Artificial Wasteland — Stratigraphy · the Pattern seam

The Knot With No End

Islamic geometric patterns are drawn as a single knotted, endless line — girih, Persian for “knot.” From the twelfth century, artisans built these patterns from just five tiles, each carrying a fixed piece of that line. Lay the tiles here yourself, turn one dial, and watch a continuous star pattern weave itself across them by the makers’ own rule. Then follow one strand: it has no end.

Look at the wall of a mosque, a madrasa, a Persian shrine, and you will not find a picture. You will find a net of interlaced lines — stars opening into stars, a pattern that seems to continue past the edge of every panel it is cut into. The line is called girih. It is drawn as strapwork: bands that pass over and under one another the way a woven cord does, and — in the ideal, unbounded pattern — never terminate. The remarkable thing is not only that these patterns are beautiful. It is how they are made, and how far the method reached.

This page is a working loom for that method. Everything you draw with it is built from the real geometry, and every number it rests on is re-derived offline — the tiles’ angles, the strapwork rule, and the reason a five-fold pattern like these can never exactly repeat.

The five tiles

In 2007 the physicists Peter J. Lu and Paul J. Steinhardt argued that by around 1200 CE Islamic craftsmen had stopped drafting these patterns line by line with compass and straightedge, and instead reconceived them as tilings of a small set of equilateral shapes they call girih tiles. There are five. Every edge is the same length; every interior angle is a multiple of 36°.

decagon (10×144°) · pentagon (5×108°) · elongated hexagon (72,144,144,72,144,144) · bowtie (72,72,216,72,72,216) · rhombus (72,108,72,108)

Each tile is decorated with the same rule. Two strands cross the midpoint of every edge, each meeting the edge at 54°. Because two adjacent tiles share an edge — and its midpoint — the strand that leaves one tile is exactly in line with the strand entering the next. The decoration is continuous across the seam without anyone lining it up by hand. That is the whole trick, and it is why the pattern reads as one endless cord rather than as separate tiles. (You can see the tile boundaries below; on a real wall you would see only the strands.)

The loom

Pick a tile, then click any glowing open edge to set it there — every edge is the same length, so any tile fits against any edge. Turn CONTACT and watch the same tiling become a different star pattern (54° is the girih value). Hide the tiles to see only the strapwork. Then hit TRACE and click a strand to follow it: within this cut patch it ends only where the pattern is cut.

PRESET
1 tile · 10 open edges
Keyboard: the loom is focusable — press T to drop the selected tile at a random open edge, R to rotate a placement, U to undo. Mouse/touch: click a glowing edge to place; in Trace mode, click a strand.

Nothing here is a picture of a girih pattern — it is the pattern, generated from the tiles you place by the same rule medieval draughtsmen used. The strand-drawing is Hankin’s “polygons-in-contact” construction: from each edge’s midpoint, send two lines in at the contact angle and let them run until they meet the lines from the neighbouring edges. The crossings are the stars. Turn the dial off 54° and you get the acute and obtuse variants a real workshop could choose between from the same layout.

One strand

Put the loom in Trace and follow a single band. It crosses a tile, passes straight through an edge-midpoint into the next tile, turns at a star-point, and keeps going. In a finite patch it stops only at the cut edge — the place where the wall, or the panel, or your screen, runs out. In the pattern the craftsman intended — the one that repeats outward forever, or (as we will see) the one that never quite repeats at all — the strand has no end. That is what girih names: a knot, an interlace, a single line with no loose ends. It is also, on a great many walls, the point: a figure without beginning or end.

Five centuries early

Here is where the geometry stops being decoration and becomes something stranger. A pattern with five-fold (or ten-fold) symmetry — the kind these tiles are built for — cannot tile the plane periodically. It is a theorem, and an old one: a repeating pattern can carry rotational symmetry only of order 1, 2, 3, 4, or 6. Never 5. The reason is arithmetic — a lattice rotation must have a whole-number trace, 2·cos(2π/n), and for n = 5 that number is 1/φ ≈ 0.618, the reciprocal golden ratio, which is irrational. (This place has that proof in two other layers, including a machine-checked one.)

So a five-fold pattern that fills the plane has only one option: to fill it without ever exactly repeating — an aperiodic, quasicrystalline order, with long-range structure but no translational period. In the West this was Roger Penrose’s discovery of 1974, and then, in real matter, Dan Shechtman’s 1982 observation of a five-fold metallic alloy — a result so heretical it took years to publish and won the 2011 Nobel Prize in Chemistry.

Lu and Steinhardt’s claim is that a girih pattern on the Darb-i Imam shrine in Isfahan, built in 1453, is — as geometry on a wall — a nearly perfect quasicrystalline Penrose pattern. Its large girih tiles are subdivided into smaller ones in a self-similar way (the inflation that generates aperiodic order), and when they mapped the design onto Penrose tiles they found only 11 defects among some 3,700 tiles — each a local slip of the kind an artisan makes repairing a wall, each correctable by nudging a few neighbours. The tessellation-with-inflation method the craftsmen were using is, in principle, capable of producing a perfect quasicrystal. They reached it, by hand and by eye, more than five centuries before the mathematics existed to describe what they had done.

What that does and doesn’t claim

It is a claim about the geometry on the wall, not about what the builders knew: nobody is saying fifteenth-century masons had a theory of quasicrystals. The reading is also contested. Emil Makovicky described a related Penrose-like pattern on the 1197 Gunbad-i Kabud tomb tower at Maragha years earlier — but that pattern is periodic, so it is not itself a quasicrystal, and Makovicky’s published comment disputes parts of Lu and Steinhardt’s account (they replied in the same 2007 issue of Science). And “nearly perfect” is doing real work: Darb-i Imam has defects and is a finite panel, not a proof of infinite aperiodicity. What is not in dispute is the tile method itself, which the fifteenth-century Topkapı Scroll lays out explicitly — the five tiles, their strapwork, and the subdivision rules.

Whether it means the infinite

It is often said that these patterns are about infinity — that a figure with no beginning and no end, extending past every frame, is a deliberate emblem of divine unity and the boundlessness of God, and that the turn to geometry follows from Islam’s reticence toward figural images in sacred space (aniconism). There is something to this, and it is a genuinely beautiful reading. It is also, the historian Gülru Necipoğlu has argued, partly a modern reading — a twentieth-century, often perennialist gloss — projected back onto craftsmen whose surviving manuals (like the Topkapı Scroll) are overwhelmingly practical: construction lines, tile sets, and the transmitted know-how of a guild. The honest position holds both at once. The metaphysical reading is real and old as a reading; the patterns were also, demonstrably, a body of hard technical knowledge, made and remade by people solving a concrete problem in the geometry of the plane — and solving it, it turns out, better than anyone knew.

You do not have to choose. Turn the dial; trace the strand; watch the star pattern that a lattice forbids appear anyway. Whatever else it is, it is a true thing you can put your hands on.

The check

Every geometric claim here is re-derived from nothing but the published angle sequences, offline, in research/the-knot-with-no-end/verify.mjs (30/30 assertions pass), and the loom above embeds a byte-identical copy of that geometry core. Verified: all five girih tiles close (a turtle walking their angles returns to its start), every edge is unit length, and every interior angle is an exact multiple of 36°; the strapwork rule (two strands crossing each edge’s midpoint at 54°) makes the strands continuous across a shared edge for any contact angle (the strand leaving one tile is exactly antiparallel to the one entering its neighbour — one straight line); the golden-ratio identities 2cos36°=φ, 2cos72°=1/φ, and pentagon diagonal/side = φ; and the crystallographic restriction — 2cos(2π/n)∈ℤ only for n∈{1,2,3,4,6}, with the forbidden five-fold value equal to 1/φ.

Named uncertainties. The 1453 Darb-i Imam reading (Lu & Steinhardt 2007) is a claim about wall geometry, is “nearly perfect” with ~11 defects in ~3,700 tiles, and is contested in the literature (Makovicky). Datings (c. 1200 for the tile method; 1197 Maragha; 1453 Darb-i Imam; Penrose 1974; Shechtman 1982, Nobel 2011) follow the cited sources. The theological reading is presented as an interpretation, with the historiographic caution (Necipoğlu) named. Nothing on this page asserts what medieval artisans consciously knew.

Sources

Connections, in words · this layer is the human-and-craft face of a story this place tells three other ways: The Shape of Five (φ as the forbidder), The Einstein Stone (aperiodicity from inflation), Seventeen and No More (the machine-checked proof that five-fold can’t repeat). Here the same √5 that forbids the crystal is the line an artisan drew by hand.