| Configuration | Vertices | Edges | Faces | Result | ||
| {5, 3, 3} | 600 | 3600 | {}@4 | 60 * 20 | {6} | part connected |
| 100 * 4 * 18 | {}@12 | 100 * 4 * 3 * 2 100 * 4 * 3 |
{3} {6/2} |
disconnected | ||
| 100 * 4 * 9 | {4} | part connected | ||||
| 1200 | disconnected | |||||
In the description of section 12 of {5, 3, 3} we need also the hexagons in section 4; they play a crucial role. This section seems to be hexagon's playroom.
First off, there is no immediate reason why the symmetry group splits the 8400 edges of section 12 in two groups, one with 7200 edges and the other with 1200 edges. The reason is that in section 4 there are only 3600 edges for 1200 hexagons. Those hexagons give hexagrams (or compound pairs of triangles) in section 12, but those cover only 7200 edges. And so there are 1200 edges left that do not form any regular pattern that I see.
But it becomes stranger. When you look at the 7200 edges covered by hexagrams, you will also find squares there, but some of the squares share edges amongst themselves, so it looks interesting if some more patterns can be found. And indeed, that is possible, each edge is part of one triangle and two squares, and we can split the edges amongs 400 configurations of 9 vertices, 18 edges, 6 triangles and 9 squares where the configurations do not share either edges or faces. This is however not all. When we look back at the hexagrams we see they form two co-planar triangles. When we try to fit that in the picture we find 100 super-configurations, each consisting of four configurations (A, B, C, D), such that the pairs A and B, A and C, B and D and C and D all have exactly three co-planar triangles.
So now the question is how to describe those super-configurations and the place of the basic configurions in them? Here again hexagons play a crucial role. Consider the following two hexagons given in coordinates (note that tau = (sqrt(5) + 1)/2, and the caret indicates exponentiation), the vertices are given in columns:
You may verify that they are indeed hexagons. (I could have scaled them by a factor tau, but I did not do so because this version leads to the actual coordinates of the vertices of one such construct in these files.)( tau^2, tau^2, tau^2, tau^2) ( tau^2, tau^2, -tau^2, -tau^2) ( tau^3, -tau , tau^2, 0) ( 0, tau^2, tau , -tau^3) ( tau , -tau^3, 0, -tau^2) (-tau^2, 0, tau^3, -tau ) (-tau^2, -tau^2, -tau^2, -tau^2) (-tau^2, -tau^2, tau^2, tau^2) (-tau^3, tau , -tau^2, 0) ( 0, -tau^2, -tau , tau^3) (-tau , tau^3, 0, tau^2) ( tau^2, 0, -tau^3, tau )
Now call the left set A and the right set B. You can verify that the vector from the origin to a vertex of A is orthogonal to the vectors from the origin to each of the vertices of B, so they are in mutually orthogonal 2D spaces (and the game can not be played in 3D).
Now superimpose on each vertex of A an hexagon of type B with that vertex as centre (just a shift of coordinates). And do the same with the roles of A and B reversed. This gives a new set of 12 hexagons, and that set describes the super-configurations I talked about earlier.
Next number the vertices of the hexagons above from 1 to 6. The vertices of the super-configuration are of the form Ai + Bj with i and j ranging from 1 to 6, this gives indeed 36 vertices. The vertices can be distributed as follows amongst the four basic configurations:
| config | i | j |
| 1 | odd | odd |
| 2 | even | even |
| 3 | odd | even |
| 4 | even | odd |
The triangles in each configuration are found by the three vertices that have either the same i or the same j. And we find (for instance) that the three triangles in configuration 1 with fixed i are coplanar with the three triangles in configuration 3 with fixed i (and come from three hexagons).
Now the squares in these configurations are not so difficult to find. Take four points Ai + Bj where both i and j have at most two permissible values for a configuration. There are nine such tuples (there are both for the pair if i's and for the pair of j's three choices). We can order them as follows:
Ai1 + Bj1, Ai1 + Bj2, Ai2 + Bj2, Ai2 + Bj1
It is clear that the distances between successive points in this list
(assuming a circular list) are equal. Next it is easy to show that that
foursome give a planar figure. The diagonals between the first and
third element meet the diagonal between the second and fourth element
(add those elements together and the sums are equal). So the four form
at least a diamond. But a planar intersection with the hypersphere that
encloses {5, 3, 3} is a circle, and so the four points are on a
circle. This is only possible if they form a square.
The following table shows how the vertices for one of the figures can be found (it is the first construct in the tables, vertices point to the canonical numbering):
| + | B1 | B2 | B3 | B4 | B5 | B6 |
| A1 | V001 | V506 | V297 | V021 | V257 | V427 |
| A2 | V226 | V122 | V190 | V573 | V109 | V159 |
| A3 | V496 | V206 | V150 | V271 | V175 | V096 |
| A4 | V024 | V264 | V438 | V004 | V519 | V304 |
| A5 | V580 | V116 | V162 | V231 | V135 | V195 |
| A6 | V266 | V178 | V097 | V497 | V211 | V139 |
But the hexagons in section 4 also are partly connected and show interesting configurations. Each edge is shared by two hexagons. And the set of 1200 hexagons can be split in 60 groups of 20 hexagons such that there is no hexagon from one group that shares an edge with an hexagon from another group. Within each group there are 20 hexagons, 60 edges and 30 vertices. At each vertex four hexagons meet, and this makes clear that it is not a 3D object. Further analysis of these objects appears to be complex. E.g. where are the centres of all those hexagons and how are the hexagons located on these centres? As there are 20 such centres, you might think that the centres form an icosahedron. That is not the case, you can again find hexagons amongst the centres! And indeed the configuration of centres is dominated by hexagons. One of the configurations can be found by taking all coordinates of the forms:
( tau^2, tau^2, +-tau^2, +-tau^2)
( 0, 0, 2.tau^2, 0)
( 0, 0, 0, 2.tau^2)
( tau^3, -tau, +-tau^2, 0)
( -tau, tau^3, 0, +-tau^2)
together with the diametrically opposite vertices (i.e. changing
the sign of all coordinate entries). But more is to be done here.