thousandth of the peak area, against a hundredth of the box.

The caveat is important and it is the same one as before. This is the model’s area, computed in a cross-section with a uniform charge per fold; a real organ has a supply whose trunks grow with the organ, so c is not constant across sizes and the growth is slower than cubic. What survives the correction is the direction, which is that a refolding body does not run into the surface-to-volume problem the way a body holding its shape does.

Which theorem was checked and how #

The closed form is checked against a search. The area is evaluated at every whole fold count up to a bound derived from the box and the thickness, the argmax is taken, and it has to sit within half a fold of S ⁄ 2(t + δ) for every channel depth drawn. The comparison is against a formula the search never sees.

The proportional fall is checked as a ratio rather than as four separate numbers. A figure that got the charge right for one δ and wrong for another would pass a check on each and fail this one.

The zero-supply case is required to be present. The comparison is against the unserviced curve, and a figure drawing only serviced ones would have no baseline; the generator refuses a list of channel depths that does not start at nothing.

And nothing here is measured from an organ. The thicknesses and channel depths are stated in units of the box, and what is computed is the shape of the trade-off, which is the same shape at any scale.

Where the model stops #

The box is a cross-section rather than a volume. The area computed is a length times a depth in two dimensions, and a real organ folds a surface into a three-dimensional space. The parabola survives the change of dimension with different constants; the reciprocal in t + δ does not depend on the dimension.

The channel is charged per fold and a real supply is shared. Two panels lying against one another might be serviced from between them, in which case one channel does for two units of surface and δ should be halved. That is a factor of two in the coefficient and it does not change the form.

The surface is taken to be uniformly useful. A fold whose inner faces are pressed together is holding area that nothing can reach, and the model counts it — the folded state is thickest exactly where the layers pile. Correcting that is a further reduction and it acts on the same term.

And a diffusion-limited surface has an effective area smaller than its geometric one, by an amount that depends on how deep the crevasses are. That is a fourth reduction, it acts against fine folding, and it is not in this arithmetic at all.

What the picture cannot show #

None of these curves shows an organ. They show what a box of a stated size will hold, given a sheet of a stated thickness with a channel of a stated depth beside it, and the essay’s claim is about what such a curve licenses.

The underdetermination is the thing the picture most conspicuously cannot resolve. Two arrangements drawing the same curve are drawn as one curve, which is the honest rendering and is also exactly what makes the point hard to see: there is nothing on the page to indicate that the line has two readings.

Nor can a figure show what the supply is for. Area serviced at a high rate and area serviced at a low one are the same area here, and an organ trading depth for throughput is solving a problem in units this model does not carry.

The idealisation, named #

The sheet is a plane of thickness t and the channel is a slab of depth δ, both uniform, both incompressible, and both charged once per fold.

The most consequential of those is uniformity. A real lining is thicker where it is folded and thinner where it is flat, and a real capillary bed is denser where the demand is. Both variations act to make the effective t + δ larger where the folds are concentrated, which penalises fine folding more than the uniform model does.

The least consequential is incompressibility, which matters not at all: nothing in the arithmetic depends on the sheet resisting the fold, only on it occupying depth.

And the whole calculation rests on the folded footprint times the mean layer count being the area of the sheet, which is the conservation identity and is exact. Everything else here is a charge against a budget; that one is a law.

Where the ladder goes next #

The correction here is a factor and the next rung is a different one entirely, because it changes what is in the box rather than what each fold costs.

A body does not fold one surface. It folds several, in the same volume, each doing a different job — and the ceiling on a single surface goes as the square of the depth it has, so sharing the depth is not a linear cost. Two surfaces in one box works that arithmetic out and finds the total across m sharers is exactly one m-th of what one of them would have reached alone, which is a much steeper price for specialisation than anything in this rung.

The habit worth carrying is about where a correction enters. A term added to the answer can be neglected when it is small; a term added to a coefficient cannot. Supply is the second kind, and so is thickness, and so is a hinge radius — which is why all three of them turn a curve over rather than shifting it, and why a model that omits any of them is not slightly optimistic but unbounded.