betaFeedback

WorksMathematics

Divisibility of Character Values of Wreath Products of Symmetric Groups

Abstract

Let GG be a finite group whose character table is integer-valued, and let KK be any fixed positive integer. We prove that almost every entry of the character table of the wreath product G≀SNG\wr S_N is divisible by KK as N→∞N\to\infty.

Introduction

Let SNS_N be the symmetric group on NN letters, and let GG be a finite group with kk conjugacy classes. By the Murnaghan–Nakayama rule [7 Theorem 2.4.7], the irreducible characters of SNS_N are integer-valued. Moreover, the number of irreducible characters of SNS_N is equal to p(N)p(N), the number of partitions of NN.

Based on extensive numerical calculations, Miller [11, 12] conjectured that, for every fixed prime power qq, the proportion of entries in the character table of SNS_N divisible by qq tends to 11 as N→∞N \to\infty. Peluse and Soundararajan [15] proved this conjecture. They also proved that, for every fixed positive integer KK, almost all entries in the character table of SNS_N are divisible by KK as N→∞N \to\infty.

These results naturally lead to the question of whether similar divisibility properties hold for other families of groups with integer-valued character tables. In this paper, we study this question for the wreath product G≀SNG \wr S_N. By the Murnaghan–Nakayama rule [7 Theorem 4.4.10] (see also [4 Proposition 2.8]), if GG has an integer-valued character table, then the character table of G≀SNG \wr S_N is also integer-valued. Thus, it is natural to investigate the divisibility properties of its character values. We also mention that, in our recent work [3], we obtained a lower bound for the number of zeros in the character table of G≀SNG \wr S_N without assuming that GG has an integer-valued character table.

Dong et al. [4] proved that, for every fixed prime pp, almost all entries in the character table of G≀SNG \wr S_N are divisible by pp. Our first result extends their theorem from primes to prime powers. It also generalizes the result of Peluse and Soundararajan [15] from symmetric groups to wreath products.

Theorem 1.1. Let GG be a finite group with an integer-valued character table, and let G≀SNG \wr S_N be the wreath product of GG with the symmetric group SNS_N. Then, for every fixed prime power qq, the proportion of entries in the character table of G≀SNG \wr S_N divisible by qq tends to 1 as N→∞N \to\infty.

The proof of Theorem 1.1 uses the combinatorial description of the irreducible representations and conjugacy classes of G≀SNG \wr S_N. If GG has kk conjugacy classes, then both the irreducible representations and the conjugacy classes of G≀SNG \wr S_N are indexed by kk-multipartitions of NN.

A kk-multipartition of a positive integer NN is a sequence λ=(λ(1),λ(2),…,λ(k))\lambda= (\lambda^{(1)}, \lambda^{(2)}, \ldots, \lambda^{(k)}), where each λ(j)\lambda^{(j)} is an integer partition, possibly empty, satisfying ∑j=1k∣λ(j)∣=N\sum_{j=1}^{k} |\lambda^{(j)}| = N. Here, ∣λ(j)∣|\lambda^{(j)}| denotes the sum of the parts of λ(j)\lambda^{(j)}.

James and Kerber [7 Corollary 4.4.4] showed that the number of irreducible representations of G≀SNG \wr S_N is given by

∑N=n1+n2+⋯+nknj≥0p(n1)p(n2)⋯p(nk).\sum_{\substack{N=n_1+n_2+\cdots+n_k\\ n_j\geq0}} p(n_1)p(n_2)\cdots p(n_k).

This sum is precisely the number of kk-multipartitions of NN. Counting kk-multipartitions according to the size of their first component, we obtain

pk(N)=∑a=0Np(a)pk−1(N−a).(1)p_k(N) = \sum_{a=0}^{N} p(a)p_{k-1}(N-a). \tag*{(1)}

Equivalently, the generating function is

∑N=0∞pk(N)zN=∏n=1∞(1−zn)−k.\sum_{N=0}^{\infty} p_k(N)z^N = \prod_{n=1}^{\infty}(1-z^n)^{-k}.

General methods for obtaining asymptotic formulas from generating functions are discussed in [1, 10, 14]. In [2], we established the following asymptotic formula for the number of kk-multipartitions of NN when kk is fixed.

Theorem 1.2 ([2 Corollary 1.4]; see also [14 Theorem 4]). Let kk be any fixed positive integer. Then,

pk(N)=(k24)k+14exp⁡(2π6Nk)2Nk+34(1+O(N−12)).p_k(N) = \left(\frac{k}{24}\right)^{\frac{k+1}{4}} \frac{\exp\left(\frac{2\pi}{\sqrt{6}}\sqrt{Nk}\right)}{\sqrt{2N}^{\frac{k+3}{4}}}\left(1+O\left(N^{-\frac{1}{2}}\right)\right).

For k=1k=1 this is the Hardy–Ramanujan [6] formula p(N)∼143Nexp⁡(2π6N)p(N) \sim\frac{1}{4\sqrt{3}N}\exp\left(\frac{2\pi}{\sqrt{6}}\sqrt{N}\right).

Let Pk(N)\mathcal{P}_k(N) denote the set of all kk-multipartitions of NN, and let pk(N)=∣Pk(N)∣p_k(N)=|\mathcal{P}_k(N)|. We now state a quantitative result from which Theorem 1.1 follows.

Theorem 1.3. Let GG be a finite group with an integer-valued character table and kk conjugacy classes. Let NN be sufficiently large, and let q=prq=p^r be a prime power satisfying

q≤10−3log⁡(N/k)(log⁡log⁡N)2.(2)q \leq10^{-3}\frac{\log(N/k)}{(\log\log N)^2}. \tag*{(2)}

Then the number of pairs (λ,μ)∈Pk(N)×Pk(N)(\lambda,\mu) \in\mathcal{P}_k(N) \times\mathcal{P}_k(N) such that q∤χμλq \nmid\chi^\lambda_\mu is

Ok(pk(N)2exp⁡(−(log⁡log⁡N)2)).O_k\left(p_k(N)^2 \exp\left(-(\log\log N)^2\right)\right).

Here, χμλ\chi^\lambda_\mu denotes the value of the irreducible character of G≀SNG \wr S_N indexed by λ\lambda on the conjugacy class indexed by μ\mu. When GG is the trivial group, we have k=1k=1, and Theorem 1.3 reduces to the result of Peluse and Soundararajan [15 Theorem 1.1].

By expressing a fixed positive integer as a product of prime powers and applying the union bound, we obtain the following corollary.

Corollary 1.4. Let GG be as in Theorem 1.1, and let KK be a fixed positive integer. Then the proportion of entries in the character table of G≀SNG \wr S_N divisible by KK tends to 11 as N→∞N \to\infty.

Our results also have applications to Weyl groups. All Weyl groups have rational-valued characters, and hence their character values are integers. The Weyl groups of types BNB_N and CNC_N are both isomorphic to Z/2Z≀SN\mathbb{Z}/2\mathbb{Z} \wr S_N. Therefore, Theorem 1.1 applies directly to these two families. The remaining infinite family of type DND_N can be treated using Clifford theory. This gives the following corollary.

Corollary 1.5. Let qq be a fixed prime power. As N→∞N \to\infty, the proportion of entries in the character table of the Weyl group of type DND_N that are divisible by qq tends to 11. Hence, the same holds for every fixed positive integer in place of qq.

Acknowledgments

J. Barman is grateful to the University Grants Commission (UGC), India, for financial support through its Fellowship Programme. K. Mahatab is supported by the ARG-MATRICS Programme (Grant No. ANRF/ARGM/2025/002540/MTR).

Preliminaries

In this section, we first discuss the conjugacy classes and irreducible characters of G≀SNG \wr S_N and their correspondence with multipartitions. We then recall some basic definitions concerning hooks, border strips, and core partitions. Finally, we state the Murnaghan–Nakayama rule for G≀SNG \wr S_N.

Wreath products, conjugacy classes and irreducible characters

Let GG be a finite group. The wreath product of GG with the symmetric group SNS_N is defined by G≀SN=GN⋊SNG \wr S_N = G^N \rtimes S_N. Its elements are of the form (g1,…,gN;σ)(g_1,\ldots,g_N;\sigma), where gi∈Gg_i \in G and σ∈SN\sigma\in S_N. The multiplication is given by

(g1,…,gN;σ)(h1,…,hN;τ)=(g1hσ−1(1),…,gNhσ−1(N);στ).(g_1,\ldots,g_N;\sigma)(h_1,\ldots,h_N;\tau) = (g_1h_{\sigma^{-1}(1)},\ldots,g_Nh_{\sigma^{-1}(N)};\sigma\tau).

The group G≀SNG \wr S_N can also be viewed as the group of N×NN \times N monomial matrices with nonzero entries in GG. Its order is ∣G∣NN!|G|^N N!.

Let c1,…,ckc_1,\ldots,c_k denote the conjugacy classes of GG, and let χ1,…,χk\chi_1,\ldots,\chi_k be its irreducible characters, where χ1\chi_1 is the trivial character. Consider an element x=(g1,…,gN;σ)x=(g_1,\ldots,g_N;\sigma) of G≀SNG\wr S_N. For a cycle (i1 i2 ⋯ im)(i_1\,i_2\,\cdots\,i_m) of σ\sigma, the conjugacy class of gimgim−1⋯gi1g_{i_m}g_{i_{m-1}}\cdots g_{i_1} in GG is called the cycle product associated with this cycle [7 (4.2.1)].

The conjugacy classes of G≀SNG\wr S_N can be described using kk-multipartitions of NN. We recall the following result of James and Kerber.

Proposition 3.1 ([7 Theorem 4.2.8]). The conjugacy classes of G≀SNG\wr S_N are indexed by the elements of Pk(N)\mathcal{P}_k(N). More precisely, the conjugacy class of xx corresponds to the kk-multipartition μ=(μ(1),…,μ(k))\mu=(\mu^{(1)},\ldots,\mu^{(k)}), where μ(j)\mu^{(j)} is the partition consisting of the lengths of those cycles of σ\sigma whose cycle products belong to cjc_j.

Let V1,…,VkV_1,\ldots,V_k be the irreducible representations of GG corresponding to the irreducible characters χ1,…,χk\chi_1,\ldots,\chi_k, respectively. We recall the following result of James and Kerber.

Proposition 3.2 ([7 Theorem 4.4.3]). The irreducible representations of G≀SNG\wr S_N are indexed by the elements of Pk(N)\mathcal{P}_k(N). Let λ=(λ(1),…,λ(k))∈Pk(N)\lambda=(\lambda^{(1)},\ldots,\lambda^{(k)})\in\mathcal{P}_k(N), and put ai=∣λ(i)∣a_i=|\lambda^{(i)}|. Then the corresponding irreducible representation is given by

Vλ=Ind⁡(G≀Sa1)×⋯×(G≀Sak)G≀SN(⨂i=1k(Sλ(i)⊗Vi⊗ai)).V^\lambda=\operatorname{Ind}_{(G\wr S_{a_1})\times\cdots\times(G\wr S_{a_k})}^{G\wr S_N}\left(\bigotimes_{i=1}^k\left(S^{\lambda^{(i)}}\otimes V_i^{\otimes a_i}\right)\right).

where Sλ(i)S^{\lambda^{(i)}} denotes the Specht module of SaiS_{a_i} corresponding to the partition λ(i)\lambda^{(i)}.

Let χλ\chi^\lambda denote the character of VλV^\lambda, and let χμλ\chi^\lambda_\mu be its value on the conjugacy class indexed by μ\mu. If GG has an integer-valued character table, that is, χi(cj)∈Z\chi_i(c_j)\in\mathbb{Z} for all 1≤i,j≤k1\leq i,j\leq k, then the character values of G≀SNG\wr S_N are also integers. In particular, χμλ∈Z\chi^\lambda_\mu\in\mathbb{Z} for all λ,μ∈Pk(N)\lambda,\mu\in\mathcal{P}_k(N) [7 Corollary 4.4.11].

Moreover, by [16 §13.1], the character table of GG is integer-valued if and only if every element g∈Gg\in G is conjugate to gag^a for every integer aa relatively prime to the order of gg.

Throughout the paper, we assume that GG is a finite group with an integer-valued character table and kk conjugacy classes.

Hooks, border strips and the Murnaghan–Nakayama rule

Let α\alpha be a partition. The hook corresponding to a box bb in the Young diagram of α\alpha consists of bb and all boxes to its right and below it. The hook length ℓ(h)\ell(h) is the total number of boxes in the hook, while its height ht⁡(h)\operatorname{ht}(h) is one less than the number of rows occupied by the hook.

The border strip, also called a rim hook, associated with hh is denoted by bs⁡(h)\operatorname{bs}(h). It consists of connected boxes along the south-east boundary of the Young diagram, starting from the rightmost box of hh and ending at its bottommost box. The border strip bs⁡(h)\operatorname{bs}(h) has ℓ(h)\ell(h) boxes and the same height as hh. Removing bs⁡(h)\operatorname{bs}(h) from α\alpha gives a partition α\bs⁡(h)\alpha\backslash\operatorname{bs}(h) of ∣α∣−ℓ(h)|\alpha|-\ell(h). Conversely, every removable border strip of α\alpha corresponds to a unique hook of α\alpha. Here, a border strip is a connected skew Young diagram containing no 2×22\times2 square.

For a positive integer tt, a partition α\alpha is called a tt-core if none of its hook lengths is divisible by tt.

Let λ=(λ(1),…,λ(k))∈Pk(N)\lambda= (\lambda^{(1)}, \ldots, \lambda^{(k)}) \in\mathcal{P}_k(N). For an index ii and a hook hh of λ(i)\lambda^{(i)}, we denote by λ\ih\lambda\backslash_i h the kk-multipartition of N−ℓ(h)N-\ell(h) obtained by removing the border strip bs⁡(h)\operatorname{bs}(h) from λ(i)\lambda^{(i)}, while leaving the other components unchanged.

Similarly, let μ=(μ(1),…,μ(k))∈Pk(N)\mu= (\mu^{(1)}, \ldots, \mu^{(k)}) \in\mathcal{P}_k(N). If μ(j)\mu^{(j)} contains a part equal to mm, then μ\jm\mu\backslash_j m denotes the kk-multipartition obtained by deleting one such part from μ(j)\mu^{(j)}.

Proposition 3.3 (Murnaghan–Nakayama rule for G≀SNG \wr S_N). Let λ,μ∈Pk(N)\lambda, \mu\in\mathcal{P}_k(N), and suppose that μ(j)\mu^{(j)} contains a part equal to mm. Then

χμλ=∑i=1kχi(cj)∑h is a hook of λ(i)ℓ(h)=m(−1)ht⁡(h)χμ\jmλ\ih.(3)\chi^\lambda_\mu= \sum_{i=1}^{k} \chi_i(c_j) \sum_{\substack{h\text{ is a hook of }\lambda^{(i)}\\ \ell(h)=m}} (-1)^{\operatorname{ht}(h)} \chi^{\lambda\backslash_i h}_{\mu\backslash_j m}. \tag*{(3)}

where the inner sum runs over all hooks of length mm in λ(i)\lambda^{(i)}. Here, χ(∅,…,∅)(∅,…,∅)=1\chi^{(\varnothing,\ldots,\varnothing)}_{(\varnothing,\ldots,\varnothing)}=1.

Proof. By the Murnaghan–Nakayama rule for wreath products [7 Theorem 4.4.10] (see also [4 Proposition 2.8]), the character value χμλ\chi^\lambda_\mu can be expressed as a sum over all rim hook decompositions of λ\lambda corresponding to μ\mu. In each decomposition, border strips are removed successively from the components of λ\lambda, with lengths equal to the parts of μ\mu in a fixed order.

If a border strip is removed from λ(i)\lambda^{(i)} corresponding to a part of μ(j)\mu^{(j)}, then it contributes the factor χi(cj)\chi_i(c_j) together with the sign (−1)ht⁡(h)(-1)^{\operatorname{ht}(h)}, where hh is the corresponding hook.

Since the rule holds for any fixed ordering of the parts of μ\mu, we choose the part mm of μ(j)\mu^{(j)} first. We then group the rim hook decompositions according to the component λ(i)\lambda^{(i)} and the hook hh of length mm removed in the first step.

After removing this border strip, the remaining decomposition corresponds to λ\ih\lambda\backslash_i h and μ\jm\mu\backslash_j m. Summing over all possible components ii and hooks hh of length mm, we obtain (3.1). □

When GG is the trivial group, we have k=1k=1, and (3.1) reduces to the classical Murnaghan–Nakayama rule [7 Theorem 2.4.7].

Permutation characters

Let λ∈Pk(N)\lambda\in\mathcal{P}_k(N), and write ai=∣λ(i)∣a_i = |\lambda^{(i)}|. Let Sλ(i)S_{\lambda^{(i)}} be the Young subgroup of SaiS_{a_i} corresponding to λ(i)\lambda^{(i)}, and put Gλ(i)=G≀Sλ(i)G_{\lambda^{(i)}} = G \wr S_{\lambda^{(i)}}. Following [4 Definition 2.9], we define the permutation module

Mλ=Ind⁡Gλ(1)×⋯×Gλ(k)G≀SN(⨂i=1kVi⊗ai).M^\lambda= \operatorname{Ind}_{G_{\lambda^{(1)}}\times\cdots\times G_{\lambda^{(k)}}}^{G\wr S_N}\left(\bigotimes_{i=1}^{k} V_i^{\otimes a_i}\right).

We denote the value of its character on the conjugacy class indexed by μ\mu by MμλM^\lambda_\mu.

Definition 1 ([4 Definition 2.10]). Let λ,μ∈Pk(N)\lambda, \mu\in\mathcal{P}_k(N). We treat each row of the components of λ\lambda and μ\mu as a distinct object, even if some rows have the same length. A row decomposition of λ\lambda by μ\mu is a map ρ\rho from the rows of μ\mu to the rows of λ\lambda such that, for each row rr of λ\lambda, the

sum of the lengths of the rows in ρ−1(r)\rho^{-1}(\mathbf{r}) equals the length of r\mathbf{r}. We denote the set of all such row decompositions by RD(λ,μ)RD(\lambda,\mu).

The following proposition gives a formula for the permutation character values.

Proposition 3.4 ([4 Proposition 2.11]). For λ,μ∈Pk(N)\lambda,\mu\in\mathcal{P}_k(N), we have

Mμλ=∑ρ∈RD(λ,μ)∏i=1k∏j=1kχi(cj)aij(ρ).M^\lambda_\mu=\sum_{\rho\in RD(\lambda,\mu)}\prod_{i=1}^{k}\prod_{j=1}^{k}\chi_i(c_j)^{a_{ij}(\rho)}.

where aij(ρ)a_{ij}(\rho) is the number of rows of μ(j)\mu^{(j)} assigned to rows of λ(i)\lambda^{(i)} by ρ\rho.

Recall that the dominance order on partitions of the same size is defined by α⪰β\alpha\succeq\beta if

∑r=1sαr≥∑r=1sβr\sum_{r=1}^{s}\alpha_r\ge\sum_{r=1}^{s}\beta_r

for every s≥1s\ge1. For kk-multipartitions λ\lambda and η\eta with equal corresponding component sizes, we write λ⪰η\lambda\succeq\eta if λ(i)⪰η(i)\lambda^{(i)}\succeq\eta^{(i)} for every 1≤i≤k1\le i\le k.

Lemma 3.5 ([4 Lemma 2.13]). For λ,η∈Pk(N)\lambda,\eta\in\mathcal{P}_k(N), we have

[Mλ:Vη]=∏i=1kKη(i),λ(i).[M^\lambda:V^\eta]=\prod_{i=1}^{k}K_{\eta^{(i)},\lambda^{(i)}}.

if ∣η(i)∣=∣λ(i)∣|\eta^{(i)}|=|\lambda^{(i)}| for every ii, and [Mλ:Vη]=0[M^\lambda:V^\eta]=0 otherwise. Here Kα,βK_{\alpha,\beta} denotes the Kostka number. In particular, the matrix ([Mλ:Vη])λ,η∈Pk(N)([M^\lambda:V^\eta])_{\lambda,\eta\in\mathcal{P}_k(N)} is unitriangular with respect to a total ordering refining the componentwise dominance order, and hence is invertible over Z\mathbb{Z}.

Corollary 3.6. Every irreducible character χλ\chi^\lambda of G≀SNG\wr S_N is an integral linear combination of the permutation characters MηM^\eta, where η∈Pk(N)\eta\in\mathcal{P}_k(N).

Proof. By Lemma 3.5, the character of MλM^\lambda satisfies

Mμλ=∑η∈Pk(N)[Mλ:Vη]χμη.M^\lambda_\mu=\sum_{\eta\in\mathcal{P}_k(N)}[M^\lambda:V^\eta]\chi^\eta_\mu.

The matrix of multiplicities is unitriangular with integer entries, so its inverse also has integer entries. Therefore, every irreducible character is an integral linear combination of the permutation characters. □\square

Counting partitions and multipartitions

We recall the following results on the number of partitions and multipartitions.

Lemma 3.7. Let k≥1k\ge1 be fixed. Then there exist positive constants Ck,ckC_k,c_k and N0(k)N_0(k) such that, for every N≥N0(k)N\ge N_0(k), the following hold.

  1. If 0≤t≤N20\le t\le\frac{N}{2}, then

pk(N−t)≤Ckpk(N)exp⁡(−πk t6N).p_k(N-t)\le C_kp_k(N)\exp\left(-\frac{\pi\sqrt{k}\,t}{\sqrt{6N}}\right).

(2) If N2<t≤N\frac{N}{2}<t\le N, then

pk(N−t)≤Ckpk(N)exp⁡(−ckN).p_k(N-t)\le C_kp_k(N)\exp\left(-c_k\sqrt{N}\right).

Proof. (1) Let NN be sufficiently large and 0≤t≤N20\le t\le\frac{N}{2}. Since N−t≥N2N-t\ge\frac{N}{2}, Lemma 1.2 gives

pk(N−t)pk(N)≤2(NN−t)k+34exp⁡(−2πk6(N−N−t)).\frac{p_k(N-t)}{p_k(N)}\le2\left(\frac{N}{N-t}\right)^{\frac{k+3}{4}}\exp\left(-\frac{2\pi\sqrt{k}}{\sqrt{6}}\left(\sqrt{N}-\sqrt{N-t}\right)\right).

Now, NN−t≤2\frac{N}{N-t}\le2 and N−N−t=tN+N−t≥t2N\sqrt{N}-\sqrt{N-t}=\frac{t}{\sqrt{N}+\sqrt{N-t}}\ge\frac{t}{2\sqrt{N}}. Therefore,

pk(N−t)≤2k+74pk(N)exp⁡(−πkt6N),p_k(N-t)\le2^{\frac{k+7}{4}}p_k(N)\exp\left(-\frac{\pi\sqrt{k}t}{\sqrt{6}N}\right),

which proves (1).

(2) Suppose that N2<t≤N\frac{N}{2}<t\le N. Since pk(n)p_k(n) is non-decreasing, we have

pk(N−t)≤pk(N−⌈N2⌉).p_k(N-t)\le p_k\left(N-\left\lceil\frac{N}{2}\right\rceil\right).

Applying part (1) with t=⌊N2⌋t=\left\lfloor\frac{N}{2}\right\rfloor, we obtain

pk(N−t)≤Ckpk(N)exp⁡(−πk6N⌊N2⌋)≤Ckpk(N)exp⁡(−ckN),p_k(N-t)\le C_kp_k(N)\exp\left(-\frac{\pi\sqrt{k}}{\sqrt{6}N}\left\lfloor\frac{N}{2}\right\rfloor\right)\le C_kp_k(N)\exp(-c_k\sqrt{N}),

for some constant ck>0c_k>0 and all sufficiently large NN. This completes the proof. □

We next recall a lemma that will be useful in the proof of Lemma 7.1.

Lemma 3.8 ([13 Lemma 5]). For integers 1≤t≤n1\le t\le n, the number of partitions of nn that are not tt-cores is at most (t+1)p(n−t)(t+1)p(n-t).

Proofs of the Main Results

In this section, we first define the relation ∼q\sim_q on multipartitions.

Definition 2. Let q=prq=p^r, where pp is a prime and r≥1r\ge1, and let μ,ν∈Pk(N)\mu,\nu\in\mathcal{P}_k(N). We say that ν\nu is obtained from μ\mu by a qq-reduction if there exist j∈{1,…,k}j\in\{1,\ldots,k\} and m≥1m\ge1 such that ν(i)=μ(i)\nu^{(i)}=\mu^{(i)} for every i≠ji\ne j, and ν(j)\nu^{(j)} is obtained from μ(j)\mu^{(j)} by replacing prp^r parts of size mm with pr−1p^{r-1} parts of size pmpm. Let ∼q\sim_q denote the equivalence relation on Pk(N)\mathcal{P}_k(N) generated by these reductions.

We now state a congruence relation for the character values of G≀SNG\wr S_N.

Proposition 4.1. If μ∼qν\mu\sim_q\nu, then

χμλ≡χνλ(modpr)\chi^\lambda_\mu\equiv\chi^\lambda_\nu\pmod{p^r}

for every λ∈Pk(N)\lambda\in\mathcal{P}_k(N).

Starting with μ∈Pk(N)\mu\in\mathcal{P}_k(N), we repeatedly apply qq-reductions whenever a part occurs at least prp^r times in some component. Each reduction decreases the total number of parts, so the process terminates.

The final multipartition does not depend on the order in which the reductions are performed. Indeed, reductions involving different components or different part sizes can be performed in either order and give the same result. If two reductions involve the same component and the same part size, they produce the same multipartition. Since each reduction decreases the total number of parts, the process eventually stops, and the final multipartition is uniquely determined.

We denote the resulting multipartition by μ~\widetilde{\mu}. By construction, each component μ~(j)\widetilde{\mu}^{(j)} is obtained from μ(j)\mu^{(j)} by the reduction procedure described in [15 Proposition 2.3]. Hence, Proposition 4.1 gives

χμλ≡χμ~λ(modpr)(4)\chi^\lambda_\mu\equiv\chi^\lambda_{\widetilde{\mu}} \pmod{p^r} \tag*{(4)}

for every λ∈Pk(N)\lambda\in\mathcal{P}_k(N).

Theorem 4.2 (Divisibility criterion). Let m1,…,mrm_1,\ldots,m_r be distinct positive integers, and let j1,…,jr∈{1,…,k}j_1,\ldots,j_r \in\{1,\ldots,k\}, not necessarily distinct. Suppose that μ∈Pk(N)\mu\in\mathcal{P}_k(N) is such that, for each 1≤i≤r1 \le i \le r, the component μ(ji)\mu^{(j_i)} contains at least pr−1p^{r-1} parts equal to mim_i. Define

T=T(m1,…,mr):={∑i=1rkimi:0≤ki≤pr−1 for all i, ki=pr−1 for some i}.\mathcal{T}=\mathcal{T}(m_1,\ldots,m_r):=\left\{\sum_{i=1}^{r} k_i m_i: 0\le k_i\le p^{r-1}\text{ for all }i,\ k_i=p^{r-1}\text{ for some }i\right\}.

If every component λ(a)\lambda^{(a)}, 1≤a≤k1\le a\le k, is a tt-core for every t∈Tt\in\mathcal{T}, then

pr∣χμλ.p^r \mid\chi^\lambda_\mu.

When r=1r=1, the result gives χμλ=0\chi^\lambda_\mu=0 if every component of λ\lambda is an m1m_1-core, as used in [4].

Proposition 4.3. Let k≥1k\ge1 be fixed. Then there exist constants N1(k)N_1(k) and C>0C>0 such that, for every N≥N1(k)N\ge N_1(k) and every prime power q=prq=p^r satisfying (1.2), the following holds. For all but at most

Ck pk(N)exp⁡(−(Nk)12q)Ck\,p_k(N)\exp\left(-\left(\frac{N}{k}\right)^{\frac{1}{2q}}\right)

multipartitions μ∈Pk(N)\mu\in\mathcal{P}_k(N), there exists a jj such that ∣μ(j)∣≥Nk|\mu^{(j)}|\ge\frac{N}{k} and μ~(j)\widetilde{\mu}^{(j)} contains rr distinct parts m1,…,mrm_1,\ldots,m_r, each occurring at least pr−1p^{r-1} times, with

pr−1mi>(1+16q)62πNklog⁡Nk(1≤i≤r).(5)p^{r-1}m_i>\left(1+\frac{1}{6q}\right)\frac{\sqrt{6}}{2\pi}\sqrt{\frac{N}{k}}\log\frac{N}{k}\qquad(1\le i\le r). \tag*{(5)}

Proposition 4.4. Let k≥1k\ge1 be fixed. Then there exist constants Ck′>0C'_k>0 and N2(k)N_2(k) such that, for every N≥N2(k)N\ge N_2(k), every 0<η≤10<\eta\le1, and every integer

t≥T0(η):=(1+η)62πNklog⁡Nk,t\ge T_0(\eta):=(1+\eta)\frac{\sqrt{6}}{2\pi}\sqrt{\frac{N}{k}}\log\frac{N}{k},

the number of multipartitions λ∈Pk(N)\lambda\in\mathcal{P}_k(N) for which at least one component λ(a)\lambda^{(a)} is not a tt-core is at most

Ck′pk(N)log⁡NNη2.C'_k p_k(N)\frac{\log N}{N^{\frac{\eta}{2}}}.

The preceding results provide the necessary tools to prove Theorem 1.3.

Proof of Theorem 1.3. Assume Propositions 4.1, 4.3, 4.4, and Theorem 4.2. Let NN be sufficiently large, and let q=prq=p^r satisfy (2). Set η=16q\eta=\frac{1}{6q}. We divide the pairs (λ,μ)∈Pk(N)×Pk(N)(\lambda,\mu)\in\mathcal{P}_k(N)\times\mathcal{P}_k(N) for which q∤χλμq\nmid\chi_{\lambda}^{\mu} into two cases.

Case 1.

Suppose that μ\mu belongs to the exceptional set in Proposition 4.3. The number of such pairs is at most

Ckpk(N)2exp⁡(−(Nk)12q).C k p_k(N)^2\exp\left(-\left(\frac{N}{k}\right)^{\frac{1}{2q}}\right).

Case 2.

Suppose that μ\mu does not belong to the exceptional set. By Proposition 4.3, there exist an index jj and distinct positive integers m1,…,mrm_1,\ldots,m_r satisfying (5). Let T=T(m1,…,mr)\mathcal{T}=\mathcal{T}(m_1,\ldots,m_r). For every t∈Tt\in\mathcal{T}, we have t≥pr−1mit\ge p^{r-1}m_i for some ii. Hence, by (5),

t>(1+16q)62πNklog⁡Nk=T0(η).t>\left(1+\frac{1}{6q}\right)\frac{\sqrt{6}}{2\pi}\sqrt{\frac{N}{k}}\log\frac{N}{k}=T_0(\eta).

Moreover,

∣T∣≤r(pr−1+1)r−1,|\mathcal{T}|\le r(p^{r-1}+1)^{r-1},

since there are rr choices for an index ii with ki=pr−1k_i=p^{r-1}, and each of the remaining r−1r-1 entries can take any integer value from 00 to pr−1p^{r-1}. Thus, there are at most r(pr−1+1)r−1r(p^{r-1}+1)^{r-1} such rr-tuples.

By Proposition 4.4, for each t∈Tt\in\mathcal{T}, the number of multipartitions λ\lambda having at least one component that is not a tt-core is at most

Ck′pk(N)log⁡NN112q.C'_k p_k(N)\frac{\log N}{N^{\frac{1}{12q}}}.

Therefore, the number of multipartitions λ\lambda having at least one component that is not a tt-core for some t∈Tt\in\mathcal{T} is at most

Ck′∣T∣pk(N)log⁡NN112q.C'_k|\mathcal{T}|p_k(N)\frac{\log N}{N^{\frac{1}{12q}}}.

For every remaining λ\lambda, all components are tt-cores for every t∈Tt\in\mathcal{T}. Since μ~(j)\widetilde{\mu}^{(j)} contains at least pr−1p^{r-1} parts equal to each mim_i, Theorem 4.2 gives

pr∣χμ~λ.p^r\mid\chi_{\widetilde{\mu}}^{\lambda}.

By (4), we also have

pr∣χμλ.p^{r} \mid\chi^\lambda_\mu.

Thus, for each non-exceptional μ\mu, there are at most

Ck′r(pr−1+1)r−1pk(N)log⁡NN112q.C'_k r(p^{r-1}+1)^{r-1}p_k(N)\frac{\log N}{N^{\frac{1}{12q}}}.

choices of λ\lambda for which q∤χμλq \nmid\chi^\lambda_\mu.

Combining the two cases, the total number of pairs (λ,μ)(\lambda,\mu) with q∤χμλq \nmid\chi^\lambda_\mu is at most

pk(N)2(Ckexp⁡(−(Nk)120q)+Ck′r(pr−1+1)r−1log⁡NN112q).(6)p_k(N)^2\left(C_k\exp\left(-\left(\frac{N}{k}\right)^{\frac{1}{20q}}\right)+C'_k r(p^{r-1}+1)^{r-1}\frac{\log N}{N^{\frac{1}{12q}}}\right). \tag*{(6)}

It remains to estimate these two terms. Write L=log⁡log⁡N\mathcal{L}=\log\log N. By (1.2), we have

q≤10−3log⁡(Nk)L2.q\leq\frac{10^{-3}\log\left(\frac{N}{k}\right)}{\mathcal{L}^2}.

Hence,

(Nk)120q=exp⁡(log⁡Nk20q)≥exp⁡(50L2)\left(\frac{N}{k}\right)^{\frac{1}{20q}}=\exp\left(\frac{\log\frac{N}{k}}{20q}\right)\geq\exp(50\mathcal{L}^2)

and

N−112q=exp⁡(−log⁡N12q)≤exp⁡(−83L2).N^{-\frac{1}{12q}}=\exp\left(-\frac{\log N}{12q}\right)\leq\exp(-83\mathcal{L}^2).

Since pr−1+1≤pr=qp^{r-1}+1\leq p^r=q and r≤log⁡qlog⁡2≤qr\leq\frac{\log q}{\log2}\leq q, we obtain

r(pr−1+1)r−1≤qr≤exp⁡((log⁡q)2log⁡2).r(p^{r-1}+1)^{r-1}\leq q^r\leq\exp\left(\frac{(\log q)^2}{\log2}\right).

Also, q≤log⁡Nq\leq\log N for sufficiently large NN, so log⁡q≤L\log q\leq\mathcal{L}. Therefore,

r(pr−1+1)r−1≤exp⁡(1.5L2).r(p^{r-1}+1)^{r-1}\leq\exp(1.5\mathcal{L}^2).

It follows that the second term in (4.3) is bounded by

Ck′exp⁡(1.5L2+L−83L2)≪kexp⁡(−L2).C'_k\exp(1.5\mathcal{L}^2+\mathcal{L}-83\mathcal{L}^2)\ll_k\exp(-\mathcal{L}^2).

Similarly, the first term is at most

Ckexp⁡(−exp⁡(50L2))≪kexp⁡(−L2).C_k\exp\left(-\exp(50\mathcal{L}^2)\right)\ll_k\exp(-\mathcal{L}^2).

Hence, the number of pairs (λ,μ)∈Pk(N)×Pk(N)(\lambda,\mu)\in\mathcal{P}_k(N)\times\mathcal{P}_k(N) satisfying q∤χμλq\nmid\chi^\lambda_\mu is

Ok(pk(N)2exp⁡(−(log⁡log⁡N)2)).O_k\left(p_k(N)^2\exp\left(-(\log\log N)^2\right)\right).

This completes the proof of Theorem 1.3. □

We now turn to the proof of Theorem 1.1.

Proof of Theorem 1.1. For any fixed prime power qq, the condition (1.2) holds for all sufficiently large NN. By Theorem 1.3, the proportion of entries in the character table that are not divisible by qq is

Ok(exp⁡(−(log⁡log⁡N)2))⟶0,as N→∞.O_k\left(\exp\left(-(\log\log N)^2\right)\right)\longrightarrow0,\qquad\text{as }N\to\infty.

This proves Theorem 1.1. □

The preceding result allows us to prove the following Corollary.

Proof of Corollary 1.4. Now let KK be a fixed positive integer with prime power factorization K=q1⋯qsK=q_1\cdots q_s, where the qiq_i are pairwise coprime. An integer is divisible by KK if and only if it is divisible by every qiq_i. Therefore, the proportion of character table entries not divisible by KK is at most

∑i=1sOk(exp⁡(−(log⁡log⁡N)2))⟶0as N→∞.\sum_{i=1}^{s}\mathcal{O}_k\left(\exp\left(-(\log\log N)^2\right)\right)\longrightarrow0 \qquad\text{as } N\to\infty.

This proves Corollary 1.4. □

Proof of Proposition 4.1

To prove Proposition 4.1, we first establish some necessary results, which are discussed below. We then use these results to prove the proposition.

Lemma 5.1. Let F∈Z[x1,…,xn]F\in\mathbb{Z}[x_1,\ldots,x_n], and write F(xp)=F(x1p,…,xnp)F(x^p)=F(x_1^p,\ldots,x_n^p). Then, for every integer r≥1r\ge1, we have

F(x)pr≡F(xp)pr−1(modpr).F(x)^{p^r}\equiv F(x^p)^{p^{r-1}}\pmod{p^r}.

Proof. First, suppose that r=1r=1. Write F(x)=∑αaαxαF(x)=\sum_{\alpha}a_\alpha x^\alpha, where aα∈Za_\alpha\in\mathbb{Z}. By the multinomial theorem, we have

F(x)p=∑αaαpxpα+pR(x),F(x)^p=\sum_{\alpha}a_\alpha^p x^{p\alpha}+pR(x),

for some R∈Z[x1,…,xn]R\in\mathbb{Z}[x_1,\ldots,x_n], since every multinomial coefficient involving at least two distinct terms is divisible by pp. By Fermat’s little theorem, aαp≡aα(modp)a_\alpha^p\equiv a_\alpha\pmod p. Hence,

F(x)p=F(xp)+pR1(x),(7)F(x)^p=F(x^p)+pR_1(x), \tag*{(7)}

for some R1∈Z[x1,…,xn]R_1\in\mathbb{Z}[x_1,\ldots,x_n].

Now let r≥2r\ge2. Taking the pr−1p^{r-1}-th power on both sides of (5.1), we obtain

F(x)pr=F(xp)pr−1+∑ℓ=1pr−1(pr−1ℓ)pℓR1(x)ℓF(xp)pr−1−ℓ.F(x)^{p^r}=F(x^p)^{p^{r-1}}+\sum_{\ell=1}^{p^{r-1}}\binom{p^{r-1}}{\ell}p^\ell R_1(x)^\ell F(x^p)^{p^{r-1}-\ell}.

For 1≤ℓ≤pr−11\le\ell\le p^{r-1}, we have

(pr−1ℓ)=pr−1ℓ(pr−1−1ℓ−1).\binom{p^{r-1}}{\ell}=\frac{p^{r-1}}{\ell}\binom{p^{r-1}-1}{\ell-1}.

Therefore,

vp((pr−1ℓ))≥r−1−vp(ℓ),v_p\left(\binom{p^{r-1}}{\ell}\right)\ge r-1-v_p(\ell),

where vp(ℓ)v_p(\ell) denotes the exponent of pp dividing ℓ\ell. Since vp(ℓ)≤ℓ−1v_p(\ell) \le\ell- 1, it follows that

vp((pr−1ℓ)pℓ)≥r−1−vp(ℓ)+ℓ≥r.v_p\left(\binom{p^{r-1}}{\ell}p^\ell\right) \ge r - 1 - v_p(\ell) + \ell\ge r.

Thus, every term in the sum is divisible by prp^r, which proves the lemma. □\square

Remark 1. Taking F(x)=aF(x)=a, where a∈Za\in\mathbb{Z}, Lemma 5.1 gives

apr≡apr−1(modpr).a^{p^r} \equiv a^{p^{r-1}} \pmod{p^r}.

When p∤ap\nmid a, this follows from Euler’s theorem, since φ(pr)=pr−1(p−1)\varphi(p^r)=p^{r-1}(p-1). When p∣ap\mid a, both terms are divisible by prp^r, since pr−1≥rp^{r-1}\ge r. The polynomial congruence in Lemma 5.1 allows us to extend this result to the sums arising in the row decomposition formula.

In the following subsection, we discuss how MμλM^\lambda_\mu can be expressed as a coefficient of a polynomial. Related results can be found in [9 Chapter I, Appendix B] and [8].

Permutation characters as coefficients

Fix λ∈Pk(N)\lambda\in\mathcal{P}_k(N). For each row of λ\lambda, introduce a variable xi,ax_{i,a}, where 1≤i≤k1\le i\le k and 1≤a≤ℓ(λ(i))1\le a\le\ell(\lambda^{(i)}). Define

xλ:=∏i=1k∏a=1ℓ(λ(i))xi,aλa(i).x^\lambda:=\prod_{i=1}^{k}\prod_{a=1}^{\ell(\lambda^{(i)})}x_{i,a}^{\lambda^{(i)}_a}.

For m≥1m\ge1 and 1≤j≤k1\le j\le k, let

Pm,j(x):=∑i=1kχi(cj)∑a=1ℓ(λ(i))xi,am∈Z[x],P_{m,j}(x):=\sum_{i=1}^{k}\chi_i(c_j)\sum_{a=1}^{\ell(\lambda^{(i)})}x_{i,a}^{m}\in\mathbb{Z}[x],

where we use the assumption that χi(cj)∈Z\chi_i(c_j)\in\mathbb{Z}. For any polynomial HH, let [xλ]H[x^\lambda]H denote the coefficient of xλx^\lambda in HH.

Lemma 5.2. For every μ∈Pk(N)\mu\in\mathcal{P}_k(N), we have

Mμλ=[xλ]∏j=1k∏m∈μ(j)Pm,j(x).M^\lambda_\mu=[x^\lambda]\prod_{j=1}^{k}\prod_{m\in\mu^{(j)}}P_{m,j}(x).

where the inner product runs over all parts of μ(j)\mu^{(j)}, counted with multiplicity.

Proof. We label the rows of λ\lambda by (i,a)(i,a), where 1≤i≤k1\le i\le k and 1≤a≤ℓ(λ(i))1\le a\le\ell(\lambda^{(i)}). Similarly, we label the rows of μ\mu by (j,b)(j,b), where 1≤j≤k1\le j\le k and 1≤b≤ℓ(μ(j))1\le b\le\ell(\mu^{(j)}). By the definition of Pm,j(x)P_{m,j}(x), we have

Pμb(j),j(x)=∑i=1kχi(cj)∑a=1ℓ(λ(i))xi,aμb(j).P_{\mu_b^{(j)},j}(x)=\sum_{i=1}^{k}\chi_i(c_j)\sum_{a=1}^{\ell(\lambda^{(i)})}x_{i,a}^{\mu_b^{(j)}}.

When we expand the product

∏j=1k∏m∈μ(j)Pm,j(x),\prod_{j=1}^{k}\prod_{m\in\mu^{(j)}}P_{m,j}(x),

we choose one term from each factor. This is equivalent to assigning each row of μ\mu to a row of λ\lambda. Thus, each choice determines a map ρ\rho from the rows of μ\mu to the rows of λ\lambda.

For each row (i,a)(i,a) of λ\lambda, let si,a(ρ)s_{i,a}(\rho) be the sum of the lengths of the rows of μ\mu assigned to (i,a)(i,a). Then the corresponding monomial is

∏i=1k∏a=1ℓ(λ(i))xi,asi,a(ρ).\prod_{i=1}^{k}\prod_{a=1}^{\ell(\lambda^{(i)})}x_{i,a}^{s_{i,a}(\rho)}.

This monomial equals xλx^\lambda if and only if

si,a(ρ)=λa(i)s_{i,a}(\rho)=\lambda_a^{(i)}

for every row (i,a)(i,a) of λ\lambda. By the definition of a row decomposition, this holds precisely when ρ∈RD(λ,μ)\rho\in RD(\lambda,\mu). For such a row decomposition ρ\rho, the corresponding coefficient is

∏i=1k∏j=1kχi(cj)aij(ρ),\prod_{i=1}^{k}\prod_{j=1}^{k}\chi_i(c_j)^{a_{ij}(\rho)},

where aij(ρ)a_{ij}(\rho) is the number of rows of μ(j)\mu^{(j)} assigned to rows of λ(i)\lambda^{(i)}.

Therefore, the coefficient of xλx^\lambda is

[xλ]∏j=1k∏m∈μ(j)Pm,j(x)=∑ρ∈RD(λ,μ)∏i=1k∏j=1kχi(cj)aij(ρ)=Mμλ,\begin{aligned} [x^\lambda]\prod_{j=1}^{k}\prod_{m\in\mu^{(j)}}P_{m,j}(x) = \sum_{\rho\in RD(\lambda,\mu)} \prod_{i=1}^{k}\prod_{j=1}^{k}\chi_i(c_j)^{a_{ij}(\rho)} \\ &= M^\lambda_\mu, \end{aligned}

where the last equality follows from Proposition 3.4. □

Example 1. Let G=Z/2ZG=\mathbb{Z}/2\mathbb{Z}, with conjugacy classes c1={1}c_1=\{1\} and c2={−1}c_2=\{-1\}. The irreducible characters of GG satisfy

χ1(c1)=χ1(c2)=χ2(c1)=1,χ2(c2)=−1.\chi_1(c_1)=\chi_1(c_2)=\chi_2(c_1)=1,\qquad\chi_2(c_2)=-1.

Consider the 22-multipartitions

λ=((4,2),(3)),μ=((2,2,1),(3,1))∈P2(9).\lambda=((4,2),(3)),\qquad\mu=((2,2,1),(3,1))\in P_2(9).

We compute MμλM^\lambda_\mu using Proposition 3.4 and Lemma 5.2.

Using Proposition 3.4. We first compute MμλM^\lambda_\mu using the row decomposition formula. The components λ(1)=(4,2)\lambda^{(1)}=(4,2) and λ(2)=(3)\lambda^{(2)}=(3) have rows of lengths 44, 22 and 33, respectively. Similarly, μ(1)=(2,2,1)\mu^{(1)}=(2,2,1) and μ(2)=(3,1)\mu^{(2)}=(3,1) have rows of lengths 22, 22, 11 and 33, 11, respectively. We treat the two rows of length 22 as distinct rows.

A row decomposition assigns each row of μ\mu to a row of λ\lambda so that the sum of the assigned row lengths equals the length of the corresponding row of λ\lambda. Each row contributes a factor of $1$, except when a row of μ(2)\mu^{(2)} is sent to λ(2)\lambda^{(2)}, in which case it contributes $-1.

We consider two cases depending on where the row of length 33 in μ(2)\mu^{(2)} is assigned.

Case 1. Suppose this row is assigned to the row of length 33 in λ(2)\lambda^{(2)}. The remaining rows have lengths 2,2,1,12,2,1,1 and must fill the rows of lengths 4,24,2 in λ(1)\lambda^{(1)}. Use the rows 21,22,11,122_1,2_2,1_1,1_2 (the two rows of length 22 are distinct). This can be done in three ways:

4=2+1+1,2=2(two ways),4=2+1+1,\qquad2=2 \qquad\text{(two ways),}
4=2+2,2=1+1(one way).4=2+2,\qquad2=1+1 \qquad\text{(one way).}

Possibility 1

212111122_1\quad2_1\quad1_1\quad1_2

22222_2\quad2_2

Possibility 2

222211122_2\quad2_2\quad1_1\quad1_2

21212_1\quad2_1

Possibility 3

212122222_1\quad2_1\quad2_2\quad2_2

11121_1\quad1_2

Each decomposition has weight −1-1. Thus, the total contribution is −3-3.

Case 2.

Suppose the row of length 33 in μ(2)\mu^{(2)} is assigned to the row of length 44 in λ(1)\lambda^{(1)}. One of the two rows of length 11 must also be assigned to this row. The row of length 33 in λ(2)\lambda^{(2)} is then filled by a row of length 22 and the remaining row of length 11. The other row of length 22 fills the row of length 22 in λ(1)\lambda^{(1)}.

There are two choices for the row of length 22 and two choices for the row of length 11. Hence, there are four decompositions. Two have weight 11, and the other two have weight −1-1. Their total contribution is therefore 00.

Thus, there are seven row decompositions, and Proposition 3.4 gives

Mμλ=−3+0=−3.M_{\mu}^{\lambda}=-3+0=-3.

Using Lemma 5.2. We now compute MμλM_{\mu}^{\lambda} using the coefficient formula. Since λ(1)\lambda^{(1)} has two rows and λ(2)\lambda^{(2)} has one row, we introduce the variables x1,1,x1,2,x2,1x_{1,1},x_{1,2},x_{2,1}. By the definition of xλx^{\lambda}, we have

xλ=x1,14x1,22x2,13.x^{\lambda}=x_{1,1}^{4}x_{1,2}^{2}x_{2,1}^{3}.

Using the character values of GG in the definition of Pm,j(x)P_{m,j}(x), we obtain

Pm,1(x)=x1,1m+x1,2m+x2,1m,P_{m,1}(x)=x_{1,1}^{m}+x_{1,2}^{m}+x_{2,1}^{m},
Pm,2(x)=x1,1m+x1,2m−x2,1m.P_{m,2}(x)=x_{1,1}^{m}+x_{1,2}^{m}-x_{2,1}^{m}.

Since μ(1)=(2,2,1)\mu^{(1)}=(2,2,1) and μ(2)=(3,1)\mu^{(2)}=(3,1), the product in Lemma 5.2 becomes

P2,1(x)2P1,1(x)P3,2(x)P1,2(x).P_{2,1}(x)^2P_{1,1}(x)P_{3,2}(x)P_{1,2}(x).

Therefore,

Mμλ=[x1,14x1,22x2,13]P2,1(x)2P1,1(x)P3,2(x)P1,2(x).M^\lambda_\mu=[x_{1,1}^{4}x_{1,2}^{2}x_{2,1}^{3}]P_{2,1}(x)^2P_{1,1}(x)P_{3,2}(x)P_{1,2}(x).

Thus, we need to find the coefficient of x1,14x1,22x2,13x_{1,1}^{4}x_{1,2}^{2}x_{2,1}^{3} in the polynomial P2,1(x)2P1,1(x)P3,2(x)P1,2(x)P_{2,1}(x)^2P_{1,1}(x)P_{3,2}(x)P_{1,2}(x). For simplicity, write x=x1,1x=x_{1,1}, y=x1,2y=x_{1,2} and z=x2,1z=x_{2,1}. Then

Mμλ=[x4y2z3](x2+y2+z2)2(x+y+z)(x3+y3−z3)(x+y−z)=−3,\begin{aligned} M^\lambda_\mu=[x^4y^2z^3](x^2+y^2+z^2)^2(x+y+z)(x^3+y^3-z^3)(x+y-z) \\ &=-3, \end{aligned}

which agrees with the result obtained from Proposition 3.4.

We are now ready to prove Proposition 4.1.

Proof of Proposition 4.1. It is enough to consider a single qq-reduction. Suppose that μ(j)=ξ∪(mpr)\mu^{(j)}=\xi\cup(m^{p^r}) and ν(j)=ξ∪((pm)pr−1)\nu^{(j)}=\xi\cup((pm)^{p^{r-1}}), for some partition ξ\xi, and ν(i)=μ(i)\nu^{(i)}=\mu^{(i)} for i≠ji\ne j. Fix λ∈Pk(N)\lambda\in\mathcal{P}_k(N), and define

A(x):=(∏1≤i≤ki≠j∏m′∈μ(i)Pm′,i(x))(∏m′∈ξPm′,j(x)).A(x):=\left(\prod_{\substack{1\le i\le k\\i\ne j}}\prod_{m'\in\mu^{(i)}}P_{m',i}(x)\right)\left(\prod_{m'\in\xi}P_{m',j}(x)\right).

Thus, A(x)∈Z[x]A(x)\in\mathbb{Z}[x] is the product corresponding to all parts of μ\mu except the prp^r selected parts of size mm in μ(j)\mu^{(j)}. It is also the product corresponding to all parts of ν\nu except the pr−1p^{r-1} selected parts of size pmpm in ν(j)\nu^{(j)}. Since

Ppm,j(x)=Pm,j(xp),P_{pm,j}(x)=P_{m,j}(x^p),

Lemma 5.2 gives

Mμλ−Mνλ=[xλ](A(x)(Pm,j(x)pr−Pm,j(xp)pr−1)).M^\lambda_\mu-M^\lambda_\nu=[x^\lambda]\left(A(x)\left(P_{m,j}(x)^{p^r}-P_{m,j}(x^p)^{p^{r-1}}\right)\right).

By Lemma 5.1, we have

Pm,j(x)pr−Pm,j(xp)pr−1∈prZ[x].P_{m,j}(x)^{p^r}-P_{m,j}(x^p)^{p^{r-1}}\in p^r\mathbb{Z}[x].

Since A(x)∈Z[x]A(x)\in\mathbb{Z}[x], it follows that

Mμλ≡Mνλ(modpr).M^\lambda_\mu\equiv M^\lambda_\nu\pmod{p^r}.

Finally, by Corollary 3.6, every irreducible character χλ\chi^\lambda can be written as an integral linear combination of permutation characters:

χλ=∑η∈Pk(N)dληMη,dλη∈Z.\chi^\lambda=\sum_{\eta\in\mathcal{P}_k(N)}d_{\lambda\eta}M^\eta,\qquad d_{\lambda\eta}\in\mathbb{Z}.

Therefore,

χμλ−χνλ=∑η∈Pk(N)dλη(Mμη−Mνη)≡0(modpr).\chi^\lambda_\mu-\chi^\lambda_\nu=\sum_{\eta\in\mathcal{P}_k(N)}d_{\lambda\eta}\left(M^\eta_\mu-M^\eta_\nu\right)\equiv0\pmod{p^r}.

This proves the proposition. □

Proof of Theorem 4.2

In this section, we extends [15 Proposition 5.2] to wreath products. We then use this proposition to prove Theorem 4.2.

Proposition 6.1. Let r,m≥1r,m\ge1, put R=pr−1R=p^{r-1}, and fix j∈{1,…,k}j\in\{1,\ldots,k\}. Suppose that μ∈Pk(N)\mu\in\mathcal{P}_k(N) and that μ(j)\mu^{(j)} contains at least RR parts equal to mm. Let ν∈Pk(N−Rm)\nu\in\mathcal{P}_k(N-Rm) be obtained from μ\mu by removing RR parts equal to mm from μ(j)\mu^{(j)}.

Suppose that λ∈Pk(N)\lambda\in\mathcal{P}_k(N) and each component λ(i)\lambda^{(i)} is an RmRm-core. Then

χμλ=p∑λ′eλ′χνλ′,\chi^\lambda_\mu=p\sum_{\lambda'}e_{\lambda'}\chi^{\lambda'}_\nu,

where eλ′∈Ze_{\lambda'}\in\mathbb{Z}. The sum runs over all kk-multipartitions λ′\lambda' of N−RmN-Rm such that each λ′(i)\lambda'^{(i)} is obtained from λ(i)\lambda^{(i)} by successively removing ni≥0n_i\ge0 border strips of length mm, where

n1+⋯+nk=R.n_1+\cdots+n_k=R.

Proof. Apply the Murnaghan–Nakayama rule (3) RR times, using the RR selected parts of size mm in μ(j)\mu^{(j)}. Since these parts belong to the same component, their cycle products lie in the same conjugacy class cjc_j. We obtain

χμλ=∑(i1,h1),…,(iR,hR)(∏s=1Rχis(cj)(−1)ht⁡(hs))χνλR.(8)\chi^\lambda_\mu=\sum_{(i_1,h_1),\ldots,(i_R,h_R)} \left(\prod_{s=1}^{R}\chi_{i_s}(c_j)(-1)^{\operatorname{ht}(h_s)}\right)\chi^{\lambda_R}_\nu. \tag*{(8)}

Here λ0=λ\lambda_0=\lambda and λs=λs−1∖ishs\lambda_s=\lambda_{s-1}\setminus i_sh_s. The sum runs over all sequences in which hsh_s is a hook of length mm in the component λs−1(is)\lambda_{s-1}^{(i_s)}.

Now group the terms in (8) according to the final multipartition λ′=λR\lambda'=\lambda_R. For each ii, let

ni=∣λ(i)∣−∣λ′(i)∣m.n_i=\frac{|\lambda^{(i)}|-|\lambda'^{(i)}|}{m}.

Thus, nin_i is the number of hooks removed from component ii, and n1+⋯+nk=Rn_1+\cdots+n_k=R.

Since removing a hook from one component does not affect the other components, a sequence leading to λ′\lambda' is determined by choosing an order for the components and a sequence of hook removals within each component. There are

(Rn1,…,nk)\binom{R}{n_1,\ldots,n_k}

ways to choose the order of the components. For each ii, let

W(i)=Wm(ni)(λ(i),λ′(i))W^{(i)}=W_m^{(n_i)}(\lambda^{(i)},\lambda'^{(i)})

denote the number of sequences of nin_i hooks of length mm leading from λ(i)\lambda^{(i)} to λ′(i)\lambda'^{(i)}, where W(i)=1W^{(i)}=1 if ni=0n_i=0. Hence, the total number of sequences leading to λ′\lambda' is

(Rn1,…,nk)∏i=1kW(i).\binom{R}{n_1,\ldots,n_k}\prod_{i=1}^{k}W^{(i)}.

Since all the selected parts belong to μ(j)\mu^{(j)}, the character factor is

∏s=1Rχis(cj)=∏i=1kχi(cj)ni.\prod_{s=1}^{R}\chi_{i_s}(c_j)=\prod_{i=1}^{k}\chi_i(c_j)^{n_i}.

which depends only on λ′\lambda'. Also, by [15 Lemma 5.4], the sign of each sequence is

∏s=1R(−1)ht⁡(hs)=∏i=1kϵm(λ(i),λ′(i)).\prod_{s=1}^{R}(-1)^{\operatorname{ht}(h_s)}=\prod_{i=1}^{k}\epsilon_m\left(\lambda^{(i)},\lambda'^{(i)}\right).

Therefore,

χμλ=∑λ′(∏i=1kχi(cj)niϵm(λ(i),λ′(i)))(Rn1,…,nk)(∏i=1kW(i))χλ′ν.\chi_{\mu}^{\lambda}=\sum_{\lambda'}\left(\prod_{i=1}^{k}\chi_i(c_j)^{n_i}\epsilon_m\left(\lambda^{(i)},\lambda'^{(i)}\right)\right)\binom{R}{n_1,\ldots,n_k}\left(\prod_{i=1}^{k}W^{(i)}\right)\chi_{\lambda'}^{\nu}.

Since χi(cj)∈Z\chi_i(c_j)\in\mathbb{Z}, it remains to prove that

p∣(Rn1,…,nk)∏i=1kW(i).p\mid\binom{R}{n_1,\ldots,n_k}\prod_{i=1}^{k}W^{(i)}.

First, suppose that at least two of the nin_i are positive. Then there is an index i0i_0 such that 0<ni0<R0<n_{i_0}<R. Since

(Rn1,…,nk)=(Rni0)(R−ni0n1,…,ni0−1,ni0^,…,nk),\binom{R}{n_1,\ldots,n_k}=\binom{R}{n_{i_0}}\binom{R-n_{i_0}}{n_1,\ldots,n_{i_0-1},\widehat{n_{i_0}},\ldots,n_k},

the multinomial coefficient is divisible by (Rni0)\binom{R}{n_{i_0}}. As R=pr−1R=p^{r-1}, we have

p∣(Rni0).p\mid\binom{R}{n_{i_0}}.

Thus, the required divisibility holds. Note that this case cannot occur when r=1r=1, since then R=1R=1.

Now suppose that only one of the nin_i is positive. Then ni0=Rn_{i_0}=R for some i0i_0, and all other nin_i are zero. In this case,

W(i0)=Wm(λ(i0),λ′(i0)).W^{(i_0)}=W_m\left(\lambda^{(i_0)},\lambda'^{(i_0)}\right).

Since λ(i0)\lambda^{(i_0)} is an RmRm-core, by [15 Lemma 5.5] gives

p∣W(i0).p\mid W^{(i_0)}.

Hence, the required divisibility holds in this case as well.

Therefore, every coefficient of χλ′ν\chi_{\lambda'}^{\nu} in the above sum is divisible by pp. This proves the proposition. □

Remark 2. The assumption that the RR cycles of length mm have their cycle products in the same conjugacy class cjc_j ensures that the factor ∏sχis(cj)\prod_s\chi_{i_s}(c_j) does not depend on the order of hook removals from different components. This condition holds in our application because qq-reductions do not move parts between components. Thus, the pr−1p^{r-1} equal parts in μ~(j)\widetilde{\mu}^{(j)} belong to the same component.

Proof of Theorem 4.2. Let R=pr−1R=p^{r-1}. For 0≤u≤r0\leq u\leq r, let μu\mu_u be obtained from μ\mu by removing RR parts equal to mim_i from μ(ji)\mu^{(j_i)} for each i>r−ui>r-u. Thus,

μu∈Pk(N−R∑i>r−umi),μ0=μ.\mu_u\in\mathcal{P}_k\left(N-R\sum_{i>r-u}m_i\right),\qquad\mu_0=\mu.

Since the mim_i are distinct, these selected parts are distinct. For 0≤u≤r0\leq u\leq r, define

Tu:={∑i=1r−ukimi:0≤ki≤R, ki=R for some i≤r−u}.\mathcal{T}_u:=\left\{\sum_{i=1}^{r-u}k_i m_i:0\leq k_i\leq R,\ k_i=R\text{ for some }i\leq r-u\right\}.

In particular, T0=T\mathcal{T}_0=\mathcal{T} and Tr=∅\mathcal{T}_r=\varnothing. We prove by induction on uu that

χμλ=pu∑λ′′eλ′′(u)χμuλ′′,eλ′′(u)∈Z,(9)\chi^\lambda_\mu=p^u\sum_{\lambda''}e^{(u)}_{\lambda''}\chi^{\lambda''}_{\mu_u},\qquad e^{(u)}_{\lambda''}\in\mathbb{Z}, \tag*{(9)}

where every component of each multipartition λ′′\lambda'' in the sum is a tt-core for all t∈Tut\in\mathcal{T}_u. For u=0u=0, this follows directly from the assumption on λ\lambda.

Now suppose that (9) holds for some u<ru<r. Let λ′′\lambda'' be a multipartition appearing in the sum, and put

m=mr−u,j=jr−u.m=m_{r-u},\qquad j=j_{r-u}.

Since Rm∈TuRm\in\mathcal{T}_u, every component of λ′′\lambda'' is an RmRm-core. Also, μu(j)\mu_u^{(j)} contains at least RR parts equal to mm. Removing RR of these parts gives μu+1\mu_{u+1}. Therefore, by Proposition 6.1,

χμuλ′′=p∑λ′′′eλ′′′χμu+1λ′′′.\chi^{\lambda''}_{\mu_u}=p\sum_{\lambda'''}e_{\lambda'''}\chi^{\lambda'''}_{\mu_{u+1}}.

where each component λ′′′(a)\lambda'''^{(a)} is obtained from λ′′(a)\lambda''^{(a)} by successively removing na≤Rn_a\leq R border strips of length mm. We now show that every component of λ′′′\lambda''' is a tt-core for all t∈Tu+1t\in\mathcal{T}_{u+1}. Fix t∈Tu+1t\in\mathcal{T}_{u+1}. For every 0≤k′≤R0\leq k'\leq R, we have

t+k′m∈Tu,t+k'm\in\mathcal{T}_u,

because tt already has a coefficient equal to RR at some index i≤r−u−1i\leq r-u-1. Hence, every component λ′′(a)\lambda''^{(a)} is a (t+k′m)(t+k'm)-core for 0≤k′≤R0\leq k'\leq R.

By [15 Lemma 5.3], removing one border strip of length mm preserves the (t+k′m)(t+k'm)-core property for 0≤k′≤R−10\leq k'\leq R-1. Repeating this argument nan_a times shows that λ′′′(a)\lambda'''^{(a)} is a (t+k′m)(t+k'm)-core for 0≤k′≤R−na0\leq k'\leq R-n_a. In particular, λ′′′(a)\lambda'''^{(a)} is a tt-core. Thus, substituting the above expression for χμuλ′′\chi^{\lambda''}_{\mu_u} in (9), we obtain (9) with u+1u+1 in place of uu.

Finally, taking u=ru=r in (9), we obtain

χμλ=pr∑λ′′eλ′′(r)χμrλ′′.\chi^\lambda_\mu=p^r\sum_{\lambda''}e^{(r)}_{\lambda''}\chi^{\lambda''}_{\mu_r}.

Since the coefficients and character values are integers, it follows that pr∣χμλp^r\mid\chi^\lambda_\mu. This completes the proof. □

Proof of Proposition 4.4

Before proving Proposition 4.4, we state and prove the following lemma, which will be useful in the proof.

Lemma 7.1. Let 1≤t≤N1 \le t \le N be integers. Then the number of λ∈Pk(N)\lambda\in\mathcal{P}_k(N) for which at least one component is not a tt-core is at most k(t+1)pk(N−t)k(t+1)p_k(N-t).

Proof. Fix a∈{1,…,k}a \in\{1,\ldots,k\}. For each n=∣λ(a)∣n=|\lambda^{(a)}|, we use Lemma 3.8 to count the multipartitions for which λ(a)\lambda^{(a)} is not a tt-core. Their number is at most

∑n=tN(t+1)p(n−t)pk−1(N−n)=(t+1)∑n′=0N−tp(n′)pk−1(N−t−n′)=(t+1)pk(N−t).\sum_{n=t}^{N}(t+1)p(n-t)p_{k-1}(N-n)=(t+1)\sum_{n'=0}^{N-t}p(n')p_{k-1}(N-t-n')=(t+1)p_k(N-t).

where the last equality follows from (1.1). Summing over all a∈{1,…,k}a \in\{1,\ldots,k\} gives the required bound. □

We now use this lemma to prove Proposition 4.4.

Proof of Proposition 4.4. Let t≥T0=T0(η)t \ge T_0=T_0(\eta). If t>Nt>N, then every component of λ\lambda is a tt-core, since its size is at most NN. Thus, we may assume that t≤Nt \le N. Let β=πk6N\beta=\frac{\pi\sqrt{k}}{\sqrt{6N}}.

Case 1. T0≤t≤N2T_0 \le t \le\frac{N}{2}.

By Lemma 7.1 and Lemma 3.7(1), the number of such multipartitions is at most

k(t+1)Ckpk(N)exp⁡(−βt).k(t+1)C_kp_k(N)\exp(-\beta t).

Let g(t)=(t+1)exp⁡(−βt)g(t)=(t+1)\exp(-\beta t). Since gg is decreasing for t≥1/βt \ge1/\beta, and

T0=(1+η)62πNklog⁡Nk≥1βT_0=(1+\eta)\frac{\sqrt{6}}{2\pi}\sqrt{\frac{N}{k}}\log\frac{N}{k}\ge\frac{1}{\beta}

for all sufficiently large NN, we have g(t)≤g(T0)g(t)\le g(T_0). Also, βT0=η+12log⁡Nk\beta T_0=\frac{\eta+1}{2}\log\frac{N}{k}. Therefore,

g(T0)=(T0+1)(Nk)−η+12≤26πNk(log⁡N)(Nk)−η+12≤2klog⁡NNη/2,g(T_0)=(T_0+1)\left(\frac{N}{k}\right)^{-\frac{\eta+1}{2}}\le2\frac{\sqrt{6}}{\pi}\sqrt{\frac{N}{k}}(\log N)\left(\frac{N}{k}\right)^{-\frac{\eta+1}{2}}\le2\sqrt{k}\frac{\log N}{N^{\eta/2}},

where we use 0<η≤10<\eta\le1 and assume that NN is sufficiently large. Hence, the number of such multipartitions is at most

2k32Ckpk(N)log⁡NNη/2.2k^{\frac{3}{2}}C_kp_k(N)\frac{\log N}{N^{\eta/2}}.

Case 2. N2<t≤N\frac{N}{2}<t\le N.

By Lemma 7.1 and Lemma 3.7(2), the number of such multipartitions is at most

k(N+1)Ckpk(N)exp⁡(−ckN)≤pk(N)log⁡NNη/2,k(N+1)C_kp_k(N)\exp(-c_k\sqrt{N})\le p_k(N)\frac{\log N}{N^{\eta/2}},

for all sufficiently large NN, since 0<η≤10<\eta\le1. This proves the proposition. □

Proof of Proposition 4.3

We first recall the following result of Peluse and Soundararajan for ordinary partitions. Its proof uses estimates for partitions into powers of pp and a saddle-point bound for the partition generating function (see [15 Sections 8–9]).

Proposition 8.1 ([15 Proposition 2.3]). There exist absolute constants N0N_0 and CC such that the following holds. Let N≥N0N \ge N_0, and suppose that

q=pr≤10−3log⁡N(log⁡log⁡N)2.q=p^r\le10^{-3}\frac{\log N}{(\log\log N)^2}.

For a partition α⊢N\alpha\vdash N, let α~\widetilde{\alpha} be the partition obtained by replacing prp^r equal parts mm with pr−1p^{r-1} equal parts pmpm. We repeat this process until no part occurs more than pr−1p^r-1 times. Then, for all but at most

Cp(N)exp⁡(−N120q)C p(N)\exp\left(-N^{\frac{1}{20q}}\right)

partitions α⊢N\alpha\vdash N, the partition α~\widetilde{\alpha} contains rr distinct part sizes m1,…,mrm_1,\ldots,m_r, each occurring at least pr−1p^{r-1} times, such that

pr−1mi>(1+16q)62πNlog⁡N,1≤i≤r.p^{r-1}m_i>\left(1+\frac{1}{6q}\right)\frac{\sqrt{6}}{2\pi}\sqrt{N}\log N,\qquad1\le i\le r.

We call a partition α⊢N\alpha\vdash N qq-exceptional if it belongs to the exceptional set in Proposition 8.1. Let bq(N)b_q(N) denote the number of qq-exceptional partitions of NN.

We now use Proposition 8.1 to prove Proposition 4.3.

Proof of Proposition 4.3. Let NN be sufficiently large so that Nk≥max⁡{N0,ee2}\frac{N}{k}\ge\max\{N_0,e^{e^2}\}, and suppose that qq satisfies (1.2). The function x↦log⁡x(log⁡log⁡x)2x\mapsto\frac{\log x}{(\log\log x)^2} is increasing for x≥ee2x\ge e^{e^2}. Hence, for every integer aa with Nk≤a≤N\frac{N}{k}\le a\le N, we have

q≤10−3log⁡Nk(log⁡log⁡N)2≤10−3log⁡Nk(log⁡log⁡Nk)2≤10−3log⁡a(log⁡log⁡a)2.q\le10^{-3}\frac{\log\frac{N}{k}}{(\log\log N)^2}\le10^{-3}\frac{\log\frac{N}{k}}{\left(\log\log\frac{N}{k}\right)^2}\le10^{-3}\frac{\log a}{(\log\log a)^2}.

Thus, Proposition 8.1 applies to partitions of aa, giving

bq(a)≤Cp(a)exp⁡(−a120q).b_q(a)\le Cp(a)\exp\left(-a^{\frac{1}{20q}}\right).

Now let μ∈Pk(N)\mu\in\mathcal{P}_k(N). Since ∑j=1k∣μ(j)∣=N\sum_{j=1}^k|\mu^{(j)}|=N, there is at least one index jj such that ∣μ(j)∣≥Nk|\mu^{(j)}|\ge\frac{N}{k}. Suppose that at least one such component μ(j)\mu^{(j)} is not exceptional in the sense of Proposition 8.1. Fix this jj, and put a=∣μ(j)∣a=|\mu^{(j)}|. By Proposition 8.1, the reduced partition μ~(j)\widetilde{\mu}^{(j)} contains rr distinct part sizes m1,…,mrm_1,\ldots,m_r, each occurring at least pr−1p^{r-1} times, such that

pr−1mi>(1+16q)62πalog⁡a≥(1+16q)62πNklog⁡Nk,p^{r-1}m_i>\left(1+\frac{1}{6q}\right)\frac{\sqrt{6}}{2\pi}\sqrt{a}\log a\ge\left(1+\frac{1}{6q}\right)\frac{\sqrt{6}}{2\pi}\sqrt{\frac{N}{k}}\log\frac{N}{k},

where the last inequality follows from a≥Nka \ge\frac{N}{k} and the fact that xlog⁡x\sqrt{x}\log x is increasing for x≥Nkx \ge\frac{N}{k}. Hence, (4.2) holds. Therefore, the exceptional multipartitions must belong to

⋃j=1kEj\bigcup_{j=1}^{k}\mathcal{E}_j

where Ej\mathcal{E}_j is the set of μ∈Pk(N)\mu\in\mathcal{P}_k(N) such that ∣μ(j)∣>Nk|\mu^{(j)}| > \frac{N}{k} and μ(j)\mu^{(j)} is exceptional in Proposition 8.1. For each fixed jj, counting according to a=∣μ(j)∣a=|\mu^{(j)}|, we obtain

∣Ej∣=∑Nk≤a≤Nbq(a)pk−1(N−a)≤Cexp⁡(−(Nk)120q)∑a=0Np(a)pk−1(N−a)=Cpk(N)exp⁡(−(Nk)120q),\begin{aligned} |\mathcal{E}_j|=\sum_{\frac{N}{k}\leq a\leq N}b_q(a)p_{k-1}(N-a)\leq C\exp\left(-\left(\frac{N}{k}\right)^{\frac{1}{20q}}\right)\sum_{a=0}^{N}p(a)p_{k-1}(N-a) \\ &=Cp_k(N)\exp\left(-\left(\frac{N}{k}\right)^{\frac{1}{20q}}\right), \end{aligned}

where the last equality follows from (1.1).

Summing over j=1,…,kj=1,\ldots,k, we conclude that the number of exceptional multipartitions is at most

Ckpk(N)exp⁡(−(Nk)120q).Ck p_k(N)\exp\left(-\left(\frac{N}{k}\right)^{\frac{1}{20q}}\right).

This proves the proposition.

Weyl groups of type DND_N

In this section, we prove Corollary 1.5. Recall that the Weyl group of type DND_N is an index-2 subgroup of the Weyl group of type BN=(Z/2Z)≀SNB_N=(\mathbb{Z}/2\mathbb{Z})\wr S_N. Here, DND_N denotes the Weyl group, not the dihedral group. The conjugacy classes of BNB_N are indexed by 2-multipartitions λ=(λ(1),λ(2))\lambda=(\lambda^{(1)},\lambda^{(2)}) of NN, where ∣λ(1)∣+∣λ(2)∣=N|\lambda^{(1)}|+|\lambda^{(2)}|=N.

Lemma 9.1 ([5 Proposition 3.4.12]). A conjugacy class of BNB_N is contained in DND_N if and only if the corresponding 2-multipartition (λ(1),λ(2))(\lambda^{(1)},\lambda^{(2)}) has an even number of parts in λ(2)\lambda^{(2)}.

Such a conjugacy class remains a single conjugacy class of DND_N, except when λ(2)=∅\lambda^{(2)}=\varnothing and all parts of λ(1)\lambda^{(1)} are even. In this case, it splits into two conjugacy classes of DND_N.

Since DND_N is a normal subgroup of BNB_N of index 2, Clifford theory gives the following description of its irreducible characters.

Lemma 9.2 ([5 §5.6.1]). Let λ=(λ(1),λ(2))\lambda=(\lambda^{(1)},\lambda^{(2)}) and λ′=(λ(2),λ(1))\lambda'=(\lambda^{(2)},\lambda^{(1)}) be 2-multipartitions of $N. Then the following hold.

(1) If λ(1)≠λ(2)\lambda^{(1)}\ne\lambda^{(2)}, then χλ↓DN\chi^\lambda\mathord{\downarrow}_{D_N} is irreducible, and

χλ↓DN=χλ′↓DN.\chi^\lambda\mathord{\downarrow}_{D_N}=\chi^{\lambda'}\mathord{\downarrow}_{D_N}.

Distinct unordered pairs {λ,λ′}\{\lambda,\lambda'\} give distinct irreducible characters of DND_N.

(2) If λ(1)=λ(2)\lambda^{(1)}=\lambda^{(2)}, then NN is even, and

χλ↓DN=χλ,++χλ,−,\chi^\lambda\mathord{\downarrow}_{D_N}=\chi^{\lambda,+}+\chi^{\lambda,-},

where χλ,+\chi^{\lambda,+} and χλ,−\chi^{\lambda,-} are distinct irreducible characters of DND_N.

Every irreducible character of DND_N arises from one of these two cases.

It follows from Lemma 9.2 that the number of irreducible characters of DND_N, which is also the number of its conjugacy classes, is

k(DN):=∣Irr⁡(DN)∣=p2(N)−d(N)2+2d(N)=p2(N)+3d(N)2,k(D_N):=|\operatorname{Irr}(D_N)|=\frac{p_2(N)-d(N)}{2}+2d(N)=\frac{p_2(N)+3d(N)}{2},

where

d(N)={p(N/2),if N is even,0,if N is odd.d(N)= \begin{cases} p(N/2), & \text{if }N\text{ is even},\\ 0, & \text{if }N\text{ is odd}. \end{cases}

We are now ready to prove Corollary 1.5.

Proof of Corollary 1.5. The number of irreducible characters of DND_N obtained by restricting χλ\chi^\lambda with λ(1)≠λ(2)\lambda^{(1)}\ne\lambda^{(2)} is p2(N)−d(N)2\frac{p_2(N)-d(N)}{2}. By Theorem 1.2, we have d(N)p2(N)⟶0\frac{d(N)}{p_2(N)}\longrightarrow0 as N→∞N\to\infty. Thus, the proportion of such irreducible characters of DND_N tends to 1.

Now, by Lemma 9.1, the conjugacy classes of BNB_N contained in DND_N remain single conjugacy classes, except for d(N)d(N) classes, each of which splits into two classes of DND_N. Since DND_N is a subgroup of BNB_N of index 2, at least half of the conjugacy classes of BNB_N are contained in DND_N. Hence, these classes form a positive proportion of the conjugacy classes of BNB_N.

By Corollary 1.4, applied with G=Z/2ZG=\mathbb{Z}/2\mathbb{Z}, the proportion of entries in the character table of BNB_N not divisible by any fixed positive integer KK tends to 0. Therefore, the same holds when we restrict to the columns corresponding to conjugacy classes contained in DND_N. Moreover, the number of split classes and the number of irreducible characters arising from split restrictions are both negligible compared with p2(N)p_2(N).

It follows that the proportion of entries in the character table of DND_N not divisible by KK tends to 0. Hence, the proportion of entries divisible by KK tends to 1 as N→∞N\to\infty. □\square

AI Declaration

The authors declare that no artificial intelligence tools or AI-assisted technologies were used in the preparation, writing, or mathematical analysis of this manuscript.

JAYANTA BARMAN, Indian Institute of Technology Kharagpur, Department of Mathematics, Kharagpur-721302, India.

Email address: [email protected], [email protected]

Kamalakshya Mahatab, Indian Institute of Technology Kharagpur, Department of Mathematics, Kharagpur-721302, India.

Email address: [email protected], [email protected]

References

  1. [1]G. E. Andrews, The theory of partitions, Encyclopedia of Mathematics and Its Applications, vol. 2, Addison-Wesley Publishing Co., Reading, MA, 1976.
  2. [2]J. Barman and K. Mahatab, Asymptotic formula for multipartitions, arXiv preprint arXiv:2607.09159 (2026).arxiv.org/abs/2607.09159
  3. [3]J. Barman and K. Mahatab, Zeros in the character table of the wreath product of the symmetric group, preprint (2026), arXiv:2609.13905.arxiv.org/abs/2609.13905
  4. [4]B. Dong, H. Graff, J. Mundinger, S. Rothstein and L. Vescovo, Almost all wreath product character values are divisible by given primes, Alebr. Comb., 6 (2023), no. 6, 1519–1531.DOI
  5. [5]M. Geck and G. Pfeiffer, Characters of finite Coxeter groups and Iwahori–Hecke algebras, London Mathematical Society Monographs, New Series, no. 21, Oxford University Press, New York, 2000.DOI
  6. [6]G. H. Hardy and S. Ramanujan, Asymptotic formulae in combinatory analysis, Proc. London Math. Soc. (2) 17 (1918), 75–115.
  7. [7]G. James and A. Kerber, The representation theory of the symmetric group, Encyclopedia of Mathematics and Its Applications, vol. 16, Addison-Wesley Publishing Co., Reading, MA, 1981.
  8. [8]A. Kerber and J. Tappe, On permutation characters of wreath products, Discrete Math., 15 (2) (1976), 151–161.DOI
  9. [9]I. G. Macdonald, Symmetric functions and Hall polynomials, 2nd ed., Oxford Mathematical Monographs, Oxford University Press, New York, 1995.DOI
  10. [10]G. Meinardus, Asymptotische Aussagen über Partitionen, Math. Z., 59 (1954), 388–398.DOI
  11. [11]A. R. Miller, On parity and characters of symmetric groups, J. Combin. Theory Ser. A, 162 (2019), 231–240.DOI
  12. [12]A. R. Miller, Congruences in character tables of symmetric groups, preprint (2019), arXiv:1908.03741.arxiv.org/abs/1908.03741
  13. [13]L. Morotti, On divisibility by primes in columns of character tables of symmetric groups, Arch. Math. (Basel), 114 (2020), no. 4, 361–365.arxiv.org/abs/1902.02625
  14. [14]M. R. Murty, The partition function revisited, in: The Legacy of Srinivasa Ramanujan, Ramanujan Math. Soc. Lect. Notes Ser., vol. 20, Ramanujan Math. Soc., Mysore, 2013, 261–279.
  15. [15]S. Peluse and K. Soundararajan, Divisibility of character values of the symmetric group by prime powers, Algebra Number Theory, 19 (2025), no. 2, 365–382.DOI
  16. [16]J.-P. Serre, Linear representations of finite groups, Graduate Texts in Mathematics, vol. 42, Springer-Verlag, New York, 1977.DOI

Paper details

vlnc:W_rxh2w7k3
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents