Sharp effective bounds for Kneser-Haken finiteness in bounded 3-manifolds and the handlebody threshold of the Freedman-Freedman periodic-curve example

Benedict and Michael Freedman (Topology, 1998) proved that for every compact 3-manifold $M$ with boundary and every integer $b>0$ there is a constant $c(M,b)$ such that any family of more than $c(M,b)$ disjoint, pairwise nonparallel, incompressible surfaces of first Betti number less than $b$, none a boundary-parallel annulus or disk, contains a parallel pair. Their proof yields only a crude bound, exponential in $b$ (of order $3^{2b-2}$), and they leave open how sharp this is; the same paper analyzes a concrete degree-one curve $\gamma$ in the solid torus whose periodic complement has stacks $X^n$ that are handlebodies for $n=1,2,3,4$, while their general argument only forces failure for some very large $n$ ("Presumably $X^n$, $n\ge 5$, are not handlebodies"). The linked open problem is to determine the optimal growth and tight value of the finiteness constant $c(M,b)$ and to settle the exact handlebody threshold of the stacks $X^n$.

Background

Classical Kneser-Haken finiteness bounds how many essentially different surfaces can sit disjointly in a 3-manifold. A properly embedded, two-sided surface $F$ in a compact 3-manifold $M$ is incompressible if it is not a 2-sphere and the inclusion $F\hookrightarrow M$ induces an injection on fundamental groups; it is boundary-incompressible if it admits no boundary compression, i.e. no embedded disk meeting $F$ in a single essential arc and $\partial M$ in the complementary arc. Two disjoint surfaces are parallel if they cobound a product region homeomorphic to $F\times[0,1]$. Kneser's theorem for 2-spheres (the engine of prime decomposition) and Haken's extension to incompressible surfaces state that for each compact, irreducible $M$ there is an integer $c(M)$ such that any family of more than $c(M)$ disjoint, pairwise nonparallel, closed incompressible surfaces contains a parallel pair. Classical proofs bound $c(M)$ linearly in the number of tetrahedra $t$ of a triangulation, with surveyed values ranging from $61t$ down to $5t$, and for closed $M$ Bachman (2004) proved $c(M)\le 2t$.

For manifolds with boundary the naive extension to surfaces that may be boundary-compressible is false: as shown by Howards (Topology and its Applications 99 (1999) 117-122), a 3-manifold whose boundary has a component of genus at least two contains arbitrarily large families of disjoint, pairwise nonparallel, incompressible, non-boundary-parallel surfaces, and in such families the complexity of the surfaces, measured by the first Betti number $b_1=\dim_{\mathbb{Q}}H_1(\cdot;\mathbb{Q})$, grows without bound. Benedict and Michael Freedman (Topology 37 (1998) 133-147) proved that with the Betti number bounded, finiteness is restored: for every compact 3-manifold $M$ with boundary and every integer $b>0$ there is a constant $c(M,b)$ such that any family of more than $c(M,b)$ disjoint, pairwise nonparallel, incompressible surfaces $F_1,\dots,F_k$ with $b_1(F_i)<b$, none of which is a boundary-parallel annulus or disk, contains a parallel pair. The constant produced by their proof is exponential in the Betti bound: for a compression body $V$ (a manifold obtained from $\partial_-V\times I$ for a closed surface $\partial_-V$ by attaching 1-handles) their estimate reads

$c(V,b)=\frac{14}{3}\,c_0(V)\,3^{2b-2}+5\,|\chi(V)|+T+b_1(V;\mathbb{Z}_2)$

where $c_0$ is the classical Kneser-Haken constant and $T$ counts torus boundary components; the authors themselves flag the estimate as crude and report that R. Stong obtained smaller bounds in specific examples.

Two structural theorems in the same paper consume this finiteness result and motivate the open quantitative questions. First, for an irreducible, noncompact, boundaryless 3-manifold $N$, the following are equivalent: $N$ contains no closed incompressible surface; $N$ admits an exhaustion by compact handlebodies (regular neighborhoods of embedded graphs) $H_1\subset\mathrm{int}\,H_2\subset\cdots$ with $N=\bigcup H_i$; and $\pi_1(N)$ is locally free, meaning every finitely generated subgroup is a free group. Second, if $M$ is compact with all boundary components tori or Klein bottles and $\theta\in H^1(M;\mathbb{Z})$ is nonzero on every boundary component, then the infinite cyclic cover $M^\theta$ (the cover corresponding to the homomorphism $\pi_1(M)\to\mathbb{Z}$ classifying $\theta$) is either a product $F\times\mathbb{R}$ (the fibered case) or contains a closed incompressible surface. As a corollary, for a curve $\gamma$ in the solid torus $S^1\times D^2$ of homological degree one (winding number one), with preimage $\tilde\gamma$ in the universal cover $\mathbb{R}\times D^2$: either $\gamma$ is isotopic to the core circle, in which case $\pi_1(\mathbb{R}\times D^2\setminus\tilde\gamma)\cong\mathbb{Z}$; or the complement $\mathbb{R}\times D^2\setminus\tilde\gamma$ contains a closed incompressible surface and its group is not even locally free. Section 1 of the paper analyzes a concrete non-core degree-one curve and the stacks $X^n$ obtained as unions of $n$ consecutive fundamental domains of the $\mathbb{Z}$-periodic complement: the authors report that $X^n$ is a handlebody for $n=1,2,3,4$, that their general argument only forces some very large uniform $n$ to fail, and they write "Presumably $X^n$, $n\ge 5$, are not handlebodies." The paper's motivation connects these questions to Kleinian groups (discrete subgroups of $\mathrm{PSL}(2,\mathbb{C})$ acting on hyperbolic 3-space) through Marden's tameness conjecture (every complete hyperbolic 3-manifold with finitely generated fundamental group is tame, i.e. homeomorphic to the interior of a compact 3-manifold) and the Ahlfors measure conjecture (the limit set of a finitely generated Kleinian group, i.e. the set of accumulation points of its action on the 2-sphere, has Lebesgue measure zero), both open in 1998 and both settled in 2004-2006.

Problem Statement

The problem is to make the Freedman-Freedman finiteness theorem for bounded 3-manifolds quantitatively sharp, and to settle its concrete test case. Both deliverables were explicitly left open by the source paper, and they are linked: the source's general argument bounds the test case's failure threshold only through the crude finiteness constant.

(1) Optimal finiteness constant. For a compact 3-manifold $M$ with boundary and an integer $b>0$, let $k(M,b)$ denote the maximal cardinality of a family of disjoint, pairwise nonparallel, properly embedded, incompressible (possibly boundary-compressible) surfaces in $M$, each of first Betti number less than $b$, none of which is a boundary-parallel annulus or disk. The goal is to determine $k(M,b)$. It has three parts, each an independent research question:

• (1a) Growth. Is the exponential dependence on $b$ of the published upper bound (of order $\frac{14}{3}c_0(M)3^{2b-2}$ plus additive terms) necessary, or does $k(M,b)$ grow subexponentially, even polynomially or linearly, in $b$ for fixed $M$?

• (1b) Upper bounds. A new upper bound on $k(M,b)$ strictly sharper than the source's published estimate quoted in the Background, with the dependence on $M$ (triangulation complexity, Heegaard genus — the minimal genus of a surface splitting $M$ into two compression bodies — or $c_0(M)$) made explicit; a bound matching the best lower bounds is desirable but not required.

• (1c) Lower-bound families. Explicit lower-bound families: concrete $M$ and $b$, together with disjoint, pairwise nonparallel, incompressible surfaces in $M$ of first Betti number less than $b$, none of which is a boundary-parallel annulus or disk, and of as large cardinality as possible.

A determination of $k(M,b)$ for all $M$ and $b$ would settle (1a)-(1c) simultaneously.

(2) The Section-1 periodic example. For the specific non-core degree-one curve $\gamma\subset S^1\times D^2$ analyzed in Section 1 of the source paper, with $\tilde\gamma$ its preimage in $\mathbb{R}\times D^2$ and $X^n$ the union of $n$ consecutive fundamental domains of the $\mathbb{Z}$-periodic complement $\mathbb{R}\times D^2\setminus\tilde\gamma$: determine exactly for which $n$ the stack $X^n$ is a handlebody. In particular, prove or disprove the authors' assertion-in-expectation that $X^n$, $n\ge 5$, are not handlebodies — that is, determine the exact set of integers $n$ for which $X^n$ is a handlebody (it is a handlebody for $n=1,2,3,4$).

Answers must preserve the exact hypotheses of the source theorem: surfaces are incompressible but may be boundary-compressible; boundary-parallel annuli and disks are excluded; the topology of the surfaces is controlled only through $b_1(F)<b$, since the statement is false without it. The broader Kleinian-group program recorded in the same paper is not part of this problem, as its target conjectures are now theorems.

References

1. Benedict Freedman and Michael H. Freedman, "Kneser-Haken finiteness for bounded 3-manifolds, locally free groups, and cyclic covers", Topology 37 (1998) 133-147, DOI 10.1016/S0040-9383(97)00007-4, https://doi.org/10.1016/S0040-9383(97)00007-4
2. Mario Eudave-Muñoz and Jeremy Shor, "A universal bound for surfaces in 3-manifolds with a given Heegaard genus", Algebraic & Geometric Topology 1 (2001) 31-37, DOI 10.2140/agt.2001.1.31, arXiv:math/0008158, https://arxiv.org/abs/math/0008158
3. David Bachman, "A note on Kneser-Haken finiteness", Proceedings of the American Mathematical Society 132 (2004) 899-902, DOI 10.1090/S0002-9939-03-07049-7, arXiv:math/0210215, https://arxiv.org/abs/math/0210215
4. Hugh N. Howards, "Limits of incompressible surfaces", Topology and its Applications 99 (1999) 117-122, DOI 10.1016/S0166-8641(98)00083-2, arXiv:math/9806101, https://arxiv.org/abs/math/9806101
5. Hugh N. Howards, "Generating disjoint incompressible surfaces", Topology and its Applications 158 (2011) 325-343, DOI 10.1016/j.topol.2010.11.008, https://doi.org/10.1016/j.topol.2010.11.008
6. Ryan Blair, John Burke, and Robin Koytcheff, "A prime decomposition theorem for the 2-string link monoid", Journal of Knot Theory and Its Ramifications 24 (2015) 1550005, DOI 10.1142/S0218216515500054, arXiv:1308.1594, https://arxiv.org/abs/1308.1594
7. Vladimir Tarkaev, "A prime decomposition theorem for string links in a thickened surface", Journal of Knot Theory and Its Ramifications 33 (2024), no. 13, 2450041, DOI 10.1142/S021821652450041X, arXiv:2403.12492, https://arxiv.org/abs/2403.12492
8. Rachel Lehman and Yo'av Rieck, "A structure theorem for bad 3-orbifolds", European Journal of Mathematics 8 (2022), DOI 10.1007/s40879-022-00546-4, arXiv:2202.00208, https://arxiv.org/abs/2202.00208
9. Ian Agol, "Tameness of hyperbolic 3-manifolds", arXiv:math/0405568 (2004), https://arxiv.org/abs/math/0405568
10. Danny Calegari and David Gabai, "Shrinkwrapping and the taming of hyperbolic 3-manifolds", Journal of the American Mathematical Society 19 (2006) 385-446, DOI 10.1090/S0894-0347-05-00513-8, arXiv:math/0407161, https://arxiv.org/abs/math/0407161
11. Jessica S. Purcell and Anastasiia Tsvietkova, "Polynomial bounds for surfaces in cusped 3-manifolds", arXiv:2311.08567 (2023, revised 2025), https://arxiv.org/abs/2311.08567
12. James W. Anderson, "Commensurability and locally free Kleinian groups", arXiv:math/9903138 (1999), https://arxiv.org/abs/math/9903138
13. Autumn E. Kent, "Bundles, Handcuffs, and Local Freedom", Geometriae Dedicata 106 (2004) 145-159, DOI 10.1023/B:GEOM.0000033837.26844.D7, https://doi.org/10.1023/B:GEOM.0000033837.26844.D7