Graphically Speaking Editor: André Stork

Beautiful Math, Part 6: Visualizing 4D Regular

Polytopes Using the Kaleidoscope Principle
Peichang Ouyang and Xinchang Wang
Jinggangshan University

Yongman Zhao
Shihezi University

S ymmetry is a basic and mysterious law that

can be widely found in natural phenomenon,
including physics, biology, and astronomy.1
The beauty of symmetrical structures appeals to
the artistic senses, and our fascination predates
In this article, we first introduce the funda-
mental root systems of 4-RPs. Then, using a
fundamental region algorithm, we present three
interesting methods to visualize 4-RPs. The em-
ployed algorithm in fact is the generalized ka-
any mathematical investigation of symmetries. leidoscope principle in high-dimensional Euclid-
Various civilizations in ancient times applied ean space. We obtain many attractive 2D or 3D
symmetry in art and architecture.2 Some people projection figures of 4-RPs. To a certain extent,
believe that God used symmetrical rules to create this preserves the metric and topology data of
this world.3,4 4-RPs so that we can identify their general sym-
Regular polygons and polyhedra are the most metrical structure. To the best of our knowledge,
basic and important symmetrical structures in 2D such figures have never appeared in the litera-
and 3D Euclidean space—R2 and R3, respectively. In ture before. These visualization methods may
higher-dimensional space, regular polytopes (RPs) provide new approaches to study complex RPs
are the generalized analog of regular polygons and in R n (n ≥ 4).
polyhedra.5,6 Perfect symmetry is their most no-
table feature, which gives RPs an aesthetic quality Fundamental Root Systems of Regular 4-RPs
that interests both mathematicians and nonmath- Four-dimensional regular polytopes (RPs) 4-RPs
ematicians. For example, 4-RP (RP in R4) 600- are the 4D analogs of regular polyhedra in three
cell consists of 600 regular tetrahedrons with 20 dimensions and regular polygons in two dimen-
meeting at each vertex. Together they form 1,200 sions. Five of them may be thought of as close an-
regular triangles, 720 edges, and 120 vertices. The alogs of the Platonic solids. Each 4-RP is bounded
edges form 72 flat regular decagons. Each vertex of by a set of 3D cells, which are all Platonic solids
the 600-cell is a vertex of six such decagons. of the same type and size. These are fitted to-
Projections are one important way to visualize gether along their respective faces in a regular
n-RPs (RP in R n, n ≥ 4). Examples include per- fashion. A 4-RP can be briefly represented as a
spective,7 orthogonal,8 and canonical projections.9 Schläfli symbol in the form {p, q, r}, where p is the
Those projections mainly aim to maintain their number of sides of each regular polygon, q is the
symmetries to the highest degree, and the corre- number of polygons meeting at each vertex, and r
sponding figures merely transmit vertex and edge is the number of regular polyhedra {p, q} meeting
information. In other words, such projections lack along each edge.5 Each {p, q, r} has a dual regu-
crucial metric or topology data. Actually, it is hard lar 4-RP {r, q, p}. They share the same symmetry
to apperceive real geometric structures from pro- group, which is usually denoted as [p, q, r]. There
jection figures. are six total 4-RPs: pentachoron or 5-cell {3, 3, 3},

92 September/October 2017 Published by the IEEE Computer Society 0272-1716/17/$33.00 © 2017 IEEE
 Table 1. Fundamental root systems with respect to symmetry group [p, q, r].*
Name Schläfli symbol Symmetry group Fundamental root systems D[p, q, r] = {v1, v2, v3, v4} Order of [p, q, r]
5-cell {3, 3, 3} [3, 3, 3] (1, –1, 0, 0), (0, 1, –1, 0), 120

(0, 0, 1, –1), (1, 0, 0, –1).

16-cell {3, 3, 4} [3, 3, 4] (1, –1, 0, 0), (0, 1, –1, 0), 384
8-cell {4, 4, 4} (0, 0, 1, –1), (0, 0, 0, 1).
24-cell {3, 4, 3} [3, 4, 3] (0, 1, –1, 0), (0, 0, 1, –1), 1,152

(0, 0, 0, 1), (0.5, –0.5, –0.5, –0.5).

600-cell {3, 3, 5} [3, 3, 5] (a, –0.5, b, 0), (–a, 0.5, b, 0), 14,400

120-cell {5, 3, 3} (0.5, b, –a, 0), (–0.5, –a, 0, b).

1+ 5 5 −1
*a = and b =
4 4

hexadecachoron or 16-cell {3, 3, 4}, octachoron ∏v k(k = 1, 2, 3, 4) are reflections associated with
or 8-cell {4, 3, 3}, icositetrachoron or 24-cell {3, v k. Then {∏v1, ∏v2 , ∏v3 , ∏v4} is a set of generators
4, 3}, hecatonicosachoron or 120-cell {5, 3, 3}, of [p, q, r]. For x 0 ∈ R 4 outside fundamental re-
and hexacosichoron or 600-cell {3, 3, 5}.5,6 gion D [p, q, r], the algorithm in Figure 1 describes
A reflection is a linear transformation P that how to find a transformation GN so that x N = GN
carries each vector to its mirror image with respect (x0) ∈ D [p, q, r]. This algorithm provides a fast way
to a fixed hyperplane P. More precisely, let P T be to transform a point into fundamental region
the orthogonal space of P. Then P(x) = x if x ∈ P D [p, q, r] symmetrically. Each time a chosen reflec-
and P(x) = – x if x ∈ P T. If v is a nonzero vector, tion ∏v k is employed on x n–1, the resulting x n =
then the reflection Pv determined by v is ∏v k (x n–1) in step 2 will get nearer to D [p, q, r], and
ultimately, a corresponding x N will fall into D [p,
q, r]. Essentially, this algorithm is the kaleidoscope
( x, v )
Πv ( x) = x − 2 v, principle in higher-dimension Euclidean space.5–7
(v, v ) For convenience, we call this the fundamental re-
gion algorithm (FRA).
where ( , ) denotes the inner product of vectors.
The fundamental region under group [p, q, r] is
a connected set, the transformed copies of which Step 1: Let n = 1.
under the action of [p, q, r] cover the entire space Step 2: For k = 1 to k = 4
without overlapping, except at the boundaries.5 The If (xn –1, vk) < 0
fundamental region of 4-RP can be sententiously Do
described using a fundamental root system.10 The {
fundamental root system with respect to [p, q, r]
( x n−1, v k )
is a set of vectors, with associated reflections that x n = Π v ( x n−1) = x n−1 − 2 vk
(v k , v k )
k

are generators of [p, q, r].

Suppose D[p, q, r] = {v1, v2, v3, v4} is a fundamental Pn = Pvk
root system with respect to [p, q, r]. We can prove Break.
that }
Step 3: If (xn, vk) < 0 for certain k ∈ {1, 2, 3, 4},
D [p, q, r] = {ξ ∈ R4  (ξ, vk) ≥ 0, ∀vk ∈ ∆[p, q, r]} n = n + 1.
Go to step 2.
is a fundamental region associated with [p, q, r]. Else
Geometrically, D [p, q, r] is a convex cone surrounded Go to step 4. (In this situation, keep in mind that (xn, vk) ≥ 0
by four hyperplanes through the origin. The normal for all k ∈ {1, 2, 3, 4}—namely, xn ∈ D [p, q, r].)
vectors of those hyperplanes constitute D[p, q, r]. For Step 4: Assume N is the repeated times of step 2. Let GN =
convenience, Table 1 lists the fundamental root PN × PN –1 × … × P2 × P1. Then xN = GN (x0) ∈ D [p, q, r].
system of the regular 4-RPs. For more details, see
earlier work.10 Figure 1. Fundamental region algorithm (FRA). This algorithm provides
Assume D[p, q, r] = {v1, v 2 , v 3, v4} is a fundamen- a fast way to transform a point into fundamental region D [p, q, r]
tal root system of symmetry group [p, q, r], and symmetrically.

IEEE Computer Graphics and Applications 93

Graphically Speaking

[3, 3, 3] [3, 3, 4] [3, 4, 3] [3, 3, 5]

[3, 3, 3] [3, 3, 4] [3, 4, 3] [3, 3, 5]

Figure 2. Solid sphere (with radius R* = 4) projections of 4-RPs based on generalized stereographic projection
F1. As the order of symmetry group [p, q, r] increases, the symmetrical structure of the corresponding sphere
becomes more complex and beautiful.

4-RPs Visualizations Based on Stereographic be a solid sphere with radius R*. The generalized
Projection and FRA stereographic projection F1 provides a novel way
For (x1, x2, x3, x4) ∈ R4 and (X1, X 2, X3) ∈ R 3, to observe 4-RPs in 3D space.
the generalized stereographic projection11 can be Figure 2 shows the solid sphere (with radius R* =
expressed as 4) projections of 4-RPs. To reveal the inner details,
one-fourth (first row) or one-eighth (second row)
( X1 , X 2 , X3 ) = Φ1 ( x1 , x2 , x3 , x4 ) = of the solid sphere has been cut off. Depending on
whether the reflection times (the index N of GN
2
= ( x1 , x2 , x3 ) . determined by FRA) are odd or even, we colored
1 + x4
the corresponding points with two different
colors. For clarity, we used eight colors to show
The corresponding inverse mapping is the cross sections of the second row of Figure 2.
We can see that, as the order of symmetry group
( x1, x2 , x3 , x4 ) = Φ1−1 ( X1, X 2 , X3 ) [p, q, r] increases, the symmetrical structure of
the corresponding sphere becomes more complex
1
=
1 + ω2
( X1 , X 2 , X3 ,1 − ω 2 ) and beautiful.
Let (x1, x2, x3) ∈ R 3 and (X1, X 2) ∈ R 2. Then,
where the familiar stereographic projection is defined as

x12 + x22 + x32

ω2 = . ( X1 , X 2 ) = Φ2 ( x1, x2 , x3 )
4
1
Geometrically, F1 is the generalized stereographic = ( x1, x2 ).
1 − x3
projection from the point (0, 0, 0, –1).
For arbitrary (X1, X 2, X3) ∈ R 3, it is easy to check Its inverse mapping is
that modulus
( x1 , x2 , x3 ) = Φ−2 1 ( X1 , X 2 )
1 ( X 1 , X 2 , X 3 ) = x1 + x 2 + x3 + x 4 = 1
Φ−1 2 2 2 2 , 1
=
x12 + x22 + 1
(2 X1, 2 X 2 , X12 + X 22 − 1).
which means that the pre-image of (X1, X 2, X3)
under F1 always lies on a 4D unit sphere. Let It is easy to check that modulus

S = {( X1 , X 2 , X3 ) ∈ R 3 | X12 + X 22 + X32 ≤ R* } Φ−2 1 ( X1 , X 2 ) = x12 + x22 + x32 = 1 ,

94 September/October 2017
 [3, 3, 3] R = 0.1 R=1 R=2 R=3 R=4 R=8

[3, 3, 4] R = 0.5 R = 0.9 R=1 R = 1.5 R = 10 R = 1,000

[3, 3, 4] R = 0.5 R = 0.6 R = 0.7 R = 0.9 R=1 R = 10

[3, 3, 5] R = 0.3 R = 0.6 R=1 R = 2.5 R = 10 R = 100

{ }
Figure 3. Planar projections of 4-RPs based on mapping (x 1 , x 2 , x 3 , x 4 ) = Φ1−1 R  Φ1−1 (X1 , X2 ) . For small and large
R that exceed certain thresholds, almost every 4-RP has the same projections (see the first and last columns),
but the projections vary sharply once R approaches 1.

which means that the pre-image of (X1, X 2) under observe 4-RPs. Let B and C be constants satisfying
F2 always lies on a 3D unit sphere. Geometrically, C > B > 0. Define a space SBC as
the stereographic projection F2 sends the points
S BC = {( x1 , x2 , x3 , x 4 ) ∈ R 4 | x 4
of unit sphere from the North Pole (0, 0, 1) onto
the plane Z = 0. = B, x12 + x22 + x32 = C 2 − B2 } .
Thus, we see that the composite mapping F2F1
projects a 4D point onto plane Z = 0. This provides Roughly speaking, this space equals a 4D plane x4 =
a 2D viewpoint to observe 4-RPs. In practice, for B, but it is a curved “plane” in the sense that point
a point (X1, X 2) ∈ R 2, we first computed the cor- (x1, x2, x3) always lies on the 3D sphere of radius
responding point C 2 − B2 .
Using space SBC and the FRA, we can show a real
( x1 , x2 , x3 , x4 ) = Φ1−1 {R Φ−2 1 ( X1 , X 2 ) } ∈ R 4 . 4-RP plane (not a projection plane) in a 3D sphere.
For example, Figure 4 demonstrates the most
Note that after multiplying R point R Φ− 
1 ( X 1 , X 2 )
1
complex symmetrical structure of {3, 3, 5}. In the
lies on the 3D sphere of radius R. Then, we used first row, we first fixed C = 1 but gradually decrease
the FRA to color (X1, X 2). Figure 3 shows the re- B. The results show that as plane x4 = B approaches
sulting figures. Compared with the simpler pro- the origin, SBC illustrates increasingly finer details.
jections of regular solids,12 the complexity of the In the second row, we fixed B = 2 but decrease C.
planar projections of 4-RPs was far beyond our In this process, we see that SBC gradually loses fine
imagination. We found that for small and large R structures, and for sufficiently small C, SBC will
that exceed certain thresholds, almost every 4-RP degenerate into a common sphere tiling with [3,
has the same projections (see the first and last col- 5] symmetry (see the last figure in Figure 4b).
umns in Figure 3), but the projections vary sharply
once R approaches 1.
In practice, we found an interesting space to

IEEE Computer Graphics and Applications 95

Graphically Speaking

(a) C = 1, B = 0.9 C = 1, B = 0.5 C = 1, B = 0.025 C = 1, B = 0.0125

(b) C = 10, B = 2 C= 7, B = 2 C= 4.5, B = 2 C= 4.1, B = 2

Figure 4. Symmetry structure of {3, 3, 5} in space S BC: (a) a fixed C = 1 and decreasing B and (b) a fixed B = 2
and decreasing C.

I n parts 1–5 of the Beautiful Math articles,13–17

we generated many beautiful patterns on the
symmetrical spaces using the dynamic system
Acknowledgments
This work was supported by the Natural Science
Foundation of China (grant 11461035), Science
method. Here, we used this method to produce and Technology Project of the Education Depart-
colored 4-RP patterns. Figure 5 displays six aes- ment of Jiangxi Province (grant GJJ160749) and
thetic patterns based on the generalized stereo- Doctoral Startup Fund of Jingangshan University
graphic projection F1. (grant JZB1303).

(a)

(b)

Figure 5. Aesthetic patterns with symmetries of (a) [3, 3, 5] and (b) [3, 3, 4]. These six patterns are based on
the generalized stereographic projection F1.

96 September/October 2017
References Xinchang Wang is an associate professor at Jinggangshan
1. B.L. Tarasov, This Amazingly Symmetrical World, Mir University’s School of Mathematics and Physics. His re-
Press, 1986. search interests include scientific visualization. Wang has a
2. P.S. Stevens, Handbook of Regular Patterns, MIT Press, PhD from the Jiangxi University of Finance and Economics.
1980. Contact him at wangxinchang11@[Link].
3. I.N. Stewart and M. Golubitsky, Fearful Symmetry: Is
God a Geometer? Blackwell, 1992. Zhao Yongman (corresponding author) is an associate
4. G. Hooft, Does God Play Dice? Iop Publishing, professor at Shihezi University’s School of Industrial En-
2007. gineering. His research interests include data analysis and
5. H.S.M. Coxeter, Regular Polytopes, Dover, 1973. mining. Yongman has a PhD in data mining from Tianjin
6. H.S.M. Coxeter, Regular Complex Polytopes, Cambridge University. Contact him at zhrym@[Link].
Univ. Press, 1974.
7. P. Mcmullen and E. Schulte, Abstract Regular Contact department editor André Stork at [Link]@igd
Polytopes, Cambridge Univ. Press, 2002. .[Link].
8. C. Séquin and J. Lanier, “Hyperseeing the Regular
Hendecachoron,” Proc. 6th Interdisciplinary Conf.
Int’l Soc. Arts, Mathematics, and Architecture (ISAMA),
2007, pp. 161–168. Read your subscriptions through
9. C.A. Reiter, “Atlas of Quasicrystalline Tilings,” Chaos the myCS publications portal at
[Link]
Solitons & Fractals, vol. 7, no. 14, 2002, pp. 937–963.
10. J.E. Humphreys, Reflection Groups and Coxeter
Groups, Cambridge Univ. Press, 1990.
11. A.F. Beardon, The Geometry of Discrete Groups,
Springer-Verlag, 1983.
12. M. Wihelm, Noneuclidean Tessellation and Their
Groups, Academic Press, 1974.
13. P.C. Ouyang and X.C. Wang, “Beautiful Math—
Aesthetic Patterns Based on Logarithmic Spirals,”
IEEE Computer Graphics and Applications, vol. 33, no.
6, 2013, pp. 21–23.

CONFERENCES
14. P.C. Ouyang, R.W. Fathauer, “Beautiful Math, Part
2, Aesthetic Patterns Based on Fractal Tilings,” IEEE
Computer Graphics and applications, vol. 34, no. 1, in the Palm of Your Hand
2014, pp. 68–75.
15. P.C. Ouyang and K.W. Chung, “Beautiful Math, Let your attendees have:
Part 3, Hyperbolic Aesthetic Patterns Based on • Conference Schedules
Conformal Mappings,” IEEE Computer Graphics and • Notifications
Applications, vol. 34, no. 2, 2014, pp. 72–79. • Networking
• And more
16. P.C. Ouyang, F. Ding, and X.C. Wang, “Beautiful
The conference program mobile app
Math, Part 4: Polygonal Aesthetic Patterns Based on works for Android devices, iPhone,
the Schwarz-Christoffel Mapping,” IEEE Computer iPad, and the Kindle Fire.
Graphics and Applications, vol. 34, no. 4, 2015, pp.
23–26.
17. P.C. Ouyang, W. Zhao, and X. Huang, “Beautiful
Math, Part 5: Colorful Archimedean Tilings from
Dynamical Systems,” IEEE Computer Graphics and
Applications, vol. 35, no. 6, 2015, pp. 90–96.
For more information please contact
Conference Publishing Services (CPS) at
cps@[Link]
Peichang Ouyang is an associate professor at Jinggang-
shan University’s School of Mathematics and Physics. His
research interests include fractal, tiling, and hyperbolic
geometry. Ouyang has a PhD in information and compu-
tational science from SunYat-Sen University. Contact him
at g_fcayang@[Link].

IEEE Computer Graphics and Applications 97