3

I was looking for all the possible ways to partition the set of faces of a cube, where partitions differing only by a rotation of the cube should count as one. I found 5 2-partitions and 9 3-partitions (and only 1 6-partition of course). Is there a way to prove these are all the possible partitions? Is there a formula to also count 4-partitions and 5-partitions (besides manually counting them)?

P.S. if we remove the constraint whereby partitions differing only by a rotation count as one, the formula for the n-partitions is simply the number of set-partitions of 6 into n subsets i.e. the stirling numbers of the second kind

  • It looks like something you could use Burnside's Lemma for, but I'm not sure it is worth it in this case. I'm only finding five 2-partitions: one 1-5 partition, two 2-4 partitions cause the size-2 set are either opposite or adjacent faces, and two 3-3 partitions cause the sets are either three faces around a vertex or an opposing pair plus another face. I also count only eight 3-partitions: two 1-1-4 partitions, two 1-2-3 partitions, and four 2-2-2 partitions of which two are a mirror image pair. Which ones am I overlooking? – Jaap Scherphuis Apr 14 '21 at 13:07
  • @JaapScherphuis yes, there are five 2-partitions (editing the question to fix the typo). There are nine 3-partitions: you are missing one 1-2-3 partition. I thought about Burnside's lemma as this problem seems to be similar to the coloring problem of the faces of the cube, but I didn't manage to apply it to this problem. – UndefinedBehavior Apr 14 '21 at 14:36
  • Ah, I see the partition I was missing now, thanks. – Jaap Scherphuis Apr 14 '21 at 14:40

1 Answers1

2

If I understand this correctly these partitions are colorings of the faces under rotational symmetry with the colors representing membership in a partition and the colors being swappable (symmetric group acting on them). This enumeration problem can be solved by Power Group Enumeration and indeed this was done for the twelve edges rather than the six faces at the following MSE link I. To obtain an answer to the present query the only change is to replace the cycle index of the edges under rotational symmetries by the one of the faces, which was computed at this MSE link II and found to be:

$$Z(G) = \frac{1}{24} (a_1^6 + 8 a_3^2 + 6 a_1^2 a_4 + 3 a_1^2 a_2^2 + 6 a_2^3).$$

With these ingredients we obtain the classification of face colorings by the number of colors with at most six colors to be

$${P_{{1}}}^{6}+2\,{P_{{1}}}^{3}{P_{{2}}}^{3}+2\,{P_{ {1}}}^{2}{P_{{2}}}^{4}+4\,{P_{{1}}}^{2}{P_{{2}}}^{2 }{P_{{3}}}^{2}+P_{{1}}{P_{{2}}}^{5}+3\,P_{{1}}{P_{{ 2}}}^{2}{P_{{3}}}^{3}\\+2\,P_{{1}}P_{{2}}{P_{{3}}}^{4 }+5\,P_{{1}}P_{{2}}{P_{{3}}}^{2}{P_{{4}}}^{2}+2\,P_ {{1}}P_{{2}}P_{{3}}{P_{{4}}}^{3}\\+2\,P_{{1}}P_{{2}}P _{{3}}P_{{4}}{P_{{5}}}^{2}+P_{{1}}P_{{2}}P_{{3}}P_{ {4}}P_{{5}}P_{{6}}.$$

We thus obtain e.g. five $2$-partitions and nine $3$-partitions, same as found by OP. We also get two $5$-partitions which is correct as well (double color adjacent or not). The Maple code for this is quite compact and relatively straightforward once the PGE algorithm is known.

with(combinat);

cube_face_cind := 1/24(a[1]^6 + 8a[3]^2 + 6a[1]^2a[4] + 3a[1]^2a[2]^2 + 6*a[2]^3);

pet_cycleind_symm := proc(n) option remember; local l;

if n=0 then return 1; fi;

expand(1/n*add(a[l]*pet_cycleind_symm(n-l), l=1..n));

end;

pet_indets2rep := proc(ip) local rep, var, deg, pos, s;

rep := []; pos := 1;

for var in indets(ip) do
    for deg to degree(ip, var) do
        rep :=
        [op(rep), [seq(s, s=pos..pos+op(1, var)-1)]];
        pos := pos + op(1, var);
    od;
od;

rep;

end;

cube_face_colorings_gf := proc(n) option remember; local idx_cols, rep, res, term_a, term_b, v_a, v_b, inst_a, len_a, len_b, p, q, parts, alldeg, cols, col, v;

if n = 1 then return P[1]^6 fi;
idx_cols := pet_cycleind_symm(n);

res := 0;

for term_b in idx_cols do
    rep := pet_indets2rep(term_b);

    for term_a in cube_face_cind do
        p := 1;

        for v_a in indets(term_a) do
            len_a := op(1, v_a);
            inst_a := degree(term_a, v_a);

            q := 0;

            for v_b in rep do
                len_b := nops(v_b);

                if len_a mod len_b = 0 then
                    q := q + len_b*
                    mul(P[col], col in v_b)
                    ^(len_a/len_b);
                fi;
            od;

            p := p*q^inst_a;
        od;

        res := res +
        lcoeff(term_a)*lcoeff(term_b)*p;
    od;
od;

res;

parts := 0;

for cols in expand(res) do
    alldeg :=
    sort(map(v -> degree(cols, v),
             [op(indets(cols))]));

    parts := parts +
    lcoeff(cols)*
    mul(P[v]^alldeg[v], v=1..nops(alldeg));
od;

parts;

end;

Addendum. OP asks what we can say about the case of the cube having no symmetries, i.e. being a strip of six slots with the group permuting the slots being the identity. The code shown above will produce the following generating function in that case:

$${P_{{1}}}^{6}+10\,{P_{{1}}}^{3}{P_{{2}}}^{3}+15\,{ P_{{1}}}^{2}{P_{{2}}}^{4}+15\,{P_{{1}}}^{2}{P_{{2} }}^{2}{P_{{3}}}^{2}+6\,P_{{1}}{P_{{2}}}^{5}\\+60\,P_ {{1}}{P_{{2}}}^{2}{P_{{3}}}^{3}+15\,P_{{1}}P_{{2}} {P_{{3}}}^{4}+45\,P_{{1}}P_{{2}}{P_{{3}}}^{2}{P_{{ 4}}}^{2}+20\,P_{{1}}P_{{2}}P_{{3}}{P_{{4}}}^{3}\\+15 \,P_{{1}}P_{{2}}P_{{3}}P_{{4}}{P_{{5}}}^{2}+P_{{1} }P_{{2}}P_{{3}}P_{{4}}P_{{5}}P_{{6}}.$$

We can verify these by inspection i.e. with five colors we must choose two slots for the double color, giving ${6\choose 2} = 15$ or with two colors one of which has two instances we also get ${6\choose 2} = 15$ or at last with two colors both of which have three instances we get $\frac{1}{2} {6\choose 3} = 10.$

Now if we are only after the count for $E_m$ being the slots so that $Z(E_m) = a_1^m$ and $Z(S_n)$ acting on $n$ colors we can actually compute a closed form. With PGE we must cover the cycles of the slot permutation $\alpha$ with cycles of the permutation of the colors $\beta$ but here we only have $m$ fixed points in $\alpha$ to cover. Therefore if $\beta$ has $q$ fixed points we get a contribution of $q^m.$ The combinatorial class of permutations with fixed points marked is

$$\def\textsc#1{\dosc#1\csod} \def\dosc#1#2\csod{{\rm #1{\small #2}}} \textsc{SET}(\mathcal{U}\times \textsc{CYC}_{=1}(\mathcal{Z}) + \textsc{CYC}_{=2}(\mathcal{Z}) + \textsc{CYC}_{=3}(\mathcal{Z}) + \textsc{CYC}_{=4}(\mathcal{Z}) + \cdots).$$

This gives the EGF

$$G(z, u) = \exp\left(uz + \frac{z^2}{2} + \frac{z^3}{3} + \frac{z^4}{4} + \cdots\right) \\ = \exp\left(uz-z + \log\frac{1}{1-z}\right) = \frac{\exp(-z)}{1-z} \exp(uz).$$

We seek to turn $u^q z^n/n!$ into $q^m z^n/n!$ and use

$$q^m = \sum_{r=1}^m q^{\underline{r}} {m\brace r}.$$

We thus obtain

$$[z^n] \left. \sum_{r=1}^m {m\brace r} \frac{\exp(-z)}{1-z} \exp(uz) z^r \right|_{u=1} = [z^n] \sum_{r=1}^m {m\brace r} \frac{z^r}{1-z}.$$

This gives the closed form

$$\bbox[5px,border:2px solid #00A000]{ \sum_{r=1}^n {m\brace r}.}$$

Here we have confirmed a formula that could have been obtained by inspection. In particular for $n=m$ we get the sequence of Bell numbers:

$$1, 2, 5, 15, 52, 203, 877, 4140, 21147, 115975, \ldots$$

Marko Riedel
  • 64,728
  • 1
    wow! very interesting! I had a superficial knowledge of Polya's theorem and didn't know PGE at all so I went and looked it up and read the original 1966 paper introducing PDE. I think I still have to understand some details but trying to apply the formulas I found in the paper to the original problem gave me the correct numbers. Thank you for making me learn something new :) – UndefinedBehavior Apr 22 '21 at 11:48
  • 1
    I actually think I have a "side question". In the case were the symmetric group acts on the color but no group (i.e the identity group) acts on the faces you get 122 possible colorings. Is it possible to use Burnside's lemma on this to get to the correct result no the question I posted? I feel it should be possible but didn't manage to do it so not sure if I just miscounted or there is something fundamental preventing the use of Burnside's lemma here. – UndefinedBehavior Apr 22 '21 at 11:56
  • 1
    I have added some material concerning the slot permutation group being the identity. You should edit your post to include a reference to the paper that you cite. I usually refer people to Harary and Palmer, Graphical Enumeration. – Marko Riedel Apr 22 '21 at 21:02