We give an explicit formula for the conductors of the primitive idempotents of fibered Burnside algebras of finite rank. This formula generalizes previous results for Burnside rings, monomial representation rings, and cyclic fibers, removing the coprimality condition required in the latter case. We also derive some consequences for cyclic fiber groups and p-groups.
Introduction
The conductors of the primitive idempotents of the rational Burnside algebra were first determined by Nicolson in 1978 [8]. In his article, Nicolson showed that, for a primitive idempotent eHB(G) of BQ(G) (indexed by the conjugacy class of H≤G), the least positive integer t such that teHB(G) belongs to B(G) is precisely [NG(H):H][H:H′]0, where [H:H′]0 denotes the product of the prime divisors of [H:H′]. Later, Kratzer & Thévenaz [6] and Dress & Vallejo [3] provided more direct proofs of this result.
In the case of the monomial representation algebra DK(G), the primitive idempotents eH,hD(G) are indexed by the conjugacy classes of pairs (H,hH′) with H≤G and h∈H, where K is a field of characteristic zero containing a primitive exp(G)-th root of unity. Müller showed in [7 §3] that the conductor of eH,hD(G) is equal to the index [NG(H,hH′):H′]. In general, [NG(H,hH′):H′]eH,hD(G) does not lie in D(G) but in DO(G), where O is the ring of integers of K.
A first generalization of the previous results in the context of fibered Burnside rings was given by Raggi-Cárdenas and the author in [4 §5]. For a positive integer n, the primitive idempotents eH,hG,n of BKn(G), where K is a field of characteristic zero containing a primitive n-th root of 1, are indexed by the conjugacy classes of pairs (H,hH[n]), with H≤G and h∈H. Here, H[n] is the minimal normal subgroup of H such that H/H[n] is an abelian group of exponent dividing n[4 §1]. When n is a divisor of exp(G) such that gcd(n,exp(G)/n)=1, the conductor of eH,hG,n is precisely [NG(H,hH[n]):H[n]][H[n]:H′]0. Although this result has restrictive hypotheses, it covers the cases of the Burnside ring (n=1) and the monomial representation ring (n=exp(G)).
In this article, we focus on fibered Burnside rings BA(G) of finite rank. As shown by Boltje and Yılmaz in [2 §3], the primitive idempotents eH,ΦG,A of the fibered Burnside algebra BKA(G) over a sufficiently large field of characteristic zero are indexed by orbits of pairs (H,Φ) under a natural action of G on the set DGA. The main result provides an explicit formula for the conductor of eH,ΦG,A, namely,
cH,ΦG,A=[NG(H,Φ):H]∣Hom(H,A)∣rA(H),
where rA(H) is an additional factor depending only on H and A. As in the previous cases, Boltje’s integrality criterion and the Möbius function play an important role in the proof. A closer examination of the values of the Möbius function allows us to remove the coprimality condition imposed in [4].
The article is organized into two sections. The first collects the necessary preliminaries on abelian groups, the Möbius function, and fibered Burnside rings. The second is devoted to the statement and proof of the main result, together with some consequences concerning cyclic fibers and p-groups.
Notation
For a positive integer n, we write n0 for its largest square-free divisor, that is, the product of its distinct prime divisors. We write SG for the poset of subgroups of a group G, and [SG] for a subset of representatives of the conjugacy classes of SG. If X is a G-set, then X/G denotes its set of orbits, and we write x=Gy for elements x,y∈X lying in the same orbit. We regard G and SG as G-sets under conjugation, writing gx instead of gxg−1. For H≤G and an abelian group A, ϕ∈Hom(H,A), and g∈G, we write gϕ for the morphism defined by (gϕ)(x)=ϕ(g−1xg), for x∈gH. For K≤H, we denote by ϕ∣K the restriction of ϕ to K. All rings R are assumed to be associative, unital, and commutative, and we use the standard notation Z, Q, R, and C for the usual number systems.
Preliminaries
We recall some facts about morphisms to abelian groups and the Möbius function of a poset of subgroups, and then review the basic properties of fibered Burnside rings that will be used throughout the paper.
Morphisms to abelian groups
For an abelian group A and a positive integer n, we denote by torn(A) the subgroup of elements of A whose orders divide n; we say that A is an n-torsion abelian group if torn(A)=A. For a finite group G, the subgroup G[n] is defined as the smallest normal subgroup T⊴G such that G/T is an n-torsion abelian group; it can be constructed as the subgroup generated by G′∪{gn∣g∈G}. By [4 Lemma 1.1], G[exp(G)]=G′, G[1]=G, H[n]≤G[n] for H≤G, and G[n]≤G[m] if m∣n. Also, there is a natural isomorphism
where n(A,G):=exp(torexp(G)(A)). By [4 Lemma 1.3], Hom(G,Cn) and G/G[n] are isomorphic, and so ∣Hom(G,Cn)∣=[G:G[n]].
Lemma 1.1. Let n be a positive integer and G be a finite abelian group. Let p be a prime divisor of ∣G[n]∣.
If p divides [G:G[n]], then G has a p-element of order greater than p;
if p does not divide [G:G[n]], then p does not divide n.
Proof. Since G is finite and abelian, the map fn:G→G given by g↦gn is an endomorphism with Im(fn)=G[n], and ∣ker(fn)∣=[G:G[n]]. Because p∣∣G[n]∣, the p-primary component Gp is non-trivial, and fn restricts to an endomorphism on Gp with image Gp[n]=G[n]∩Gp=1. If p∣[G:G[n]], then p divides exp(G/G[n]), which divides n. Then there is some g∈Gp such that 1=gn∈Gp[n]; since p∣n, it follows that g has order greater than p. On the other hand, if p∤[G:G[n]], then Gp∩ker(fn)=1, hence fn is an automorphism on Gp and so n is coprime to p.
If A is a finite abelian group, we consider a decomposition A=Cd1×Cd2×⋯×Cdr into cyclic factors. For any finite group G, there is a natural isomorphism Hom(G,A)≅∏i=1rHom(G,Cdi), and for each H≤G, the following diagram involving the restriction maps commutes:
by orthogonality relations for ker(resHG). Together with assertion 1, this proves 2. □
The Möbius function of the poset of subgroups
We consider SG as a poset with respect to inclusion. The Möbius functionμ:SG×SG⟶Z is defined recursively as follows: for H≤G, μ(H,H)=1, and for K=H,
μ(K,H)=−K<T≤H∑μ(K,T).
Equivalently, μ is defined as the inverse of the zeta function in the incidence algebra of SG. It is easy to see that μ(K,H)=−1 if K is a maximal subgroup of H. We write μ(G) instead of μ(1,G). Moreover, if N⊴G, then μ(N,G)=μ(G/N) since SG/N and {T∈SG∣N≤T} are isomorphic as posets. The following proposition collects some properties of μ proved by Hawkes, Isaacs and Özaydin in [5].
If N⊴G, let K(N,G) denote the set of complements of N in G. Then
μ(G)=μ(G/N)X∈K(N,G)∑μ(X,G).
If G is a p-group of order pr, then μ(G)=0 if and only if G is an elementary abelian p-group. In this case, μ(G)=(−1)rp(2r).
If H≤G, then [NG(H):H] divides μ(H,G)[G:HG′]0.
Next, we establish a vanishing criterion for the Möbius function in terms of the indices determined by the subgroups HG[n]. This will be useful in the proof of our main result.
Lemma 1.4. Let H⊴G and n be a positive integer. If [G:HG[n]] and [HG[n]:HG′] are not coprime, then μ(H,G)=0.
Proof. The quotient G/HG′ is abelian and (G/HG′)[n]=HG[n]/HG′. By part 1 of Proposition 1.3, we have μ(H,G)=μ(HG′,G)⋅T for some integer T. Since [G:HG[n]] and [HG[n]:HG′] have a common prime divisor p, it follows from Lemma 1.1 that G/HG′ has a p-element of order greater than p. Consequently, there is a subgroup HG′≤M≤G such that G/M is cyclic of order p2. Applying part 1 of Proposition 1.3 again, μ(HG′,G)=μ(M,G)⋅U for some integer U. Finally, μ(M,G)=μ(G/M)=0 by part 2 of Proposition 1.3, and the result follows. □
Fibered Burnside rings
Given a finite group G and an abelian group A, an A-fibered G-set is a G×A-set on which A acts freely and with finitely many orbits. The category of A-fibered G-sets and G×A-equivariant functions, denoted by gsetA, is a symmetric semiring category for disjoint unions and a tensor product ⊗A. The A-fibered Burnside ring of G is defined as the Grothendieck ring of the category gsetA, denoted by BA(G).
The following presentation of BA(G) is often more convenient for computations: let MGA be the set of A-monomial pairs or A-subcharacters of G, that is, pairs of the form (H,ϕ) where H≤G and ϕ∈Hom(H,A); the group G acts on the set MGA by conjugation, and the stabilizer and orbit of a pair (H,ϕ) are denoted by NG(H,ϕ) and [H,ϕ]G, respectively. As the transitive A-fibered G-sets are in one-to-one correspondence with the orbits of MGA, the ring BA(G) is the free abelian group on MGA/G, with multiplication determined on the basis by
and with [G,1]G acting as multiplicative identity; MGA/G is called the monomial basis of BA(G). It is known that BA(G)≅Btorexp(G)(A)(G), and moreover, BA(G) has finite rank if and only if torexp(G)(A) is finite. Throughout this paper, we assume that torexp(G)(A) is finite.
For each H≤G, there is a restriction map resHG:BA(G)⟶BA(H) defined on the basis by
resHG([K,ψ]G)=HgK∈H\G/K∑[H∩gK,gψ∣H∩gK]H.
as well as an induction map indHG:BA(H)⟶BA(G), [K,ψ]H↦[K,ψ]G. For H≤G and g∈G, there is also a g-conjugation map cg,H:BA(H)⟶BA(gH), given by [K,ψ]H↦g[K,ψ]H=[gK,gψ]gH. The collection of the rings BA(H) for H≤G together with induction, restriction and conjugation maps is a Green functor over G in the usual sense.
The ordinary Burnside ring B(G) embeds naturally into BA(G) via the ring homomorphism [G/H]↦[H,1]G, admitting a retraction [H,ϕ]G↦[G/H]. There is also a surjective ring homomorphism πG:BA(G)⟶ZHom(G,A) mapping [K,ϕ]G to ϕ if K=G, and to 0 otherwise; this map admits a section ϕ↦[G,ϕ]G.
The group G acts on the product of group rings ∏H≤GZHom(H,A) through the conjugation maps, and the ghost ring of BA(G) is defined as the ring of fixed points
BA(G):=(H≤G∏ZHom(H,A))G.
The product map ρ=∏H≤GπH∘resHG is known as the mark morphism and embeds BA(G) into its ghost ring. It is explicitly given on the basis by
ρ([K,ψ]G)=gK∈G/K,H≤gK∑gψ∣HH≤G.
We write BRA(G) for the scalar extension of BA(G) to a commutative ring R, regarded as an R-algebra, and BRA(G) for the corresponding extension of its ghost ring. The extensions of the maps defined in this section are denoted the same. Boltje’s Integrality Criterion[1 Theorem 2.2] characterizes the image of the extended mark morphism in BRA(G) as follows. Let
y=ϕ∈Hom(H,A)∑aH,ϕϕH≤G
be an element of BRA(G). Then y belongs to ρ(BRA(G)) if and only if, for every (K,ψ)∈MGA, the congruence
holds. Here, (K,ψ)≤(H,ϕ) means that K≤H and ψ=ϕ∣K. This defines a partial order on MGA respecting the action of G.
Let K be a field of characteristic zero that is a splitting field for every Hom(H,A), with H≤G. As noted in [2 §3], this is equivalent to requiring that K contains a root of unity of order exp(torexp(G)(A)). We consider the set DGA consisting of pairs (H,Φ) where H≤G and Φ∈Hom(Hom(H,A),K×); again, G acts on this set by conjugation and the stabilizer and orbit of a pair (H,Φ) are denoted by NG(H,Φ) and [H,Φ]G, respectively. If A=Cn, then DGCn can be identified with the set of pairs (H,hH[n]), H≤G and h∈H. For each (H,Φ), the composite sH,ΦG,A:=Φ∘πH∘resHG is a ring homomorphism from BA(G) to K. By [2], two of such maps sH,ΦG,A and sK,ΨG,A are equal if and only if (H,Φ)=G(K,Ψ). Moreover, these maps account for all elements of Homring(BA(G),K), and BKA(G) and ∏[H,Φ]G∈DGA/GK are isomorphic as K-algebras.
The primitive idempotents of BKA(G) are in bijection with the orbits of DGA. More precisely, for each (H,Φ)∈DGA there is a primitive idempotent eH,ΦG,A determined by
as showed in [2]. If O is the ring of algebraic integers of K, the conductor of a primitive idempotent eH,ΦG,A is the least positive integer t such that teH,ΦG,A belongs to BOA(G), denoted by cH,ΦG,A. In the case of cyclic fiber, we write eH,hG,Cn instead of eH,eνhG,Cn, and CH,hG,n for its conductor. The conductors do not depend on the choice of K, since the coefficients of the primitive idempotents with respect to the monomial basis are algebraic numbers.
As in [7], the behavior of certain idempotents with respect to induction and restriction will play a role in the proof of the main result. The following lemma collects the required properties. The proof is straightforward and follows the same ideas as the corresponding results in [4] and [7], so we omit it.
In the following, A is an abelian group such that torexp(G)(A) is finite. Let K be a field of characteristic zero that is a splitting field for every Hom(H,A), with H≤G, and let O be its ring of algebraic integers. Consider the primitive idempotent of BQ(G) corresponding to the G-th mark,
eGB(G)=H∈[SG]∑aH[G/H],
with aH∈Q. The following characterization of a primitive idempotent eG,ΦG,A is a generalization of that presented by Müller in [7 §3].
Lemma 2.1. Let Φ∈Hom(Hom(G,A),K×). Then
eG,ΦG,A=∣Hom(G,A)∣1λ∈Hom(G,A)∑Φ(λ−1)xλ.(3)
where
xλ:=H∈[SG]∑aH[H,λ∣H]G.
Proof. It is straightforward that πK([H,λ∣H]G)=ϕK(G/H)λ∣K for K≤G, hence
by row orthogonality relations for the group Hom(G,A), proving that
∣Hom(G,A)∣eG,ΦG,A=λ∈Hom(G,A)∑Φ(λ−1)xλ.
For a positive integer n, we set
rn(G):=gcd(n,[G[n]:G′]0)[G[n]:G′]0,
that is, rn(G) is the product of the prime divisors of [G[n]:G′] which are coprime to n. In general, we set
rA(G):=rexp(torexp(G)(A))(G).
This notation is useful, since the exponent of torexp(G)(A) may change as G varies.
Theorem 2.2. Let G be a finite group, and let A be an abelian group such that torexp(G)(A) is finite. Let K be a field of characteristic zero that is a splitting field for every Hom(H,A), with H≤G. For (H,Φ)∈DGA, the conductor of the primitive idempotent eH,ΦG,A of BKA(G) is
cH,ΦG,A=[NG(H,Φ):H]∣Hom(H,A)∣rA(H).(4)
Proof. Let
torexp(G)(A)=Cd1×Cd2×⋯×Cdt
be a decomposition of torexp(G)(A) into invariant factors, where di∣di+1 for i=1,…,t−1, and dt=exp(torexp(G)(A)). We first consider the case H=G, and set
which lies in BOA(G) by Lemma 2.1. We test Boltje’s integrality criterion on y. Because the K-th entry of y is zero for any proper subgroup K, we only need to check the congruences for (K,ϕ) such that NG(K,ϕ)=G and ϕ∈Im(resKG), that is, K⊴G and ϕ=λ∣K for some λ∈Hom(G,A). For such a pair, we have
for some λ0∈(resKG)−1(ϕ). Since there is a natural identification Hom(L,A)≅Hom(L,torexp(G)(A)) for L≤G, we have
∣ker(resKG)∣=i=1∏t[G:KG[di]]
by Lemma 1.2, while Φ(λ0−1) is a root of unity. Moreover, every [G:KG[di]] is a divisor of [G:KG[dt]]. If [G:KG[dt]] and [KG[dt]:KG′] are not coprime, then μ(K,G)=0 by Lemma 1.4. On the other hand, if gcd([G:KG[dt]],[KG[dt]:KG′])=1, every prime divisor of [KG[dt]:KG′] is a divisor of [G[dt]:G′] that does not divide dt by Lemma 1.1, so it is a divisor of rdt(G)=rA(G). Hence, μ(K,G)∣Hom(G,A)∣rA(G) is multiple of μ(K,G)[G:KG′], and therefore a multiple of [NG(K):K]=[G:K], by part 3 of Proposition 1.3. Since NG(K,ϕ)=G, Boltje’s integrality criterion implies that
rA(G)y=ρ(∣Hom(G,A)∣rA(G)eG,ΦG,A)∈ρ(BOA(G)),
hence ∣Hom(G,A)∣rA(G)eG,ΦG,A∈BOA(G). For general (H,Φ), we have that
belongs to BOA(G) since ∣Hom(H,A)∣rA(H)eH,ΦH,A belongs to BOA(H). This shows that [NG(H,Φ):H]∣Hom(H,A)∣rA(H) divides cH,ΦG,A. Conversely, note that torexp(H)(A) is also finite, and consider a decomposition
torexp(H)(A)=Cf1×Cf2×⋯×Cfs
into invariant factors, where fi∣fi+1 for i=1,…,s−1 and fs=exp(torexp(H)(A)). Note that the coefficient of [H,1]G in eH,ΦG,A with respect to the monomial basis is ([NG(H,Φ):H]∣Hom(H,A)∣)−1, therefore [NG(H,Φ):H]∣Hom(H,A)∣ divides cH,ΦG,A. We now prove that rA(H)=rfs(H) divides
u(H,Φ):=[NG(H,Φ):H]∣Hom(H,A)∣cH,ΦG,A.
Let p be a prime divisor of [H[fs]:H′] that does not divide fs. For any K≤H such that [H:K]=p, we have [H:KH[fs]]=1, hence [H:KH[fi]]=1 for each i, and so
ker(resKH)=i=1∏s[H:KH[fi]]=1,
proving that resKH:Hom(H,A)→Hom(K,A) is injective. Let
Ip={K≤H∣[H:K]=p}.
We have ∣Ip∣≡1(modp), and Ip splits into equivalence classes under the relation =G. Therefore, there is some K∈Ip such that p does not divide
m=#{L∈Ip∣L=GK}.
For such K, the coefficient of [K,1]G in [NG(H,Φ):H]∣Hom(H,A)∣eH,ΦG,A with respect to the monomial basis is
Since u(H,Φ)(−pm) must be an integer, we conclude that p must divide u(H,Φ), completing the proof. □
Corollary 2.3. Let G be a finite group, and let n be a positive integer. For (H,hH[n])∈DGCn, the conductor of eH,hG,n is
cH,hG,n=[NG(H,hH[n]):H[n]]rn(H).(5)
For both the Burnside ring and the monomial representation ring, the order of a group can be easily recovered from their conductors [7 Proposition 3.7]. This also holds for all fibered Burnside rings with cyclic fiber group.
Proposition 2.4. Let n be a positive integer. For a finite group G, we have
Proof. Take a characteristic zero field K containing a primitive n-th root of unity. The conductors of the primitive idempotents of BK(G) do not depend on the field K, and [NG(H,hH[n]):H[n]]rn(H) divides ∣G∣ for every pair (H,hH[n]). As cH,hG,n=∣G∣ is reached at least once, this proves both equalities. □
The previous argument fails for more general fibers, as shown in the following example.
Example 2.5. Let G=C4, H≤C4 be the subgroup of order 2 and A=C2×C2. Then NG(H,1)=NG(H)=C4, ∣Hom(H,C2×C2)∣=4, and rC2×C2(H)=1. Thus, we have
The idempotents of B(G) whose conductors are exactly ∣G∣ play a central role in the proof of the existence of normalizing isomorphisms, with important consequences on the isomorphism problem of the Burnside ring [9]. In the case of D(G), we see that G is abelian if and only if all conductors are equal to ∣G∣[7]. More generally, we can characterize pairs (H,hH[n]) for which cH,hG,n=∣G∣.
Proposition 2.6. Let n be a positive integer. For (H,hH[n])∈DGCn, cH,hG,n=∣G∣ if and only if H is a normal abelian subgroup such that H[n] has square-free order coprime to n and hH[n]∈Z(G/H[n]).
Proof. Note that [NG(H,hH[n]):H[n]]rn(H)=[G:1] implies NG(H,hH[n])=G, and so NG(H)=G and hH[n]∈Z(G/H[n]). Then [G:H[n]]rn(H)=[G:1] implies [H[n]:1]=rn(H), the latter is a divisor of [H[n]:H′], thus implying H′=1 and proving the assertion about the order of H[n]. Conversely, suppose that H is normal and abelian, and that H[n] has square-free order coprime to n. For hH[n]∈Z(G/H[n]), we have NG(H,hH[n])=G, and, since H′=1, we have rn(H)=∣H[n]∣. Consequently, cH,hG,n=[G:H[n]]∣H[n]∣=∣G∣. □
Specializing to p-groups, we get the following.
Corollary 2.7. For a prime number p, let P be a p-group and let q be a power of p. For (H,hH[q])∈DPCq, cH,hP,q=∣P∣ if and only if H is a normal abelian subgroup such that H[q]=1 and h∈Z(P).
Proof. This follows immediately from the previous result, since in this case $r^q(H)=1. □
References
[1]R. Boltje, Integrality conditions for elements in ghost rings of generalized Burnside rings. Preprint, 2004. https://boltje.math.ucsc.edu/publications/p04z.pdf
[2]R. Boltje, D. Yılmaz, The A-fibered Burnside ring as A-fibered biset functor in characteristic zero. Algebras and Representation Theory, 24(5):1359–1385, 2021.arxiv.org/abs/1909.06668
[3]A. W. M. Dress, E. Vallejo, A simple proof for a result by Kratzer and Thévenaz concerning the embedding of the Burnside ring into its ghost ring. Journal of Algebra, 154(2):356–359, 1993.DOI
[4]B. García, A. G. Raggi-Cárdenas, On fibered Burnside rings, fiber change maps and cyclic fiber groups. Journal of Pure and Applied Algebra, 229(6), article 107961, 2025. https://doi.org/10.1016/j.jpaa.2025.107961
[5]T. Hawkes, I. M. Isaacs, M. Özaydin, On the Möbius function of a finite group. Rocky Mountain Journal of Mathematics, 19(4):1003–1034, 1989. https://doi.org/10.1216/RMJ-1989-19-4-1003
[6]C. Kratzer, J. Thévenaz, Fonction de Möbius d’un groupe fini et anneau de Burnside. Commentarii Mathematici Helvetici, 59:425–438, 1984. https://doi.org/10.5169/seals-45403
[7]M. Müller, On the isomorphism problem for the ring of monomial representations of a finite group. Journal of Algebra, 333(1):427–457, 2011. https://doi.org/10.1016/j.jalgebra.2011.02.013
[8]D. M. Nicolson, The orbit of the regular G-set under the full automorphism group of the Burnside ring of a finite group G. Journal of Algebra, 51:288–299, 1978.DOI