The channel grows with what it feeds

Standing up beats lying down by eight ends by naming the assumption the whole factor rests on: a wall’s members do not share the budget that pays them. A wall’s height is the clearance whatever the wall count is, so walls compete only for footing, and the quadratic penalty a stack of plies pays disappears.

A body cannot keep that assumption, because a surface has to be reached as well as fitted. Whatever feeds a wall runs beside the wall, which is to say it is charged against the pitch — the one budget the comb was winning on. And the amount of feeding needed is not a constant.

A wall serves more the taller it is #

The asymmetry is in one sentence and everything follows from it.

A wall of height $c$ carries two faces, so it serves $2c$ of surface per unit of its length. Whatever supplies that surface — a vessel, a duct, a nerve, a coolant line — has to be sized for what it serves, so its width is some share of it: $\sigma = \beta c$. It runs beside the wall, so the pitch cannot be less than $\tau + \beta c$, and the comb reaches

$$f_{\text{walls}} = \frac{2c}{\tau + \beta c} \longrightarrow \frac{2}{\beta}$$

as the clearance grows. A supplied comb has a ceiling that no clearance can pass. Doubling the depth doubles both the surface a wall carries and the channel it needs, and the two divide out.

A ply is in a different position. It serves a fixed area — the base, once — however deep the stack is, so its channel is a constant $\beta$ rather than a growing one, and it is charged against the clearance alongside the ply’s own thickness. That is exactly the accounting the surface has to be supplied already established, where supply and sheet enter the optimum on identical terms:

$$f_{\text{plies}} = \frac{c}{4(\tau + \beta)}$$

which keeps growing linearly for ever.

The table is the saturation happening. At a wall height of 0.25 the channel is 0.0125 against a sheet of 0.01, so the pitch is a little over twice the sheet and the comb reaches 22.2. At a height of 4 the channel is 0.2 — twenty times the sheet — the pitch is essentially all channel, and the comb reaches 38.1. Everything above a certain height is the walls carrying plumbing rather than surface.

Where the crossover is #

Setting the two expressions equal and cancelling gives a clearance in closed form:

$$c^* = \frac{7\tau}{\beta} + 8$$

which is pleasant enough to be worth checking rather than trusting. At a sheet of 0.01 and a channel a twentieth of what it serves, the crossover is at 9.40 — where both architectures reach about 6.7 times the base, the comb with its ceiling of 40 nine-tenths spent and the stack still climbing.

The crossover moves surprisingly little. A channel a fiftieth of what it serves puts it at 11.5; a channel a fifth puts it at 8.35. Ten times the plumbing cost moves the crossing by a quarter, because the $8$ in the formula dominates once $7\tau/\beta$ is small — and $7\tau/\beta$ is small whenever the channel is wide compared with the sheet, which for anything that has to be fed it is.

So the crossover is at a clearance of a little over eight times nothing in particular. It is not a length; it is a pure number, because the clearance in the formula is measured in the same units as the sheet’s thickness and the channel’s share is dimensionless. A structure is a comb or a stack according to whether its clearance is more or less than about eight, and eight is measured in sheet thicknesses only through the small correction. That is a much sharper statement than the comparison it comes from.

The ceiling matters more than the crossing #

The crossover is the headline and it is the less useful of the two results, because a structure near it is getting very little from either architecture.

At the crossover both architectures deliver about 6.7 times their base. The comb’s ceiling is 40. So a comb working well is working at a clearance far below the crossover, where it is holding four or five times what the stack would — and a comb working near the crossover has already given up five sixths of what it could reach and is about to be overtaken.

That reframes what the number is for. The crossover does not mark where a body should switch architecture; it marks where a body has already lost the argument for combs. A structure that needs more surface than $2/\beta$ times its base cannot get it from standing walls at any depth, and its only remaining move is the one the stack offers: go deeper, and pay the constant channel per ply.

Two ways to be short of surface #

Put the two failure modes side by side, because they ask for opposite remedies.

A stack that is short of surface is short because its clearance is small: the ceiling is linear in the clearance, so the remedy is depth, and the cost of depth is whatever else wanted the same depth. Two surfaces in one box prices that exactly — $m$ surfaces sharing one depth reach between them one $m$-th of what one of them would have reached alone — so a body adding depth for one surface is taking it from another on very steep terms.

A comb that is short of surface is short for a different reason and depth will not help it. Its ceiling has no clearance in it at all. The only lever is $\beta$ — the share of the served surface that the channel takes — so the remedy is a better-plumbed wall: a narrower vessel for the same flow, a shorter diffusion path, a supply that branches rather than one that runs straight. That is the opposite of the lever four materials, four optima identified for a folded sheet, where the material’s hinge radius set the best fold count and nothing about the supply entered; a comb has no hinge and no fold count, and the quantity that decides it is the one that essay had no reason to name. A comb improves by improving its supply and a stack improves by finding room, and no amount of either does the other’s job.

That is a testable difference in principle. A lineage under selection for more surface should respond in the two cases in visibly different ways — deeper in one, better-plumbed in the other — and which way it responds says which architecture it is in without anyone having to decide from a picture.

The same eight, twice, for different reasons #

The crossover’s leading term is 8 and the unsupplied advantage is 8, and it is worth saying that these are the same number arriving twice rather than a coincidence.

The advantage is $8 = 2 \times 4$: two faces on a wall, four for the parabola a stack’s plies pay. The crossover is where a growing charge has eaten that advantage, and since the charge enters the comb’s denominator in proportion to the clearance, the clearance at which it has eaten a factor of eight is eight — plus a correction for the sheet, which is the $7\tau/\beta$. The eight in the crossover is the eight in the advantage, spent.

That makes the formula easier to trust and easier to use. A structure whose clearance is much less than eight of anything is deep in comb country; one whose clearance is much more is in stack country; and the sheet’s thickness only matters through a term that is negligible for anything whose plumbing is wider than its wall.

What a lining looks like from here #

Nothing in this account measures an organ, and the field’s rule is that a figure here draws a geometry this repository computed rather than a specimen. What the model does give is a reading — a way of looking at a described structure and saying which economy it is in.

A structure with short members standing off a base, each carrying its own supply, is a comb near the bottom of its range and doing well: it is getting most of a factor of eight and has not yet spent it. A structure with tall members is a comb at its ceiling, where further height buys almost nothing, and the honest description is that it has stopped being a way of getting surface and become a way of getting reach. And a structure that has abandoned standing members for layers is one whose surface requirement passed $2/\beta$ — which the model says can happen for no reason except that the requirement grew.

How much surface fits in a body opened this line of argument with the observation that folding buys unlimited surface only for a sheet of no thickness. The supplied comb adds a second and more awkward limit: even with no thickness at all, a supplied comb’s ceiling is $2/\beta$, because setting $\tau$ to zero removes the sheet from the pitch and leaves the channel. Thickness turned the first curve over; supply caps the second outright, and it does so without any help from the material.

It also puts a boundary where nothing grown has a seam says one belongs. A comb’s walls each meet the base along a line, and that line is where the channel enters. A stack’s plies do not; their supply crosses the clearance to reach them. So the two architectures differ not only in what they are charged but in where their plumbing has to be, and a structure grown rather than assembled has to route it along surfaces that already exist.

What a nest does to a supplied structure #

Nesting was already a losing move for packed levels and a wash for combs, and supply does not rescue it.

Every packed level costs a factor of four unsupplied. Supplied, each level’s members carry their own channels, so the effective thickness at every level is the sheet plus the channel and the loss per level becomes $4(\tau+\beta)/\tau$ — larger, and larger by the same factor at every level, so the nest falls off faster and still telescopes. Nothing about supply changes the sign of that argument.

And the unsupplied comb-in-comb result — that a nest creeps up to a single comb and never reaches it — survives for the same reason. The inner combs are the ones whose walls grow shortest, so their channels are the narrowest, but the outer comb’s walls are the ones carrying every inner wall’s supply as well as their own. Whatever the accounting does in detail, it cannot turn a bound into a gain.

The eight, revised #

Standing up beats lying down by eight ends with the factor stated in its bare form and a warning that supply had not been charged. Charged, the factor becomes

$$\frac{f_{\text{walls}}}{f_{\text{plies}}} = \frac{8(\tau+\beta)}{\tau + \beta c}$$

which is eight when $\beta = 0$ and falls through one at the crossover. The comb’s advantage is therefore real, large, and spent: it is large exactly where the clearance is small, which is where neither architecture holds much surface in absolute terms.

So the two essays together say something narrower and more useful than either alone. Standing the members up is worth a factor of eight in the geometry and buys a ceiling instead of a slope in the plumbing. A shallow surface that is expensive to supply should be a comb; a deep surface that is cheap to supply should be a stack; and a structure that wants a very large multiple of its base has no choice but the second, whatever the geometry says.

What the curves cannot show #

The supply model is one parameter and a body’s is not.

It cannot show branching. A channel that splits as it rises serves a wall’s upper surface with less cross-section at the top than at the bottom, so the constant share $\beta$ is a worst case, and a well-branched wall behaves as though $\beta$ were smaller in a way the model has no term for. Every real vascular supply branches, so every real comb sits somewhere between this account and a more generous one.

It cannot show that a channel is inside the wall rather than beside it. Where the supply runs within the member — which a villus’s core is — it is charged against the member’s own thickness rather than against the pitch, and the arithmetic becomes the stack’s rather than the comb’s. That is the same charge in a different budget, and it is the single modelling choice the whole crossover turns on.

And it cannot show what the surface is for. Area is being counted as though every square of it were equally useful, and an organ’s surface is not: a gut absorbs different things at different rates along its length, a gill’s efficiency depends on the flow past it, and a heat exchanger’s surface is worth what the temperature difference across it is worth. None of that is geometry.

What the model assumes #

A channel is a fixed share of the surface it serves. That is the crudest possible sizing rule. It has no transport law in it, no pressure drop, no diffusion length, and no dependence on what is being carried.

A wall’s channel runs beside it and a ply’s runs across the clearance. This is the asymmetry the whole result is, and it is a modelling choice rather than a derivation.

Supply and structure are charged to the same budgets as before — pitch for the comb, clearance for the stack — with no third budget for junctions, roots or the plumbing that connects the channels to each other.

And every member is supplied independently. A comb whose walls share a channel between each pair would halve $\beta$ exactly and move the crossover accordingly; nothing here forbids it and nothing here computes it.

How the numbers were checked #

The crossover is checked against both curves rather than derived and drawn. At each of the four channel widths the two expressions are evaluated at $7\tau/\beta + 8$ and required to agree to within a part in a billion of the comb’s ceiling, so an algebraic slip in either would show up as a crossing in the wrong place.

The saturation is checked as an approach, not as a limit. The figure requires the supplied comb to have reached between ninety and a hundred per cent of $2/\beta$ by the right-hand edge — enough to show the flattening, not so much that the curve is a horizontal line pretending to be a result.

The monotone direction of the crossover in $\beta$ is required, so a fatter channel must bring the crossing in. That is the claim the whole comparison rests on and it is checked on the computed values rather than read from the drawing.

And the supplied share of the bare ceiling is required to fall at every wall height, which is the statement that the channel is the term growing with height and the sheet is not.

Still open: a channel that knows where it is going #

The model’s sizing rule is the weakest thing in it, and the improvement is a well-posed problem rather than a wish.

A real supply is a tree: a trunk at the base of the wall, branches into the wall, and twigs at the surface. If the tree carries a flow proportional to the area it serves and each branching obeys a fixed rule, its cross-section at height $h$ is set by the surface above $h$ rather than by the whole wall — so the channel tapers, its average width is a fraction of $\beta c$, and the comb’s ceiling rises by the reciprocal of that fraction. How large a fraction is a question about the branching rule, and it decides whether a comb’s ceiling is a real limit or an artefact of assuming a pipe.

The computation needs one thing this account does not have: a law relating a channel’s cross-section to the flow it carries. With one, the taper follows, the average follows, the ceiling follows, and the crossover moves to wherever it belongs. Without one, everything above is a bound rather than an estimate — which is the honest description of it.

Sideways from here, the same question is being asked of a fold rather than of a wall. A nest pays four a level ends by pointing at the dead fraction a level carries upward, and a tapering supply is precisely a dead fraction that shrinks with height. Whether the packed nest’s factor of four survives a tapering supply is the matching computation on the other architecture, and neither is harder than the other.

The habit worth carrying is about what a cost is proportional to. When a benefit and its cost grow with the same variable, the ratio has a ceiling and the variable is not a lever. Finding the ceiling is then more useful than finding the crossover, because a structure anywhere near the crossover has already spent most of what the arrangement could give it.