A polyhedron is said to be uniform if its faces are regular polygons and its vertices are all alike, that is, for any two vertices there is a symmetry of the whole polyhedron that moves one vertex into the other. It is regular if it is uniform and all the faces are alike, that is, for any two faces there is a symmetry of the whole polyhedron that moves one face into the other.
There are five regular convex polyhedra. They were known to Plato, and are often called "the Platonic solids". They are familiar to most people interested in geometry: the tetrahedron, the cube, the octahedron, the dodecahedron, and the icosahedron. Other convex uniform polyhedra are often called "Archimedean solids" or "Archimedean polyhedra." There are two infinite families of these, namely the prisms and antiprisms, and thirteen others.
A prism has two n-gonal faces; one of these faces may be obtained from the other by translation in the direction perpendicular to the plane of the face. Between the n-gonal faces is a belt of n squares.
An antiprism also has two n-gonal faces; one may be obtained from the other by translation as before, followed by a rotation through π/n radians. Between these is a belt of 2n equilateral triangles, having edges in common with the first and second n-gons alternately.
If a convex uniform polyhedron is enclosed in a sphere concentric with the polyhedron, then the faces of the polyhedron may be projected onto the sphere. The result is a tessellation of the sphere; it is "uniform" in that each face is a regular spherical polygon and the vertices are all alike in the same sense as for polyhedra.
The symmetry group of a uniform polyhedron or tessellation can be a kaleidoscope group of rotations and reflections, or the even subgroup, which consists of the rotations and translations. (These are products of even numbers of generating reflections.) For a polyhedron or spherical tessellation, the kaleidoscope group must be one of the spherical groups. The other kaleidoscope groups are also symmetry groups of uniform tessellations, of the Euclidean plane or the hyperbolic plane.
The reflections in two edges of the fundamental triangle generate a dihedral group of rotations and reflections; the rotations are centered at the intersection of the two sides. Any other point V is moved by the elements of this group into the vertices of a polygon. If the angle of the intersection is π/p, then the polygon has p sides if V is on the edge defining one of the two reflections; in this case, the polygon is regular. If V is elsewhere, then the polygon has 2p sides; it is regular if (and only if) V is equidistant from the two edges that define the reflections.
We can assemble such polygons into uniform tessellations. A systematic way of doing this is called Wythoff's construction. In this construction, we select a point V in the fundamental triangle, and generate regular polygons from V as described in the preceding paragraph. The images of these polygons under all the symmetries of the kaleidoscope group (or the rotational subgroup) make up our tessellation or polyhedron. There are several ways to select V, and for each of them a concise notation.
Let the angles of the fundamental triangle be π/p, π/q, π/r; then (p q r) is a notation for the Kaleidoscope group. The vertices of the fundamental triangle are labeled P, Q, R.
The regular and quasi-regular cases
Special cases arise if one or more of the numbers p,q, r is equal to 2.
Regular tessellations p | q 2
The dihedron 2 | q 2
A bundle of sticks: p | 2 2
Quasi-regular tessellations 2 | q r
The truncated and rhombic cases
Truncated tessellations 2 q | r
Rhombic tessellations p q | 2
The prisms 2 q | 2
The great-rhombic case
The Archimedean polyhedra 2 3 4 | and 2 3 5 | have traditionally been called the "truncated cuboctahedron" and "truncated icosidodecahedron." Actual truncation of a quasi-regular polyhedron would produce rectangular faces, not squares. For this reason, the names "great rhombicuboctahedron" and "great rhombicosidodecahedron" are becoming more current.
The snub case
The choice of V for this construction is more difficult than for the other cases. The condition that it generate equilateral triangles can be expressed as requiring it to be at an intersection of two conics. Thus its coordinates are the roots of quartic equations, and in general are not constructible with ruler and compass.
The Archimedean polyhedra | 2 3 4 and | 2 3 5 have traditionally been called the "snub cube" and "snub dodecahedron;" recently people have started calling them "snub cuboctahedron" and "snub icosidodecahedron." When they are displayed by the kaleidoscope applet, the snub triangles can have a different color from the triangles that are centered on kaleidoscope vertices.
The antiprisms | 2 2 r
The snub tetrahedron | 2 3 3
Duality is an important relation between different tessellations or different polyhedra. In it, the vertices of one figure correspond to faces of the other, and vice versa.
In the case of polyhedra related to a spherical kaleidoscope, duality can be defined in terms of polarity. Let S be a sphere. For a point P on S, the polar of P is the plane through P tangent to S; and P is the pole of that plane. If P is outside of S, then the polar of P passes through the points of contact of tangents to S passing through P. If a plane intersects S in a circle, then the tangents to S at the points of intersection are concurrent at the pole of the plane. For a point inside the sphere, or a plane which does not intersect it, polarity can be determined by the rule that if P is on the polar plane of Q, then Q is on the polar plane of P. The polar line of a line L is another line, which is the intersection of all the polars of points on L, and also the join of all the poles of planes passing through L.
When a polyhedron has a center of symmetry, it is natural to construct its dual by means of polarity with respect to a sphere which shares the same center. The vertices of the dual polyhedron are the poles of the face planes of the original polyhedron. Two vertices of the dual are joined by a dual edge if the corresponding faces of the original meet in a edge. Then the former edge is on the dual line of the latter. The face planes of the dual polyhedron are the polars of the vertices of the original one.
Let us take a regular polyhedron for an example. The midpoints of its edges lie on a sphere, which we can use to dualize the polyhedron. The dual edges are perpendicular bisectors of the edges of the original polyhedron. Over each face of the original, dual edges meet at the dual vertex. The edges coming from an original vertex have their midpoints arranged in a circle, and in fact they form a regular polygon; the dual edges also form a regular polygon which is a face of the dual polyhedron. If the original polyhedron is p | q 2, which has q-sided faces with p at each vertex, then the dual has p-sided faces with p at each vertex; therefore it is q | p 2.
The other uniform polyhedra can be dualized in a similar way. Because the vertices of the original, uniform polyhedron are all alike, they lie on a common sphere, which is the circumsphere. The edges are of equal length; therefore, for any vertex V, the vertices connected to it by edges are equidistant from it; so, they lie on a small circle on the circumsphere, and also on the plane containing that circle. The midpoints of the edges that connect to V also lie on a circle; they are the corners of a polygon called the vertex figure. Also, we see that the midpoints of all the edges are equidistant from the center of the circumsphere, so they lie on a common sphere, the midsphere.
We construct the dual polyhedron with respect to the midsphere. The dual edges are tangent to the midsphere, and those which correspond to the edges of an original face meet in a point over the center of that face; of course, this is the pole of that facial plane. For the edges that connect to the original vertex V, the dual edges lie in the plane of the vertex figure, and in fact are tangent to the circle in which the vertex figure is inscribed. They are the edges of the dual face.
The dual uniform polyhedra are symmetrical, just as are the uniform polyhedra, but in a different way. The faces are not, in general, regular polygons. But they are all congruent, and for any two there is a symmetry of the whole polyhedron which takes one into the other. The vertices of the dual polyhedron are not all alike in general. But at each vertex, the facial angles are equal, and the dihedral angles between faces are all equal.
Wythoff's construction of uniform polyhedra can be extended to the dual uniform polyhedra. This may be the best way of seeing that dualization can be extended to uniform tessellations. Given a uniform tessellation, the dual vertices are at the centers of the original faces. The dual edges are perpendicular bisectors of the original edges. Each original vertex is the in-center of the corresponding dual face, that is, it is the center of a circle tangent to the edges of the dual face. The dual Wythoff construction starts from the fundamental triangle PQR of the Kaleidoscope group, and a particular point V in that triangle; but this time, V is to be the in-center of a face of the tessellation. Points like the corners of the fundamental triangle are the vertices. For a polyhedron, V is the pole of a facial plane.
The dual regular and dual semi-regular cases
Dual regular tessellations D(p | q 2)
Dual quasi-regular tessellations D(2 | q r)
The dual truncated and dual rhombic cases
Dual truncated tessellations D(2 q | r)
Dipyramidal tessellations D(2 q | 2)
The Kaleidoscope as dual uniform tessellation
The dual snub case
The trapezohedra D(| 2 2 r)
The Wythoff construction may be found in the paper by H. S. M. Coxeter and others, listed below. In this paper, one can find out about a much wider class of uniform polyhedra, most of which are non-convex. Fr. Wenninger's book Polygonal Models gives photographs of cardboard models of these, and advice on their construction. His book Dual Models treats of the non-convex dual uniform polyhedra and some of the stellations of uniform polyhedra. Cromwell's book is an interesting recent survey of the geometry of polyhedra. Finally, Mathematical Models by Cundy and Rollett is a book dear to my heart, full of instructions and advice on the construction of these models and other interesting topics.
Coxeter, H. S. M; Longuet-Higgins, M. S.; Miller, J. C. P: Uniform Polyhedra, Philos. Trans. Roy. Soc. London Set. A 246(1954),401-450.
Cromwell, Peter R.: Polyhedra, Cambridge, Cambridge University Press, 1997.
Cundy, H. Martyn; Rollett, A. P.: Mathematical Models, Stradbroke (England), Tarquin Publications, 1997.
Wenninger, Fr. Magnus J., OSB: Dual Models, Cambridge, Cambridge University Press, 1983.
Wenninger, Fr. Magnus J., OSB: Polygonal Models, Cambridge, Cambridge University Press, 1971.