Four-manifolds, two-complexes and the quadratic bias invariant

Kreck and Schafer produced the first examples of stably diffeomorphic closed smooth 4-manifolds which are not homotopy equivalent. They were constructed by applying the doubling construction to 2-complexes over certain finite abelian groups of odd order. By extending their methods, we formulate a new homotopy invariant on the class of 4-manifolds arising as doubles of 2-complexes with finite fundamental group. As an application we show that, for any $k \ge 2$, there exist a family of $k$ closed smooth 4-manifolds which are all stably diffeomorphic but are pairwise not homotopy equivalent.

💡 Research Summary

The paper by Ian Hambleton and John Nicholson develops a new homotopy invariant for a class of smooth closed 4‑manifolds that arise as “doubles” of finite 2‑complexes with finite fundamental group. The construction builds on the earlier work of Kreck and Schafer, who showed that two such manifolds can be stably diffeomorphic (i.e., become diffeomorphic after connected sum with a number of S²×S² factors) without being homotopy equivalent. Their examples used a “bias invariant” defined modulo 2 for doubles whose fundamental group is a finite abelian group of odd order.

The authors generalize this bias invariant to a “quadratic bias invariant” β_Q. Starting from a finite group G and a minimal finite 2‑complex X with π₁(X)≅G (minimal meaning χ(X) is as small as possible), the classical bias invariant β(X) lives in an abelian group B(G) = (ℤ/m)×/⟨±D(G)⟩, where m is the exponent of the torsion part of H₂(G;ℤ) and D(G) is the image of a homomorphism Aut(G)→(ℤ/m)×/±1. By analysing the decomposition of H₂(G;ℤ) into cyclic summands, the authors introduce a multi‑scaled hyperbolic form and its unitary isometry group. This yields a natural subgroup N(G)⊂B(G) (the “quadratic bias obstruction”), and the quotient B_Q(G)=B(G)/N(G) becomes the target for the new invariant. The map q:B(G)→B_Q(G) is canonical, and β_Q(M(X)) is defined to be q(β(X)). Theorem A proves that β_Q is indeed a homotopy invariant of the double M(X).

The paper then investigates the structure of B_Q(G). When G is “efficient” (i.e., its minimal Euler characteristic satisfies χ_min(G)=1+d(H₂(G;ℤ))) and H₂(G;ℤ)≅(ℤ/m)^d with d≥3, Theorem B shows that B_Q(G)≅(ℤ/m)×±(ℤ/m)×2·D(G). If G is not efficient, the quadratic bias group collapses to zero. Finite abelian groups are always efficient, while Swan’s non‑efficient example (ℤ/7)^3⋊ℤ/3 illustrates the other case.

Armed with β_Q, the authors answer a recent question of Kasprowski‑Powell‑Ray by constructing, for any integer k≥2, a family of k smooth closed 4‑manifolds M₁,…,M_k that are all stably diffeomorphic but pairwise non‑homotopy‑equivalent (Theorem C). The fundamental groups can be taken to be G=(ℤ/m)^d with d≥3 odd and m having sufficiently many distinct prime factors; variations with extra cyclic factors (ℤ/t) are also possible. These manifolds can be chosen stably parallelizable, extending the original Kreck‑Schafer examples (which dealt with k=2 and prime‑power groups).

The authors also treat non‑abelian groups. For G=Q₈×(ℤ/p)^3 with p≡1 (mod 8), they construct minimal 2‑complexes X and Y that are not homotopy equivalent, yet have the same Euler characteristic. Consequently, the doubles M(X) and M(Y) are stably diffeomorphic but not homotopy equivalent (Theorem D). This provides the first known example of a non‑abelian finite group without periodic cohomology for which such a phenomenon occurs, and it demonstrates that β_Q can be computed in non‑abelian settings.

The relationship between β_Q and the classical quadratic 2‑type Q(M)=