Folding that gets built

There is a class of engineering problem with a characteristic shape: something must be large when it is working and small when it is being carried, and the transition has to happen once, correctly, without supervision.

Folding answers it, and the reason is not primarily the packing ratio.

The four canonical cases #

Solar arrays. A spacecraft’s power scales with array area, and the array has to fit inside a fairing. The Miura-folded array on Japan’s Space Flyer Unit in 1995 is the origin story, and the pattern has been used repeatedly since.

Airbags. Packed for a decade in a steering wheel, inflated in thirty milliseconds. The fold pattern determines the order in which the bag emerges, which determines whether it inflates toward the occupant correctly or catches on itself.

Stents. A mesh tube threaded through an artery at a few millimetres diameter and expanded in place. Kaori Kuribayashi’s origami stent graft uses a modified Yoshimura pattern for exactly the radial collapse behaviour that pattern has.

Starshades. A telescope occulter tens of metres across, launched in a tube two and a half metres wide, which must deploy to a precise shape because its purpose is a sharp shadow.

The ratio, and what it means #

The bar chart states packing ratios and they deserve unpacking, because they are ratios of area rather than of length and that is a large difference.

A Miura array packing to six per cent of its deployed area is contracting by about a factor of four in each direction. That sounds modest and is not, because the two contractions multiply — which is the whole point of the negative Poisson’s ratio.

An airbag at four per cent is doing something different again: it is not a folded surface reaching a flat packed state but a fabric bag folded into a cavity, and the ratio is volumetric.

A starshade at nine per cent is limited by the fairing diameter rather than by the fold — a 26-metre disc into a 2.5-metre tube is a factor of ten in diameter, and the fold pattern has to achieve it while keeping the deployed edge accurate to a fraction of a millimetre.

So the ratios are not comparable to one another in any strict sense. What they share is the shape of the requirement.

In the units a fairing is specified in #

Ratios of area are the wrong currency for three of these four cases, and converting them says which.

Linear contraction is the square root of the area ratio, so the bar chart reads: the Miura array at 6% is 4.1× in each direction, the airbag at 4% is 5.0×, the starshade at 9% is 3.3×.

The starshade is where that conversion breaks, and instructively. Its real requirement is a 26-metre disc into a 2.5-metre fairing — a linear factor of 10.4, which for an isotropic fold would be an area ratio of 0.9% rather than 9%.

The discrepancy is not an error in either number. It is the signature of a fold that is deliberately anisotropic: the starshade does not pack into a small disc, it packs into a tall bundle, spending length to buy diameter because length is the dimension the fairing is generous about. The quoted area ratio is a summary of a shape that was never chosen for its area.

Which is the general form of the mismatch #

That is worth generalising, because it is what makes these four numbers incomparable in the way the essay says they are.

A fairing constrains two extents, and a packing ratio is their product. A pattern can satisfy the product while failing an extent, and a pattern can miss the product badly while fitting easily — the accordion is the standing example, packing eight-fold by area while leaving one dimension at full length.

So the honest specification for a deployable is never a percentage. It is a stowed diameter and a stowed length, checked separately, and the ratio is a number written afterwards for people who want one figure. Every one of these four programmes was specified the second way and is reported the first.

What folding is bad at #

The list of applications is impressive and it is worth stating the complement, because folding is not a general answer.

Repeated cycling. A crease is damage. Fold and unfold a sheet a thousand times and the crease fails. Deployables that must operate repeatedly use hinges and linkages, not folds.

Load-bearing in the packed state. A folded structure is compliant along its degree of freedom, which is exactly what makes it deployable and exactly what makes it useless as a structural member while packed.

Precision in the deployed state. A fold pattern deploys to a shape determined by its geometry and its hinges, and hinge tolerances accumulate. Anything needing optical precision — a mirror, rather than a shade — generally cannot be folded and is segmented instead.

Thick material. Thickness breaks the mechanism, and past a certain thickness-to-span ratio no accommodation technique recovers it.

Why one degree of freedom is the product #

The packing ratios are impressive and they are not the hardest part. Several folding schemes pack better than the Miura.

What matters is that a one-degree-of-freedom mechanism has exactly one path. Push it and it goes along that path or it does not go at all. There is no partial configuration that is not on the way to the full one, and nothing to jam in.

A mechanism with many freedoms can deploy incorrectly — bind, deploy partially, or find a configuration nobody anticipated. For a satellite there is no second attempt and no engineer present.

So the reliability is the product, and the packing ratio is what makes the reliability affordable.

Actuation follows from the freedom count #

One degree of freedom needs one actuator, and that is a large practical simplification.

A multi-freedom deployable needs coordinated actuation — several motors that must move in a defined relationship, with the failure of any one leaving the structure in an undefined state. That is where deployable structures historically failed, and the failures were expensive.

A single-freedom mechanism can be driven by one motor, by a spring, or by the release of stored strain energy. Some designs use nothing at all: the packed state is held by a restraint, and cutting it lets the structure spring open along its only available path.

Self-folding, and folding without hands #

At small scales there are no hands, and the fold has to happen by itself. That has produced a distinct branch of the subject.

The approach is to build the actuation into the material. A sheet with a shape-memory polymer at the crease lines folds when heated. A sheet with differential swelling layers folds when wetted. A sheet with printed resistive traces folds when current is applied.

In every case the crease pattern is the design and the material is the motor, and the mechanism must be rigid-foldable because the panels are stiff by construction.

This is where microscale folding lives — assembling three-dimensional structures out of flat lithography, which is the only way to make anything three-dimensional at that scale, since lithography is inherently planar.

What each application constrains #

The pattern is chosen for the application, and the constraints differ sharply.

Solar arrays want maximum packing and absolute reliability, and can tolerate a slow deployment. They also want the deployed surface flat, which rules out patterns that leave residual curvature.

Airbags want a specific deployment order rather than a good ratio, and the timescale is milliseconds. The fold is designed around the sequence in which the fabric leaves the housing.

Stents want radial collapse, biocompatibility and a smooth outer surface, and the material is a metal mesh rather than a sheet. Layer count is irrelevant and surface finish is critical.

Starshades want the deployed shape accurate to a fraction of a millimetre across tens of metres, because the optical performance depends on the edge profile. Packing is almost secondary.

So there is no general-purpose deployable fold. There is a small vocabulary of patterns and a lot of application-specific selection.

Thickness is the recurring obstacle #

Every one of these is built from something with a depth, and the zero-thickness assumption fails immediately.

Solar panels are several millimetres thick and rigid. Airbag fabric is thin and there are many layers. Stent struts are hundreds of microns. In each case the crease pattern has to be modified — offset hinges, membrane hinges, tapered panels — before anything can be made.

That is why the engineering literature on origami is so heavily weighted toward thickness accommodation. The geometry is understood; getting a real material to follow it is the work.

The scale range is unusual #

One thing worth noticing about this list: it spans about six orders of magnitude.

A microscale self-folding device is measured in tens of microns. A stent is millimetres. An airbag is tens of centimetres. A solar array is metres. A starshade is tens of metres.

The same geometry serves all of them, because the constraints are geometric rather than physical — the sheet does not stretch, the paper is conserved, the mechanism has one freedom. None of that has a length scale in it.

The things that do have a length scale — thickness, gravity, actuation, material behaviour — change completely across the range, and they are exactly the things the crease pattern does not describe.

Testing something that works once #

An engineering problem specific to deployables, and one the geometry contributes nothing to.

A deployment mechanism operates a single time, in an environment that cannot be reproduced, after years of storage. Every test is therefore a test of a different article than the one that will fly, because testing consumes it.

The responses are all statistical or indirect. Test many articles and infer reliability. Test sub-assemblies rather than the whole. Test in gravity offload rigs that approximate weightlessness badly. Model everything and validate the model on the tests that were possible.

That is where the single degree of freedom earns its keep a second time. A one-path mechanism has far fewer states to test, and a model of it is far more credible, than a mechanism that could deploy several ways.

Storage, and what a crease does over years #

Something the crease pattern says nothing about: a fold held for a long time is not the same as a fold just made.

An airbag sits folded for a decade at temperatures ranging over eighty degrees, and the fabric takes a set. A solar array is stowed for the launch campaign and the cruise. In both cases the material relaxes, the creases sharpen, and the friction between layers changes.

That can help or hurt. A set crease deploys more readily along its intended line; but a fabric that has taken a permanent set may resist unfolding where it has bonded to itself, and airbag design includes surface treatments specifically to stop layers sticking.

None of this is geometry, and all of it decides whether the geometry gets a chance to work.

The pattern is the smallest part #

Worth stating plainly at the end of a page about applications: the fold pattern is a small fraction of a deployable structure.

A solar array is panels, hinges, latches, dampers, a restraint system, release actuators, harness routing, thermal control and a test programme. The crease pattern determines the kinematics and essentially nothing else.

What the pattern buys is that the kinematics are simple — one input, one path, a predictable sequence — which makes everything else tractable. A structure with complicated kinematics needs complicated everything else.

So the right claim for folding in engineering is modest and real. It does not build the structure. It makes the structure’s motion something an engineer can reason about, and that is why it keeps being chosen.

Where the model stops #

Geometry, not engineering. Everything on this page is about the fold pattern. An actual deployable is mostly latches, restraints, dampers, thermal design and testing, none of which the geometry touches.

Packing ratios are approximate. The figures quoted are representative rather than measured from any particular device, and they vary considerably within each category.

Deployment is dynamic. The kinematics give a sequence of shapes. A real deployment has momentum, and several failures have been dynamic rather than kinematic.

The figure is a comparison, not data. The bars show orders of magnitude to make the point that the ratios are large. They are not a survey.

Nothing here is about cost. A folded deployable is usually chosen because nothing else fits, not because it is cheap, and the manufacturing complexity is substantial.

The pattern vocabulary is tiny #

A last observation, and the most surprising thing about the engineering literature.

Essentially every folded structure that has been built uses one of about five patterns: the Miura fold, the Yoshimura pattern, the waterbomb tessellation, a simple accordion, or a wrapping fold. The variations are in the thickness accommodation, the actuation and the materials.

That is not for want of alternatives. There are infinitely many flat-foldable tessellations. What narrows it is the conjunction of requirements: the pattern must tile, must be rigid-foldable, must have one degree of freedom, must pack well, and must tolerate thickness accommodation.

Each condition is a strong filter and they are close to independent, so the survivors are few. The field spends its effort on making those few work in real materials rather than on finding more, which is a reasonable allocation and slightly disappointing to anybody who came for the geometry.

Why it took twenty-five years #

The Miura fold was published in 1970 and flew in 1995, and the gap is the whole story of turning geometry into hardware.

The geometry was finished immediately; nothing about the pattern changed in that time. What took a quarter of a century was everything else: qualifying panel hinges for the thermal cycling of orbit, designing a restraint that holds through launch vibration and releases reliably, proving the deployment in gravity when the mechanism was designed for none, and building confidence in a mechanism that operates once.

That ratio — a few years of geometry, decades of engineering — recurs. Origami stents were proposed in the early 2000s and are still not routine. Self-folding materials have been demonstrated in laboratories for fifteen years with very few products.

It is worth knowing when reading claims about origami engineering. The geometry is usually settled and the demonstrations are usually real; the distance from a working demonstration to a deployed device is where the time goes, and it is not a distance the crease pattern shortens.

What a folded structure is not #

A short list of things folding does not provide, because the enthusiasm in this area runs ahead of the capability.

It does not provide strength. A folded structure is compliant along its degree of freedom by construction, and that is the opposite of structural.

It does not provide precision in the deployed state. Hinge tolerances accumulate across a large array, and anything needing optical accuracy is segmented and actively adjusted instead.

It does not provide reversibility. Creases are damage, and a structure that must cycle repeatedly uses linkages.

And it does not provide simplicity. A folded deployable has more parts than a rigid one, not fewer; what it has is a motion that can be reasoned about, which is a different and more specific benefit.

The requirement, stated once #

Stripping the applications back, they share a single sentence.

Something must be large when working and small when carried, the change must happen once, and nobody will be there to help.

Every item on this page is an instance. What varies is the scale, the material and which of the three clauses dominates — a stent cares most about the middle one, a starshade about the first, a spacecraft array about the last.

Folding answers the sentence because a crease pattern converts a size change into a motion, and a motion with one degree of freedom is something that can be made to happen reliably. Alternatives — telescoping, inflating, assembling — each answer some of the sentence and not all of it.

That is the whole case, and it is worth having in one place because the literature usually arrives at it through a specific application and never says it plainly.

The ladder from here #

Later rungs: the Miura array on the Space Flyer Unit. Airbag folding and the sequence problem. Origami stents. The starshade and its precision requirements. Self-folding materials. Microscale assembly from planar lithography. Deployment dynamics. Restraint and release mechanisms. Testing a mechanism that only works once. And the question of what folding is not good for, which is anything needing to deploy repeatedly.

The Miura-folded array flew in 1995, twenty-five years after Miura published the pattern, and the gap is entirely about qualifying a mechanism that has to work exactly once at a distance of several hundred kilometres.