Tricolorability: Proving the Trefoil Knot Is Knotted

Here is the oldest trap in topology. Two loops of string lie in front of you. One is a plain circle. The other has been knotted, the ends fused so you can't cheat. They are made of the same stuff; they can be pushed and slid and stretched. Can you ever prove that one of them cannot be turned into the other?

Wiggling is no proof. If you try a million rearrangements and none of them unknots the thing, that is a million failures, not a theorem — the next wiggle might have done it. To prove a knot is knotted you need a quantity that survives every legal move at once: something you can read off the picture, that never changes no matter how the string is pushed around. A number the knot can't shake off. Mathematicians call such a thing an invariant, and the most childlike one is the most astonishing: you can compute it with three crayons.

The rule of three crayons

Draw the knot flat, as a shadow — a closed curve that dives under itself at each crossing. The picture falls into arcs: the unbroken strokes that run from one dive to the next. Now colour every arc with one of three crayons, obeying a single law at each crossing:

the three strands meeting at a crossing must be all the same colour, or all three different.

That's it. A colouring that breaks the rule at even one crossing doesn't count. Using a single crayon for the whole thing always works — call those the boring colourings; there are exactly three of them. The real question is whether the knot admits a colouring that genuinely uses more than one colour. If it does, it is tricolourable — and that property cannot be created or destroyed by any rearrangement of the string. (The three crossing-moves that generate every rearrangement — the Reidemeister moves — each preserve the count of valid colourings exactly. The proof is below.)

So here is the whole argument in one breath: a plain loop can never be coloured with more than one crayon — it has no crossings to satisfy, so every arc is forced to agree. If your tangled loop can be three-coloured, it is not a plain loop in disguise. It is knotted, and you have proved it. Try it:

The plain loop is one single arc. There is nothing to satisfy and nothing to vary — three boring colourings, full stop, so the readout reads 3 and refuses to climb. The trefoil reads 9: the three boring ones, plus six that wheel all three colours around its three crossings. Nine is not three. That single inequality, between two integers a child could count, is a complete proof that the trefoil is a real knot — the simplest one there is, and the first object ever proven knotted (Heinrich Tietze and, with this exact colouring argument, Ralph Fox decades later).

Why a count is a proof

Any two diagrams of the same knot are related by a finite sequence of three local moves, proven sufficient by Kurt Reidemeister in 1927. So an invariant only has to survive these three:

Check the colouring rule against each move and the number of valid colourings comes out identical on both sides — a twist adds no constraint that wasn't forced, a poke's two new crossings cancel, a slide just shuffles which arc is which. (Walk through the algebra in the note below.) Because the count is unchanged by every move, and every rearrangement is a sequence of moves, the count is the same for every picture of a given knot. It belongs to the knot, not to the drawing. Two knots with different counts are genuinely, provably different. That is the whole magic: a property of pictures that turns out to be a property of knots.

The rule, exactly — and why it's really arithmetic mod 3

Name the crayons 0, 1, 2. "All same or all different" at a crossing is exactly the equation 2·(over) ≡ (under₁) + (under₂) (mod 3), where over is the strand passing on top and under₁, under₂ are the two arcs it separates. (Same colour: 2c = c + c ✓. All different, e.g. 0,1,2 in some order: 2·0 = 1+2 = 3 ≡ 0 ✓.) The valid colourings are the solutions of one such linear equation per crossing over the integers mod 3 — a linear system. Its solution set is a vector space over the field of three elements, so the number of colourings is always a power of three: 3 (only boring) or 9, 27, … A Reidemeister move changes the diagram's equations by an invertible substitution, which can't change how many solutions there are. That invariance is the proof, and it generalises to any number of crayons n — the count of n-colourings is a knot invariant for every n, a fact we lean on for the figure-eight below.

The thing three crayons cannot see

Now the catch, because honesty is the point. An invariant proves two knots are different when their values differ — but equal values prove nothing. And the tricolouring count has a blind spot you can hold in your hand. Take the trefoil and its mirror image — flip every crossing so each over-strand dives under instead. Your right hand and your left. It is a theorem that no amount of wiggling turns one into the other: the trefoil is chiral, the simplest knot that differs from its own reflection. But both trefoils have exactly nine colourings. To three crayons, the mirror is invisible.

To catch chirality you need a sharper instrument — one that knows left from right. In 1984 Vaughan Jones found it (and won the Fields Medal partly for it): a polynomial, read off the diagram by a recipe that splits each crossing two ways and bookkeeps the result. It is still just counting, only weighted. Here it is, computed live for both hands of the trefoil:

The two polynomials are mirror twins — every exponent flipped in sign, t → 1/t — and crucially not the same. If the two trefoils were secretly one knot, their Jones polynomials would have to match. They don't. So they don't. The blind spot of the crayons is exactly where the polynomial sees.

And the knot the crayons let walk free

One more honesty, the other direction. Tricolouring can fail to convict a guilty knot. Meet the figure-eight, the next knot after the trefoil — four crossings, unmistakably tangled. Try to three-colour it and you'll find you can't do better than boring: its tricolouring count is 3, the same as a plain loop's. To three crayons, the figure-eight looks innocent.

But it is knotted — and a fourth crayon won't help; you need five. With five colours it lights up (twenty-five valid colourings), and its Jones polynomial is nowhere near the unknot's. No single invariant is the whole truth; each is a lens with a resolution, and the craft is knowing which lens to reach for. Colour the figure-eight yourself, then switch to five crayons and watch it confess:

What is actually being shown

Every number on this page is computed in your browser, twice over and from two independent starting points. The two diagrams you colour are real projections: the trefoil is the curve (sin t + 2 sin 2t, cos t − 2 cos 2t, −sin 3t) and the figure-eight a similar space curve, each sampled, intersected with itself, and broken at every crossing according to which strand is higher — so the picture you colour is a true shadow of a knot, not an illustration. The colouring rule is enforced against the crossings that geometry produces.

The Jones polynomials and the colouring counts are recomputed a second, independent way from the knots' standard planar-diagram codes — the same recipe checked offline against the published literature (Rolfsen / KnotInfo): the right trefoil's −t⁻⁴ + t⁻³ + t⁻¹, the figure-eight's palindromic t⁻² − t⁻¹ + 1 − t + t², and the determinant cross-check det = |V(−1)| giving 3 and 5. The self-check at the top runs that whole battery live; the offline run is research/knots/verify.mjs, 16/16, logged in research/knots/facts.md. If any value here disagreed with the mathematics, the banner would not be green.

A knot can't be talked out of being knotted. But it can be counted out — and the count never lies.

That is the whole shape of the thing. To prove two tangles are the same you must produce the moves — a constructive, finite, honest demonstration. To prove they are different you reach instead for an invariant, a number the string carries with it through every contortion, and let two unequal integers do what no amount of wiggling ever could. Three crayons convict the trefoil. A polynomial separates its two hands. Five crayons catch the figure-eight the three let walk. Each tool has a reach and a blind spot, and naming both — out loud, with the check shown — is the only way the proof is worth anything.