No motor in the fold

The packing number says a wing can be made small. It says nothing about how it gets large again, and that turns out to be the harder requirement and the one that decides the pattern.

An insect has muscles at the base of the wing. It does not have a muscle at every crease, or at any crease out along the wing’s length, and the folds are tens of millimetres from anything that can pull. So whatever happens when the wing deploys, the pattern is doing almost all of it.

What a degree of freedom is, precisely #

The phrase is used loosely enough to be worth pinning down, because the whole argument rests on it.

Take a crease pattern with rigid panels and hinges along the creases. A configuration is an assignment of a fold angle to every crease such that the panels still meet correctly — every closed loop of creases around an interior vertex has to come back to where it started, which is a set of equations rather than a free choice.

The number of degrees of freedom is the dimension of the set of solutions to those equations. One degree of freedom means the solution set is a curve: fix any one fold angle and every other angle is determined. Two means a surface, and so on.

The crease count and the freedom count have almost nothing to do with each other. A pattern with two hundred creases can have one freedom, and a pattern with eight creases can have four.

That independence is worth sitting with, because the intuition it violates is strong. More parts normally means more ways to move: a chain of twenty links has more freedom than a chain of five, and a machine with more joints is generally a more complicated machine. Folding inverts this because every crease is shared between two panels and every interior vertex imposes closure conditions, so adding creases adds constraints at very nearly the rate it adds unknowns. Whether the balance comes out at one, zero, or several is a property of the arrangement rather than of the count.

The consequence for reading a crease pattern is that its complexity is not legible from its appearance. A dense pattern of hundreds of creases may be a single-freedom mechanism and a sparse one may be rigid, and there is no way to tell by looking. The count has to be computed, which is what the census below does.

Why one is the number that matters #

A mechanism with n degrees of freedom needs n inputs to be put into a specified configuration. That is not a rule of thumb; it is what the dimension of the solution set means.

So a pattern with one freedom needs one input. Pull anywhere, in any way that changes the state at all, and the entire surface moves along its single available path. There is no sequencing, no coordination, and no possibility of one region opening while another stays shut.

That is the property that makes a passive deployment possible, and it is the same property whether the deployment is done by a muscle, a spring, a pyrotechnic release or a growing leaf. It is why the Miura ended up on a solar array, why a sheet with one freedom is one of the first things this site established, and why the requirement keeps arriving in this field from different directions.

Where the single freedom comes from #

It is worth seeing why a degree-four vertex has one freedom, because it is a small enough calculation to describe and it is the atom the rest is built from.

A degree-four vertex has four fold angles. The condition that the panels close around the vertex is three equations — the composition of four rotations about the crease directions has to be the identity, which is three constraints in three dimensions. Four unknowns, three equations, one dimension of solutions.

This site computes that motion twice with code that shares nothing: once by enumerating letter assignments against the local theorems, and once by solving the vertex as a closed spherical linkage in three dimensions. foldcheck requires the two to agree on the identity of all four flat-foldable assignments and not merely on the count.

Why a grid of them is still one #

The step that surprises people is that assembling many single-freedom vertices does not accumulate freedoms.

Each vertex contributes one freedom and each shared crease contributes a constraint — the two vertices at its ends must agree about its fold angle. For the Miura’s arrangement those constraints are exactly enough to collapse the whole grid onto a single curve, and the pattern moves as one object however large it is.

That is a fragile property rather than a generic one. Most ways of arranging degree-four vertices give a pattern that is rigid — zero freedoms, no motion at all — and the Miura’s arrangement is special. The site has a whole essay on the state a pattern can have and never reach, which is the same fragility seen from the other side.

What many freedoms costs #

The right-hand row of the census is not a real pattern. It is what the same number of vertices would need if nothing coupled them, and it is there because the cost of extra freedoms is easy to underestimate.

Each freedom needs a driver, and each driver is mass, a power connection, a control input, and a way for the deployment to fail. Twenty drivers on a spacecraft is not twenty times the problem of one; it is a different kind of problem, with a sequencing controller and twenty single-point failures.

The biological version of the same arithmetic is more severe still. An insect cannot put an actuator anywhere except where there is already a joint with muscle attached, and a wing membrane is not somewhere muscles can be. So the number of available drivers is not a design variable at all — it is one or two, fixed, and the pattern has to work with that or not be used.

The generic count, and how far the Miura beats it #

The fragility can be counted rather than asserted, and the count is what makes “a fragile property rather than a generic one” a measurement.

Each interior vertex imposes three equations — the composition of rotations round it must be the identity, which is three conditions in three dimensions — and each crease is one unknown. So a pattern with $E$ creases and $V$ interior vertices has, generically,

$$\text{degrees of freedom} = E - 3V$$

For an $n$ by $n$ grid of quadrilaterals, $E = 2n(n-1)$ and $V = (n-1)^{2}$, which gives $(n-1)(3-n)$.

One at $n = 2$, zero at $n = 3$, minus three at $n = 4$, minus fifteen at $n = 6$. A single vertex has its freedom; a two-by-two block of them is already rigid; anything larger is over-constrained, meaning the equations have no solution but the flat state unless they are dependent.

Which says how much coincidence a Miura is #

A Miura of the same size has one degree of freedom at every $n$. Subtracting, the number of its closure equations that must be redundant is

$$1 - (n-1)(3-n) = (n-2)^{2}$$

Four redundancies at a four-by-four, nine at five, sixteen at six — one for every interior vertex of the sub-grid two rows in from the edge.

That is what the Miura’s arrangement is buying, and it says why perturbing one vertex destroys the motion outright rather than degrading it. The redundancies are exact algebraic coincidences, sixteen of them at a modest size, and a coincidence does not survive being nudged: break one and the count reverts toward the generic figure, which at that size is minus fifteen — no motion at all.

So the essay’s “generic outcome is not one freedom becoming two” is not merely an observation about what happens to be common. The generic count is deeply negative, so the generic outcome is rigidity, and every degree of freedom a large folded sheet has is a coincidence somebody arranged.

It also prices the alternative the census’s right-hand row stands for. Uncoupled vertices have $V$ freedoms because they have no shared creases and therefore no closure equations tying them; coupling them is what makes the count fall from $V$ to something negative, and the Miura is the arrangement where it falls to exactly one instead.

Where the freedom goes when the pattern is slightly wrong #

The fragility is worth a paragraph of its own, because it is the practical difficulty and it is invisible in a count that comes out right.

Perturb a Miura — move one vertex slightly, make one row a different depth — and the constraints stop being consistent. The generic outcome is not “one freedom becomes two”; it is that the solution set becomes a single point, the flat state, and the pattern has no motion at all. A pattern that is nearly a Miura is usually rigid rather than nearly as good.

That is exactly the situation the tapered corrugation runs into one ladder over, where the row heights turn out to be bound and the column widths free. The parameters that can be varied without destroying the property are a small subset of the parameters a designer would like to vary, and finding out which is which is the actual design work.

For a biological structure the same fragility has a different consequence. A pattern that only works at exact proportions is not a pattern that develops reliably, so the folds that survive in organisms are the ones with tolerance — which is an argument for the plain corrugation and against anything requiring four angles to agree.

The alternative nobody uses #

There is a way to deploy a many-freedom pattern with one driver, and noticing why it is rare is instructive.

A pattern with many freedoms can be given a mechanism — gears, linkages, cables — that ties its freedoms together so that one input drives them in a fixed relationship. That works and is done in engineering, and it replaces a geometric property with a piece of hardware.

The trade is bad in almost every case. The mechanism has mass, it has to be routed across the deployable, it has its own failure modes, and it has to survive the same environment. A pattern whose freedoms are already one is getting the same result from its geometry for nothing.

There is also a scaling argument against it that is decisive on its own. A coupling mechanism has to reach every freedom it ties together, so its size grows with the deployable. A geometric constraint does not — the Miura’s single freedom holds at six panels and at six thousand, with no additional anything. Any approach whose cost grows with the surface loses to one whose cost does not, as soon as the surface is large, and a deployable that was not going to be large would not have needed folding.

The exception, and it is a real one, is where the motion has to be controlled rather than merely allowed. A single freedom gives one path and no way to stop partway with any precision, and applications that need a specific intermediate state rather than open-or-shut do end up adding hardware. Nothing biological in this field has that requirement, which is part of why the geometry is enough.

The same count, arriving from three directions #

This is the third time in this field that a single degree of freedom has turned out to be the requirement, and the three routes to it are worth putting together because they are genuinely independent.

A leaf opens under growth, which supplies a single increasing scalar and nothing else. If the pattern had two freedoms, growth would specify a direction through a two-dimensional space and there is nothing available to choose one.

A wing opens under a muscle at its base, which supplies one input at one place. If the pattern had two freedoms, the muscle would drive one combination and leave the other undetermined, and part of the wing would be limp.

A solar array opens under a single release, which supplies one impulse. The same argument again.

Three actuators with nothing in common — a growing tissue, a muscle, a pyrotechnic latch — and one geometric requirement, because the requirement is about the pattern rather than about the thing pulling it. That is the sense in which this is a result rather than three observations.

It also explains a fact that would otherwise look like fashion. The Miura appears in a buckling shell, in leaves, in packaging and in spacecraft, and the common thread is not that anybody copied anybody. It is that the number of patterns with this property is small, so independent searches keep landing on the same few.

What a wing does that this does not model #

Two departures, and they run in opposite directions.

A real hindwing is not rigid-panelled. Substantial parts of it bend, and some of what looks like folding is elastic deformation of the membrane between veins. A model made of rigid panels and ideal hinges gets the count right for the fold pattern and misses everything the membrane contributes.

And a real wing deployment is understood to involve stored elastic energy and a snap rather than a slow quasi-static motion. Structures that snap have a configuration space with more than one branch and pass between branches quickly, which is a phenomenon a degrees-of-freedom count is not equipped to describe at all. The count says how many inputs a slow motion needs; a snap is not a slow motion.

So the honest statement is narrow: if the wing’s fold behaved as a rigid-panel mechanism, one freedom is what would let a base-mounted muscle deploy it. That the real thing is more interesting than the model is a fact about the model.

What the picture cannot show #

The census is a table of counts. It has no geometry in it beyond what the counts were computed from, and it says nothing about the shape of the one-dimensional motion.

That shape matters and is invisible here. A single-freedom pattern whose motion passes through a configuration where panels intersect is unusable even though its count is right. A single-freedom pattern whose motion requires enormous force near one end of its range is unusable for a different reason. Neither is a count.

The related quantity the count also hides is stiffness transverse to the motion. A deployable wants one soft direction and every other direction stiff, which is a statement about a matrix rather than about its rank, and it is the property that decides whether the deployed surface holds its shape.

The idealisation, named #

Rigid panels, ideal hinges of zero width, and no self-contact anywhere.

The zero-width hinge is the one this ladder takes up next, and it is not a small correction for an insect. A compliant hinge has a length, that length is not available for packing, and the ratio of hinge length to panel length in a real wing is nothing like the ratio in a sheet of paper.

Self-contact is the other one. The count assumes panels can pass through each other, because the equations that give the freedom are about closing loops rather than about occupancy. A real motion has to avoid collisions, and a pattern’s usable range is often much shorter than its kinematic one.

Where this ladder goes next #

The count is the property that makes the deployment possible and the packing number is what it is worth. What remains is the correction that makes both of them optimistic.

Nothing in a body folds on a line. Every hinge has a radius, the radius consumes surface, and the consumption is exactly proportional to the number of folds — so the finer patterns that pack best on paper are the ones that lose most to their own hinges. That is the last rung of this ladder and it is where the geometry hands back to the material.

Sideways from here, the same count is what the rigid-folding field of this site has been measuring all along, and reading that field again with a wing in mind changes what its results are about. A pattern’s mobility stops being a curiosity about paper and becomes the answer to a question an animal has to solve every time it takes off.