Swan modules and homotopy types after a single stabilisation

Tommy Hofmann, John Nicholson · 2025

Read Swan modules and homotopy types after a single stabilisation by Tommy Hofmann, John Nicholson (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-21975.txt.

Introduction

Tommy Hofmann and John Nicholson resolve, in a 2025 preprint, a problem that has stood since a famous 1979 list in topology. Swan modules are a special family of modules over group rings. The open problem asked whether a Swan module can be stably free without being free. The authors show one can. That single example settles the question. It also unlocks results about shapes built from cells, called CW-complexes. Two such shapes may become the same after adding spheres, yet the count of spheres needed matters.

In dimensions congruent to three modulo four, they build finite complexes that match only after several stabilizations, never after one. That answers a question posed by Dyer. A side question of Plotnick about automorphisms also falls. As a further payoff, they give a short proof that some group has periodic cohomology without the matching free period, dodging a heavy obstruction computation.

Catalog record

CREATORTommy Hofmann, John Nicholson
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE89 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-21975

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Swan modules and homotopy types after a single stabilisation Tommy Hofmann, John Nicholson Abstract . We study Swan modules, which are a special class of projective modules over integral group rings, and their consequences for the homotopy classification of CW-complexes. We show that there exists a non-free stably free Swan module, thus resolving Problem A4 in the 1979 Problem List of C. T. C. Wall. As an application we show that, in all dimensions n ≡ 3 n\equiv 3 mod 4 4 , there exist finite n n -complexes which are homotopy equivalent after stabilising with multiple copies of S n S^{n} , but not after a single stabilisation. This answers a question of M. N. Dyer. We also resolve a question of S. Plotnick concerning Swan modules associated to group automorphisms and, as an application, obtain a short and direct proof that there exists a group with k k -periodic cohomology which does not have free period k k . In contrast to the original proof our R. J. Milgram, our proof circumvents the need to compute the Swan finiteness obstruction. 2020 Mathematics Subject Classification Primary 55P15; Secondary 20C05, 57Q12 1. Introduction A Swan module over a finite group G is a \mathbb{Z}G -module defined by a left ideal (N,r)\leq\mathbb{Z}G where N=\sum_{g\in G}g and r\in\mathbb{Z} is an integer coprime to |G| . These modules were introduced by Swan [ Swa60b ] , who showed that (N,r) is projective and is determined up to isomorphism by the value of r\bmod|G| . Consequently, we often write r\in(\mathbb{Z}/|G|)^{\times} . Swan modules play an important role in topology through the classification of spherical space forms [ Swa60a , TW71 , LT73 ] and more general CW-complexes [ Joh03 , Nic24a , Nic24b ] . They have also found applications in algebraic number theory, particularly through determining the Galois module structure of the ring of integers of a number field [ Tay81 , Tay84 , GRRS99 ] . A classical question, which arose in work of Dyer [ Dye76 , Note (c) p276] and Plotnick [ Plo82 , p98] , is whether every stably free Swan module is free. A finite group G for which this holds is said to have the weak cancellation property [ Joh04 , p582] . The case where G has periodic cohomology, which is most relevant for applications to topology, appeared in Wall’s problem list [ Wal79a , Problem A4] (see also [ Joh04 , Section 2] , [ Nic24b , Question 7.2] ). We will show: Theorem A . There exists a finite group G and r\in(\mathbb{Z}/|G|)^{\times} such that (N,r) is a non-free stably free \mathbb{Z}G -module. Furthermore, G can be taken to have periodic cohomology. In particular, we can take G=Q_{56} to be the quaternion group of order 56 and r=15 . The main step in the proof is showing that (N,15) is not free. This is achieved as a result of extensive computations of unit groups of rings which arise in two pullback squares (see Eqs. 1 and 3 ). We say two finite n -complexes X and Y are stably equivalent if X\vee kS^{n}\simeq Y\vee kS^{n} for some k\geq 0 . Is one stabilisation always enough, i.e. does k=1 always suffice? This question, with various conditions on the CW-complexes, appeared in [ Dye79 , Problem (c)] , [ HAMS93 , p124] and [ Nic21c , Problem B2] . The analogous question for simple homotopy equivalence ( \simeq_{s} ) appeared in [ Dye81 , Question 1] . Using the examples from A , we will show: Theorem B . Let n\geq 1 with n\equiv 3\,\,\text{mod}\,4 . Then there exist finite n -complexes X and Y such that X\vee 2S^{n}\simeq Y\vee 2S^{n} but X\vee S^{n}\not\simeq Y\vee S^{n} . Furthermore, we can assume that X and Y have finite fundamental group, (n-1) -connected universal covers, and satisfy X\vee 2S^{n}\simeq_{s}Y\vee 2S^{n} . The fact that G can be taken to be finite is of interest since it was shown by Browning [ Bro78 , Theorem 5.4] (see also [ Nic24b , Corollary 4.7] ) that, for n\geq 2 even, one stabilisation is enough for finite n -complexes with finite fundamental group and (n-1) -connected universal covers. We show that this also holds when n\equiv 1\,\,\text{mod}\,4 ( 3.7 ). Hence the dimension restrictions in B are optimal for finite n -complexes satisfying these conditions. The examples in B are obtained by taking X to be the closed n -manifold S^{n}/Q_{56} and defining Y=X_{15} via the following general construction (see 3.3 for more details): Construction 1.1 . Let G be a finite group and let (N,r) be a stably free Swan module for some r\in(\mathbb{Z}/|G|)^{\times} . Given a (connected) finite n -complex X with fundamental group G and (n-1) -connected universal cover, we construct a finite n -complex X_{r} as follows: 1. Let m:\mathbb{Z}\to\mathbb{Z} be multiplication by a representative of r^{-1} , let \varepsilon:C_{0}(\widetilde{X})\twoheadrightarrow H_{0}(\widetilde{X})\cong\mathbb{Z} be the standard map and define m^{*}(C_{*}(\widetilde{X}))=(C_{n}(\widetilde{X})\xrightarrow{\partial_{n}}\cdots\xrightarrow{\partial_{2}}C_{1}(\widetilde{X})\xrightarrow{(\partial_{1},0)}C_{0}(\widetilde{X})\times_{\varepsilon,m}\mathbb{Z}). We have C_{0}(\widetilde{X})\times_{\varepsilon,m}\mathbb{Z}\cong(I_{G},r^{-1})\oplus\mathbb{Z}G^{k}\cong(N,r)\oplus\mathbb{Z}G^{k} for some k\geq 0 by [ Nic24b , Proposition 3.4 & Lemma 4.12] . This is stably free so, by adding extensions of the form \mathbb{Z}G\xrightarrow{\cong}\mathbb{Z}G , this is chain homotopy equivalent to an exact sequence C_{*}^{\prime} of free \mathbb{Z}G -modules. 2. Since n\geq 3 , there exists a finite n -complex X_{r} with fundamental group G such that C_{*}(\widetilde{X}_{r}) is chain homotopy equivalent to C_{*}^{\prime} (see [ Joh12 , Corollary 8.27] ). The complex X_{r} has fundamental group G , (n-1) -connected universal cover, \pi_{n}(X_{r})\cong\pi_{n}(X) as \mathbb{Z}G -modules, and k -invariant k^{n+1}(X_{r})=r\cdot k^{n+1}(X)\in\mathbb{Z}/|G| ( \cong H^{n+1}(G;\pi_{n}(X)) ). To prove that X=S^{n}/Q_{56} and Y=X_{15} are not homotopy equivalent, it remains to show that 1,15\in(\mathbb{Z}/56)^{\times} do not coincide under the combined action of \Aut(Q_{56}) and \Aut_{\mathbb{Z}G}(\pi_{n}(X)) on the k -invariants. Our proof uses that \pi_{n}(X)\cong\mathbb{Z} and so \Aut_{\mathbb{Z}G}(\pi_{n}(X))=\{\pm\id\} . Note that \Aut_{\mathbb{Z}G}(\pi_{n}(X)) is typically hard to compute so, whilst this construction applies to any X and stably free Swan module (N,r) , showing that X_{r}\not\simeq X is difficult in general. We now give two consequences of B . The first follows directly from the properties of 1.1 listed above. Corollary 1.2 . Let n>2 be even. Then there exist homotopically distinct finite n -complexes X and Y with fundamental group Q_{56} and (n-1) -connected universal covers such that \pi_{n}(X)\cong\pi_{n}(Y) as \mathbb{Z}Q_{56} -modules. The same holds for n=2 provided Q_{56} has the D2 property. Such examples have only previously been known for finite abelian fundamental groups [ SD79 ] . The case n=2 would require part (2) of 1.1 to apply in that setting, which does so if and only if Q_{56} has the D2 property (see, for example, [ Nic24b , Proposition 5.1] ). The D2 property for quaternion groups was studied by the authors in [ HN25 ] . In the notation given there, it would suffice to construct a finite presentation \mathcal{P} for Q_{56} such that \Psi(\mathcal{P})=[(N,15)] . We next consider applications to manifolds. It is an open problem to determine whether there exist closed smooth 2n -manifolds M and N such that M\#2(S^{n}\times S^{n})\cong N\#2(S^{n}\times S^{n}) are diffeomorphic but M\#(S^{n}\times S^{n})\not\cong N\#(S^{n}\times S^{n}) , with particular interest in the case where n=2 and M and N are homeomorphic. This question admits many variations for

Read the complete text