This project enumerates all distinct 3D shapes obtainable by orthographic projection of each convex regular 4-polytope, such that every projected edge lies on a default-color zometool axis (B / Y / R / G in the icosahedral system).
The 8 polytopes documented below are the regulars + non-Wythoff corpus.
For the full Wythoff family (47 unique convex uniform 4-polytopes including
all of the rectified / truncated / cantellated / runcinated descendants),
see WYTHOFF_SWEEP.md and the inheritance-free
corpus-completeness audit in INHERITANCE_FREE_SWEEP.md.
| Polytope | Verts | Edges | Distinct shapes | Notes |
|---|---|---|---|---|
| 5-cell {3,3,3} | 5 | 10 | 4 | saturated rng=3; matches Richter/Vorthmann 2007 |
| 8-cell {4,3,3} | 16 | 32 | 1 inf family + 3 sporadic | inf family = split cuboids parametrized by ℤ[φ]-Pythagorean triples (a:b) with a²+b²=c²∈ℤ[φ]; sporadics = cube, rhombic dodec, phi-oblique (V=16, kernel support 3) |
| 16-cell {3,3,4} | 8 | 24 | 6 | saturated rng=4 |
| 24-cell {3,4,3} | 24 | 96 | 3 | saturated rng=3 |
| 120-cell {5,3,3} | 600 | 1200 | 1 | saturated rng=3 (330-ball image) |
| 600-cell {3,3,5} | 120 | 720 | 1 | saturated rng=4 (75-ball image) |
| Snub 24-cell | 96 | 432 | 2 | cell-first Tₕ, vertex-first D₃d |
| Grand antiprism | 100 | 500 | 2 | vertex-first D₂ₕ, ring-first D₅d |
(See per-polytope RESULTS.md files under output/<polytope>/ for details.)
For each polytope P:
n ∈ ℤ[φ]⁴ with |a|, |b| ≤ rng for each
component a + bφ, deduplicated by sign-flip-per-axis and decreasing
magnitude.|cos∠| between projected
edge directions are among the 40 allowed values for default-axis pairs,
then attempt to align two non-parallel edges to a chosen default-axis
pair and verify all remaining edges land on default axes.See METHODOLOGY.md for caveats about empirical-vs-formal saturation.
zomeable-4polytopes\
├── lib\
│ ├── polytopes.py vertex/edge specifications for all 8 polytopes
│ ├── search_engine.py generic kernel-direction search
│ ├── run_search.py CLI runner: python lib/run_search.py <name> [rng]
│ ├── emit_generic.py generic <Polytope4d>-based emitter
│ ├── emit_vzome.py hand-built emitter (24-cell)
│ ├── emit_8cell.py 8-cell emitter
│ ├── emit_snub24.py snub 24-cell emitter
│ └── emit_grand_antiprism.py grand antiprism emitter
├── tools\
│ ├── fingerprint_audit.py cross-polytope consistency check
│ └── patch_origin.py idempotent origin-ball cleanup
├── docs\
│ ├── SUMMARY.md this file
│ ├── METHODOLOGY.md saturation & open questions
│ ├── UNIFORM_PLAN.md extension to all 47 convex uniform 4-polytopes
│ ├── WYTHOFF_SWEEP.md Wythoff-family corpus: methodology, taxonomy, per-polytope tables
│ └── INHERITANCE_FREE_SWEEP.md 38-hour matrix sweep auditing the Wythoff corpus for completeness
└── output\
├── 5cell\ RESULTS.md + 4 .vZome files
├── 8cell\ CLASSIFICATION.md + 9 .vZome files
├── 16cell\ RESULTS.md + 6 .vZome files
├── 24cell\ ZOMEABLE_PROJECTIONS.md + 3 .vZome files
├── 120cell\ RESULTS.md + 1 .vZome file
├── 600cell\ RESULTS.md + 1 .vZome file
├── snub_24cell\ RESULTS.md + 2 .vZome files
├── grand_antiprism\ RESULTS.md + 2 .vZome files
├── <wythoff descendants>\ 39 folders (e.g. rectified_5cell, bitruncated_8cell,
│ omnitruncated_120cell, …), each with RESULTS.md
│ plus 0–3 .vZome files — see WYTHOFF_SWEEP.md
└── wythoff_sweep_manifest.json provenance ledger for the 69 Wythoff-descendant .vZome files
Two distinct criteria show up; the 8-cell case forced us to make the distinction explicit:
c_i can be rotated
so each unit-vector lies on a default axis.|c_i|/|c_j| is in ℚ(φ).The search engine reports criterion 1. Criterion 2 is verified at emit
time by the per-strut classify_direction check in emit_vzome.py
(every directed strut must match a (B/Y/R/G)i_proto * GF(...)).
Most polytopes satisfy 1 ⇔ 2 because their geometry forces the
length ratios to be in ℚ(φ) automatically. The 8-cell breaks this:
the search finds support-3 kernels (e.g. n = (1,−1,1,0)) whose 4
image vectors all align with default axes but have length ratios
√2/√3 ∉ ℚ(φ) or √(2+φ) ∉ ℚ(φ) — so they exist as zometool
sculptures but cannot be expressed in vZome’s standard icosahedral
field. We do not emit such cases.
lib/emit_8cell.py’s emit_skew_zonotope(..., norm=...)).1/(a + bφ) = ((a+b) − bφ) / (a²+ab−b²).
Useful for normalization when the longest edge is itself a non-integer
ZZ[φ] value like 2φ−1 = √5.classify_direction on
every directed strut. A match means the strut lies on a default axis
AND the length factor is in ℤ[φ]; a None result means the model is
not vZome-embeddable.SCALE = GF(2,2)
(= 2+2φ, the zometool blue-strut length unit). Smaller SCALEs put
balls too close; larger ones make models unwieldy.All non-hand-built files use vZome’s <Polytope4d> element with a
specific Coxeter group + wythoff index + quaternion. The quaternion q
is derived from the kernel direction n of the projection by
q = conj(n) = (n_w, -n_i, -n_j, -n_k) (vZome’s projection is
v ↦ imag(v · q), kernel = span(conj(q))).
| Polytope | Group / wythoff | Files |
|---|---|---|
| 5-cell | A4 / 1000 | 5cell/5cell_*.vZome (4 files) |
| 16-cell | B4/C4 / 1000 | 16cell/16cell_*.vZome (6 files) |
| 24-cell | F4 / 1000 | 24cell/*.vZome (3 hand-built) |
| 120-cell | H4 / 1000 | 120cell/120cell_H4_to_H3.vZome |
| 600-cell | H4 / 0001 | 600cell/600cell_H4_to_H3.vZome |
Caveat: kernel directions in our search embedding may differ from
vZome’s internal embedding (vZome uses a Coxeter group’s standard
realization). The emitted files were built using each polytope’s
standard ℤ[φ] embedding, which matches vZome for {16,24,120,600}-cells
but may need reframing for the 5-cell. Each file should be opened in
vZome and verified to render with default-color struts; shapes that fail
need either (a) a different quaternion in the same family, or (b) a
hand-built <ShowPoint>/<JoinPointPair> model (see
24cell/search/emit_vzome.py for the pattern).
.vZomes: ✓ verified by user..vZomes (12 files): emitted via emit_polytope4d.py,
pending user verification in vZome.Independent prior work corroborating our enumerations:
5cell/RESULTS.md for the file-by-file mapping.16cell_antiprism_*.vZome
files; truncating any one of them to its edge midpoints yields the
same 24-cell zometool model, an independent consistency check.