Regular Polytopes: The Dimensional Ladder of Perfection

The Classification Problem

How many perfectly regular shapes exist? The answer depends on the dimension — and the answer is surprising.

In 2D, infinitely many: one regular polygon for every integer $n \geq 3$ . In 3D, exactly five: the Platonic solids, known since antiquity. In 4D, there are six — three more than you’d expect, with no analogues in any other dimension. In every dimension 5 and above, exactly three.

The constraint that produces this pattern is Schläfli’s criterion. A regular polytope $\{p, q, r, \ldots\}$ must satisfy a positive-definiteness condition on its Gram matrix — the same kind of condition that governs the Niggli cone in crystallography. The number of solutions to this condition is finite in each dimension, and dimension 4 is the unique case where the constraint surface admits exceptional solutions.

24-Cell

{3,4,3}

24 vertices, 96 edges* exists only in 4D

drag to orbit · scroll to zoom

Why Dimension 4 Is Exceptional

In 3D, three Platonic solids form a family (tetrahedron, cube, octahedron — the simplicial/cubical/cross-polytope series) and two are exceptional (dodecahedron, icosahedron — the fivefold symmetry objects). These two have no direct generalization to higher dimensions.

In 4D, the three-family pattern continues (5-cell, tesseract, 16-cell) and three exceptionals appear:

The 24-cell $\{3,4,3\}$ — 24 vertices, 96 edges. Self-dual (its dual is another 24-cell). Its vertices are the 24 roots of the $D_4$ root system, the same root system that governs triality in Lie theory. No analogue exists in any other dimension because $D_n$ root systems only form a regular polytope when $n = 4$ .

The 600-cell $\{3,3,5\}$ — 120 vertices, 720 edges. The 4D analogue of the icosahedron, built from 600 tetrahedra. Its vertices include the 120 elements of the binary icosahedral group, connecting polytope geometry to finite group theory.

The 120-cell $\{5,3,3\}$ — 600 vertices, 1200 edges. The dual of the 600-cell. Built from 120 dodecahedral cells, each a regular dodecahedron. The densest of all regular polytopes.

For dimensions $n \geq 5$ , only the simplex, hypercube, and cross-polytope survive. The constraint surface becomes too tight for exceptional solutions.

Projection from Higher Dimensions

To render an $n$ -dimensional object on a 2D screen, we project twice: first from $n$ D to 3D, then from 3D to 2D via a standard camera.

The $n$ D to 3D projection uses successive perspective projections — collapsing one dimension at a time. For a point $\mathbf{p} \in \mathbb{R}^n$ , each collapsed dimension $k$ applies:

$p_i' = \frac{d_k}{d_k - p_k} \cdot p_i \quad \text{for } i < k$

where $d_k$ is the “view distance” in dimension $k$ . The product of all scale factors $\prod d_k / (d_k - p_k)$ encodes the total projection depth — how deep a point sits in the higher-dimensional structure. This drives the depth-attenuated brightness: nearer structures glow bright white, deeper structures fade to blue.

Moving the projection depth slider from perspective toward orthographic increases all $d_k$ simultaneously, flattening the projection so that internal structure becomes more visible.

Rotation in Higher Dimensions

In 3D, rotation happens around an axis. In 4D and above, rotation happens in a plane. A Givens rotation in the $(i,j)$ plane applies a 2D rotation to coordinates $i$ and $j$ while leaving all others fixed.

A 4D object has $\binom{4}{2} = 6$ rotation planes: $XY, XZ, XW, YZ, YW, ZW$ . The visualization auto-rotates in two planes simultaneously at speeds in the golden ratio $1 : \varphi$ , producing quasi-periodic motion that never exactly repeats. This is what gives higher-dimensional projections their hypnotic quality — the structure breathes and morphs continuously without cycling.

Duality

Every regular polytope has a dual obtained by swapping vertices with cells. The dual of a dual is the original.

Polytope	Dual	Self-dual?
Tetrahedron	Tetrahedron	yes
Cube	Octahedron	no
Dodecahedron	Icosahedron	no
5-Cell	5-Cell	yes
Tesseract	16-Cell	no
24-Cell	24-Cell	yes
120-Cell	600-Cell	no

The 24-cell’s self-duality is remarkable: it is one of only two self-dual regular polytopes in 4D (alongside the 5-cell), and its vertex/cell symmetry reflects the triality of $D_4$ .

Connection to Root Systems

The vertices of several regular polytopes coincide with root systems from Lie theory:

Octahedron = $B_2$ roots (6 vertices)
Cuboctahedron = $A_3$ roots (12 vertices, not regular but Archimedean)
24-cell = $D_4$ roots (24 vertices)
600-cell includes the 120 elements of the binary icosahedral group $2I$ , related to $E_8$ via the McKay correspondence

The $D_4$ connection is the deepest: the 24-cell is the only regular polytope whose vertices form the complete root system of a simple Lie algebra. The triality automorphism of $D_4$ — which cyclically permutes the three 8-dimensional representations — manifests geometrically as the 24-cell’s three-fold symmetric projection.

Built with TypeScript + raw WebGL2. Bloom post-process via separable Gaussian blur. No external 3D dependencies.