The snub 24-cell {3,4,3}/2 is one of two non-Wythoff convex uniform 4-polytopes. It has 96 vertices, 432 edges, 480 faces (96 triangles + 384 faces from icosahedral cells), 144 cells (24 icosahedra + 120 tetrahedra), and symmetry group of order 1152 (the snub-24 symmetry group, related to D₄ < F₄ by index 6).
It is built by alternation of the truncated 24-cell, so its edge structure is fundamentally non-root: edges live in directions outside the F₄ root system. This makes it the prime candidate (along with the grand antiprism) for genuinely new zomeable orthographic projections beyond those of the 6 regular 4-polytopes.
Empirical search on direction grid up to rng=4 finds exactly 2 distinct zomeable orthographic projections, each verified strict- orthographic by the M·Mᵀ = s²·I test:
For front-visible variants that remove geometry hidden from the positive side of
each projection kernel, see
../snub_24cell_hidden_cells/RESULTS.md.
| File | Kernel direction | 3D balls | 4D edges → 3D struts | Strut counts |
|---|---|---|---|---|
snub_24cell_cell_first.vZome |
(1, 0, 0, 0) | 60 | 432 → 228 | B:72 Y:84 R:72 |
snub_24cell_vertex_first.vZome |
(φ², φ, 1, 0) | 69 | 432 → 312 | B:90 Y:102 R:120 |
Both projections use all three default zometool colors (blue, yellow, red) — a richer color signature than any of the regular 4-polytope projections, reflecting the snub’s symmetry breaking that mixes pentagonal (red), tetrahedral (yellow), and rectilinear (blue) features.
Kernel: (1, 0, 0, 0). This direction is the center of one of the 24 icosahedral cells of the snub 24-cell. It is also exactly a vertex of the removed inscribed 24-cell (recall: snub 24-cell = 600-cell minus a 24-cell, and the removed 24-cell vertices become the icosahedral cell centers).
Projecting along this axis squashes the icosahedral cell at +e₁ and the antipodal one at −e₁ onto the 3D origin. 96 vertices → 60 distinct 3D points (36 collapses; 12 zero-length antipodal-edge collapses).
Found in 80 distinct kernel directions at rng=4 (the orbit of the cell-center direction under the snub-24’s symmetry group).
3D point group: Tₕ (pyritohedral, order 24). The proper rotation group is the chiral tetrahedral T (order 12: E, 8C₃, 3C₂); inversion lifts this to Tₕ. Despite the kernel (1,0,0,0) having a 4-fold cubic stabilizer in the parent 24-cell, the snub’s 24+24 tetrahedral cell- pair pattern breaks the 4-fold to a 3-fold + 2-fold, giving the same symmetry as a pyritohedron.
Relation to 24-cell projections. The kernel (1, 0, 0, 0) plays three different roles depending on which polytope it is applied to:
| Polytope | What (1,0,0,0) is | Resulting 3D shape |
|---|---|---|
| 24-cell long-root (verts ±eᵢ±eⱼ) | a short-root direction = octahedral cell center | cuboctahedron (18 balls) |
| 24-cell short-root (verts ±eᵢ and (±½)⁴, the inscribed 24-cell removed from the 600-cell) | a vertex direction | rhombic dodecahedron (15 balls) |
| snub 24-cell | a cell center (icosahedral cell) | the 60-ball shape above |
So the snub’s cell-first projection does not coincide with the zomeable cell-first projection of the 24-cell (the 18-ball cuboctahedron lives in the long-root embedding). Instead, the snub’s cell centers are positioned at the vertices of the removed short-root 24-cell, so the snub’s cell-first kernel coincides with that 24-cell’s vertex-first kernel. The two 3D shadows along (1,0,0,0) — the snub’s 60-ball figure and the rhombic-dodecahedron of the inscribed short-root 24-cell — are different shapes; they share only the kernel direction, not a containment relation at canonical ϕ-power scales.
Kernel: (φ², φ, 1, 0). After normalization: (φ/2, 1/2, 1/(2φ), 0), which is exactly a snub 24-cell vertex (it is one of the even permutations of (φ, 1, 1/φ, 0)/2).
Projecting along a vertex direction collapses that vertex and its antipode onto the 3D origin, with the remaining 94 vertices giving 67 more distinct 3D points (after 27 further collapses; 6 of which are zero-length edge collapses). Total 69 balls.
Found in 20 distinct kernel directions at rng=4.
3D point group: D₃d (antiprismatic, order 12). The proper rotation group is D₃ (order 6: E, 2C₃, 3C₂) with the principal 3-fold aligned along the kernel; central inversion lifts it to D₃d.
Where are the deleted 24-cell vertices? Of the 24 deleted 4D vertices
(the inscribed 24-cell removed to form the snub), 18 land exactly on
kept-vertex shadow positions and only 6 visible holes remain. All 6
visible holes lie on the outer equator (axial coord along the kernel
= 0) at the maximum perp radius 1.0, occupying 6 of the 30 evenly-spaced
slots on the unit circle (the other 24 are kept balls). This is the same
camouflage effect seen in the grand antiprism vertex-first projection
(see grand_antiprism/RESULTS.md): vertex-first projections of any
diminished 600-cell hide most of the deleted vertices via overlap with
kept-vertex shadows.
Since snub-24 vertices = 600-cell vertices \ inscribed-24-cell vertices, every snub edge is a 600-cell edge. So every zomeable kernel of the 600-cell is automatically zomeable for the snub.
The 600-cell has exactly one zomeable orthographic shape (one H₄ orbit of kernel directions, the H₄→H₃ projection with 75 distinct balls). Both kernels (1, 0, 0, 0) and (φ², φ, 1, 0) belong to this single H₄ orbit and give the same 75-ball figure for the 600-cell.
Under the snub’s smaller symmetry group (order 1152, an index-12 sub-orbit structure of H₄), the single H₄ orbit splits into two orbits — precisely matching the two snub shapes (60 and 69 balls). Empirically:
| Polytope | n=(1,0,0,0) | n=(φ²,φ,1,0) |
|---|---|---|
| 600-cell | 75 balls (cell/vertex/face/edge — all related under H₄) | 75 balls (same shape) |
| snub-24 | 60 balls (cell-first) | 69 balls (vertex-first) |
| 24-cell long-root | 18 balls (cuboctahedron, cell-first) | MISS (kernel not zomeable) |
| 24-cell short-root | 15 balls (rhombic dodec, vertex-first) | HIT (15 balls) |
So starting from the 24-cell alone is insufficient to generate the snub’s projections (24-cell edges ≠ snub edges). The right parent is the 600-cell, and the snub’s two shapes arise from the symmetry- breaking split of the 600-cell’s single H₄ orbit. There are no additional snub projections beyond these two — the search is saturated.
In retrospect, both snub 24-cell zome models can be built directly from the well-known H₄→H₃ zome model of the 600-cell, without any 4D computation:
Start from the standard 75-ball, vertex-first/cell-first
projection of the 600-cell (file 600cell/600cell_H4_to_H3.vZome
in this repo). This single 3D figure is the only zomeable
projection of the 600-cell.
Identify an inscribed regular 24-cell among the 120 vertices of the 600-cell. There are 25 of them. After projection, they fall into 2 H₃-orbits, distinguished by whether the chosen 24-cell contains the vertex on the projection axis (±n):
| Orbit | The inscribed 24-cell shadow … | Strategy |
|---|---|---|
| (a) on-axis | ±n is one of its 24 vertices; shadow is a 15-ball rhombic dodecahedron, completely disjoint from the snub’s 60 balls. | “cell-first” snub |
| (b) off-axis | ±n not in it; shadow is 24 balls of which 18 coincide with snub-vertex projections and only 6 are exclusive to the inscribed 24-cell. | “vertex-first” snub |
Delete all balls exclusive to the chosen inscribed 24-cell (15 balls for orbit (a); 6 balls for orbit (b)) and every strut incident to a deleted ball. Where a ball is shared (orbit (b)), keep the ball but delete the struts that ran from it to a deleted neighbour.
Result: a valid snub 24-cell zome model.
This is a clean post-hoc construction; we only saw it after running the exhaustive 4D search and noticing the 600-cell relation. The saturation guarantee from the search ensures that no third snub zome model exists.
A photograph of the snub 24-cell cell-first zome model (60 balls, icosahedral outer hull) appears in:
The vertex-first projection does not appear there.
By the rep-theory argument applied to all Wythoff uniforms, the
zomeable kernels of every Wythoffian uniform 4-polytope (47 of them)
must be a subset of the kernels for the regular 4-polytopes
(empirically confirmed for A₄ in docs/UNIFORM_PLAN.md).
The snub 24-cell, being non-Wythoff, escapes this constraint. Its edges include directions that are not in any F₄-root orbit. The 2 projections found here are thus not realizable as projections of any of the 6 regular 4-polytopes.
| rng | dirs | hits | distinct shapes | ||
|---|---|---|---|---|---|
| 3 | 20474 | 60 | 2 | ||
| 4 | 135750 | 100 | 2 |
Saturated at rng=3. Both projections are exactly orthographic (M·Mᵀ eigenvalues all equal to 1.000000000, residual ~1e-15).
These two shapes use all three default colors with 60+ ball counts. A natural follow-up: do they decompose into recognisable pieces (e.g. icosahedral or rhombicuboctahedral substructures)? Visual inspection in vZome may reveal nested polyhedra.
➡️ Open this page on GitHub Pages to interact with the 3D models below (the embeds only render when this file is served via GitHub Pages, not in github.com’s markdown preview).