Zomeable 4-polytopes

Uniform 4-polytopes: zomeable projection prospects

Status: planning. Drafted 2026-05-02.

This document inventories the ~47 convex uniform 4-polytopes and identifies which are worth searching for new zomeable orthographic projections to 3D.

TL;DR

Most uniform 4-polytopes are Wythoffian (built by ringing a Coxeter diagram). For these, edge directions are constrained to the union of root orbits of the parent Coxeter group, which means their zomeable projections are necessarily subsets of the projections we already enumerated for the regular 4-polytopes. No new shapes arise.

The genuinely new opportunities are the two non-Wythoffian uniform 4-polytopes and the prismatic/duoprism families.

Recommended priority order:

  1. Snub 24-cell (D₄-symmetric, 5 edge orbits). High value, moderate cost.
  2. Grand antiprism (non-Wythoff, partial H₄). High value, moderate cost.
  3. Pentagonal duoprism {5}×{5} and a couple of small duoprisms. Moderate value (educational; trivially zomeable in some cases), low cost.
  4. (Skip) All Wythoff uniforms — automated batch run only as sanity check.

1. The root-orbit argument (why Wythoff uniforms yield nothing new)

In the Wythoff construction with ringed nodes S ⊆ {0,1,…,n}, edges of the resulting polytope come in |S| orbits under the symmetry group W. The k-th orbit consists of W-translates of a simple-root direction associated with ring node k.

If all simple roots of W lie in a single W-orbit (i.e. W is simply laced or has connected Coxeter diagram with all single bonds, OR the diagram is non-crystallographic but uniform-orbit like H₄), then every uniform polytope in family W has edges in the same direction-orbit as the regular {p,q,r} of that family. Identical zomeable kernel set.

If W has two root orbits (B₄, F₄), then a uniform polytope’s edges live in the union of those two orbits. A kernel n is zomeable for the polytope iff n is zomeable for both orbits simultaneously — i.e., the intersection of the zomeable kernels for the two regular polytopes that exhibit those orbits. Strictly fewer, never more.

Coxeter family # root orbits Regular polytopes exhibiting each orbit
A₄ (5-cell) 1 5-cell
B₄ (8/16-cell) 2 16-cell (long), 8-cell (short)
F₄ (24-cell) 2 24-cell long-root, short-root 24-cell
H₄ (120/600-cell) 1 120-cell, 600-cell (same orbit)

Corollary. For each Wythoff uniform 4-polytope, the set of zomeable kernels is contained in the kernels we already enumerated for the regular polytopes of its family. We expect no new 3D shapes from Wythoff uniforms, only restrictions or duplicates.

This is testable: an automated batch run on all Wythoff uniforms should yield only shapes already in our database. (Useful as a sanity check on the search engine and the database, but not as a new-shape generator.)

1.1 Empirical confirmation: bitruncated 5-cell (2026-05-02)

Built the bitruncated 5-cell (30 vertices, 60 edges in 2 Wythoff orbits) in the same R⁴ embedding as the regular 5-cell (sym coordinates = permutations of (0,0,1,2,2), mapped via T(w) = Σ wᵢ vᵢ). Ran search at rng=3:

  5-cell bitruncated 5-cell
hits 40 40
distinct shapes (rng=3) 3 3
kernel set diff — 0 new, 0 missing

Every one of the 40 zomeable kernel directions is identical between the two polytopes. The 3D shapes change (4-ball → 18-ball, 5-ball → 30-ball ×2) because the bitruncated projection has more vertices, but the projection direction is in 1-to-1 correspondence. This empirically confirms the representation-theory argument: A₄ Wythoff uniforms share the 5-cell’s zomeable kernel set exactly. (Script: reality_check_bitrunc5.py.)

2. Inventory of Wythoff uniform 4-polytopes (47 total)

Listed by family with edge-orbit count. EO = “edge orbits” = number of ringed nodes. All edges of these polytopes are in the corresponding direction orbits of the parent group.

A₄ family (9 polytopes)

All have edges in the single A₄ root orbit (20 vectors e_i − e_j). Zomeable kernels = same as 5-cell (4 distinct shapes).

Symbol Name EO
{3,3,3} 5-cell 1
r{3,3,3} rectified 5-cell 1
t{3,3,3} truncated 5-cell 2
rr{3,3,3} cantellated 5-cell 2
2t{3,3,3} bitruncated 5-cell 2
t₀,₃ runcinated 5-cell 2
tr{3,3,3} cantitruncated 5-cell 3
t₀,₁,₃ runcitruncated 5-cell 3
t₀,₁,₂,₃ omnitruncated 5-cell 4

B₄ family (15 polytopes)

Edges in long-root orbit (24 vectors), short-root orbit (8 vectors), or both. Zomeable kernels = subset of 16-cell ∪ 8-cell projections.

Notable mixed-orbit cases: truncated 8-cell (rings 2,3 → mixes orbits), runcinated tesseract, omnitruncated tesseract.

F₄ family (9 polytopes, including 24-cell families)

Edges in F₄ long-root orbit, short-root orbit, or both. Zomeable kernels = subset of long-root-24-cell ∪ short-root-24-cell.

H₄ family (15 polytopes)

All edges in single H₄ root orbit (120 vectors). Zomeable kernels = same as 120/600-cell, i.e. the unique H₄→H₃ Coxeter projection only.

This is a strong rep-theory statement: there is essentially one linear H₄-equivariant map H₄ → H₃ up to symmetry, so every uniform H₄ polytope has at most one zomeable projection up to symmetry — the same one.

3. The genuinely new candidates (non-Wythoff)

3.1 Snub 24-cell {3,4,3}/2 — TOP PRIORITY

3.2 Grand antiprism — SECOND PRIORITY

3.3 Duoprism / antiprism-prism families (infinite, picky selection)

These are mainly educational / pretty rather than novel-shape-finding.

4. Proposed work plan

Phase A — automated sanity check (low effort)

Run existing search_engine over all 9 A₄ uniforms, verify zomeable kernels are subsets of the 5-cell kernels. Quick regression test. Could similarly hit a representative subset of B₄/F₄/H₄ uniforms.

Optional: write the batch driver and run overnight.

Phase B — snub 24-cell

  1. Add snub_24cell to polytopes.py (vertices + edges).
  2. Run search_engine at rng=3,4. Identify zomeable kernels.
  3. For each kernel, classify projected 3D shape and emit .vZome.
  4. Document in snub_24cell/RESULTS.md.

Phase C — grand antiprism

Same workflow; smaller in vertex count but irregular structure means edge enumeration is the careful step.

Phase D — duoprisms (optional decorative)

Pick {5}×{5}, {10}×{10}, {3}×{8}, emit one each as illustration.

5. Open uncertainties (good to confirm before starting)

6. Summary recommendation