Panels instead of paper

Paper cheats. Fold a bird base and watch the paper between the creases — it bows, it twists slightly, it absorbs a little of the motion by bending where no crease was drawn.

Sheet metal does not do that, and neither does a silicon solar cell, a glass panel or a rigid composite. For those, a crease pattern has to work with flat faces and hinges only, at every instant of the motion, and that is a much harder requirement.

The definition #

A crease pattern is rigid-foldable if there is a continuous motion from the flat sheet to the folded state in which every face remains planar and undeformed throughout, with all the movement in the creases.

Two words carry the weight.

Continuous — there must be a path, not merely two endpoints. A pattern whose flat state and folded state are both fine but which cannot get from one to the other is not rigid-foldable.

Throughout — every intermediate configuration must be valid. It is not enough for the faces to be planar at the ends.

Why a vertex is a mechanism #

Take a single degree-four vertex. It has four creases, so four dihedral angles. Those are not independent: the four faces must close up around the vertex at every moment, which imposes constraints.

Working through the spherical trigonometry, three constraints emerge, leaving one free parameter. Set any one fold angle and the other three follow.

That is what makes rigid origami buildable. A one-degree-of-freedom mechanism has one input and one path, and a machine with one input is something an engineer can control.

Degree-six and higher vertices generally have more freedoms, which is why almost every engineered fold pattern uses degree-four vertices exclusively.

Why most patterns fail #

The stricter condition rules out a great deal, and the reason is usually the same.

Paper accommodates by bending. A squash fold, a petal fold, a rabbit-ear — all of the standard manoeuvres of traditional folding involve a moment where the paper is neither flat nor at a crease, and the sheet is briefly a curved surface.

With rigid panels there is nowhere for that to go. The mechanism binds: the faces would have to interpenetrate or deform, and neither is available.

So the bird base is flat-foldable and not rigid-foldable. Nor is the frog base, nor the waterbomb base in general, nor essentially any traditional model. The Miura fold and the Yoshimura pattern are rigid-foldable, which is precisely why they are the ones in orbit.

The constraint count #

The one-degree-of-freedom claim can be counted rather than asserted, and doing so shows why degree four is special.

A vertex of degree $n$ has $n$ fold angles. The faces must close around the vertex, which is a condition on a spherical polygon: the $n$ arcs on the unit sphere, with the sector angles as arc lengths, must form a closed loop. Closing a spherical polygon is three scalar conditions.

So the freedom count is $n - 3$. For $n = 4$ that is one. For $n = 6$ it is three, and for $n = 8$ it is five.

That is the whole reason engineered patterns use degree-four vertices almost exclusively. A degree-six vertex is a three-freedom mechanism, which needs three coordinated inputs and can reach configurations nobody intended. One input and one path is what makes a deployable trustworthy.

Why paper is a bad guide #

Anyone whose intuition for folding comes from paper will systematically overestimate what is rigid-foldable, and it is worth knowing why.

Paper is thin enough that bending it costs very little energy — bending stiffness goes as the cube of thickness, so a sheet a tenth of a millimetre thick is a thousand times more willing to bend than one a millimetre thick. A folder pressing a base into shape is unknowingly spending that bending allowance constantly.

Watch the paper during a squash fold and the facets visibly curve. The fold works because of that curving, not despite it. Replace the paper with card and the same manoeuvre becomes difficult; replace it with aluminium and it becomes impossible.

So the traditional repertoire is a catalogue of manoeuvres that exploit compliance, and none of it transfers. Engineering had to start again from the small set of patterns that never needed the compliance in the first place.

What the compliance is worth, in numbers #

“Paper cheats” has a size, and putting one on it says why the intuition transfers so badly.

Bending stiffness goes as the material’s modulus times the cube of its thickness. For paper that is roughly $3 \times 10^{9}$ pascals against a tenth of a millimetre; for an aluminium panel, $69 \times 10^{9}$ against a millimetre. The product is $3 \times 10^{-3}$ against $69$.

An aluminium panel is about twenty thousand times stiffer in bending than a paper facet.

That is the allowance a folder spends without noticing. A squash fold works because the facets bow by a fraction of a millimetre under finger pressure; the same manoeuvre in panels would need a force twenty thousand times larger, which is not a harder version of the same operation but a different operation that nobody performs.

So the traditional repertoire is not merely inapplicable to panels. It is calibrated to a compliance four orders of magnitude away, which is why almost none of it survives the translation and why engineering had to start from the small set of patterns that never spent it.

Why degree four rather than six #

The freedom count also runs at the level of a whole pattern, and there it says something the single-vertex count does not.

A vertex of degree $n$ imposes three closure conditions on $n$ fold angles. Across a pattern where every vertex has degree $n$ and every crease is shared, the creases number $nV/2$, so the pattern’s freedom is

$$\frac{nV}{2} - 3V \;=\; V\left(\frac{n}{2} - 3\right)$$

At degree four that is $-V$: over-constrained, with $V$ more equations than unknowns. At degree six it is exactly zero — marginally rigid. At degree eight it is $+V$, one freedom per vertex.

Which inverts the usual reason #

That makes the engineering choice look stranger and better than the essay’s account suggests.

Degree four is the most over-constrained option, and a generic pattern of degree-four vertices has no motion at all. What the Miura has is not slack but a symmetry that makes $V+1$ of its equations redundant, leaving exactly one freedom — and that is a coincidence arranged rather than a property inherited.

Degree eight would give freedoms in abundance, one per vertex, with no coincidence needed. It is rejected for exactly that reason: $V$ freedoms need $V$ actuators, and a mechanism with a hundred freedoms has a hundred ways to deploy wrongly.

So degree four is chosen because it is over-constrained, not despite it. The engineering wants a system with no freedom at all and then one put back deliberately, which is a far more controlled object than one with freedom to spare.

Self-locking, and the flat state #

There is a subtlety about the endpoints worth flagging, because it catches people.

A rigid-foldable pattern’s flat state is a singular configuration — the faces are all coplanar, and at that instant the mechanism has more freedom than it does anywhere else, because there is nothing distinguishing one direction of folding from the other.

That means a rigid mechanism at its flat state can bifurcate: it can start folding in more than one way. For a deployable structure that is a hazard, and real designs avoid the fully flat state or bias it with springs.

It also means that “rigid-foldable to flat” and “rigid-foldable” are slightly different claims, and the literature is not always careful about which one is meant.

What the extra requirement buys #

Rigid-foldability is restrictive, and everything it buys follows from panels being real objects.

Thickness can be accommodated. A rigid mechanism has hinges at defined lines, and a hinge can be moved off the mid-surface to make room for the material. A pattern that relies on bending has no such option.

Stiffness is available. Rigid panels carry load. A folded paper structure is floppy; a folded panel structure can be a load-bearing shell.

The motion is predictable. One degree of freedom means one path, and a deployment that has one path either happens or does not — there is no partial state to get stuck in.

Actuation is simple. One freedom needs one actuator. Multi-freedom mechanisms need coordinated actuation, which is where deployables historically failed.

The two questions are independent #

It is tempting to think rigid-foldability implies flat-foldability or the reverse. Neither holds.

A pattern can be rigid-foldable and never reach a flat state — many deployable mechanisms fold from flat to a curved shell and stop, and there is no flat folded configuration at all.

And a pattern can be flat-foldable and not rigid-foldable, which is the common case and covers most of traditional origami.

The overlap — patterns that are both — is small, highly structured, and contains essentially everything that has ever been manufactured.

Testing it #

Deciding rigid-foldability is harder than deciding the local flat-folding conditions, and there is no equivalent of Kawasaki and Maekawa.

The standard approach is numerical. Write the constraint equations for every vertex, differentiate to get the Jacobian of the constraint system, and examine its null space: the dimension of the null space is the number of instantaneous degrees of freedom. Track that along a candidate motion.

That gives a local answer at each configuration, and stitching those into a global statement about the whole motion is delicate — the mechanism can lose or gain freedom at particular configurations, which is exactly what happens at the flat state.

So the situation mirrors flat-foldability: local questions are tractable, the global question is not, and practice proceeds by computation rather than by theorem.

Compliance, used deliberately #

The strict dichotomy between rigid panels and bending paper is softening, and the middle ground is where a good deal of current engineering sits.

A compliant mechanism allows a controlled amount of flexing and uses it as part of the motion. Rather than a hinge, a thin section of the same material bends; rather than forbidding facet deformation, the design allocates a specific amount of it and computes the consequences.

That buys two things. It removes the hinges, which are the expensive, heavy and failure-prone parts of any mechanism. And it allows patterns that are not strictly rigid-foldable, because a little bending in the right place lets the mechanism through a configuration that would otherwise bind.

The cost is that the analysis becomes elastic rather than kinematic — forces and energies rather than positions — and the clean one-degree-of-freedom statement becomes an approximation about stiffness ratios.

Most self-folding materials work this way, and so does a good deal of deployable hardware that is described as rigid origami and is not quite.

Bifurcation, and choosing a branch #

The flat state’s extra freedom is not merely a technicality; it is a design problem with standard solutions.

At the flat configuration a degree-four vertex can begin folding along either of two branches — the assignment is not yet determined by the geometry, and both are locally valid. A mechanism released from flat may pick either.

Real designs bias the choice. Pre-folding leaves a small residual angle so the mechanism starts off the singular point. Springs at the hinges push toward the intended branch. Hard stops make the wrong branch mechanically unavailable. Asymmetric hinges — different stiffness in the two directions — do it passively.

None of that is in the kinematics, and all of it is in every real device. The geometry says a path exists; making sure the mechanism takes that path and not its mirror image is engineering.

Where the two conditions overlap #

It is worth naming the intersection, because that small set is where the practical subject lives.

A pattern that is both flat-foldable and rigid-foldable can be built from panels and reaches a completely flat packed state — which is the ideal for a deployable, since flat packs best.

The Miura fold is in that set. So is the Yoshimura pattern. So are a handful of related tessellations, and essentially nothing else that anybody has found useful.

That is a very small vocabulary for a field with this much activity, and it explains why the same few patterns appear again and again in the engineering literature. It is not a lack of imagination; it is that the intersection of two strong conditions is small.

Where the model stops #

Perfectly rigid panels. Real panels flex a little, which is one of the idealisations doing useful work, and that flex is sometimes load-bearing — several “rigid” deployables rely on a small amount of compliance to pass through tight configurations.

Ideal hinges. A hinge is treated as a line about which two panels rotate freely. Real hinges have thickness, friction, backlash and a finite range.

Zero thickness, again. The kinematics are for a zero-thickness surface. Thickness breaks the mechanism and restoring it is a design discipline of its own.

No self-intersection check. The kinematic analysis says the faces stay planar; it does not by itself check that they avoid one another, which for a large pattern is a separate and expensive computation.

The figure poses one vertex. The motion drawn is computed from a single parameter, and the projection is a simple painter’s algorithm. The positions are right; the rendering does not resolve which face is in front where they nearly touch.

What a folded panel structure is good for #

Beyond deploying, rigid folding gives a way of making stiff shells out of flat stock, and that is a manufacturing argument rather than a kinematic one.

Flat material is cheap, arrives in sheets, and can be cut, printed and populated while flat. A structure that is assembled flat and then folded into shape inherits all of that — which is why folded construction turns up in electronics packaging, in furniture that ships flat, and in microscale devices where planar processing is the only processing there is.

The folded state can also be far stiffer than the flat one. A corrugated sheet resists bending along one axis enormously better than the flat sheet it came from, at the same weight, and every cardboard box is an application of that.

So there are two distinct reasons to fold something rigid: because it has to get smaller, and because folding is a way of making a stiff three-dimensional object out of a flat cheap one. The second is quieter and much more widespread.

The vocabulary is small, and that is the finding #

A striking feature of the engineering literature, and worth stating as a result rather than as an observation.

Almost every folded structure that has been built uses one of a handful of patterns: the Miura fold, the Yoshimura pattern, the waterbomb tessellation, a plain accordion, or a wrapping fold.

That is not a failure of imagination. It is the intersection of several strong conditions. A useful pattern must tile, must be rigid-foldable, must have one degree of freedom, must pack well enough to be worth the trouble, and must tolerate thickness accommodation. Each condition removes most of what survives the previous one.

So the field’s effort goes into making those few patterns work in real materials rather than into finding more, and the research output is overwhelmingly about hinges, thickness and manufacturing rather than about geometry. That is what a mature engineering subject looks like, and it arrived about twenty years after the geometry did.

Checking it computationally #

There is no rigid-foldability analogue of Kawasaki, so the practical test is numerical, and knowing its shape explains what can and cannot be claimed.

Write the closure constraints at every vertex — the spherical polygon conditions — as a system of equations in the fold angles. Differentiate to get the Jacobian. The dimension of its null space at a given configuration is the number of instantaneous degrees of freedom there.

Track that along a candidate motion and a picture emerges: a mechanism with one freedom throughout is rigid-foldable along that path; one whose null space collapses somewhere binds there; one whose null space jumps has a bifurcation.

What that does not give is a global statement. It is a local test repeated along a path that was guessed, and a pattern could have a valid motion the search never found. Like flat-foldability, the local question is tractable and the global one is not.

The ladder from here #

Later rungs: the degree-four constraint equations derived. The Jacobian test and instantaneous degrees of freedom. Bifurcation at the flat state. Rigid-foldability of the Miura, proved. Patterns that are rigid-foldable but not flat-foldable. Thickness accommodation techniques. Actuation and self-folding. Compliant mechanisms, where a little bending is allowed on purpose. And the computational question of deciding rigid-foldability, which is open in general.

Almost every traditional origami model is impossible to build out of anything stiffer than paper, and almost every folded structure that has flown is a tessellation of degree-four vertices. The two populations barely overlap.