The first relative k k -invariant
Anthony Conway, Daniel Kasprowski · 2025
Read The first relative k k -invariant by Anthony Conway, Daniel Kasprowski (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-2510-18796.txt.
Introduction
Anthony Conway and Daniel Kasprowski define a new obstruction in algebraic topology. The work is motivated by classifying four-dimensional manifolds with boundary. Their object is the relative k-invariant. It belongs to a pair of spaces. One space sits inside the other. The invariant measures whether a certain map can be extended. The map lives on the smaller space and on a two-stage approximation of the larger one.
The authors show the invariant is exactly the obstruction to a section. They prove it is complete in a comparison problem too. Take two pairs of spaces and a map between the smaller members. The invariants decide whether that map extends over the first space's three-dimensional skeleton. Set identifications on the first two homotopy groups must hold. Complete obstructions are prized in this field. Nothing is lost when passing to the invariant.
Catalog record
| CREATOR | Anthony Conway, Daniel Kasprowski |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 96 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-18796
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
The first relative k k -invariant Anthony Conway, Daniel Kasprowski Abstract . Motivated by work on the homotopy classification of 4 4 -manifolds with boundary, we define a relative k k -invariant for pairs of spaces that are homotopy equivalent to CW pairs. We show that for such a pair ( X , Y ) (X,Y) with Postnikov 2 2 -type X → P 2 ( X ) X\to P_{2}(X) , the relative k k -invariant is the obstruction to the existence of a section B π 1 ( X ) → P 2 ( X ) B\pi_{1}(X)\to P_{2}(X) extending Y ↪ X → P 2 ( X ) Y\hookrightarrow X\to P_{2}(X) . Given CW pairs ( X 0 , Y 0 ) (X_{0},Y_{0}) and ( X 1 , Y 1 ) (X_{1},Y_{1}) , as well as a map h : Y 0 → Y 1 h\colon Y_{0}\to Y_{1} , we also prove that relative k k -invariants provide a complete obstruction to constructing a map X 0 ( 3 ) ∪ Y 0 → X 1 X_{0}^{(3)}\cup Y_{0}\to X_{1} that extends h h and induces given isomorphisms on π 1 \pi_{1} and π 2 \pi_{2} . 1. Introduction Given a space X with Postnikov tower \ldots\to P_{2}(X)\to B\pi_{1}(X) , the (first) k -invariant of X is the obstruction k_{X}\in H^{3}(B\pi_{1}(X);\pi_{2}(X)) to finding a section of P_{2}(X)\to B\pi_{1}(X) . This invariant, first introduced in [ 2 ] , is a mainstay in algebraic topology, but also plays a role in geometric topology and, more specifically, in the (homotopy) classification of manifolds. For example, the homotopy type of a closed oriented 4 -manifold X with finite fundamental group is often determined by its quadratic 2 -type (\pi_{1}(X),\pi_{2}(X),k_{X},\lambda_{X}) , where \lambda_{X} denotes the equivariant intersection form of X ; see e.g. [ 3 ] . Motivated by upcoming work on the homotopy classification of 4 -manifolds with nonempty boundary [ 1 ] , the present article is concerned with relative k -invariants and collects several results on the topic that might be known to experts but that we have not been able to find in the literature. Statement of results Let (X,Y) be a pair of spaces that is homotopy equivalent to a CW pair, let c\colon X\to P_{2}(X) be the Postnikov 2 -type of X , and given a map \nu\colon X\to B\pi_{1}(X) , we write M(\nu|_{Y}) for the mapping cylinder of \nu|_{Y} . The relative k -invariant refers to a cohomology class k_{X,Y}^{\nu}\in H^{3}(M(\nu|_{Y}),Y;\pi_{2}(X)). The definition of this class is described precisely in Section 5 , but for the moment we note that k_{X,Y}^{\nu} is the obstruction to the existence of a dashed map making the following diagram commute: {\lx@inpgf@ignorespaces Y} {\lx@inpgf@ignorespaces P_{2}(X)} {\lx@inpgf@ignorespaces B\pi_{1}(X)} {\lx@inpgf@ignorespaces B\pi_{1}(X).} \scriptstyle{\lx@inpgf@ignorespaces c|_{Y}} \scriptstyle{\lx@inpgf@ignorespaces\nu|_{Y}} \scriptstyle{\lx@inpgf@ignorespaces\operatorname{id}} Up to isomorphism, k_{X,Y}^{\nu} is independent of the choice of the map \nu ; see Remark 5.4 . The reason for which we write H^{3}(M(\nu|_{Y}),Y;\pi_{2}(X)) instead of H^{3}(B\pi_{1}(X),Y;\pi_{2}(X)) is to make it apparent that we choose a model of B\pi_{1}(X) that contains Y as a subspace. Also, k_{X,\emptyset}^{\nu} agrees with k_{X} . The following result (the absolute version of which can be found in [ 2 , 4 ] ) plays a crucial role in upcoming work on the homotopy classification of 4 -manifolds with boundary [ 1 ] . Theorem 1.1 . Let (X_{0},Y_{0}) and (X_{1},Y_{1}) be pairs of spaces that are homotopy equivalent to CW pairs, and let c_{1}\colon X_{1}\to P_{2}(X_{1}) be the Postnikov 2 -type of X_{1} . For a map h\colon Y_{0}\to Y_{1} , an isomorphism u\colon\pi_{1}(X_{0})\to\pi_{1}(X_{1}) with u\circ(\iota_{0})_{*}=(\iota_{1})_{*}\circ h , and an u -equivariant homomorphism F\colon\pi_{2}(X_{0})\to\penalty\ \pi_{2}(X_{1}) , the following assertions are equivalent: • there is a map c_{0}\colon X_{0}\to P_{2}(X_{1}) such that (c_{0})_{*}=u\colon\pi_{1}(X_{0})\to\pi_{1}(P_{2}(X_{1}))\cong\pi_{1}(X_{1}),\quad c_{1}\circ h\simeq c_{0}|_{Y_{0}},\quad\text{and} (c_{0})_{*}=F\colon\pi_{2}(X_{0})\to\pi_{2}(P_{2}(X_{1}))\xleftarrow{(c_{1})_{*},\cong}\pi_{2}(X_{1}); • the relative k -invariants satisfy (u,h)^{*}(k_{X_{1},Y_{1}}^{\nu_{1}})=F_{*}(k_{X_{0},Y_{0}}^{\nu_{0}})\in H^{3}(M(\nu_{0}|_{Y_{0}}),Y_{0};\Res_{u}\pi_{2}(X_{1})) for every \nu_{i}\colon X_{i}\to B\pi_{1}(X_{i}) that induces the identity on \pi_{1} for i=0,1 . The map (u,h)^{*} is defined in Lemma 5.1 . Here, a homomorphism F\colon\pi_{2}(X_{0})\to\pi_{2}(X_{1}) is called u -equivariant if F(\gamma x)=u(\gamma)F(x) for every x\in\pi_{2}(X_{0}) and \gamma\in\pi_{1}(X_{0}) . Equivalently, F is a \mathbb{Z}[\pi_{1}(X_{0})] -linear map \pi_{2}(X_{0})\to\Res_{u}\pi_{2}(X_{1}) , where given a \mathbb{Z}[\pi_{1}(X_{1})] -module H , we write \Res_{u}H for the \mathbb{Z}[\pi_{1}(X_{0})] -module whose underlying abelian group is H , but with \pi_{1}(X_{0}) -action given by \gamma\cdot x=u(\gamma)x . Remark 1.2 . The reason we work with spaces that are merely homotopy equivalent to CW pairs (instead of working with CW pairs as is common in the literature) is to ensure that our results apply to 4 -dimensional topological manifolds. If (X_{0},Y_{0}) and (X_{1},Y_{1}) are CW pairs, Theorem 1.1 admits the following reformulation. Corollary 1.3 . Let (X_{0},Y_{0}) and (X_{1},Y_{1}) be CW pairs, and let \iota_{j}\colon Y_{j}\to X_{j} be the inclusion for j=0,1 . For a group isomorphism u\colon\pi_{1}(X_{0})\to\pi_{1}(X_{1}) , a map h\colon Y_{0}\to Y_{1} satisfying u\circ(\iota_{0})_{*}=(\iota_{1})_{*}\circ h , and an u -equivariant homomorphism F\colon\pi_{2}(X_{0})\to\pi_{2}(X_{1}) , the following assertions are equivalent: • there is a map f\colon X_{0}^{(3)}\cup Y_{0}\to X_{1} that extends h and induces u and F ; • the relative k -invariants satisfy (u,h)^{*}(k_{X_{1},Y_{1}}^{\nu_{1}})=F_{*}(k_{X_{0},Y_{0}}^{\nu_{0}})\in H^{3}(M(\nu_{0}|_{Y_{0}}),Y_{0};\pi_{2}(X_{1})) for every \nu_{i}\colon X_{i}\to B\pi_{1}(X_{i}) that induces the identity on \pi_{1} for i=0,1 . Proof. By Theorem 1.1 , it suffices to show that there exists a map f\colon X_{0}^{(3)}\cup Y_{0}\to X_{1} as in this corollary if and only if there exists a map c_{0}\colon X_{0}\to P_{2}(X_{1}) as in Theorem 1.1 . In one direction, starting from f , since the inclusion X_{0}^{(3)}\cup Y_{0}\to X_{0} induces a homotopy equivalence on Postnikov 2 -types, the required c_{0} arises as the composition X_{0}\to P_{2}(X_{0})\simeq P_{2}(X_{0}^{(3)}\cup Y_{0})\xrightarrow{P_{2}(f)}P_{2}(X_{1}). It remains to construct f from c_{0} . For this, choose a model for P_{2}(X_{i}) that is obtained from X_{i} by attaching cells of dimension \geq 4 . Homotoping c_{0} if necessary (using cellular approximation), we assume that it is cellular. Restricting this map to X_{0}^{(3)} leads to a map f^{\prime}\colon X_{0}^{(3)}\to X_{1}^{(3)}\hookrightarrow X_{1} that induces u and F and such that f^{\prime}|_{Y_{0}^{(3)}}\simeq\iota_{Y_{1}}\circ h|_{Y_{0}^{(3)}} . Since Y_{0}^{(3)}\hookrightarrow X_{0}^{(3)} , the homotopy extension property shows that f^{\prime} is homotopic to a map f^{\prime\prime}\colon X_{0}^{(3)}\to X_{1}^{(3)}\hookrightarrow X_{1} that induces u and F and such that f^{\prime\prime}|_{Y_{0}^{(3)}}=\iota_{Y_{1}}\circ h|_{Y_{0}^{(3)}} . Using h to extend f^{\prime\prime} over the rest of Y_{0} yields the required f . ∎ Organisation Section 2 collects some conventions and lemmas on chain complexes. Section 3 introduces the relative k -invariant of a chain map. Section 4 focuses on the relative k -invariant of CW pairs, whereas Section 5 defines the relative k -invariant in general and proves our main results. Section 6 collects some additional properties of relative k -invariants. Conventions Unless specified otherwise, spaces are assumed to path-connected and equipped with a basepoint. Maps are assumed to be basepoint preserving and if (X,Y) is a pair, the basepoint is assumed to lie in Y ; if Y is empty, we assume that x\in X . Throughout this paper, for every space X , we choose a model B\pi_{1}(X) and assume that all maps X\to B\pi_{1}(X) that we consider induce the identity on fundamental groups. The mapping cylinder of a map f\colon X\to Y is denoted M(f) , whereas the mapping cone of a chain map g\colon C\to D is denoted \operatorname{Cone}(g)