This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Generating functions for fixed points of the Mullineux map

David J. Hemmer Department of Mathematical Sciences
Michigan Technological University
1400 Townsend Drive
Houghton, MI 49931, USA
[email protected]
Abstract.

Mullineux defined an involution on the set of ee-regular partitions of nn. When e=pe=p is prime, these partitions label irreducible symmetric group modules in characteristic pp. Mullineux’s conjecture, since proven, was that this “Mullineux map” described the effect on the labels of taking the tensor product with the one-dimensional signature representation. Counting irreducible modules fixed by this tensor product is related to counting irreducible modules for the alternating group AnA_{n} in prime characteristic. In 1991, Andrews and Olsson worked out the generating function counting fixed points of Mullineux’s map when e=pe=p is an odd prime (providing evidence in support of Mullineux’s conjecture). In 1998, Bessenrodt and Olsson counted the fixed points in a pp-block of weight ww. We extend both results to arbitrary ee, and determine the corresponding generating functions. When ee is odd but not prime the extension is immediate, while ee even requires additional work and the results, which are different, have not appeared in the literature.

2020 Mathematics Subject Classification:
Primary 05E10, Secondary 20C30

1. Introduction

Let SnS_{n} denote the symmetric group on nn letters. Recall that a partition λ=(λ1,λ2,,λs)n\lambda=(\lambda_{1},\lambda_{2},\ldots,\lambda_{s})\vdash n is ee-regular if no part repeats ee or more times. Let kk be an algebraically closed field of characteristic pp. The irreducible kSnkS_{n} modules are labelled by pp-regular partitions [Jam78, Chapter 11] and are denoted {Dλλ is p-regular}\{D^{\lambda}\mid\lambda\textrm{ is p-regular}\}. Define an involution PP on the pp-regular partitions by DλsgnDP(λ).D^{\lambda}\otimes\operatorname{sgn}\cong D^{P(\lambda)}. In [Mul79], Mullineux defined a combinatorial map λme(λ)\lambda\rightarrow m_{e}(\lambda) on the set of ee-regular partitions. He conjectured for e=pe=p a prime that mp=Pm_{p}=P. Almost twenty years later, Kleshchev [Kle96] described PP and Ford and Kleshchev [FK97] proved the conjecture by proving Kleshchev’s description matched Mullineux’s. The bijection mem_{e} is defined combinatorially on ee-regular partitions for ee arbitrary, and can be interpreted similarly using irreducible representations of a certain Hecke algebra [Mat99, Chapter 6.3]. For e=2e=2 the sign representation is trivial and the Mullineux map is the identity. Combinatorial properties of this map have inspired much research, often independent of representation theoretic applications.

Let AnA_{n} be the alternating group and let the characteristic of kk be an odd prime pp. One can count irreducible kAnkA_{n} modules in two ways: group theoretically by counting pp-regular conjugacy classes of AnA_{n}, or by restricting modules from SnS_{n} using Clifford theory and counting in terms of fixed points of the map PP. In 1991 Andrews and Olsson, using Olsson’s work in [Ols92], counted the fixed points of mpm_{p}:

Theorem 1.1.

[AO91, Propositions 2,3] Let p>2p>2 be prime. The number of fixed points of mpm_{p} is the number of partitions of nn with distinct odd parts, none of which are divisible by pp.

This answer agreed with the known representation theoretic count for fixed points of PP, providing evidence for the as-yet-unproven Mullineux conjecture by showing mpm_{p} had the expected number of fixed points. Later Bessenrodt and Olsson refined this by computing fixed points in an arbitrary pp-block of weight w:

Theorem 1.2.

[BO98, Theorem 3.5] Let pp be an odd prime, and μnpw\mu\vdash n-pw be a self-conjugate pp-core. For ww even the number of λn\lambda\vdash n with mp(λ)=λm_{p}(\lambda)=\lambda and having pp-core μ\mu is given by the number of (p1)/2(p-1)/2-tuples of partitions with total weight w/2w/2. For ww odd there are no such partitions.

Our main results are to extend Theorems 1.1 and 1.2 to arbitrary ee. For ee odd but not prime it is a simple observation that the original proofs carry over, while for ee even additional work is needed. The author would like to acknowledge his colleague William Keith for useful discussions about generating functions.

2. The Mullineux Map, ee-cores and ee-weights

For a partition λ=(λ1,λ2,,λs)\lambda=(\lambda_{1},\lambda_{2},\ldots,\lambda_{s}) with Young diagram [λ][\lambda], define the rim of λ\lambda to be the boxes along the southeast edge of the diagram, i.e. boxes (i,j)[λ](i,j)\in[\lambda] with (i+1,j+1)[λ](i+1,j+1)\not\in[\lambda]. Now consider a subset of the rim defined as follows, and called the ee-rim. Starting at the top right of the rim, take the first ee elements on the rim. Then move to the rightmost element of the rim in the next row, and take the next ee elements. Continue until the final row is reached, observing that the final segment may contain fewer than ee boxes.

For example if λ=(7,7,7,4,4,1,1)31\lambda=(7,7,7,4,4,1,1)\vdash 31 and e=5e=5 we have the e-rim:

\ytableausetup

centertableaux, boxsize=.5cm \ytableaushort *7,7,4,4,1,1 *[*(green)]6+1,6+1,4+3,3+1,4,1,1

Figure 2.1. The 55-rim of λ=(7,7,7,4,4,1,1)\lambda=(7,7,7,4,4,1,1).

Let a1a_{1} be the number of boxes and r1r_{1} be the number of rows in the ee-rim, so in our example a1=12a_{1}=12, r1=7r_{1}=7.

To define the Mullineux symbol Ge(λ)G_{e}(\lambda), remove the ee-rim, and then calculate the ee-rim of what remains to determine (a2,r2)(a_{2},r_{2}). Continue this process until all boxes are removed. Assemble these numbers in an array, called the Mullineux symbol of λ\lambda:

Ge(λ)=(a1a2akr1r2rk).G_{e}(\lambda)=\left(\begin{array}[]{cccc}a_{1}&a_{2}&\cdots&a_{k}\\ r_{1}&r_{2}&\cdots&r_{k}\\ \end{array}\right). (2.1)

Notice that (a1,a2,,ak)(a_{1},a_{2},\ldots,a_{k}) is also a partition of the same integer nn. For example from Figure 2.2 we see that:

G5(7,7,7,4,4,1,1)=(12854274332).G_{5}(7,7,7,4,4,1,1)=\left(\begin{array}[]{ccccc}12&8&5&4&2\\ 7&4&3&3&2\\ \end{array}\right).
\ytableausetup

centertableaux, boxsize=.5cm \ytableaushort *7,7,4,4,1,1 *[*(green)]6+1,6+1,4+3,3+1,4,1,1 *[*(red)]5+1,3+3,3+1,0+3 *[*(blue)]2+3,2+1,2+1,0+3 *[*(orange)]1+1,1+1,0+2 *[*(purple)]0+1,0+1

Figure 2.2. Calculating the Mullineux symbox G5(7,7,7,4,4,1,1)G_{5}(7,7,7,4,4,1,1)

Now define ϵi=0\epsilon_{i}=0 if eaie\mid a_{i} and ϵi=1\epsilon_{i}=1 otherwise, and set si=airi+ϵi.s_{i}=a_{i}-r_{i}+\epsilon_{i}. Then we have Mullineux’s conjecture (now theorem):

Proposition 2.1.

Let λn\lambda\vdash n be ee-regular with Mullineux symbol

Ge(λ)=(a1a2akr1r2rk).G_{e}(\lambda)=\left(\begin{array}[]{cccc}a_{1}&a_{2}&\cdots&a_{k}\\ r_{1}&r_{2}&\cdots&r_{k}\\ \end{array}\right).

Then the Mullineux symbol for me(λ)m_{e}(\lambda) is:

Ge(meλ)=(a1a2aks1s2sk).G_{e}(m_{e}\lambda)=\left(\begin{array}[]{cccc}a_{1}&a_{2}&\cdots&a_{k}\\ s_{1}&s_{2}&\cdots&s_{k}\\ \end{array}\right).

It is easy to reconstruct λ\lambda from Ge(λ)G_{e}(\lambda) so Proposition 2.1 gives a combinatorial description of the Mullineux map. For example with λ\lambda as in Figure 2.2

G5(m5(7,7,7,4,4,1,1))=(12854265221)G_{5}(m_{5}(7,7,7,4,4,1,1))=\left(\begin{array}[]{ccccc}12&8&5&4&2\\ 6&5&2&2&1\\ \end{array}\right)

and m5(7,7,7,4,4,1,1)=(12,9,4,2,2,2).m_{5}(7,7,7,4,4,1,1)=(12,9,4,2,2,2). Thus a fixed point of mem_{e} will have a Mullineux symbol of the form:

Ge(λ)=(a1a2aka1+ϵ12a2+ϵ22ak+ϵk2)G_{e}(\lambda)=\left(\begin{array}[]{cccc}a_{1}&a_{2}&\cdots&a_{k}\\ \frac{a_{1}+\epsilon_{1}}{2}&\frac{a_{2}+\epsilon_{2}}{2}&\cdots&\frac{a_{k}+\epsilon_{k}}{2}\\ \end{array}\right) (2.2)

where aia_{i} is even if and only if eaie\mid a_{i} if and only if ϵi=0\epsilon_{i}=0.

In his original paper Mullineux described necessary and sufficient conditions for such an array to arise as the Mullineux symbox of an ee-regular partition:

Proposition 2.2.

[Mul79, Theorem 3.6] An array as in (2.1) is the Mullineux symbol of an ee-regular partition of n=ain=\sum a_{i} if and only if:

  • (i)

    0riri+1e0\leq r_{i}-r_{i+1}\leq e;

  • (ii)

    riri+1+ϵi+1aiai+1riri+1+ϵi+1+er_{i}-r_{i+1}+\epsilon_{i+1}\leq a_{i}-a_{i+1}\leq r_{i}-r_{i+1}+\epsilon_{i+1}+e;

  • (iii)

    ri=ri+1eair_{i}=r_{i+1}\Rightarrow e\mid a_{i};

  • (iv)

    riri+1=eeair_{i}-r_{i+1}=e\Rightarrow e\nmid a_{i};

  • (v)

    0akrk<e0\leq a_{k}-r_{k}<e

  • (vi)

    1 rke\leq r_{k}\leq e and if rk=er_{k}=e then akrk>0.a_{k}-r_{k}>0.

Now we can apply Proposition 2.2 to arrays of the form (2.2), i.e. fixed points. The conditions on the rir_{i} in Proposition 2.2 can easily be translated to conditions on the aia_{i}. Thus we can enumerate fixed points simply by counting the suitable partitions (a1,a2,,ak)n(a_{1},a_{2},\ldots,a_{k})\vdash n.

Definition 2.3.

Define e(n)\mathcal{M}_{e}(n) to be the set of partitions (a1,a2,,ak)n(a_{1},a_{2},\ldots,a_{k})\vdash n satisfying:

  • (i)

    2aieai2\mid a_{i}\Longleftrightarrow e\mid a_{i}

  • (ii)

    0aiai+12e0\leq a_{i}-a_{i+1}\leq 2e

  • (iii)

    If ai=ai+1a_{i}=a_{i+1} then aia_{i} is even.

  • (iv)

    If aiai+1=2ea_{i}-a_{i+1}=2e then aia_{i} is odd.

  • (v)

    ak<2ea_{k}<2e.

Then we have:

Proposition 2.4.

[AO91, Proposition 1] Let pp be an odd prime. The number of partitions λn\lambda\vdash n fixed by the Mullineux map is equal to the cardinality of p(n)\mathcal{M}_{p}(n).

The proof of Proposition 2.4 goes through for ee arbitrary, so to enumerate Mullineux fixed points we need to understand the set e(n)\mathcal{M}_{e}(n). The criteria defining e(n)\mathcal{M}_{e}(n) are not easily translated into a generating function. The main theorem (Theorem 2) of [AO91] gives a remarkable enumeration of partitions with difference conditions like this. Observe for pp odd that p(n)\mathcal{M}_{p}(n) is a set of type P2(A;N,n)P_{2}(A;N,n) from that paper, where N=2pN=2p and A={1,3,5,p2,p+2,2p1}A=\{1,3,5,\ldots p-2,p+2,\ldots 2p-1\}. The paper gives a bijection with a set P1(A;2p,n)P_{1}(A;2p,n), which in this case is just partitions with distinct odd parts not divisible by pp, giving Theorem 1.1.

2.1. ee-cores and ee-weights

A rim e-hook of λ\lambda is ee consecutive boxes in the rim which, when removed, leave a Young diagram of a partition. Given a partition λn\lambda\vdash n, there is a unique partition λ(e)\lambda_{(e)}, called the ee-core of λ\lambda, and obtained by removing rim ee-hooks from λ\lambda until none remain. The number of such hooks removed is called the ee-weight ww of λ\lambda, so λ(e)new\lambda_{(e)}\vdash n-ew. This is all described in [JK81]. In [GKS90, Bijection 2] there is a bijection between ee-core partitions and vectors n=(n0,n1,,ne1)e\vec{n}=(n_{0},n_{1},\ldots,n_{e-1})\in\mathbb{Z}^{e} where i=0e1ni=0\sum_{i=0}^{e-1}n_{i}=0. Given the vector n\vec{n}, the corresponding ee-core is a partition of

e2i=0e1ni2+i=0e1ini.\frac{e}{2}\sum_{i=0}^{e-1}n_{i}^{2}+\sum_{i=0}^{e-1}in_{i}. (2.3)

The ee-weight can be obtained by subtracting (2.3) from jaj\sum_{j}a_{j} and dividing by ee.

Bessonrodt and Olsson worked out how to determine n\vec{n} directly from the Mullineux symbol of a Mullineux fixed point (they state it for pp an odd prime but the proof works for any ee).

Proposition 2.5.

[BO98, p. 235] Suppose λn\lambda\vdash n with me(λ)=λm_{e}(\lambda)=\lambda, ee-core λ(e)\lambda_{(e)} and Gp(λ)G_{p}(\lambda) is as in (2.2). Then the vector n\vec{n} corresponding to λ(e)\lambda_{(e)} is:

nj=#{iaiϵi2jmode}#{iaiϵi2jmode}.n_{j}=\#\{i\mid\frac{a_{i}-\epsilon_{i}}{2}\equiv j\mod e\}-\#\{i\mid\frac{-a_{i}-\epsilon_{i}}{2}\equiv j\mod e\}. (2.4)
Remark 2.6.

It is clear from (2.4) that n\vec{n}, and thus λ(e)\lambda_{(e)}, depends only on the set of nonzero residues mod ee of the aia_{i}. The aia_{i} which are divisible by ee (and hence even with ϵi=0\epsilon_{i}=0), do not contribute to n.\vec{n}.

3. Main Results

In Table 1 we list the number of fixed points under the Mullineux map when e=4e=4 for 1n201\leq n\leq 20 and ee-weights 0w50\leq w\leq 5. Weight zero partitions are ee-cores and mem_{e} acts as conjugation on ee-cores, so the first column of this table enumerates self-conjugate 4-core partitions, which is sequence A053692 in the Online Encyclopedia of Integer Sequence (OEIS) [OEI]. Each successive column is shifted down by e=4e=4 and multiplied by the corresponding entry in the sequence {1,1,3,4,9,12,23}\{1,1,3,4,9,12,23\ldots\}, which is A002513 in the OEIS, counting partitions of nn with even parts of two colors, also known as “cubic partitions”. Our main results generalize Theorems 1.1 and 1.2 to arbitrary ee and explain the structure of this table.

Table 1. Mullineux fixed points for e=4e=4 by weight
nn w=0w=0 w=1w=1 w=2w=2 w=3w=3 w=4w=4 w=5w=5
1 1
2 1
3 0
4 1
5 1 1
6 1 1
7 1 0
8 1 1
9 0 1 3
10 0 1 3
11 2 1 0
12 0 1 3 4
13 1 0 3 4
14 1 0 3 4
15 1 2 3 0
16 2 0 3 4
17 0 1 0 4 9
18 1 1 6 4 0
19 1 2 0 4 9
20 0 0 3 0 9 12

We fix some standard generating function notation. Define the qq-Pochhammer symbol:

(a;q):=k=0(1aqk)(a;q)_{\infty}:=\prod_{k=0}^{\infty}(1-aq^{k})

and the Ramanujan χ\chi function:

χ(q):=(q,q2)=(1+q)(1+q3)(1+q5).\chi(q):=(-q,q^{2})_{\infty}=(1+q)(1+q^{3})(1+q^{5})\cdots.

Observe that χ(q)\chi(q) is the generating function counting partitions of nn into distinct odd parts.

Let mfe(n)mf_{e}(n) be the number of ee-regular partitions of nn fixed by the Mullineux map and let

MFe(q):=n=0mfe(n)qnMF_{e}(q):=\sum_{n=0}^{\infty}mf_{e}(n)q^{n}

be the corresponding generating function. Our first result determines this generating function:

Theorem 3.1.
  • (a)

    [Ols92, for e prime] Let ee be odd. Then mfe(n)mf_{e}(n) is the number of partitions of nn into distinct odd parts not divisible by ee. Thus:

    MFe(q)=χ(q)χ(qe)=(1+q)(1+q3)(1+q5)(1+qe)(1+q3e)(1+q5e)=k oddek(1+qk).MF_{e}(q)=\frac{\chi(q)}{\chi(q^{e})}=\frac{(1+q)(1+q^{3})(1+q^{5})\cdots}{(1+q^{e})(1+q^{3e})(1+q^{5e})\cdots}=\prod_{\begin{subarray}{c}k\text{ odd}\\ e\nmid k\end{subarray}}(1+q^{k}). (3.1)
  • (b)

    Let ee be even. Then mfe(n)mf_{e}(n) is the number of partitions of nn into parts which are either odd or are odd multiples of ee, and such that the odd parts are all distinct. Thus:

    MFe(q)=χ(q)χ(qe)=(1+q)(1+q3)(1+q5)(1qe)(1q3e)(1q5e).MF_{e}(q)=\frac{\chi(q)}{\chi(-q^{e})}=\frac{(1+q)(1+q^{3})(1+q^{5})\cdots}{(1-q^{e})(1-q^{3e})(1-q^{5e})\cdots}. (3.2)

Notice that when e=2e=2 that MF2(q)MF_{2}(q) simplifies to k odd11qk\prod_{k\text{ odd}}\frac{1}{1-q^{k}}, the generating function counting partitions with odd parts, which is known to be the same as that for distinct parts. Here, the Mullineux map is trivial and we obtain mf2(n)mf_{2}(n) is the number of two-regular partitions, i.e. partitions with distinct parts, as expected.

If we look at the corresponding alternating series, there is a nice common description of the two generating functions:

Corollary 3.2.

For ee arbitrary:

MFe(q)=n=0(1)nmfe(n)qn=k=11+qek1+qk.MF_{e}(-q)=\sum_{n=0}^{\infty}(-1)^{n}mf_{e}(n)q^{n}=\prod_{k=1}^{\infty}\frac{1+q^{ek}}{1+q^{k}}. (3.3)
Proof.

For both the ee even and ee odd case, MFe(q)=χ(q)χ(qe)MF_{e}(-q)=\frac{\chi(-q)}{\chi(-q^{e})}, which simplifies to the expression above. ∎

For e=3,4,5,6e=3,4,5,6 the generating function MFe(q)MF_{e}(-q) corresponds to the sequences A098884, A261734, A133563 and A261736 respectively in the OEIS [OEI]. As of this writing there is no mention of the Mullineux map in any of these entries!

3.1. Mullineux fixed points of a given ee-weight

We also give a result for all ee generalizing Theorem 1.2, counting Mullineux fixed points by ee-weight.

Definition 3.3.

Let mfe,w(n)mf_{e,w}(n) be the number of λn\lambda\vdash n with ee-weight ww and me(λ)=λm_{e}(\lambda)=\lambda.

It is well-known that the Mullineux map when restricted to ee-core partitions simply takes the transpose or conjugate partition, so determining mfe,0(n)mf_{e,0}(n) means counting self-conjugate ee-cores. These were enumerated by Garvan, Kim and Stanton.

Definition 3.4.

Let sce(n)sc_{e}(n) (=mfe,0(n)=mf_{e,0}(n)) denote the number of self-conjugate ee-core partitions of n and let

SCe(q):=n=0sce(n)qn.SC_{e}(q):=\sum_{n=0}^{\infty}sc_{e}(n)q^{n}.

Garvan, Kim and Stanton determined SCe(q)SC_{e}(q):

Theorem 3.5.

[GKS90, 7.1a,b] The generating function SCe(q)SC_{e}(q) is:

SCe(q))={(q,q2)(q2e,q2e)e/2if e is even(q,q2)(q2e,q2e)e12(qe,q2e)if e is odd.SC_{e}(q))=\begin{cases}(-q,q^{2})_{\infty}(q^{2e},q^{2e})_{\infty}^{e/2}&\mbox{if }e\mbox{ is even}\\[10.0pt] \frac{(-q,q^{2})_{\infty}(q^{2e},q^{2e})_{\infty}^{\frac{e-1}{2}}}{(-q^{e},q^{2e})}&\mbox{if }e\mbox{ is odd.}\end{cases}

It turns out the for each ee-core partition of newn-ew , we have the same number of Mullineux fixed point partitions of nn with that ee core, a number determined by ww. These numbers also have nice generating functions, again depending on the parity of ee. The next definition gives the sequences of “column multipliers” that we observed in Table 1:

Definition 3.6.

Let ee be even. Define fe(n)f_{e}(n) by:

n=0fe(n)qn\displaystyle\sum_{n=0}^{\infty}f_{e}(n)q^{n} :=\displaystyle:= k=11(1q2k)e/21(1q2k1)\displaystyle\prod_{k=1}^{\infty}\frac{1}{(1-q^{2k})^{e/2}}\frac{1}{(1-q^{2k-1})}
=\displaystyle= 1(q2,q2)e/2(q,q2).\displaystyle\frac{1}{(q^{2},q^{2})_{\infty}^{e/2}(q,q^{2})_{\infty}}.

Let ee be odd. Define ge(n)g_{e}(n) by:

n=0ge(n)qn\displaystyle\sum_{n=0}^{\infty}g_{e}(n)q^{n} :=\displaystyle:= k=11(1qk)(e1)/2\displaystyle\prod_{k=1}^{\infty}\frac{1}{(1-q^{k})^{(e-1)/2}}
=\displaystyle= 1(q,q)(e1)/2\displaystyle\frac{1}{(q,q)_{\infty}^{(e-1)/2}}

Notice that both fe(n)f_{e}(n) and ge(n)g_{e}(n) enumerate certain tuples of partitions with total weights adding up to nn. With this notation, we can state our theorem enumerating Mullineux fixed points of a given weight. This result for ee an odd prime is Theorem 3.5 of [BO98].

Theorem 3.7.
  • (a)

    Suppose ee is even. Then:

    mfe,w(n)=fe(w)sce(new).mf_{e,w}(n)=f_{e}(w)sc_{e}(n-ew).
  • (b)

    Suppose ee is odd. Then mfe,w(n)mf_{e,w}(n) is zero unless ww is even in which case

    mfe,w(n)=g(w2)sce(new).mf_{e,w}(n)=g(\frac{w}{2})sc_{e}(n-ew).

Theorem 3.7 lets us determine a single two-variable generating function that keeps track of Mullineux fixed points by nn and the weight ww. We will need a small reindexing given by:

Lemma 3.8.

Suppose A(x)=i=0aixiA(x)=\sum_{i=0}^{\infty}a_{i}x^{i} and B(x)=j=0bjqj.B(x)=\sum_{j=0}^{\infty}b_{j}q^{j}. Then:

A(qex)B(q)=n,wawbnewxwqn.A(q^{e}x)B(q)=\sum_{n,w}a_{w}b_{n-ew}x^{w}q^{n}.
A(q2ex2)B(q)=nwevenaw2bnewxwqn.A(q^{2e}x^{2})B(q)=\sum_{n}\sum_{w\textrm{even}}a_{\frac{w}{2}}b_{n-ew}x^{w}q^{n}.
Proof.

Simply reindex the sum with a substitution n=j+ewn=j+ew and (for the second term), t=w/2t=w/2. ∎

Now define a two-variable generating function:

Definition 3.9.

Let

MFe(x,q):=n=0w=0mfe,w(n)xwqn.MF_{e}(x,q):=\sum_{n=0}^{\infty}\sum_{w=0}^{\infty}mf_{e,w}(n)x^{w}q^{n}.

Lemma 3.8 applied to the expressions in Theorem 3.7, and using the generating functions from Definition 3.6 and Theorem 3.5, proves the following:

Theorem 3.10.

The generating function MFe(x,q)MF_{e}(x,q) is:

MFe(x,q))={(q,q2)(q2e,q2e)e/2(q2ex2,q2ex2)e/2(qex,q2ex2)if e is even(q,q2)(q2e,q2e)e1/2(q2ex2,q2ex2)(e1)/2(qe,q2e)if e is oddMF_{e}(x,q))=\begin{cases}\frac{(-q,q^{2})_{\infty}(q^{2e},q^{2e})_{\infty}^{e/2}}{(q^{2e}x^{2},q^{2e}x^{2})_{\infty}^{e/2}(q^{e}x,q^{2e}x^{2})_{\infty}}&\mbox{if }e\mbox{ is even}\\[10.0pt] \frac{(-q,q^{2})_{\infty}(q^{2e},q^{2e})_{\infty}^{e-1/2}}{(q^{2e}x^{2},q^{2e}x^{2})_{\infty}^{(e-1)/2}(-q^{e},q^{2e})}&\mbox{if }e\mbox{ is odd}\par\end{cases}

4. Proof of the main results

Theorems 1.1 and 1.2 are stated for pp an odd prime, because that is where the original representation theory motivation comes from. However there is nothing in either proof that makes use of primality, so the ee odd case of Theorems 3.1 and 3.7 should be attributed to those authors. Thus we will consider only the case where ee is even.

When ee is odd notice that Definition 2.3(i) gives that if eaie\mid a_{i} then actually N=2eaiN=2e\mid a_{i}. This is not the case for ee-even, and this means that e(n)\mathcal{M}_{e}(n) in this case is not dealt with by the bijections in [AO91]. However in [Bes95], Bessenrodt gave a vast generalization that includes this case. So in her notation let N=2eN=2e for ee even. Choose the sets A={e}A^{\prime}=\{e\} and A′′={1,3,5,,2e1}A^{\prime\prime}=\{1,3,5,\ldots,2e-1\} with A=AA′′.A=A^{\prime}\cup A^{\prime\prime}. One can check that the conditions for e(n)\mathcal{M}_{e}(n) are precisely those defining Bessenrodt’s set P2(A,A′′;N,n).P_{2}(A^{\prime},A^{\prime\prime};N,n). Bessenrodt gives a bijection with a set P1(A,A′′;N,n)P_{1}(A^{\prime},A^{\prime\prime};N,n), which are all partitions with parts congruent mod 2e2e to elements of AA and repeating parts must be congruent to elements of AA^{\prime}. This is precisely the set of partitions with all parts odd or odd multiples of ee with the odd parts distinct. This proves Theorem 3.1(b).

Finally, to prove Theorem 3.7(b), we need to count Mullineux fixed points in a given block. We will follow the idea of [BO98, Theorem 3.5], although the proof is slightly more complicated since ee is even.

Definition 4.1.

Define κ(r,s)\kappa(r,s) to be the number of tuples (γ1,γ2,,γr;τ)(\gamma^{1},\gamma^{2},\ldots,\gamma^{r};\tau) where the γi\gamma^{i} and τ\tau are partitions, τ\tau has odd parts, and |τ|+2i|γi|=s.|\tau|+2\sum_{i}|\gamma^{i}|=s.

Theorem 4.2.

Let ee be even, let μ\mu be a self-conjugate ee core with w=n|μ|ew=\frac{n-|\mu|}{e}. Then:

κ(e2,w)=#{λnλ=me(λ),λ(e)=μ}.\kappa\left(\frac{e}{2},w\right)=\#\{\lambda\vdash n\mid\lambda=m_{e}(\lambda),\lambda_{(e)}=\mu\}.

Consider the generating function in (3.6). The first term is counting e/2e/2 tuples of partitions with even parts and the second term counts partitions with odd parts. We can equally well count e/2e/2 tuples of arbitrary partitions but then double all the parts. So Theorem 4.2 immediately implies Theorem 3.7(b), and all that remains is to prove Theorem 4.2.

For the next result we will need the following observation, which is easy to see in [Bes95] (recalling that N=2eN=2e.)

Lemma 4.3.

Suppose λn\lambda\vdash n is a Mullineux fixed point with Ge(λ)G_{e}(\lambda) as in (2.2) and corresponding partition (a1,a2,,ak)P2(A,A′′;N,n).(a_{1},a_{2},\ldots,a_{k})\in P_{2}(A^{\prime},A^{\prime\prime};N,n). Suppose the image partition under Bessenrodt’s bijection is (b1,b2,bs)P1(A,A′′;N,n)(b_{1},b_{2},\ldots b_{s})\in P_{1}(A^{\prime},A^{\prime\prime};N,n). Then both partitions have the same set of nonzero residues mod 2e2e (and thus also mod ee.)

Thus we can use the bib_{i}’s to calculate n\vec{n} and the corresponding ee-core and ee-weight as in Section 2.1. Recall that Bessenrodt’s P1(A,A′′;N,n)P_{1}(A^{\prime},A^{\prime\prime};N,n) in this case are partitions of nn with all parts either odd or odd multiples of ee, and the odd parts must be distinct. Henceforth we will consider these as pairs of partitions {(c1,c2,,ck),eμ}\{(c_{1},c_{2},\ldots,c_{k}),e\mu\} where (c1,c2,,ck)(c_{1},c_{2},\ldots,c_{k}) has distinct odd parts and μ\mu has odd parts. Notice that the set of nonzero residues mod ee in (c1,c2,,ck)(c_{1},c_{2},\ldots,c_{k}) is the same as in (b1,b2,bt)(b_{1},b_{2},\ldots b_{t}), since the odd multiples of ee all have ee-residue zero To complete the proof we will need to use the theory of tt-bar cores and tt-bar quotients.

4.1. t-bar cores

For partitions with distinct parts (sometimes called “bar partitions”), Morris introduced [Mor65] the notion of a tt-bar core. These played a role in Humphrey’s determination [Hum86] of blocks of projective representations of symmetric groups in characteristic pp similar to the role played by pp-cores in the theory for SnS_{n}. We will follow the combinatorics introduced in [MY86], where a version of James’ abacus notation is given for computing tt-bar cores. Note this is different from the more well-known James’ abacus display for arbitrary partitions using sets of β\beta-numbers.

Given a partition λ\lambda with distinct parts, we form an abacus with runners lying north to south labelled {0,1,2,,t1}\{0,1,2,\ldots,t-1\}. The bead positions are labelled as below:

01t2t1tt+12t22t1\begin{array}[]{ccccc}0&1&\cdots&t-2&t-1\\ t&t+1&\cdots&2t-2&2t-1\\ \vdots&\vdots&&\vdots&\vdots\end{array}

To display λ\lambda on the abacus place a bead on the abacus corresponding to each part. For example if λ=(23,21,17,13,11,9,7)\lambda=(23,21,17,13,11,9,7) and t=6t=6 we obtain the display in Figure 4.1, where we have also included the labels on the runners.

012345
Figure 4.1. Abacus display for λ=(23,21,17,13,11,9,7)\lambda=(23,21,17,13,11,9,7) and t=6t=6.

From λ\lambda we can obtain its tt-bar core, denoted λ~(t)\tilde{\lambda}_{(t)}, by performing two types of operations on the abacus. The first slides a bead on a runner up one into a vacant spot (beads reaching the location zero disappear). This corresponds to replacing a part λi\lambda_{i} in the partition by λit\lambda_{i}-t if λit\lambda_{i}-t is not already a part, or removing a part λi=t\lambda_{i}=t. Alternately you can remove two beads at positions aa and tat-a in the first row. This corresponds to removing two parts of the partition equal to aa and tat-a. Either operation has the effect of removing a tt-bar (which we will not define) from the partition and reducing the number being partitioned by tt. Eventually there are no further operations possible, and you reach the tt-bar core λ~(t)\tilde{\lambda}_{(t)}. The total number of such operations is well-defined and called the tt-bar weight. So in our example we obtain λ~(6)=(9,5,3)\tilde{\lambda}_{(6)}=(9,5,3) with abacus display:

012345
Figure 4.2. Abacus display for λ~(6)=(9,5,3)\tilde{\lambda}_{(6)}=(9,5,3)

Much of the work done on tt-bar cores has assumed tt is odd, because the combinatorics is not as nice when tt is even. For example if tt is even a partition may be a tt-bar core without being a 2t2t-bar core (e.g. for λ\lambda= (3,1)(3,1) and t=2t=2 one can remove a 44-bar but not a 22-bar). Also the number of partitions with a given tt-bar core and tt bar weight may depend on the choice of tt-bar core and not just the weight. Neither of these pathologies happens for ordinary tt-cores and rim tt-hooks. However, as Olsson observes in [Ols87, p.,235], these difficulties do not arise if there are no beads on runner t/2t/2. This will be the case in our situation, where tt will equal 2e2e for ee even, and the partitions will have only odd parts.

Remark 4.4.

Suppose t=2et=2e for ee even and let λ\lambda have distinct odd parts. Then the abacus display for λ\lambda with tt runners has beads only on runners with odd labels and, in particular, no beads on runner 0 or e=t/2e=t/2. Moreover all partitions with distinct odd parts and the same set of residues mod 2e2e will have the same 2e2e-bar core as λ\lambda.

We now consider the tt-bar quotient, following [MY86]. Assume t=2et=2e for ee even, and λ\lambda has distinct odd parts. Then we see that the tt-bar quotient is a sequence of e/2e/2 partitions, one for each pair of runners {(i,ti)i=1,3,5,,e1}\{(i,t-i)\mid i=1,3,5,\ldots,e-1\}. The actual construction is in [MY86, Theorem 2.2] but we will need only the count:

Lemma 4.5.

[MY86, Theorem 2.(2)] The number of partitions with distinct odd parts having the same 2e2e-bar core as λ\lambda and qq-bar weight w~\tilde{w} is the number of e/2e/2 tuples of partitions of total weight w~\tilde{w}.

Proof of Theorem 4.2: We are now ready to prove Theorem 4.2 by constructing a bijection between the Mullineux fixed points on the right and a set which has cardinality κ(e2,w)\kappa(\frac{e}{2},w) on the left.

Suppose λn\lambda\vdash n has ee weight ww and self-conjugate core λ(e).\lambda_{(e)}. Apply Bessenrodt’s bijection to get a pair {τ,eγ}\{\tau,e\gamma\} where τ=(c1,c2,,ck)\tau=(c_{1},c_{2},\ldots,c_{k}) has distinct odd parts and γ\gamma has all odd parts, and the nonzero residues mod 2e2e for τ\tau are the same as for λ\lambda. In particular all the τ\tau have the same 2e2e-bar core. As in [BO98], we can check that that 2e2e-bar core τ~(2e)\tilde{\tau}_{(2e)} is the same size as the ee-core of λ\lambda, i.e. that |λ(e)|=|τ~2e|.|\lambda_{(e)}|=|\tilde{\tau}_{2e}|. Equation 2.3 gives us |λ(e)||\lambda_{(e)}| in terms of the nin_{i}. Now the argument proceeds just as on the top of page 238 of [BO98].

Recall in the ee odd case that for λ\lambda of even weight ww and all the corresponding partitions had the same 2e2e-bar weight w/2w/2. In this situation, with ee even, the 2e2e-bar weight of τ\tau can be less, with the “difference” made up for by eγ.e\gamma.

The next step in the bijection is to calculate the 2e2e-bar quotient of τ.\tau. Since τ\tau has distinct odd parts, its 2e2e-bar quotient is a tuple (ρ1,ρ2,,ρe/2)(\rho^{1},\rho^{2},\cdots,\rho^{e/2}) with 2e2e-bar weight w~=i|ρi|\tilde{w}=\sum_{i}|\rho^{i}| and

τ|τ~2e|+2ew~.\tau\vdash|\tilde{\tau}_{2e}|+2e\tilde{w}.

Now we can describe the bijection that proves Theorem 4.2. Given λ\lambda as above we send λ\lambda to the tuple (ρ1,ρ2,,ρe/2;eγ)(\rho^{1},\rho^{2},\cdots,\rho^{e/2};e\gamma). The calculations above shows that

w\displaystyle w =\displaystyle= n|λ(e)|e\displaystyle\frac{n-|\lambda_{(e)}|}{e}
=\displaystyle= |τ|+e|γ||τ~2e|e\displaystyle\frac{|\tau|+e|\gamma|-|\tilde{\tau}_{2e}|}{e}
=\displaystyle= |τ||τ~2e|e+|γ|\displaystyle\frac{|\tau|-|\tilde{\tau}_{2e}|}{e}+|\gamma|
=\displaystyle= 2i|ρi|+|γ|.\displaystyle 2\sum_{i}|\rho^{i}|+|\gamma|.

So (ρ1,ρ2,,ρe/2;eγ)(\rho^{1},\rho^{2},\cdots,\rho^{e/2};e\gamma) is in the set enumerated by κ(e/2,w).\kappa(e/2,w).

The final necessary observation is that any choice of 2e2e-bar quotient concentrated on the odd runners together with eγe\gamma satisfying the weight condition w=2i|ρi|+|γ|w=2\sum_{i}|\rho^{i}|+|\gamma| will correspond to an original λ\lambda in the correct block. This is again the observation from [BO98] that adding or removing a 2e2e bar does not change the vector n^\hat{n}, nor does the choice of eγe\gamma. So the map is surjective, completing the proof.

References

  • [AO91] George E. Andrews and Jørn B. Olsson. Partition identities with an application to group representation theory. J. Reine Angew. Math., 413:198–212, 1991.
  • [Bes95] Christine Bessenrodt. Generalisations of the Andrews-Olsson partition identity and applications. Discrete Math., 141(1-3):11–22, 1995.
  • [BO98] C. Bessenrodt and J. B. Olsson. On residue symbols and the Mullineux conjecture. J. Algebraic Combin., 7(3):227–251, 1998.
  • [FK97] Ben Ford and Alexander S. Kleshchev. A proof of the Mullineux conjecture. Math. Z., 226(2):267–308, 1997.
  • [GKS90] Frank Garvan, Dongsu Kim, and Dennis Stanton. Cranks and tt-cores. Invent. Math., 101(1):1–17, 1990.
  • [Hum86] John F. Humphreys. Blocks of projective representations of the symmetric groups. J. London Math. Soc. (2), 33(3):441–452, 1986.
  • [Jam78] Gordon James. The representation theory of the symmetric groups. Number 682 in Lecture Notes in Mathematics. Springer-Verlag, 1978.
  • [JK81] Gordon James and Adalbert Kerber. The representation theory of the symmetric group, volume 16 of Encyclopedia of Mathematics and its Applications. Addison-Wesley Publishing Co., Reading, Mass., 1981.
  • [Kle96] A. S. Kleshchev. Branching rules for modular representations of symmetric groups. III. Some corollaries and a problem of Mullineux. J. London Math. Soc. (2), 54(1):25–38, 1996.
  • [Mat99] Andrew Mathas. Iwahori-Hecke algebras and Schur algebras of the symmetric group, volume 15 of University Lecture Series. American Mathematical Society, Providence, RI, 1999.
  • [Mor65] A. O. Morris. The spin representation of the symmetric group. Canadian J. Math., 17:543–549, 1965.
  • [Mul79] G. Mullineux. Bijections of pp-regular partitions and pp-modular irreducibles of the symmetric groups. J. London Math. Soc. (2), 20(1):60–66, 1979.
  • [MY86] A. O. Morris and A. K. Yaseen. Some combinatorial results involving shifted Young diagrams. Math. Proc. Cambridge Philos. Soc., 99(1):23–31, 1986.
  • [OEI] OEIS Foundation Inc. The On-Line Encyclopedia of Integer Sequences. Published electronically at http://oeis.org.
  • [Ols87] Jørn B. Olsson. Frobenius symbols for partitions and degrees of spin characters. Math. Scand., 61(2):223–247, 1987.
  • [Ols92] Jørn B. Olsson. The number of modular characters in certain pp-blocks. Proc. London Math. Soc. (3), 65(2):245–264, 1992.