Closed manifold surgery obstructions and the Oozing Conjecture

We complete the description of surgery obstructions up to homotopy equivalence for closed oriented manifolds with finite fundamental group. New examples are presented of non-trivial obstructions for Arf invariant product formulas in codimensions $\geq 4$, which give counterexamples to the well-known ‘‘Oozing Conjecture’’ from the 1980’s.

💡 Research Summary

The paper “Closed manifold surgery obstructions and the Oozing Conjecture” by I. Hamblen and Ö. Üngün provides a comprehensive analysis of surgery obstruction groups for closed oriented manifolds whose fundamental groups are finite, with a particular focus on 2‑groups. The authors first recall the surgery exact sequence and the role of the L‑theory assembly map A_π : Bπ₊∧L₀(ℤ) → L₀(ℤπ). While the Farrell–Jones conjecture gives powerful information for infinite groups, it offers no guidance for finite groups. By exploiting transfer to the 2‑Sylow subgroup and a detailed study of group homology, the authors completely determine the assembly map for finite groups up to homotopy equivalence (Theorem E).

Two families of universal homomorphisms are introduced: I_s^j : H^j(π;ℤ_(2)) → L_s^j(ℤπ) and κ_s^j, κ_h^j : H^j(π;ℤ/2) → L_s^{j+2}(ℤπ), L_h^{j+2}(ℤπ) respectively. The κ‑maps encode the Arf‑invariant product formulas that appear when one forms products of a degree‑one normal map from a Kervaire manifold K^{4n+2} with an arbitrary closed manifold M^k. The classical “Oozing Conjecture” (circa 1980) asserted that all co‑dimension‑k Arf invariants vanish for k ≥ 4, i.e. κ_h^j = 0 for j ≥ 4. The authors show that this conjecture is false in general.

The paper proceeds with a systematic study of these homomorphisms for abelian and “basic” 2‑groups (cyclic, dihedral, quaternion, semi‑dihedral). Theorem A establishes that I_s^j = 0 for j > 0, κ_s^0 and κ_s^1 are split injective, κ_s^2 vanishes for the basic groups, κ_s^j = 0 for j > 2 unless π is quaternion, and for quaternion groups κ_s^3 is injective while κ_s^j = 0 for j ≥ 4. Moreover, the kernel of κ_s^2 coincides with the kernel of κ_h^2, and their images are isomorphic.

Next, the authors examine the 2‑adic reduction ρ:ℤ→ℤ̂₂ and the induced maps \bar κ_s^j : H^j(π;ℤ/2) → L_s^{j+2}(ℤ̂₂π). Theorem B shows that \bar κ_s^0 and \bar κ_s^1 are split injective, \bar κ_s^j is non‑zero only when j = 2^ℓ (ℓ ≥ 1), and \bar κ_s^{2r+2} = \bar κ_s^4 ∘ s_r where s_r is the dual of the iterated squaring operation in cohomology. This yields a precise description of the image of κ′_j (the weakly simple version) in terms of the 2‑adic L‑groups and the abelianization of π.

A key technical advance is the identification of the obstruction κ′_4 with a purely group‑homological map λ₄(π) = δ ∘ β ∘ s^: H⁴(π;ℤ/2) → H¹(Wh′(ℤ̂₂π)). Here β is induced by the Bockstein, s^ is the dual of the Steenrod square, and δ is the coboundary from the short exact sequence 0 → Wh′(ℤ̂₂π) → I(ℤ̂₂π) → π_ab → 0. Theorem E proves that κ′_4(π) = 0 iff λ₄(π) = 0, and that κ′_4 vanishes on the image of integral homology. Using extensive GAP computations, the authors exhibit a 2‑group of order 16 384 for which λ₄ is non‑zero, thereby producing a concrete counterexample to the Oozing Conjecture: κ′_4 ≠ 0, and consequently κ_s^4 ≠ 0. This also shows that κ_h^4 can be non‑zero while κ_s^4 may vanish, answering a question raised by Kasprowski, Nicholson, and Véselá.

Further, Theorem C and Corollary D give conditions under which κ_s^2 is non‑zero while κ′_2 = 0, demonstrating that the kernels of κ_s^2 and κ_h^2 can differ. Explicit examples include the non‑abelian 2‑group SG