042 — The Tile Comes Back Mirrored
agylövés & direction: Lysarith Penrose foundation: Claude Opus 5 (desktop code) topology, instrument, verification & build: GPT-5.6 Sol (Codex terminal) Status: built → agylovesek/042-the-tile-comes-back-mirrored/
The agylövés, verbatim
„az egyik kedvenc elképzelésem hogy tegyél penrose-t möbiusra”
Translation: “one of my favourite ideas is to put Penrose on a Möbius strip.”
The sentence first entered the Kert as bud 042 on 2026-08-23. The bud deliberately did not decide whether the seam can close. It recorded three possibilities and one instruction: measure it; do not choose by argument.
What the mathematics said before anything was built
Write a strip as coordinates (u,v), with 0 ≤ u ≤ 2π around it and −w ≤ v ≤ w across it. A cylinder closes by (0,v) ~ (2π,v). A Möbius strip closes by (0,v) ~ (2π,−v). The transition on the local frame is therefore
J = diag(1, −1), so det J = −1.
The centreline v=0 is a closed loop: the carried point returns exactly. Its transverse basis vector does not. After one loop it is its own negative; after two it returns. For n half-twists the holonomy is (−1)^n. This is the orientation obstruction, stated without asking a picture to witness it.
The Penrose ingredient is the decorated rulebook. Standard presentations do not rely on the two bare shapes alone: arrows or equivalent vertex markings say which neighbouring edges may meet. The Oxford construction notes that the two triangle shapes really split into reflected marked types and that edge arrows must agree. N. G. de Bruijn's algebraic treatment likewise works with arrowed rhombus patterns. Those are the marks this piece carries around the strip.
The fork the first wording hid
The first tempting sentence was: “Penrose matching rules are chiral, therefore they cannot close on a Möbius strip.” That was too large. Penrose tilings may themselves have reflection symmetry, and a rulebook may admit reflected decorated prototiles. The obstruction is not “Penrose” by itself. It is the attempt to keep one global handed orientation convention on a non-orientable surface.
That correction became the second control in the piece:
- one-handed rulebook: odd half-twist → one orientation conflict on the core loop;
- mirror admitted: the returning decorated tile is another legal letter, so this particular obstruction disappears;
- even half-twists: orientation returns without enlarging the alphabet.
This does not construct a complete edge-to-edge Penrose tiling of a compact Möbius strip. It identifies the first question such a construction must answer.
What was built
The surface is the standard parametrisation
M(u,v)=((R+v cos(nu/2)) cos u, (R+v cos(nu/2)) sin u, v sin(nu/2)).
A generation-five patch of the same φ-exact Robinson-triangle subdivision used by piece 030 is sampled as an annulus and mapped over the strip. A white decorated tile travels along the centreline. The visible arrow follows the local width vector, so the return is read from the geometry rather than animated as a costume change. The seam panel independently computes the parity of n.
Drag the surface, move the journey slider, compare zero through three half-twists, and then change the rulebook by admitting the mirror. The picture, the travelling frame and the integer parity are three views of the same transition.
The handle
Run the browser-and-numerical handle:
python agylovesek/042-the-tile-comes-back-mirrored/verify.py
It checks the edge identification numerically for zero through five half-twists; the determinant parity; one-lap reversal and two-lap recovery; the one-handed and mirror-admitted rulebooks; and the inherited Penrose subdivision at φ. It also reads the page back so the instrument cannot silently stop exposing the controls its claim depends on.
Run the independent symbolic handle in Wolfram Language:
agylovesek/042-the-tile-comes-back-mirrored/verify.wl
It proves the parity-dependent boundary identification and transition determinant for integer n, checks odd/even orientation and two-lap recovery, tests whether the two rulebook alphabets are closed under reflection, and independently verifies the φ substitution identities. Run through the stateless Wolfram kernel by GPT-5 (Sol · Codex terminal · OpenAI) on 2026-09-15, all nine checks returned True and the final line was claim holds: True.
What is claimed, and what is not
Claimed: any orientation-sensitive decoration transported once around the core of a strip with an odd number of half-twists returns mirrored. A one-handed rulebook must either expose a conflict or be enlarged to include the reflected hand.
Not claimed: a classification of all Penrose tilings on non-orientable surfaces; a proof that no finite Penrose-like patch can satisfy some chosen boundary convention; or a physically rigid tiling of the curved embedding. The displayed triangles are a texture carried by the parametrisation, not flat wooden rhombs bent without strain.
What would falsify the measured claim: an odd n for which the transition determinant is +1, or a point M(0,v) that fails to equal M(2π,−v). Both are tested directly.
Sources for the rulebook, not for the result
Alexander F. Ritter, Oxford Masterclasses in Geometry: Penrose tilings — marked reflected triangle types and arrow matching.
N. G. de Bruijn, Algebraic theory of Penrose's non-periodic tilings of the plane (1981) — arrowed rhombus patterns and the pentagrid construction.