[Paper Review] On Embedding Circle-Bundles in Four-Manifolds

Peter Ozsváth, Zoltán Szabó|ArXiv.org|Jun 26, 2000

Geometric and Algebraic Topology10 references3 citations

TL;DR

This paper establishes a topological obstruction to splitting complex surfaces of general type along twisted circle bundles over Riemann surfaces using Seiberg-Witten theory. It proves that if the Euler number |n| of the bundle satisfies |n| ≥ 2g−1, such a splitting is impossible when both pieces have positive b₂⁺, highlighting a sharp distinction from elliptic surfaces where such splittings do exist under similar conditions.

ABSTRACT

We demonstrate an obstruction to finding certain splittings of four-manifolds along sufficiently twisted circle bundles over Riemann surfaces, arising from Seiberg-Witten theory. These obstructions are used to show a non-splitting result for algebraic surfaces of general type.

Motivation & Objective

• To identify topological obstructions to splitting four-manifolds along circle bundles over Riemann surfaces using Seiberg-Witten invariants.
• To clarify the difference in splitting behavior between complex surfaces of general type and elliptic surfaces.
• To establish a sharp vanishing theorem for Seiberg-Witten invariants under specific geometric constraints.
• To demonstrate that minimal surfaces of general type cannot be decomposed along highly twisted circle bundles when both pieces have positive b₂⁺.
• To provide a framework linking equivariant Morse theory and Seiberg-Witten invariants via moduli space analysis.

Proposed method

• The authors use Seiberg-Witten invariants as homogeneous polynomial maps on the algebra A(X) = Z[U] ⊗ Λ*(H₁(X;Z)) to analyze the splitting of four-manifolds.
• They define the vanishing theorem (Theorem 2.1) which states that the sum of Seiberg-Witten invariants over each δH¹(Y;Z)-orbit vanishes when |n| ≥ 2g−1.
• The proof relies on analyzing the moduli space of solutions to the Seiberg-Witten equations, focusing on the ends of the flow to reducible configurations.
• The argument uses dimension counts to rule out certain configurations, particularly those with negative square classes in the cohomology of the boundary Y.
• The authors exploit the structure of basic classes on blow-ups of general type surfaces, showing that only two such classes exist: c₁ = −K_X and its conjugate.
• They apply the vanishing theorem to minimal surfaces of general type by showing that the orbit structure of basic classes under δH¹(Y;Z) leads to contradiction if a splitting exists.

Experimental results

Research questions

• RQ1 Under what conditions can a four-manifold be split along a circle bundle Y(n,g) with b₂⁺(X₁), b₂⁺(X₂) > 0?
• RQ2 Why do complex surfaces of general type fail to admit splittings along highly twisted circle bundles (|n| ≥ 2g−1), while elliptic surfaces do?
• RQ3 What role does the Seiberg-Witten invariant play in obstructing such splittings, and how does it differ from standard vanishing theorems?
• RQ4 How do the basic classes of blow-ups of general type surfaces behave under the action of δH¹(Y;Z) for twisted circle bundles?
• RQ5 Can the equivariant Floer homology framework be used to interpret the spectral sequence differential in the moduli space analysis?

Key findings

• Theorem 1.1 establishes that no complex surface of general type can be split along Y(n,g) with |n| ≥ 2g−1 and both b₂⁺(X₁), b₂⁺(X₂) > 0.
• The obstruction arises from a vanishing theorem (Theorem 2.1) on Seiberg-Witten invariants, which forces the sum of invariants over δH¹(Y;Z)-orbits to vanish.
• For minimal surfaces of general type, only two basic classes exist: c₁ = −K_X and its conjugate, and their orbit structure under δH¹(Y;Z) leads to contradiction if a splitting exists.
• The hypothesis |n| ≥ 2g−1 is sharp: when |n| ≤ 2g−2 or b₂⁺(X₂) = 0, such splittings do exist via blow-ups of embedded complex curves.
• Elliptic surfaces with b₂⁺ > 3 can be split along Y(1,1) and Y(n,1) for any n > 0, showing a fundamental difference in splitting behavior.
• The proof reveals that the difference between two basic classes in the same orbit must have square zero, which fails under the given constraints, contradicting the existence of such orbits.