4-manifolds with a given boundary
Anthony Conway, Daniel Kasprowski
TL;DR
This work develops a comprehensive obstruction-theoretic framework for classifying 4-manifolds with boundary up to homotopy and, in favorable cases with good fundamental groups, up to homeomorphism. Central to the approach are the Postnikov 2-type and a pair of obstructions: a primary obstruction detected by comparing pulled-back relative intersection forms, and a secondary obstruction b(c0,c1) valued in Hermitian forms modulo a 𝒢-action. The authors establish conditions under which these obstructions vanish, and crucially prove a main technical result that allows arbitrary adjustments to the secondary obstruction via controlled homotopies, effectively enabling the vanishing of the obstruction in broad group-theoretic settings. They then connect these homotopy classifications to homeomorphism classifications through surgery theory, obtaining concrete results for various π (including finite, dihedral,BS groups, and 3-manifold groups) and recovering known simply-connected and Z-case outcomes as special cases. The methodology blends Postnikov theory, relative k-invariants, twisted (co)homology, and pushouts of Poincaré pairs, yielding a robust framework for extending boundary data to the entire 4-manifold and for deciding when boundary homeomorphisms extend to interior homeomorphisms under spin/non-spin and Kirby–Siebenmann constraints.
Abstract
This paper studies the homotopy and homeomorphism classifications of $4$-manifolds with boundary. Given $4$-manifolds $X_0$ and $X_1$ with fundamental group $π$, we consider the problem of extending a homotopy equivalence $h \colon \partial X_0 \to \partial X_1$ to a homotopy equivalence $X_0 \to X_1$. We solve this problem in broad settings for a class of groups that includes free groups, finite cyclic groups, finite dihedral groups, solvable Baumslag-Solitar groups, and many $3$-manifold groups. When the fundamental group is additionally assumed to be good, we use surgery theory to list situations when a homeomorphism $h\colon\partial X_0 \to\partial X_1$ extends to a homeomorphism $X_0 \to X_1$. The outcome recovers results of Boyer in the simply-connected case and work of the first author and Powell when $π\cong \mathbb{Z}$ and the $\partial X_i$ have torsion Alexander module.
