Origami & Folding
what one sheet can do
Axioms and construction

Folding beats the compass, by exactly one degree

Straightedge and compass solve quadratics. A single fold solves cubics. That one-step difference settles two problems Greek geometry could not, and leaves a third exactly as impossible as it was.
15 min read 6 figures Folding beats the compassOne sheet, no cuts

Trisecting an arbitrary angle with straightedge and compass is impossible. That is not a statement about anybody’s ingenuity — it was proved by Wantzel in 1837, and the proof is about which numbers those two instruments can reach.

A sheet of paper does it in one fold.

Trisecting 63° with one foldAbe's construction. Two horizontal creases give a reference; then a single fold carries the corner onto the lower one at the same moment as it carries the point above onto the ray. The two creases that result divide the angle into exact thirds — a construction provably out of reach of straightedge and compass.hh/263°42.0°21.0°a third of 63° is 21.00° — the fold found it, nothing was drawn at a thirdthe corner reaches the lower crease and the marked point reaches the ray at the same instantmountainvalley
Fig. 1 Abe’s construction. Two horizontal creases give a reference; then one fold carries the corner onto the lower crease at the same instant as it carries the point above onto the ray. The creases that result divide the angle into exact thirds — and the angles marked are computed from the folded positions, not drawn at a third.

The gap between those two facts is one degree of polynomial, and it is worth understanding precisely, because “origami is more powerful than Euclid” is usually said far too loosely.

What a compass reaches

Start with the rational numbers. A straightedge draws lines through known points; a compass draws circles about known points. Every new point is the intersection of two such objects, and intersecting lines and circles means solving equations of degree at most two.

So each construction step adjoins at most a square root. After finitely many steps the numbers available are those reachable by repeated square roots, and every one of them lies in a field extension of the rationals whose degree is a power of two.

[Q(α):Q]=2k.[\mathbb{Q}(\alpha) : \mathbb{Q}] = 2^k.

That is the whole characterisation, and it is what makes the impossibility proofs work. To show something cannot be constructed, show its degree is not a power of two.

Trisecting 60° requires solving 8x3−6x−1=08x^3 - 6x - 1 = 0, which is irreducible over the rationals and therefore of degree three. Three is not a power of two. Done.

What a fold reaches

Axiom 6 — the fold that carries two points onto two lines at once — behaves differently, and the reason is geometric before it is algebraic.

A fold that places a point PP onto a line ℓ\ell has its crease tangent to the parabola with focus PP and directrix ℓ\ell. That is a restatement of the definition of a parabola: the crease is the perpendicular bisector of PP and its image, and every point on it is equidistant from PP and from ℓ\ell.

Axiom 6 asks for a crease that does this for two point-and-line pairs at once — a common tangent to two parabolas. Two conics have three common tangents in general, and finding them is a cubic.

What each toolset can reachThe numbers each construction method can produce. A compass generates square roots and so reaches degrees that are powers of two; a fold solves cubics and reaches products of powers of two and three. The gap between the two rows contains the three classical impossible problems.straightedge alonerationalno new numbers at allstraightedge and compassdegree 2^k√2, the regular 17-gonone fold at a timedegree 2^a 3^b∛2, the trisected angle, the regular 7-goneverything past here is out of reach of Euclid's toolsdoubling the cube, trisecting the angle and the regular heptagon all live in the gap
Fig. 2 What each toolset can produce. The compass climbs by square roots and reaches degrees that are powers of two; the fold climbs by square and cube roots together and reaches products of powers of two and three. Everything in the gap is out of reach of Euclid’s instruments and inside reach of a crease.

So folding adjoins cube roots as well as square roots, and the reachable numbers are those in extensions of degree 2a3b2^a 3^b. That is a strictly larger set, by exactly one prime.

Trisection, worked

The construction on this page is Abe’s, and it is worth following because it makes the cubic visible.

Set the angle at the corner of the sheet, between the bottom edge and a ray. Make a horizontal crease at some height hh, and another at h/2h/2. Now make one fold that simultaneously:

  • carries the corner onto the crease at h/2h/2, and
  • carries the point at height hh on the left edge onto the ray.

Two points, two lines, one fold. Axiom 6.

When it is unfolded, the image of the corner and the image of the midpoint mark two rays from the corner, and those rays trisect the angle. The figure above computes the fold by root-finding on exactly that pair of conditions, and then checks that the resulting angles are a third and two-thirds — to better than a millionth of a degree, or the build fails.

The height hh is arbitrary. It sets the scale of the reference and does not affect the answer, which is a good sign that the construction is doing something structural rather than something coincidental.

Doubling the cube, in one crease

The second classical problem is the cube root of two, and folding produces it about as directly as anything could.

Doubling the cube in one foldMesser's construction. Dividing the sheet into thirds and folding one corner onto an edge divides that edge in the ratio one to the cube root of two — the classical problem Greek geometry could not solve, obtained from a single crease.1∛2∛2 = 1.259921, and the point sits at 0.442493one crease, for a problem Greek geometry could not solve
Fig. 3 Messer’s construction. The square is divided into thirds, and then one fold places a corner onto an edge. The point where it lands divides that edge in the ratio one to the cube root of two.

Divide the sheet into thirds — which folding does exactly — and then fold so that one corner lands on the left edge while the third-mark lands on the top edge. The left edge is divided in the ratio 1:231 : \sqrt[3]{2}.

Delian legend has it that the oracle at Delos demanded an altar of twice the volume, and that Greek geometry spent centuries failing. The failure was real and the reason is now understood; the resolution takes a piece of paper and about fifteen seconds.

What does not fall

Squaring the circle is the third classical problem, and folding does not touch it.

Constructing a square of the same area as a given circle requires π\sqrt{\pi}, and π\pi is transcendental — Lindemann proved it in 1882. A transcendental number is not the root of any polynomial with rational coefficients, of any degree at all.

So the fold’s advantage is irrelevant here. Climbing from degree 2 to degree 6, or to degree 60, or to any finite degree, does not help against a number that satisfies no polynomial. The compass and the crease are equally helpless, and for the same reason.

This is the sharpest available statement of what folding buys: one prime, not a general escape.

The neusis connection

There is a much older construction that also trisects, and comparing it explains what axiom 6 really is.

Archimedes trisected an angle with a marked ruler — a straightedge with two notches at a fixed distance. Slide the ruler until the two marks land simultaneously on a circle and a line, and the resulting angle is a third of the original. The technique is called neusis, or verging, and Greek geometers used it while regarding it as inferior.

Neusis is exactly as powerful as folding: both solve cubics, both trisect, both double the cube, and both fail at squaring the circle. Axiom 6 is a neusis construction in disguise — the fold slides until two incidences occur at once, which is what sliding a marked ruler does. The same simultaneity appears whenever two conditions have to be satisfied at once.

The Greeks had the power and declined to use it, on grounds of purity. The classification of what is and is not constructible follows from a decision about which instruments count, made for aesthetic reasons two and a half thousand years ago, and everything since has been about the consequences.

Two folds at once, and where it stops being folding

If one fold reaches cubics, more folds reach further, and the hierarchy has been worked out.

Alperin and Lang showed that constructions with several simultaneous folds — creases made at the same moment, with alignment conditions relating all of them — reach quintics, and with enough simultaneous folds, polynomials of arbitrary degree.

That is a genuine result and it is also where the subject stops describing paper. One crease at a time is what hands do. Three creases made simultaneously with shared alignments is a formal device, and no folder has ever performed one — though collapsing a base comes closer than it looks.

The seven ways to specify a foldThe Huzita–Hatori axioms: every fold that can be specified by bringing existing points and lines into coincidence. The list is complete — six were catalogued in 1991, the seventh in 2001, and no eighth exists. The sixth is the one a compass cannot reach.axiom 1through two pointslinearaxiom 2point onto pointlinearaxiom 3line onto linelinearaxiom 4through a point, square to a linelinearaxiom 5point onto a line, through a pointquadraticaxiom 6two points onto two linescubicaxiom 7point onto a line, square to a linelinearthe degree each axiom can solve — one of them is why paper beats the compass
Fig. 4 The seven single-fold operations. Everything on this page lives inside this list, and the multi-fold hierarchy lives outside it — a larger theory that no longer corresponds to an action anybody takes.

Why the exactness matters in practice

Everything above is about what is possible in principle. There is a separate and more immediately useful point: the constructions are exact, and estimating is not.

A folder dividing a square into fifths by eye is out by perhaps a per cent. That is invisible, and it stops being invisible once the fifths are used as references for further creases, because errors compound with every alignment.

An exact construction contributes no error of its own. The only inaccuracy is the folder’s hands, and it does not accumulate in the same way, which is what makes a design on a grid foldable. For a design on a grid of sixty-fourths, the difference between exact and estimated references is the difference between a model that closes and one that does not.

Dividing a square into 3, exactlyHaga's theorem. Folding one corner onto a point part-way along an opposite edge produces exact rational divisions elsewhere on the sheet — so a square can be divided into any whole number of parts by folding alone, with no measurement and no accumulated error.3 equal partsestimated by eyethe left-hand divisions are exact — a consequence of the fold, not of carethe right-hand ones are a guess, and the error compoundsvalleymountain
Fig. 5 Exact thirds on the left, estimated thirds on the right. The left-hand construction is Haga’s theorem — one fold, no measurement — and the difference between the two panels is what compounds.

Who found it, and when

Margherita Piazzolla Beloch published the key result in 1936: that a single fold satisfying two incidence conditions extracts cube roots. It is the first appearance of the cubic in paper folding, and it was almost entirely overlooked for fifty years.

Hisashi Abe gave the trisection in the 1970s. Peter Messer gave the cube-root construction in 1986. Huzita catalogued the operations in 1991. Roger Alperin connected the whole thing to field theory in 2000, and Lang settled the completeness and the multi-fold hierarchy shortly after.

So the mathematics is recent even though the craft is ancient, and the ordering is unusual: the constructions came first and the theory that explains why they work came decades later.

What the extra degree is worth

It is easy to read “cubics instead of quadratics” as a modest technical gain. It is not, and the regular polygons make that concrete.

Gauss characterised which regular nn-gons a compass can construct: those whose odd prime factors are distinct Fermat primes. The heptagon fails, the nonagon fails, the 11-gon and 13-gon fail. The first few constructible ones are 3, 4, 5, 6, 8, 10, 12, 15, 16, 17 — and then a long gap.

With folding, the criterion loosens to allow factors of three as well, and the heptagon, the nonagon, the 13-gon and the 19-gon all become reachable. The set of constructible regular polygons roughly doubles in density.

That is what one prime buys. Not a marginal improvement to a few constructions, but a different answer to a classification question that had stood since 1801.

The seven ways to specify a foldThe Huzita–Hatori axioms: every fold that can be specified by bringing existing points and lines into coincidence. The list is complete — six were catalogued in 1991, the seventh in 2001, and no eighth exists. The sixth is the one a compass cannot reach.axiom 5point onto a line, through a pointquadraticaxiom 6two points onto two linescubicthe degree each axiom can solve — one of them is why paper beats the compass
Fig. 6 The two axioms that do the interesting work. Axiom 5 places a point on a line through a fixed point, which is a quadratic — the compass can do it. Axiom 6 places two points on two lines at once, which is a cubic, and nothing in Euclid’s toolkit reaches it.

Why nobody noticed for so long

Paper folding is old and the mathematics is not, and the gap is worth an explanation.

Recreational folding in Europe and Japan goes back centuries. Friedrich Fröbel put folding into kindergartens in the 1830s specifically as geometry teaching. Sundara Rao published Geometric Exercises in Paper Folding in 1893, which reaches a good deal and stops short of the cubic.

What was missing was the question. Constructibility was a live problem in the nineteenth century and it was studied with compasses, because that was the instrument the classical formulation named. Nobody asked what a different instrument would reach, and when Beloch did ask, in 1936, she published in a journal that geometers read and folders did not.

The subject needed somebody who was fluent in both, and there were very few such people until origami societies and mathematics departments started overlapping in the 1980s.

The construction is checked, not drawn

A note about the figure at the top, because it is the site’s method in miniature.

The trisection is not drawn at a third. The generator sets up the two incidence conditions, finds the fold by root-finding, reflects the two points across it, measures the angles the images make with the base edge, and then asserts that they are a third and two-thirds to within a millionth of a degree. If they are not, the build stops.

That assertion has already earned its place. The first version of the construction used the wrong pair of incidence conditions, and the figure would have shipped a picture labelled “trisection” that trisected nothing. The check caught it, the conditions were derived properly by testing which pair actually produces thirds, and the picture is now a demonstration rather than a claim.

It also has to choose between roots. Axiom 6 generally has three solutions and the equation has two here; one of them places the image on the backward extension of the ray rather than on the ray. The generator selects geometrically — the image must land on the half-line that bounds the angle — and not by checking which root gives the answer it wants, which would be circular.

Where the model stops

Exactness is mathematical. The trisection is exact as geometry and approximate as an action. A crease made by aligning two points by eye is good to a fraction of a millimetre, which is a hundredth of a degree at the scale of a sheet — better than most people expect and not exact.

One fold at a time. The whole argument assumes a sequence of single creases on a flat sheet. Simultaneous folds reach further and are not physical; folding an already-folded sheet is a different problem the axioms do not cover.

The sheet is finite. Several of these constructions want a crease whose line leaves the square. The figures clip to the paper, which is honest about the sheet and hides that the construction sometimes needs the part that is not there.

Degree is not the only obstruction. A number of degree six is reachable in principle, which does not mean a short construction exists. The theory says what is possible, not what is convenient, and some foldable constructions take a great many steps, each one a crease that stays in the finished sheet.

The trisection figure shows one solution. Axiom 6 generally has three, and the figure selects the one whose image lands on the ray rather than on its backward extension. That selection is geometric and stated in the code, and the picture does not show the roots it discarded.

The instrument decides the theorem

The last thing worth taking from all this is not about folding at all.

“Impossible” in classical geometry is always relative to a toolkit. Trisection is impossible with straightedge and compass, possible with a marked ruler, possible with a fold, possible with a conic section, and possible with a piece of string. None of those facts contradicts the others, and the word “impossible” is doing different work in each.

What Wantzel proved is a statement about a specific algebraic closure. The Greeks chose their instruments for reasons that were partly practical and largely aesthetic, and two thousand years of effort went into a problem that their own contemporaries could already solve by neusis — while regarding the solution as not counting.

Folding is a reminder that the classification was contingent. Change the instrument and the map of the possible redraws itself, and the new map is no less rigorous than the old one.

The ladder from here

Later rungs: Wantzel’s theorem and the degree argument in full. Parabolas as the geometry of axiom 6. Beloch’s fold, derived. The regular heptagon and nonagon, both foldable. Neusis, and the Greek objection to it. The multi-fold hierarchy. Constructible numbers as a field. Angle quintisection. And the practical question of accuracy, which decides whether any of this survives contact with a sheet of paper.

Wantzel proved trisection impossible in 1837, at the age of twenty-three, and the result was so completely absorbed that it is often attributed to Gauss. He also proved the impossibility of doubling the cube in the same paper, and died at thirty-three.

Where this goes next

axiom 1through two pointslinearaxiom 2point onto pointlinearaxiom 3line onto linelinearaxiom 4through a point, square to a linelinearaxiom 5point onto a line, through a pointquadraticaxiom 6two points onto two linescubicaxiom 7point onto a line, square to a linelinearthe degree each axiom can solve — one of them is why paper beats the compass One fold at a time, and there are exactly seven of them A fold is specified by bringing points and lines into coincidence. There are seven ways to do that, the list is provably complete, and one of the seven does something no compass can. 3 equal partsestimated by eyethe left-hand divisions are exact — a consequence of the fold, not of carethe right-hand ones are a guess, and the error compoundsvalleymountain Dividing without measuring A square can be divided into any whole number of equal parts by folding alone — exactly, with no ruler, and with no error to accumulate. The construction is one fold and a theorem nobody expected. Levery point within L is spentthe flapLthe circle is not a metaphor — it is the paper the flap consumesso designing a base is packing circles A flap costs a circle A flap of a given length uses up every point of the sheet within that distance of it. Two flaps whose circles overlap are asking for the same paper twice — and that one observation turned origami design from an art into an algorithm.