Coset posets of infinite groups

Kai-Uwe Bux; Cora Welsch

doi:10.1515/jgth-2019-0162

Article Publicly Available

Coset posets of infinite groups

Kai-Uwe Bux and Cora Welsch

Published/Copyright: March 10, 2020

Published by

Become an author with De Gruyter Brill

Submit Manuscript Author Information Explore this Subject

From the journal Journal of Group Theory Volume 23 Issue 4

Abstract

We consider the coset poset associated with the families of proper subgroups, proper subgroups of finite index and proper normal subgroups of finite index. We investigate under which conditions those coset posets have contractible geometric realizations.

Let X be a set, and let 𝒰 = ( U α ) α ∈ A be a family of subsets. H. Abels and S. Holz associate three simplicial complexes to 𝒰 , namely:

𝒳 ⁢ ( 𝒰 ) denotes the complex of finite subsets of X contained within some U α .
𝒩 ⁢ ( 𝒰 ) is the nerve of the family 𝒰 , i.e., a subset σ ⊆ A is a simplex of 𝒩 ⁢ ( 𝒰 ) provided that the intersection U σ : = ⋂ α ∈ σ U α is non-empty.
𝒫 ⁢ ( 𝒰 ) is the geometric realization of the poset { U α ∣ α ∈ A } , where the order is given by inclusion. Thus, simplices of 𝒫 ⁢ ( 𝒰 ) are finite ⊂ -chains in { U α ∣ α ∈ A } .

Abels–Holz show [1, Theorem 1.4]

𝒳 ⁢ ( 𝒰 ) and 𝒩 ⁢ ( 𝒰 ) are homotopy equivalent.
𝒳 ⁢ ( 𝒰 ) , 𝒩 ⁢ ( 𝒰 ) and 𝒫 ⁢ ( 𝒰 ) are all homotopy equivalent provided that the family 𝒰 is closed with respect to non-empty intersection, i.e., the intersection U ∩ V also belongs to { U α ∣ α ∈ A } if U , V ∈ { U α ∣ α ∈ A } have non-empty intersection U ∩ V ≠ ∅ .

Note that 𝒳 ⁢ ( 𝒰 ) depends only on the set { U α ∣ α ∈ A } . Hence, the homotopy type of 𝒩 ⁢ ( 𝒰 ) does not depend on the particular indexing of this collection of subsets. We shall therefore restrict ourselves to the case of self-indexing families, where α = U α , i.e., A is a set of subsets of X.

Let 𝒦 be a set of subgroups of a fixed group G, and let Cosets 𝒦 be the corresponding collection of cosets. We put

𝒳 𝒞 ( 𝒦 ) : = 𝒳 ( Cosets 𝒦 ) , 𝒩 𝒞 ( 𝒦 ) : = 𝒩 ( Cosets 𝒦 ) , 𝒫 𝒞 ( 𝒦 ) : = 𝒫 ( Cosets 𝒦 ) .

Specializing the above result of Abels–Holz to these families, one obtains the following consequence.

Fact 1.

Let G be a group, and fix a set 𝒦 of subgroups of G.

𝒳 ⁢ 𝒞 ⁢ ( 𝒦 ) and 𝒩 ⁢ 𝒞 ⁢ ( 𝒦 ) are homotopy equivalent.
If 𝒦 is closed with respect to finite intersections, then 𝒳 ⁢ 𝒞 ⁢ ( 𝒦 ) , 𝒩 ⁢ 𝒞 ⁢ ( 𝒦 ) and 𝒫 ⁢ 𝒞 ⁢ ( 𝒦 ) are homotopy equivalent.

We shall be interested in the family ℋ ⁢ ( G ) of all proper subgroups in G, the family ℋ fi ⁢ ( G ) of all its proper subgroups of finite index and the family ℋ nor , fi ⁢ ( G ) of normal subgroups from ℋ fi ⁢ ( G ) . All three families are closed with respect to finite intersections. We study the topology of the coset nerves 𝒩 𝒞 ( G ) : = 𝒩 𝒞 ( ℋ ( G ) ) , 𝒩 𝒞 fi ( G ) : = 𝒩 𝒞 ( ℋ fi ( G ) ) and 𝒩 𝒞 nor , fi ( G ) : = 𝒩 𝒞 ( ℋ nor , fi ( G ) ) .

In particular, we shall investigate for which groups these nerves are contractible. For the families ℋ fi ⁢ ( G ) and ℋ nor , fi ⁢ ( G ) , we have a satisfying answer to this problem.

Theorem \ref{equiv}.

The coset nerve N ⁢ C fi ⁢ ( G ) is contractible if and only if the collection H fi ⁢ ( G ) has infinitely many maximal elements.

Theorem \ref{normal-equiv}.

The coset nerve N ⁢ C nor , fi ⁢ ( G ) is contractible if and only if the collection H nor , fi ⁢ ( G ) has infinitely many maximal elements.

For the coset nerve 𝒩 ⁢ 𝒞 ⁢ ( G ) associated to the family of all proper subgroups, we have only partial results, which we present in Section 3.

It follows from Fact 1 that the coset nerve 𝒩 ⁢ 𝒞 ⁢ ( G ) is homotopy equivalent to the coset poset 𝒫 𝒞 ( G ) : = 𝒫 𝒞 ( ℋ ( G ) ) which is the set of all proper subgroups of G and their cosets, ordered by inclusion.

The coset poset was introduced for finite groups by K. S. Brown in [2]. He considered the Euler characteristic of the coset poset since it is connected to the probabilistic zeta function. The probabilistic zeta function is the reciprocal of the probability P ⁢ ( G , s ) that a randomly chosen ordered s-tuple from a finite group G generates G. Hall gave a formula for P ⁢ ( G , s ) as a finite Dirichlet series. In view of that formula, one can evaluate P ⁢ ( G , s ) at an arbitrary complex number s. Brown proved S. Bouc’s observation that P ⁢ ( G , - 1 ) can be interpreted as the negative reduced Euler characteristic of the coset poset 𝒫 ⁢ 𝒞 ⁢ ( G ) . This result motivated Brown to study the homotopy type of the coset poset, which (as Brown said) raised more questions than it answered.

The question of simple connectivity was studied by D. A. Ramras in [3], and among other results, he proved that the coset poset 𝒫 ⁢ 𝒞 ⁢ ( G ) is contractible if G is an infinitely generated group. The question of contractibility was answered by J. Shareshian and R. Woodroofe in [4]. They proved that if G is a finite group, the coset poset 𝒫 ⁢ 𝒞 ⁢ ( G ) is not contractible.

This motivated the second author of this paper to study the homotopy type of the coset poset of a finitely generated infinite group, focusing on the contractibility in her PhD thesis [5]. In her thesis, she was more interested in some special subsets of the coset posets than the coset poset itself, namely the finite index coset poset 𝒫 𝒞 fi ( G ) : = 𝒫 𝒞 ( ℋ fi ( G ) ) . She proved contractibility for many groups and provided nearly all the examples of Section 1.2. Moreover, the thesis showed that there are some groups with non-contractible finite index coset poset. By her examples, she was led to conjecture Theorem 5, which we are now able to prove using an extended version of her cone argument.

Although Theorem 5 formally implies the result of Shareshian–Woodroofe that, for finite G, the coset nerve 𝒩 ⁢ 𝒞 ⁢ ( G ) is not contractible, our proof actually makes use of their result. In order to prove Theorem 12, we show the corresponding result for the family of proper normal subgroups, which might be of independent interest.

Theorem \ref{SW-analogon}.

If G is finite, N ⁢ C nor ⁢ ( G ) is not contractible.

1 The family of finite index subgroups

In this section, we characterize those groups G for which 𝒩 ⁢ 𝒞 fi ⁢ ( G ) is contractible. It will turn out that 𝒩 ⁢ 𝒞 fi ⁢ ( G ) is contractible if and only if the poset ℋ fi ⁢ ( G ) of proper finite index subgroups has infinitely many maximal elements. See Theorem 5 for the complete statement.

Observation 2.

Let K and L be two co-final families of subgroups in G, i.e., for any subgroup H ∈ K , there is a subgroup M ∈ L containing H, and vice versa.

Then N ⁢ C ⁢ ( K ) and N ⁢ C ⁢ ( L ) are homotopy equivalent via the chain

𝒩 ⁢ 𝒞 ⁢ ( 𝒦 ) ≃ 𝒳 ⁢ 𝒞 ⁢ ( 𝒦 ) = 𝒳 ⁢ 𝒞 ⁢ ( ℒ ) ≃ 𝒩 ⁢ 𝒞 ⁢ ( ℒ ) . ∎

The family ℋ fi max ⁢ ( G ) of maximal elements in ℋ fi ⁢ ( G ) is co-final in ℋ fi ⁢ ( G ) , and for the corresponding coset nerve 𝒩 ⁢ 𝒞 max , fi ⁢ ( G ) , we obtain the homotopy equivalence

(1.1) 𝒩 ⁢ 𝒞 fi ⁢ ( G ) ≃ 𝒩 ⁢ 𝒞 max , fi ⁢ ( G ) .

From this, we obtain one direction of our characterization.

Corollary 3.

Suppose that H fi ⁢ ( G ) has only finitely many maximal finite index subgroups, i.e., the family H max , fi ⁢ ( G ) = { M 1 , … , M n } is finite. Then the Frattini subgroup Φ : = M 1 ∩ ⋯ ∩ M n is a normal subgroup of finite index in G, and the coset nerve N ⁢ C fi ⁢ ( G ) is homotopy equivalent to N ⁢ C ⁢ ( G / Φ ) . In particular, N ⁢ C max , fi ⁢ ( G ) is not contractible.

Proof.

The subgroups M i are in 1–1 correspondence to the maximal subgroups of the finite group G / Φ , and the projection G → G / Φ induces an isomorphism of nerves 𝒩 ⁢ 𝒞 max , fi ⁢ ( G ) ≅ 𝒩 ⁢ 𝒞 max ⁢ ( G / Φ ) . The homotopy equivalence (1.1) implies

𝒩 ⁢ 𝒞 fi ⁢ ( G ) ≃ 𝒩 ⁢ 𝒞 max , fi ⁢ ( G ) ≅ 𝒩 ⁢ 𝒞 max ⁢ ( G / Φ ) ≃ 𝒩 ⁢ 𝒞 ⁢ ( G / Φ ) .

As G / Φ is a nontrivial finite group, its coset nerve is not contractible as shown by Shareshian–Woodroofe [4]. ∎

For the converse, we consider complementary subgroups. We call a proper subgroup M < G a complement of the subgroup H ≤ G if G = H ⁢ M . Equivalently, one can say that M intersects each coset of H. Given a collection H 1 , … , H n , we say that M is a common complement of the H i if it is a complement of their intersection H 1 ∩ ⋯ ∩ H n . Note that a complement M of a subgroup H cannot contain H because then G = H ⁢ M = M , while we assume the complement M to be a proper subgroup of G.

We can now state the key argument of this paper.

Proposition 4 (Cone construction).

Let K be a family of subgroups in G such that any finitely many subgroups H 1 , … , H n ∈ K have a common complement M ∈ K . Then the coset nerve N ⁢ C ⁢ ( K ) is contractible.

Proof.

Let Y be a finite subcomplex of 𝒩 ⁢ 𝒞 ⁢ ( 𝒦 ) . Thus, there are finitely many subgroups H 1 , … , H n ∈ 𝒦 such that Y ≤ 𝒩 ⁢ 𝒞 ⁢ ( { H 1 , … , H n } ) ≤ 𝒩 ⁢ 𝒞 ⁢ ( 𝒦 ) . By hypothesis, there is a common complement M for the subgroups H 1 , … , H n . Then the coset 1 ⁢ M is a vertex in 𝒩 ⁢ 𝒞 ⁢ ( 𝒦 ) whose star contains the whole subcomplex Y. Therefore, Y can be contracted within 𝒩 ⁢ 𝒞 ⁢ ( 𝒦 ) .

Since spheres are compact, any element of any homotopy group can be realized within a finite subcomplex of 𝒩 ⁢ 𝒞 ⁢ ( 𝒦 ) . Thus, all homotopy groups of 𝒩 ⁢ 𝒞 ⁢ ( 𝒦 ) are trivial and 𝒩 ⁢ 𝒞 ⁢ ( 𝒦 ) is contractible by Whitehead’s theorem. ∎

The main result of this section is an easy consequence.

Theorem 5.

For any group G the following are equivalent:

Every proper subgroup of finite index in G has a complement that is also of finite index.
The nerve 𝒩 ⁢ 𝒞 fi ⁢ ( G ) is contractible.
The poset ℋ fi ⁢ ( G ) of proper finite index subgroups in G has infinitely many maximal elements.

Proof.

First, we show that (i) implies (ii). Let H 1 , … , H n be proper finite index subgroups of G. Their intersection H 1 ∩ ⋯ ∩ H n is a proper finite index subgroup of G, which by hypothesis (i) has a complement M ∈ ℋ fi ⁢ ( G ) . This is a common complement of the collection H 1 , … , H n . By the cone principle, we conclude that 𝒩 ⁢ 𝒞 fi ⁢ ( G ) is contractible.

That (ii) implies (iii) has been proven in Corollary 3.

For the remaining implication, assume that ℋ fi max ⁢ ( G ) is infinite. Let H be a proper subgroup of finite index in G. We have to find a complementary subgroup M ∈ ℋ fi ⁢ ( G ) . Since a complement to a subgroup of H will also be a complement to H, we may assume without loss of generality that H is normal in G. Then, for any M ≤ G , the product HM is a subgroup of G. Since only finitely many maximal subgroups contain the finite index subgroup H (they correspond to the maximal subgroups of the finite group G / H ), we can choose a maximal subgroup M of finite index not containing H. Then G = M ⁢ H since the maximal subgroup M is a proper subgroup of MH. Thus, we have found a complement for H. Condition (i) follows. ∎

1.1 Inheriting a contractible coset nerve

Proposition 6.

Let G be a group with contractible coset nerve N ⁢ C fi ⁢ ( G ) , and let H ≤ G be a subgroup of finite index. Then its coset nerve N ⁢ C fi ⁢ ( H ) is also contractible.

Proof.

We use the characterization by the existence of complements. Let K be a proper finite index subgroup of H. Then it is a proper finite index subgroup of G. Since 𝒩 ⁢ 𝒞 fi ⁢ ( G ) is contractible, there is a finite index complement M for K in G. Then M ∩ H is a complement for K in H. ∎

Observation 7.

Let π : G ~ → G be an epimorphism of groups, and let H ≤ G be a maximal subgroup. Then the preimage π - 1 ⁢ ( H ) is a maximal subgroup of G ~ . Hence, N ⁢ C fi ⁢ ( G ~ ) is contractible provided that N ⁢ C fi ⁢ ( G ) is contractible.∎

1.2 Examples: groups with contractible coset nerve

A group G is called indicable if it admits an epimorphism onto the infinite cyclic group ℤ . Since ℤ has infinitely many maximal subgroups, its coset nerve is contractible, and G inherits this property by Observation 7.

Hence, the following groups all have contractible coset nerves:

Free groups.
G * ℤ , G × ℤ , G ⋊ ℤ for any group G.
Free abelian groups.
HNN extensions.
Orientation preserving Fuchsian groups of genus at least 1. This can be seen by abelianizing the presentation.
Other Fuchsian groups, provided the genus is at least 2.
Baumslag–Solitar groups 〈 a , b ∣ b ⁢ a m = a n ⁢ b 〉 .
Thompson’s group F has ℤ × ℤ as its maximal abelian quotient.
Artin groups.
Pure braid groups (finite index subgroups of indicable groups are indicable).

As a variation of this trick, we can use the infinite dihedral group D ∞ instead of ℤ . Thus, all of the following groups also have a contractible coset nerve:

Semi-direct products ℤ ⋊ G map to D ∞ with full image or image ℤ .
Infinite Coxeter groups 〈 s 1 , … , s n ∣ s i 2 , ( s i ⁢ s j ) m i ⁢ j 〉 , where all m i ⁢ j are even and m 1 , 2 = ∞ . Killing all s i for i ≥ 3 , we recognize D ∞ as a quotient.

We can generalize from the infinite dihedral group to other euclidean crystallographic groups.

Observation 8.

Let G be a crystallographic group in R d . Then G has finite index in a semi-direct product G ~ = Z d ⋊ Q , where the finite quotient Q acts on the lattice Z d via linear automorphisms. Hence, the rescaled lattices H m : = m Z d are invariant sublattices. Thus, the group H m ⋊ Q is a subgroup of index m d in G ~ . In particular, for each prime number p, the group G ~ has a subgroup of index p d . Thus, the index of a maximal subgroup containing H p ⋊ Q is a p-power. It follows that G ~ has infinitely many maximal subgroups of finite index.

By Theorem 5, the coset nerve of G ~ is contractible. By Proposition 6, this carries over to the coset nerve of the finite index subgroup G ≤ G ~ .

Dealing with SL n ⁡ ( ℤ ) requires a new idea.

Observation 9.

Let G be a finitely generated group that has infinitely many finite index normal subgroups with simple quotient. Then any proper finite index subgroup H ≤ G has a finite index complement. In particular, the coset nerve N ⁢ C fi ⁢ ( G ) is contractible.

Proof.

G acts by left multiplication on the finite set G / H . Let K be the kernel of this action. Thus, K is a normal subgroup of finite index in G. For any normal subgroup N of G, the product KN is normal in G. If the quotient G / N is simple and if N does not contain K, then G = K ⁢ N = H ⁢ N .

As K has finite index, it is contained only within finitely many normal subgroups N of G. Thus, there are (infinitely many) normal subgroups N that are complementary to K and hence to H. ∎

Examples of groups whose coset nerve is contractible by this argument include:

SL n ⁡ ( ℤ ) , Aut ⁡ ( F n ) and Out ⁡ ( F n ) .
Symplectic groups and mapping class groups.

1.3 Examples: groups with non-contractible coset nerves

As infinite simple groups do not have any proper subgroups of finite index, they can be used to construct some easy examples of groups with non-contractible coset nerve.

Observation 10.

Suppose the infinite group G has a simple subgroup S. Let N be the normal subgroup generated by S. Then G and G / N have the same finite quotients since a projection homomorphism π : G → Q is trivial on S (by simplicity) and hence is trivial on N.

Hence, G has only finitely many normal subgroups of finite index provided that G / N has only finitely many normal subgroups of finite index. As each such normal subgroup is contained in only finitely many subgroups of G, the whole family H fi ⁢ ( G ) is finite in this case, and the coset nerve N ⁢ C fi ⁢ ( G ) is not contractible.

This applies for instance to any group of the form

infinite simple * ( finite × infinite simple ) .

Of course, this theme allows for many variations. More interesting are residually finite examples. The examples we know are p-groups.

Observation 11.

Let p be a prime. Any finitely generated p-group G has only finitely many maximal subgroups. In particular, N ⁢ C fi ⁢ ( G ) is not contractible.

Proof.

We claim that every maximal finite index subgroup in G has index p. Since G is finitely generated, there are only finitely many subgroups of index p.

A maximal finite index subgroup H ≤ G contains a finite index subgroup N which is normal in G. Hence, the maximal subgroup H corresponds to a maximal subgroup of the finite p-group G / N . Therefore, H has index p in G. ∎

Thus, the first Grigorchuk group or the Gupta–Sidki groups have non-contractible coset nerves.

2 The family of finite index normal subgroups

In this section, we consider the family

ℋ nor , fi ( G ) : = { N ◁ G ∣ N has finite index in G }

of proper normal subgroups that have finite index in G. Let 𝒩 ⁢ 𝒞 nor , fi ⁢ ( G ) be the coset nerve for this family. We shall prove the exact analogue of Theorem 5. For this, we need the analogue of the result of Shareshian–Woodroofe [4], which we shall prove as Theorem 20 in the appendix.

Theorem 12.

For any group G, the following are equivalent:

𝒩 ⁢ 𝒞 nor , fi ⁢ ( G ) is contractible.
ℋ nor , fi ⁢ ( G ) has infinitely many maximal elements.
Any proper finite index subgroup H ≤ G has infinitely many complementary normal subgroups of finite index in G.
Each proper normal subgroup of finite index in G has a complementary normal subgroup of finite index in G.

Proof.

To see that (i) implies (ii), we argue the contrapositive. So assume that G has only finitely many maximal proper normal subgroups M 1 , … , M n of finite index. Their intersection Φ ~ : = M 1 ∩ ⋯ ∩ M n is normal and of finite index. By correspondence, the coset nerves 𝒩 ⁢ 𝒞 nor , fi ⁢ ( G ) and 𝒩 ⁢ 𝒞 nor , fi ⁢ ( G / Φ ~ ) are isomorphic. By Theorem 20 from the appendix, the latter is not contractible.

Now we assume (ii) and show (iii). So let H ≤ G be a proper subgroup of finite index. As before, we consider the action of G on the finite set G / H . The kernel K of the action is a proper normal subgroup of finite index in G. Any maximal proper normal subgroup N ◁ G not containing K is a complement to K and thus to H. Since K is only contained in finitely many such N, there are infinitely many complementary subgroups in ℋ nor , fi ⁢ ( G ) .

Clearly, (iii) implies (iv). The remaining implication from (iv) to (i) follows immediately from the cone construction (Proposition 4) applied to the family ℋ nor , fi ⁢ ( G ) . ∎

Under the equivalent conditions of Theorem 12, each proper finite index subgroup has a complement. Thus, we obtain the following.

Corollary 13.

If N ⁢ C nor , fi ⁢ ( G ) is contractible, then so is N ⁢ C fi ⁢ ( G ) .

For finitely generated G, we also have a characterization via finite simple quotients.

Observation 14.

If G is finitely generated, the following are equivalent:

ℋ nor , fi ⁢ ( G ) has infinitely many maximal elements.
G has infinitely many pairwise non-isomorphic finite simple quotients.

Proof.

The direction from (ii) to (i) is obvious: for each finite simple quotient, the kernel of the projection is a maximal finite index normal subgroup.

For the converse, we just observe that, for a finite (simple) group Q, there are only finitely many homomorphisms G → Q as G is finitely generated. Hence, each such possible quotient arises from at most finitely many normal subgroups. ∎

3 The family of all proper subgroups

In the poset of proper finite index subgroups, the maximal elements form a co-final family. Our analysis in that case crucially relied on that feature. For finitely generated groups G, the poset of all proper subgroups behaves in the same way.

Lemma 15.

Let G be a finitely generated group. Then any proper subgroup of G is contained in a maximal proper subgroup. In particular, the families H ⁢ ( G ) and H max ⁢ ( G ) are co-final, and we have the corresponding homotopy equivalence N ⁢ C ⁢ ( G ) ≃ N ⁢ C 𝑚𝑎𝑥 ⁢ ( G ) .

Proof.

By Zorn’s lemma, it suffices to see that any ascending union of proper subgroups in G is a proper subgroup. However, this follows from the existence of a finite generating set for G. If an ascending union contained all the generators, this would happen already at some stage along the underlying ascending chain of subgroups. ∎

Observation 16.

If a maximal subgroup M ≤ G has only finitely many conjugate subgroups in G, it is of finite index in G.

Proof.

The normalizer of N ⁢ ( M ) of M has finite index in G as it is a stabilizer of a G-action on a finite set. Moreover, we have the inclusions M ≤ N ⁢ ( M ) ≤ G . As M is maximal, it follows that M = N ⁢ ( M ) (and therefore has finite index in G) or that M is normal in G.

If M is normal, it follows that the trivial group is a maximal subgroup in the quotient G / M . Hence, the quotient is cyclic of prime order. Again, M has finite index in G. ∎

Since the collection of maximal subgroups is closed with respect to taking conjugates, we infer the following.

Corollary 17.

If a finitely generated group G has only finitely many maximal subgroups, they all are of finite index in G, and the coset nerve N ⁢ C ⁢ ( G ) is homotopy equivalent to N ⁢ C fi ⁢ ( G ) . Both are non-contractible.∎

We conclude this section with two results towards contractibility.

Observation 18.

If every proper subgroup of G is contained in a proper finite index subgroup of G, then H ⁢ ( G ) and H fi ⁢ ( G ) are co-final. In this case, we have the homotopy equivalence N ⁢ C ⁢ ( G ) ≃ N ⁢ C fi ⁢ ( G ) , and contractibility of N ⁢ C fi ⁢ ( G ) implies contractibility of N ⁢ C ⁢ ( G ) .∎

A slightly more involved argument is required in the following situation. We call a subgroup H of G extreme if it is of finite index or not even contained in a subgroup of finite index. So extreme subgroups are either very large or very small.

Proposition 19.

Let G be a group, and assume additionally that every finite collection H 1 , … , H n ≤ G of proper extreme subgroups has a common complement. Then the coset nerve N ⁢ C ⁢ ( G ) is contractible.

Proof.

We consider the family 𝒦 of proper extreme subgroups of G. This family is obtained from ℋ ⁢ ( G ) by removing all subgroups of infinite index that are contained in a proper finite index subgroup of G. Clearly, 𝒦 and ℋ ⁢ ( G ) are co-final, wherefore 𝒩 ⁢ 𝒞 ⁢ ( 𝒦 ) ≃ 𝒩 ⁢ 𝒞 ⁢ ( G ) .

By the cone construction (Proposition 4), it suffices to find for any finite collection H 1 , … , H n ∈ 𝒦 a common complement in 𝒦 . By hypothesis, the H i have a common complementary subgroup M ≤ G . If M ∈ 𝒦 , there is nothing to be argued. Otherwise, M has infinite index in G and is contained in a proper finite index subgroup, which then can be used as a common complement for the H i . ∎

Communicated by Dessislava H. Kochloukova

Funding source: Deutsche Forschungsgemeinschaft

Award Identifier / Grant number: SPP 2026

Award Identifier / Grant number: SFB 878

Funding statement: Financial support by the DFG through the programs SPP 2026 and SFB 878 is gratefully acknowledged.

A Appendix: The coset nerve for proper normal subgroups in a finite group

In this appendix, we consider the family

ℋ nor ( G ) : = { N ⊴ G ∣ N ≠ G }

of proper normal subgroups of G. Let 𝒩 ⁢ 𝒞 nor ⁢ ( G ) denote the coset nerve associate to ℋ nor ⁢ ( G ) . This section is devoted to a proof of the following analogue to the result of Shareshian–Woodroofe [4].

Theorem 20.

If G is finite, N ⁢ C nor ⁢ ( G ) is not contractible.

Proof.

If the group G is trivial, the family of proper normal subgroups is empty. Hence, we assume that G is nontrivial.

The family ℋ nor ⁢ ( G ) is closed with respect to intersections. Hence, we shall consider the associated order complex 𝒫 ⁢ 𝒞 nor ⁢ ( G ) instead of the homotopy equivalent coset nerve. Our argument uses many ideas of Brown [2, Section 8].

Inducting on the complexity of G, we shall show that 𝒫 ⁢ 𝒞 nor ⁢ ( G ) is a nontrivial wedge of spheres. Simple groups make up the base of the induction. Fortunately, simple groups are easily understood since the only proper normal subgroup is the trivial subgroup.

Claim 21.

If G is simple, P ⁢ C nor ⁢ ( G ) is a finite discrete set, which we shall identify with G itself. It is 0-spherical and not contractible (since G is nontrivial).

Now assume that G is not simple. Let M ◁ G be a minimal normal subgroup. Let π : G → G / M denote the canonical projection.

Claim 22.

Assume that G contains no proper normal subgroup that surjects onto the quotient G / M . Then the map

q : 𝒫 ⁢ 𝒞 nor ⁢ ( G ) → 𝒫 ⁢ 𝒞 nor ⁢ ( G / M ) , g ⁢ N ↦ π ⁢ ( g ⁢ N )

is a homotopy equivalence. Thus, sphericity and non-contractibility of P ⁢ C nor ⁢ ( G ) is inherited from P ⁢ C nor ⁢ ( G / M ) .

To see this, consider the intersection-closed family 𝒦 : = { N ◁ G ∣ M ⊆ N } . The elements of 𝒦 are in 1–1 correspondence to the proper normal subgroups of G / M . Therefore, 𝒫 ⁢ 𝒞 𝒦 ⁢ ( G ) and 𝒫 ⁢ 𝒞 nor ⁢ ( G / M ) are isomorphic. On the other hand, 𝒦 is co-final in ℋ nor ⁢ ( G ) since by hypothesis NM is a proper normal subgroup of G for any proper normal N ◁ G . This proves Claim 22.

It remains to deal with the case that there is a proper normal subgroup N ~ ◁ G that surjects onto G / M . Note that the intersection N ~ ∩ M is normal in G. As N ~ is proper, but the product N ~ ⁢ M is all of G, we conclude that N ~ ∩ M ≠ M . As M is a minimal normal subgroup in G, we find N ~ ∩ M = { 1 } and G = M × N ~ . In particular, we can identify N ~ with G / M . We call a coset gH in G large if it surjects onto N ~ . We call it small otherwise. Note that the coset gH is small if and only if 1 ⁢ H is small.

We consider the following families:

ℋ large ( G ) : = { N ◁ G ∣ 1 N is large } ,

ℋ small ( G ) : = { N ◁ G ∣ 1 N is small } ,

ℒ : = { M × L ∣ L ◁ N ~ } .

Note that ℋ small ⁢ ( G ) and ℒ are closed with respect to intersections. Moreover, ℒ ⊆ ℋ small ⁢ ( G ) is co-final in ℋ small ⁢ ( G ) . More precisely, for N ∈ ℋ small ⁢ ( G ) , we have N ⊆ M × π N ~ ⁢ ( N ) ∈ ℒ . Hence, we have the homotopy equivalence

(A.1) 𝒫 ⁢ 𝒞 ⁢ ( ℋ small ⁢ ( G ) ) ≃ 𝒫 ⁢ 𝒞 ⁢ ( ℒ ) ≅ 𝒫 ⁢ 𝒞 nor ⁢ ( N ~ ) .

Now consider a large normal subgroup L ◁ G . We claim that the projection onto N ~ restricts to an isomorphism on L. The reason is that M is minimal, and because of this, L ∩ M is trivial or all of M. However, the latter possibility is excluded since L is a proper subgroup of G. This argument shows the following.

Claim 23.

If L is a large proper normal subgroup of G, then G = M × L .

Consequently, all large proper normal subgroups have the same cardinality. Thus, they are mutually incomparable with respect to inclusion.

We now see the structure of the coset poset 𝒫 ⁢ 𝒞 nor ⁢ ( G ) . Vertices from ℋ small ⁢ ( G ) form the bottom part whose order complex is homotopy equivalent to 𝒫 ⁢ 𝒞 nor ⁢ ( N ~ ) . Even more is true. The coset 1 ⁢ N ~ is large. Its link is spanned by small cosets contained in 1 ⁢ N ~ . Therefore, this link is isomorphic to 𝒫 ⁢ 𝒞 nor ⁢ ( N ~ ) , and homotopy equivalence (A.1) is a deformation retraction of the bottom part 𝒫 ⁢ 𝒞 ⁢ ( ℋ small ⁢ ( G ) ) onto this link.

The vertices from ℋ large ⁢ ( G ) lie above the bottom part. The order complex 𝒫 ⁢ 𝒞 nor ⁢ ( G ) is the union of the bottom part and the stars of the large vertices. Each such star is just the cone over the link of the vertex, and we may consider these stars independently because two large vertices are never joined by an edge. Moreover, the link of a large coset gL is its descending link, i.e., all vertices in the link are cosets that are contained in gL.

Claim 24.

The link of a large vertex gL is isomorphic to P ⁢ C nor ⁢ ( L ) .

This just follows from the fact that a coset gN is contained in gL if and only if N ≤ L .

Now we build 𝒫 ⁢ 𝒞 nor ⁢ ( G ) by adding the stars of large vertices, one by one, to the bottom part 𝒫 ⁢ 𝒞 ⁢ ( ℋ small ⁢ ( G ) ) . Adding 1 ⁢ N ~ as the first large vertex, we cone of its descending link. At this point, we obtain a contractible space. Adding each of the remaining large vertices (and there are at least the other cosets of N ~ ) amounts to wedging on the suspension of its link, i.e., wedging on a copy of Σ ⁢ ( 𝒫 ⁢ 𝒞 nor ⁢ ( N ~ ) ) . Thus, we have argued the following.

Claim 25.

Assume that there is a proper normal subgroup N ~ ◁ G that surjects onto G / M . Then P ⁢ C nor ⁢ ( G ) is homotopy equivalent to a nontrivial wedge of copies of the suspension Σ ⁢ ( P ⁢ C nor ⁢ ( N ~ ) ) .

Note that the theorem follows by induction from Claims 21, 22, and 25. ∎

Acknowledgements

We would like to thank Russ Woodroofe for suggesting that a proof of Theorem 20 could be carried out using ideas from Brown [2] rather than following the much steeper path of Shareshian–Woodroofe [4]. We also thank Benjamin Brück, Linus Kramer and Russ Woodroofe for helpful comments on a preliminary version of this paper.

References

[1] H. Abels and S. Holz, Higher generation by subgroups, J. Algebra 160 (1993), 310–341. 10.1006/jabr.1993.1190Search in Google Scholar

[2] K. S. Brown, The coset poset and probabilistic zeta function of a finite group, J. Algebra 255 (2000), 989–1012. 10.1006/jabr.1999.8221Search in Google Scholar

[3] D. A. Ramras, Connectivity of the coset poset and the subgroup poset of a group, J. Group Theory 8 (2005), 719–746. 10.1515/jgth.2005.8.6.719Search in Google Scholar

[4] J. Shareshian and R. Woodroofe, Order complexes of coset posets of finite groups are not contractible, Adv. Math. 291 (2016), 758–773. 10.1016/j.aim.2015.10.018Search in Google Scholar

[5] C. Welsch, On coset posets, nerve complexes and subgroup graphs of finitely generated groups, PhD Thesis, Münster, 2018. Search in Google Scholar

Received: 2019-11-09

Revised: 2020-01-30

Published Online: 2020-03-10

Published in Print: 2020-07-01

Articles in the same Issue

https://doi.org/10.1515/jgth-2019-0162