| Configuration | Vertices | Edges | Faces | Cells | Result | |||
| {3, 3, 5} | 120 | 1200 | {}@2 | 2400 | {3} | 5 * 5 * 24 | {3, 4} | 5 * 5 * {3, 4, 3} |
| [75] * 24 | {4} | [75] * 8 | {4, 3} | [75] * {4, 3, 3} | ||||
| 200 | {6} | disconnected | ||||||
| [75] * 24 | {}@4 | [75] * 32 | {3} | [75] * 16 | {3, 3} | [75] * {3, 3, 4} | ||
| [75] * 4 | {3, 4} | disconnected | ||||||
| [75] * 6 | {4} | disconnected | ||||||
| 200 * 6 | {}@6 | 200 * 2 200 |
{3} {6/2} |
disconnected | ||||
All basic information comes from the first line, where 25 {3, 4, 3}'s are found, the other lines derive from the compounds found in a basic {3, 4, 3}. The 25 {3, 4, 3}'s can be grouped in five sets of five such that each set of polytopes covers the vertices exactly once. Actually this grouping can be done in two different ways. The numbering in the data files is such that a subsection of member m in set s is indicated as (s.m). Interchanging the roles of those two numbers give the other grouping. As the hexagons found are disconnected, they do not play a role in the further discussion.
The symmetry group is transitive over the sets of polytopes, however, the rotation group is not. Within the rotation group the different groupings do not meet, and within a certain grouping there are only even permutations between the sets of polytopes. So there is a left-handed grouping and a right-handed grouping, and we find here two different sub-compounds.
Still more interesting is what happens with the other compounds (that derive from the compounds found in each {3, 4, 3}. The situation in the set of {4, 3, 3}'s is the same as in the {3, 3, 4}'s (as will become clear), so I will focus on the latter. Ten obvious compounds of 15 {3, 3, 4}'s can be found by taking a subcompound of five {3, 4, 3}'s and replacing each {3, 4, 3} by its compound of three {3, 3, 4}'s. As before, the rotation splits those in two orbits of five. However it is possible to find 280 compounds of 15 {3, 4, 3}'s, and this way we find a number of essentially different compounds.
First let us give the {3, 3, 4} numbers: (s.m.i), where s and m are as before and where i ranges from 1 to 3.
Next let's take a left-handed compound and a right-handed compound of the above ten. Suppose the left-handed compound with s = 1, and the right-handed compound with m = 1. So one set is (1.m.i) and the other set is (s.1.i). Obviously the subset (1.1.i) is shared. However, it appears that the subset (1.m.i) for some i and m > 1 has the same vertices as the subset (s.1.i) with s > 1. So we can in the first compound replace one or two of those subsets by a subset from the second compound and get a new compound (if we replace all three we get the second compound). So for each of the 25 possible pairings we get six new compounds, for a total of 150 compounds. Because groups of {3, 3, 4}'s in a grouping of {3, 4, 3} remain grouped in a similar grouping under the symmetry group it is easy to see that this set of 150 is distinctive under the symmetry group, and they form a new single orbit under that group. Under the rotation group there is a distinction between replacing only one or two of the subsets, so there the set of 150 is split in two subsets of 75. And so we find two essentially new compounds, a left-handed version and a right-handed version.
NB: although in the data files I use numbers as above, you will not find that the subsets act as shown above for fixed i. It is impossible to chose a number consistently for each {3, 3, 4} so that it works for all pairings. You have to find out what quadruples can be replaced by what other quadruples, but they can be found.
But that is indeed not all. When we look at the left-handed groups of {3, 4, 3}'s we can take three of them for a total of 15 {3, 4, 3}'s, covering the vertices of the {3, 3, 5} three times. It is possible to select a {3, 3, 4} from each of the {3, 4, 3}'s such that that set of 15 is again a compound. And actually this selection can be done in six different ways. (FURTHER ANALYSIS NEEDED HERE.) As there are ten possible ways to select three subgroups we get 60 compounds, and doing the same with right-handed groups we get a total of 120 new compounds. Here again this set of 120 forms a single orbit under the symmetry group and splits in two orbits of 60 under the rotation group, so also here a left-handed and a right-handed compound, and again essentially different from the other compounds.
If we replace a {3, 3, 4} by its complement in a {3, 4, 3} we get a double covering of the vertices. And so, if we do this for all of the {3, 3, 4}'s in a compound we get a double covering of the vertices of the {3, 3, 5}. But these complements cover the same vertices as the {4, 3, 3}'s. So each compound discussed above leads to a compound of the {4, 3, 3}'s with a double covering (a single covering is not possible).
All compounds found this way are also cell-regular. That is another reason why the compounds of the {4, 3, 3}'s act similar, they are just the reciprocals of the compounds found above.
To summarise what I do now know the following table (and it holds both for compounds of {3, 3, 4}'s as for compounds of {4, 3, 3}'s):
| # | count | remarks |
| 1L 1R | 5 5 |
compounds of {3, 4, 3} |
| 2L 2R | 75 75 |
replacing four elements |
| 3L 3R |
60 60 |
three out of five |
Although each {3, 4, 3} contains 16 hexagons, there are not 400 hexagons in section 2 but only 200. The reason is that the {3, 4, 3} with number (i, j) shares an hexagon with each of the 16 {3, 4, 3}'s with the first index number not equal to i and the second not equal to j. Obviously the same applies to the triangles in section 6.