TheoremComplete

Papakyriakopoulos' Sphere Theorem and Loop Theorem

Theorem6.2Loop Theorem and Sphere Theorem

(Loop Theorem, Papakyriakopoulos 1957) Let MM be a 3-manifold with boundary component SS. If the induced map π1(S)π1(M)\pi_1(S) \to \pi_1(M) is not injective (i.e., there is a loop in SS that is null-homotopic in MM but not in SS), then there exists a properly embedded disk DMD \subset M with DS\partial D \subset S such that [D]0[\partial D] \neq 0 in π1(S)\pi_1(S).

(Sphere Theorem) If MM is a compact orientable 3-manifold with π2(M)0\pi_2(M) \neq 0, then there exists an embedded 2-sphere S2MS^2 \hookrightarrow M representing a nontrivial element of π2(M)\pi_2(M).


Proof (Key Ideas)

Proof

Proof of the Loop Theorem (following Papakyriakopoulos' tower construction, simplified by Stallings).

Step 1: Singular disk. The hypothesis gives a map f:(D2,D2)(M,S)f: (D^2, \partial D^2) \to (M, S) with fD2f|_{\partial D^2} non-null-homotopic in SS. The map ff may have self-intersections.

Step 2: Tower construction. Consider the singularities of ff (the set where ff is not injective). The idea is to "unstack" self-intersections by passing to covering spaces.

Define the tower: M0=MM_0 = M, and inductively, if the image of the disk in MiM_i has self-intersections, let M~i+1Mi\tilde{M}_{i+1} \to M_i be an appropriate covering space where the disk lifts with fewer self-intersections. Precisely: if fi:D2Mif_i: D^2 \to M_i has a self-intersection, there exists a loop in MiM_i along which fif_i intersects itself. Take the covering corresponding to the subgroup that kills this loop.

Step 3: Tower terminates. The tower M0M1M_0 \leftarrow M_1 \leftarrow \cdots terminates after finitely many steps because each covering increases the topology of the boundary (measured by its genus or complexity). Since we are working with compact manifolds, this cannot continue indefinitely.

Step 4: Embedded disk at the top. At the top of the tower MkM_k, the lifted disk is embedded (no more self-intersections to resolve). We have an embedded disk DkMkD_k \subset M_k with Dk\partial D_k non-trivial on Mk\partial M_k.

Step 5: Push down. Project DkD_k back down through the tower. At each stage, the projection may re-introduce intersections, but a cut-and-paste argument (surgery on the disk) allows us to find an embedded disk at each level while preserving the non-triviality of the boundary curve. After finitely many such operations, we obtain an embedded disk DMD \subset M with [D]0[\partial D] \neq 0 in π1(S)\pi_1(S).

Sphere Theorem. Similar tower argument: a non-trivial element of π2(M)\pi_2(M) is represented by a map f:S2Mf: S^2 \to M. The tower construction eliminates self-intersections, yielding an embedded S2S^2 representing a non-trivial class. An additional argument using equivariant techniques ensures the embedded sphere represents the original non-trivial element (not just some non-trivial element). \square


ExampleApplications to Knot Theory

Asphericity of knot complements: For a knot KS3K \subset S^3, the sphere theorem implies that π2(S3K)=0\pi_2(S^3\setminus K) = 0 (any embedded sphere in the complement bounds a ball, since S3KS^3\setminus K is irreducible). Thus knot complements are aspherical (K(π,1)K(\pi,1) spaces), and the knot group determines all the homotopy theory. Incompressibility of Seifert surfaces: The loop theorem implies that a minimal-genus Seifert surface is π1\pi_1-injective (incompressible) in the knot complement.

RemarkDehn's Lemma and the Disk Theorem

Dehn's lemma (proved by Papakyriakopoulos 1957, correcting Dehn's flawed 1910 proof) states: if f:D2M3f: D^2 \to M^3 is a map with fD2f|_{\partial D^2} an embedding and ff an embedding near D2\partial D^2, then fD2f|_{\partial D^2} extends to an embedding of D2D^2. This is closely related to the loop theorem and was historically the first of these results to be proved. Together, the loop theorem, sphere theorem, and Dehn's lemma are the foundational tools of classical 3-manifold topology, enabling the theory of incompressible surfaces and Haken manifolds.