betaFeedback

WorksMathematics

Nonhomeomorphic closed aspherical four-manifolds with the same homotopy type

Abstract

We construct closed connected aspherical topological four-manifolds that are homotopy equivalent but not homeomorphic, disproving the homeomorphism-existence formulation of the Borel conjecture in dimension four. Their common fundamental group is word-hyperbolic. One of the manifolds also has a self-homotopy equivalence not homotopic to a homeomorphism.

Introduction

A connected manifold is aspherical if its universal cover is contractible. Its fundamental group then determines its homotopy type. The topological Borel conjecture commonly asks whether every homotopy equivalence between closed aspherical manifolds is homotopic to a homeomorphism [2, 20]. The weaker homeomorphism-existence formulation asks only whether homotopy-equivalent closed aspherical manifolds are homeomorphic. Our main result disproves this formulation in dimension four.

Theorem 1.1. There exist closed connected aspherical topological four-manifolds MM and NN that are homotopy equivalent but not homeomorphic. Their common fundamental group is word-hyperbolic.

The conclusion excludes every homeomorphism between the two manifolds, without prescribing a homotopy class. The construction also yields a separate obstruction for a self-map of one manifold.

Theorem 1.2. There exist a closed connected aspherical topological four-manifold MM and a homotopy equivalence f:M→Mf : M \to M such that no homeomorphism M→MM \to M is homotopic to ff.

We prove this assertion separately and use the same MM in Theorem 1.1. A nonhomeomorphic pair alone would not imply the self-map assertion.

The homeomorphism-existence question already appears in Borel’s letter to Serre of May 2, 1953, phrased for compact manifolds that are classifying spaces [5]. Rigidity holds classically in dimensions at most two, and for closed orientable aspherical three-manifolds by geometrization and three-dimensional rigidity [19], Theorem 0.7 and Section 5. In higher dimensions, Hsiang and Wall proved the homeomorphism-existence statement for homotopy tori in dimensions at least five [17], and Farrell and Jones established it for closed negatively curved manifolds in those dimensions [11]. Bartels and Lück proved the prescribed-class formulation in dimensions at least five for a class of groups containing word-hyperbolic groups and groups acting geometrically on finite-dimensional CAT(0)CAT(0) spaces [2], Theorem A.

In dimension four, these rigidity theorems use further geometric input. Bartels and Lück’s four-dimensional application assumes that the fundamental group is good in the sense of Freedman [2], Proposition 0.3. For closed aspherical topological four-manifolds with torsion-free word-hyperbolic fundamental group, Khan proves rigidity up to topological ss-cobordism [18], Theorem 1.22 and Corollary 1.23; the product theorem for five-dimensional topological ss-cobordisms in Freedman–Quinn assumes a good fundamental group [12], Theorem 7.1A. The word-hyperbolicity in Theorem 1.1 therefore lies within the group-theoretic range of these rigidity results, while the dimension is essential to their topological conclusion. Recent examples of Davis, Hayden, Huang, Ruberman, and Sunukjian are closed aspherical smooth four-manifolds that are homeomorphic but not diffeomorphic [10], Theorem 1.1. Their proof uses Davis reflection and a characteristic-subgroup argument to force a hypothetical diffeomorphism between these manifolds to lift to covers distinguished by the genera of smoothly embedded homologically essential surfaces [10]. Our conclusion concerns homeomorphism in the topological category.

The proof uses two companion constructions. The marked tensor obstruction of [MT] supplies plumbing blocks, marked grope bodies, and algebraic potentials associated to three-dimensional cuts. The reflection and realization results of [PD] turn the absence of a marked filling into the absence of a closed manifold model for an aspherical Poincaré complex. We also use its universal-development and support arguments. We state the imported results at their points of use and verify their hypotheses for the chambers constructed here. The principal inputs are the marked algebraicity theorem [MT], Theorem 5.1, the controlled filling implication [PD], Sections 3–5, and the support construction [PD], Lemma 3.3.

The construction

Start with a finite collared Poincaré chamber (P,S)(P,S) of dimension four, where PP has the homotopy type of a finite graph and its fundamental group has an epimorphism to Z\mathbb{Z}. Write SdS_d for the boundary in the connected cyclic cover of degree dd. A marked filling of SdS_d is a compact four-manifold with this identified boundary and the prescribed graph homotopy type, including the boundary map. The double cover S2S_2 has a marked filling, whereas SdS_d has none for odd d≥3d \ge3.

This parity comes from two colors of plumbing caps. Each grope has four terminal caps grouped into two red–blue pairs. In the double cover, choosing red caps at one vertex and blue caps at the other makes the selected sheets disjoint and gives the filling. For odd d≥3d \ge3, a hypothetical filling of SdS_d instead produces formal data on a cycle of dd vertices with five red and five blue edges between each neighboring pair. The endpoint-assignment theorem would assign every edge to one endpoint so that each vertex receives at most one edge in some color. Choose one such color at each vertex. Adjacent vertices cannot choose the same color, since together they would receive at most two of the five edges of that color between them. The choices would alternate around the cycle, impossible when the cycle is odd.

We formulate the endpoint-assignment theorem for arbitrary finite loopless multigraphs. The cyclic-cover criterion works in every manifold dimension. It converts a double-cover model and eventual nonexistence of odd-cover models into a self-homotopy equivalence not represented by any homeomorphism. The proof composes a lifted homeomorphism with a large deck translation and imposes no finite-order condition on that homeomorphism. This last distinction separates the criterion from finite-group Nielsen realization obstructions such as [4], Theorem 1.6.

To construct the pair, use the reflection method of Davis [8, 9] in the Poincaré-chamber form of [PD]. A full reflection of PP has no manifold model in its cyclic covers of odd degree d≥3d \ge3. We quotient its elementary abelian group of chamber labels by a binary relation space that acts freely and forces every simplicial symmetry preserving the relations to be trivial. The quotient is a finite aspherical complex QQ with the same odd-cover obstruction.

There are two ways to replace the chambers of Q2Q_2 by the filling of S2S_2. In one, every boundary identification is the same. In the other, it is twisted by the double-cover deck involution according to a nonzero linear functional ℓ\ell on the label group A. Both gluings are closed manifold models of Q2Q_2. They realize two subgroups of outer automorphisms of the common group Π\Pi: the horizontal subgroup A on MM and the graph subgroup BℓB_\ell on NN. These lie in A ×⟨δ⟩≤Out⁡(Π)\times\langle\delta\rangle\leq\operatorname{Out}(\Pi), where δ\delta is the cyclic involution. The cyclic-cover criterion forbids δ\delta on MM and at the same time proves Theorem 1.2. The subgroups A and BℓB_\ell are not conjugate in Out⁡(Π)\operatorname{Out}(\Pi): they contain different numbers of elements admitting finite-order lifts to Aut⁡(Π)\operatorname{Aut}(\Pi).

The remaining step controls all possible homotopy classes of a homeomorphism between the models. We prove that Out⁡(Π)\operatorname{Out}(\Pi) is finite: a uniform bound on square grids makes the support cubulation hyperbolic, and Poincaré duality with the homological Shapiro lemma excludes the virtually cyclic splittings forced by an infinite outer automorphism group. A direct rigidity argument for the right-angled Coxeter group of the infinite lifted triangulation then gives

NOut⁡(Π)(A)=A×⟨δ⟩.N_{\operatorname{Out}(\Pi)}(\mathrm{A}) = \mathrm{A} \times\langle\delta\rangle.

The normalizer equality and the absence of δ\delta make A a Sylow 2-subgroup of the outer automorphisms realized on MM. A homeomorphism from MM to NN would transport BℓB_\ell into that realized group and force the forbidden conjugacy. This proves Theorem 1.1.

Section 2 gives the two realization criteria. Section 3 states the chamber data and derives the reflected odd-cover obstruction. Sections 4 and 5 construct the two models and identify their outer actions. Sections 6 and 7 prove the group-theoretic statements and the main theorem. Finally, Sections 8–10 prove the chamber data, including the general tensor theorem and the odd-cover obstruction.

All manifolds are connected unless otherwise indicated. Boundaries have collars, and cornered constructions are rounded. Homotopies of markings are part of the data when boundary maps are prescribed. We use singular cohomology unless a differential-form model is specified, and write ΓjF\Gamma_j F for the lower central series of a group FF.

Cyclic covers and realized symmetries

Two criteria organize the passage from a filling obstruction to the homeomorphism conclusions. The first forbids a prescribed outer automorphism from being induced by a homeomorphism. The second uses finite symmetry groups to distinguish two manifold models without prescribing the homotopy class of a possible homeomorphism.

A criterion from cyclic covers

We first isolate the passage from manifold existence in finite covers to a homotopy equivalence that cannot be represented by a homeomorphism. This passage is independent of the four-dimensional construction. A closed topological nn-manifold model of a space is a closed, connected topological nn-manifold equipped with a homotopy equivalence to that space.

Theorem 2.1 (Cyclic-cover criterion). Let QQ be a connected finite aspherical CW complex, put G=π1(Q)G=\pi_1(Q), and let w ⁣:G→Zw\colon G\to\mathbb{Z} be an epimorphism. For every positive integer dd, write

Gd=w−1(dZ)andQd⟶QG_d = w^{-1}(d\mathbb{Z}) \qquad\text{and}\qquad Q_d \longrightarrow Q

for the associated connected cyclic cover. Fix an integer n≥1n\geq1. Suppose that Q2Q_2 has a closed topological nn-manifold model MM, but that QdQ_d has no such model for every sufficiently large odd integer $d." Choose γ∈G\gamma\in G with w(γ)=1w(\gamma)=1, and use the model equivalence to identify π1(M)\pi_1(M) with G2G_2. The outer automorphism of G2G_2 represented by

α(g)=γgγ−1\alpha(g)=\gamma g\gamma^{-1}

is induced by a self-homotopy equivalence of MM. No homeomorphism of MM induces this outer automorphism. In particular, that self-homotopy equivalence is not homotopic to a homeomorphism.

Proof. The manifold MM is aspherical, since it is homotopy equivalent to Q2Q_2. Topological manifolds have CW homotopy type [21]. The classification of maps between Eilenberg–Mac Lane spaces therefore realizes α\alpha by a map f:M→Mf:M\to M, and realizes its inverse by a homotopy inverse to ff; see [16], Proposition 1B.9.

Suppose that a homeomorphism h:M→Mh:M\to M induces the outer class of α\alpha. We will construct closed manifold models for all sufficiently large odd-degree covers of QQ. Put K=ker⁡wK=\ker w, and let

X=K\M~X=K\backslash\widetilde{M}

be the infinite cyclic cover of MM, where M~\widetilde{M} is its universal cover. We use left deck actions throughout. The element γ2∈G2\gamma^2\in G_2 induces a deck transformation TT of XX, and TT generates its deck group over MM.

Choose a lift u:M~→M~u:\widetilde{M}\to\widetilde{M} of hh that intertwines the deck action by exactly α\alpha:

ugu−1=γgγ−1(g∈G2).ugu^{-1}=\gamma g\gamma^{-1}\qquad(g\in G_2).

Indeed, any lift induces an automorphism in the prescribed outer class. Composing that lift with a deck transformation removes the inner discrepancy. Since α(K)=K\alpha(K)=K, the map uu descends to a homeomorphism h′:X→Xh':X\to X. Moreover, α(γ2)=γ2\alpha(\gamma^2)=\gamma^2, so (2.1) gives

h′T=Th′.h'T=Th'.

We next show that a sufficiently large translation composed with h′h' acts freely and cocompactly on XX. Represent the homomorphism w/2:G2→Zw/2:G_2\to\mathbb{Z} by a map M→S1M\to S^1 and lift that map to obtain a continuous height function

τ:X⟶R,τ(Tx)=τ(x)+1.\tau:X\longrightarrow\mathbb{R},\qquad\tau(Tx)=\tau(x)+1.

This height is proper. To see this, choose a compact set C⊂XC\subset X whose TT-translates cover XX, using compactness of MM, and choose a,b∈Ra,b\in\mathbb{R} with τ(C)⊂[a,b]\tau(C)\subset[a,b]. A bounded closed height band τ−1([A,B])\tau^{-1}([A,B]) is a closed subset of

⋃A−b≤j≤B−aTjC,\bigcup_{A-b\leq j\leq B-a}T^jC,

which is a finite union of compact sets.

By (2.2), the continuous function x↦τ(h′x)−τ(x)x\mapsto\tau(h'x)-\tau(x) is TT-invariant. It is therefore bounded, say in absolute value by c≥0c\geq0. For an integer r>c+1r>c+1, put

Fr=Trh′.F_r=T^rh'.

For every x∈Xx\in X we have

0<r−c≤τ(Frx)−τ(x)≤r+c.0<r-c\leq\tau(F_rx)-\tau(x)\leq r+c.

Consequently no nonzero power of FrF_r fixes a point. Iterating the lower bound shows that only finitely many translates of any compact set can meet that set; thus the Z\mathbb{Z}-action generated by FrF_r is properly discontinuous. Finally, along every orbit the height tends to +∞+\infty in positive time and to −∞-\infty in negative time. The first nonnegative height on such an orbit belongs to [0,r+c][0,r+c], by the upper bound in (2.3). The compact band τ−1([0,r+c])\tau^{-1}([0,r+c]) therefore meets every orbit. This proves cocompactness.

It follows that

Yr=X/⟨Fr⟩Y_r=X/\langle F_r\rangle

is a closed connected topological nn-manifold. Its universal cover is M~\widetilde{M}, so YrY_r is aspherical. To identify its fundamental group, lift FrF_r to

F~r=γ2ruon M~.\widetilde{F}_r=\gamma^{2r}u\quad\text{on }\widetilde{M}.

Equation (2.1) gives

F~rkF~r−1=γ2r+1kγ−(2r+1)(k∈K).(1)\widetilde{F}_r k\widetilde{F}_r^{-1}=\gamma^{2r+1}k\gamma^{-(2r+1)}\quad(k\in K). \tag*{(1)}

The deck group over YrY_r is generated by KK and F~r\widetilde{F}_r. Its quotient by KK is infinite cyclic, since its action on XX is generated freely by FrF_r. Hence (1) identifies it with

π1(Yr)≅K⋊Ad⁡γ2r+1Z≅G2r+1.\pi_1(Y_r)\cong K\rtimes_{\operatorname{Ad}_{\gamma^{2r+1}}}\mathbb{Z}\cong G_{2r+1}.

For the second isomorphism, send the generator of Z\mathbb{Z} to γ2r+1\gamma^{2r+1} and act identically on KK. This is bijective because every element of G2r+1G_{2r+1} has a unique expression kγ(2r+1)jk\gamma^{(2r+1)j} with k∈Kk\in K and j∈Zj\in\mathbb{Z}.

Both YrY_r and Q2r+1Q_{2r+1} are Eilenberg–Mac Lane spaces, so this group isomorphism is realized by a homotopy equivalence. Taking rr sufficiently large contradicts the assumed absence of odd-degree manifold models.

Remark 2.2. The equivalence ff in Theorem 2.1 satisfies f2≃id⁡Mf^2\simeq\operatorname{id}_M. Indeed, α2=Ad⁡γ2\alpha^2=\operatorname{Ad}_{\gamma^2} is inner because γ2∈G2\gamma^2\in G_2, and unbased homotopy classes of maps between aspherical CW-type spaces are classified by homomorphisms modulo inner automorphisms of the target. The proof nevertheless imposes no order condition on a hypothetical realizing homeomorphism. Its only geometric input is the bounded height displacement of a lift commuting with the deck translation.

A criterion for distinct manifold models

Let XX be a connected topological manifold and choose an isomorphism μX:π1(X)→≅Π\mu_X:\pi_1(X)\xrightarrow{\cong}\Pi to a group Π\Pi. Basepoint paths affect an induced automorphism only by an inner automorphism, so this marking defines a subgroup

RX=im⁡(Homeo⁡(X)⟶Out⁡(Π)).R_X=\operatorname{im}(\operatorname{Homeo}(X)\longrightarrow\operatorname{Out}(\Pi)).

Changing the marking conjugates RXR_X in Out⁡(Π)\operatorname{Out}(\Pi). More generally, if XX and YY are so marked and f:X→Yf:X\to Y is a homeomorphism, the outer isomorphism ρ\rho defined by μYf∗μX−1\mu_Yf_*\mu_X^{-1} satisfies

RY=ρRXρ−1.R_Y=\rho R_X\rho^{-1}.

This follows by transporting self-homeomorphisms through ff.

References

  1. [1]M. Artin, Algebraic approximation of structures over complete local rings, Publ. Math. Inst. Hautes Études Sci. 36 (1969), 23–58. doi:10.1007/BF02684596.DOI
  2. [2]A. Bartels and W. Lück, The Borel conjecture for hyperbolic and CAT(0)-groups, Ann. of Math. (2) 175 (2012), no. 2, 631–689. doi:10.4007/annals.2012.175.2.5.DOI
  3. [3]M. Bestvina and M. Feighn, Stable actions of groups on real trees, Invent. Math. 121 (1995), 287–321. doi:10.1007/BF01884300.DOI
  4. [4]J. Block and S. Weinberger, On the generalized Nielsen realization problem, Comment. Math. Helv. 83 (2008), no. 1, 21–33. doi:10.4171/CMH/116.DOI
  5. [5]A. Borel, Letter to Jean-Pierre Serre, May 2, 1953, reproduced with commentary by A. Ranicki in The birth of the Borel conjecture, September 22, 2016. https://www.maths.ed.ac.uk/~v1ranick/surgery/borel.pdf.
  6. [6]M. R. Bridson and A. Haefliger, Metric spaces of non-positive curvature, Grundlehren der mathematischen Wissenschaften, vol. 319, Springer-Verlag, Berlin, 1999. doi:10.1007/978-3-662-12449-9.DOI
  7. [7]V. Chepoi, Graphs of some CAT(0) complexes, Adv. Appl. Math. 24 (2000), no. 2, 125–179. doi:10.1006/aama.1999.0677.DOI
  8. [8]M. W. Davis, Groups generated by reflections and aspherical manifolds not covered by Euclidean space, Ann. of Math. (2) 117 (1983), no. 2, 293–324. doi:10.2307/2007079.DOI
  9. [9]M. W. Davis, The geometry and topology of Coxeter groups, London Mathematical Society Monographs Series, vol. 32, Princeton University Press, Princeton, NJ, 2008.
  10. [10]M. Davis, K. Hayden, J. Huang, D. Ruberman, and N. Sunukjian, Exotic aspherical 4-manifolds, Ann. of Math. (2) 204 (2026), no. 2, 777–794. doi:10.4007/annals.2026.204.2.6.DOI
  11. [11]F. T. Farrell and L. E. Jones, Compact negatively curved manifolds (of dim ≠ 3, 4) are topologically rigid, Proc. Natl. Acad. Sci. USA 86 (1989), no. 10, 3461–3463. doi:10.1073/pnas.86.10.3461.DOI
  12. [12]M. H. Freedman and F. Quinn, Topology of 4-manifolds, Princeton Mathematical Series, vol. 39, Princeton University Press, Princeton, NJ, 1990.
  13. [13]M. Gutierrez and A. Kaul, Automorphisms of right-angled Coxeter groups, Int. J. Math. Math. Sci. 2008 (2008), Article ID 976390, 10 pp. doi:10.1155/2008/976390.DOI
  14. [14]M. Gutierrez, A. Piggott, and K. Ruane, On the automorphisms of a graph product of abelian groups, Groups Geom. Dyn. 6 (2012), no. 1, 125–153. doi:10.4171/GGD/153.DOI
  15. [15]A. J. S. Hamilton, The triangulation of 3-manifolds, Quart. J. Math. Oxford Ser. (2) 27 (1976), no. 1, 63–70. doi:10.1093/qmath/27.1.63.DOI
  16. [16]A. Hatcher, Algebraic topology, Cambridge University Press, Cambridge, 2002.
  17. [17]W.-C. Hsiang and C. T. C. Wall, On homotopy tori II, Bull. London Math. Soc. 1 (1969), no. 3, 341–342. doi:10.1112/blms/1.3.341.DOI
  18. [18]Q. Khan, Homotopy invariance of 4-manifold decompositions: Connected sums, Topology Appl. 159 (2012), no. 16, 3432–3444. doi:10.1016/j.topol.2012.08.005.DOI
  19. [19]M. Kreck and W. Lück, Topological rigidity for non-aspherical manifolds, Pure Appl. Math. Q. 5 (2009), no. 3, 873–914. https://arxiv.org/abs/math/0509238.
  20. [20]W. Lück, Survey on aspherical manifolds, in European Congress of Mathematics, European Mathematical Society, Zürich, 2010, 53–82. doi:10.4171/077-1/4.DOI
  21. [21]J. Milnor, On spaces having the homotopy type of a CW-complex, Trans. Amer. Math. Soc. 90 (1959), no. 2, 272–280. doi:10.1090/S0002-9947-1959-0100267-4.DOI
  22. [22]J. Paixão and J. Spreer, Random collapsibility and 3-sphere recognition, arXiv:1509.07607v1 [math.GT], September 25, 2015.DOI
  23. [23]F. Paulin, Outer automorphisms of hyperbolic groups and small actions on R-trees, in Arboreal group theory (Berkeley, CA, 1988), R. C. Alperin (ed.), Mathematical Sciences Research Institute Publications, vol. 19, Springer-Verlag, New York, 1991, 331–343. doi:10.1007/978-1-4612-3142-4_12.DOI
  24. [24]P. Przytycki and J. Świątkowski, Flag-no-square triangulations and Gromov boundaries in dimension 3, Groups Geom. Dyn. 3 (2009), no. 3, 453–468. doi:10.4171/GGD/66.DOI
  25. [25]E. Spanier, Singular homology and cohomology with local coefficients and duality for manifolds, Pacific J. Math. 160 (1993), no. 1, 165–200. doi:10.2140/pjm.1993.160.165.DOI
  26. [26]J. Stallings, Group theory and three-dimensional manifolds, Yale University Press, New Haven and London, 1971.
  27. [27]C. A. Weibel, An introduction to homological algebra, Cambridge Studies in Advanced Mathematics, vol. 38, Cambridge University Press, Cambridge, 1994.DOI
  28. [28][MT] OpenAI, A marked tensor obstruction to four-dimensional disk embedding, OpenAI Math Release preprint OAI:A-marked-tensor-obstruction-to-four-dimensional-disk-embedding-September-24-2026, 2026.
  29. [29][PD] OpenAI, A PD₄ group without an aspherical manifold model, OpenAI Math Release preprint OAI:A-PD4-group-without-an-aspherical-manifold-model-September-24-2026, 2026.

Paper details

vlnc:W_zhv8cdq6
Version
v1 · latest
Published
License
LicenseRef-OpenAI-public
Usage

views · 21

downloads · 2

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

unreleased internal OpenAI model

View statement
Looking for related papers…
Contents