Zomeable 4-polytopes

duoprism_4_10 inf-family question — corrected: infinite family exists

2026-05-29 correction

The boundedness verdict below is incorrect. The focused snap/signature probe was a false negative caused by the finite alignment/scale search; it did not constitute a constructive classification of the inf family.

The emitted representatives are:

output/duoprisms/duoprism_4_10/duoprism_4_10_inf_family_a5_b12.vZome
output/duoprisms/duoprism_4_10/duoprism_4_10_inf_family_a8_b15.vZome

For {4} x {10}, the preserved decagon can use blue edge axes in a plane perpendicular to a red axis. The decagon edge length is 1/phi for the unit radius model, and the blue/red length ratio is 2/sqrt(5), which lies in Q(phi) because sqrt(5)=2phi-1. Therefore the tesseract support-2 arithmetic condition transfers:

c = sqrt(a^2+b^2) in Q(phi).

For any such (a,b,c), scale the blue decagon edge frame by c and put the four square layers on the red height axis with scalar

beta = phi * |blue| / |red| = 2phi/sqrt(5) = (4+2phi)/5.

The square edges have height differences proportional to a+b and a-b, so they remain Q(phi) multiples of that red axis. Example: (a,b,c)=(3,4,5) gives a valid zome-axis {4} x {10} projection, and the usual infinite Q(phi) Pythagorean triples give infinitely many projective ratios.

The old rng tables remain useful as diagnostics for that particular probe, but the “bounded at 5 shapes” interpretation is not valid. The old text is retained below only as historical context.

Superseded TL;DR

Superseded: duoprism_4_10 was previously thought not to be an inf-family analog of the tesseract. The finite (a,b,0,0) snap probe saturated at 5 signatures, but that saturation was not a proof of boundedness.

Background

PR #5 finished the rng=4 prismatic sweep and reported duoprism_4_10 as the sole real gainer: rng=3 had 2 shapes, rng=4 found 5 (+3). The user noted that all five rng=4 kernels lie in the (a,b,0,0) plane — same structural form as the tesseract’s (a,b,0,0) inf-family — and asked whether duoprism_4_10 is also inf-family.

The full rng=5 prismatic sweep on duoprism_4_10 was OOM-killed by Windows commit-limit (~40 GB working set in gen_dirs(5) for 4D directions). A focused (a,b,0,0)-only ground-truth probe was developed instead to answer the question definitively.

Probe method

ongoing_work/probes/duoprism_4_10_snap_sig.py:

  1. Enumerate Z[φ] directions (a, b, 0, 0) for a, b ∈ [-rng..rng].
  2. For each direction n, project the 40 duoprism vertices to 3D via projection_matrix(n).
  3. Run _try_align (axis classification). If pass, snap the projected vertices to the zome ball lattice via _snap_coords. Dedup balls.
  4. Compute the robust shape signature: 5-decimal SHA-256 hash of the sorted scale-normalised squared distance matrix, plus the multiset of normalised edge lengths. This signature is the same one used by tools/dedup_corpus_by_shape.py.

The shape signature is preferred over the fingerprint used in the production sweep’s emit pipeline because the fingerprint’s bucket size grows with rng — at rng ≥ 5 it can collapse two genuinely distinct shapes into one hash. The signature does not have this defect.

Results

rng n_dirs aligned snapped distinct sigs new at rng elapsed
4 3280 3280 160 5 5 113 s
5 7320 7320 236 5 0 244 s
6 14280 14280 328 5 0 480 s
7 25312 25312 444 5 0 854 s
8 41760 41760 576 5 0 1388 s

The five SHA-256 hashes (top 16 hex chars of the d²-prefix) are identical at all five rngs:

2f14904a74911329
6ff7569b1fa4c243
7321c2c23935e90d
7e1e9efa27cc9ddb
9e455cd5c5a4fafd

The finite probe’s snap-count kept growing with rng (160 → 236 → 328 → 444 → 576) because more directions hit a valid kernel quantum, but every new kernel hash-collided with an existing probe signature. This looked like a bounded-family signature, but it only described the subset reached by that probe’s snap/alignment branch.

For comparison, the tesseract inf-family at rng=4 already produces dozens of distinct shapes and keeps growing without saturation.

Output JSON

Each holds the full sig list and probe metadata.

Superseded verdict

Superseded: duoprism_4_10 is not bounded at 5 shapes. The exact construction above gives a genuine infinite family; the finite probe only measured the subset reachable by its snap/alignment branch.

Wider duoprism_4_q census (from rng=4 sweep manifest)

q n_shapes (rng=4) notes
5 0 pentagon zome-compatible but no joint snap
6 3 finite-probe signatures at rng=5; exact infinite family now known
7 0 heptagon non-zomeable
8 0 octagon non-zomeable
9 0 nonagon non-zomeable
10 5 finite-probe signatures at rng=8; exact infinite family now known
11 0 hendecagon non-zomeable
12 0 dodecagon non-zomeable in this geometry
13–20 0 all non-zomeable

The finite-probe pattern was: zomeable q-gons (5,6,10) yielded small apparent shape counts when paired with the square; the rest yielded zero. q=5 remains interesting because the pentagon IS zome-compatible yet the duoprism emitted zero in that sweep — this is a joint-axis constraint, not a per-polygon limitation.