The Borel Conjecture on Topological Rigidity of Aspherical Manifolds

The Borel Conjecture on Topological Rigidity of Aspherical Manifolds

OpenAI's claim

Two closed aspherical topological four-manifolds are claimed homotopy equivalent but not homeomorphic, sharing a word-hyperbolic fundamental group; one also admits a self-equivalence outside every homeomorphism homotopy class. The cited manuscript argues for that result; its relation to this problem is specified below.

Mathematical statement of the claimed result

Mathematical claim 1.

OpenAI claims closed connected aspherical topological four-manifolds $M,N$ with a common word-hyperbolic fundamental group such that \[ M\simeq N\quad\text{but}\quad M\not\cong N\text{ by any homeomorphism}. \] Aspherical means that the universal cover is contractible; $\simeq$ denotes homotopy equivalence. A second theorem claims a topological four-manifold that is aspherical, closed and connected $M$ and a self homotopy equivalence $f:M\to M$ for which no homeomorphism is homotopic to $f$. These are four-dimensional topological counterexamples; the first excludes homeomorphism existence in any homotopy class.

Outline of the argument

Manuscript 1 (family 320, 2026-10-04): Reflected marked chambers retain an odd-cover filling obstruction, and carefully chosen binary label relations eliminate unwanted chamber symmetries. Hyperbolicity and Poincare duality make the outer automorphism group finite, with a controlled normalizer of the realized chamber action. A hypothetical homeomorphism would transport another finite action into that realized subgroup and force an excluded conjugacy.

Formal verification

Family 320 has no Lean scope page in this snapshot; no formalization of these manuscript results is listed.

This site has not executed the Lean files or checked their dependencies. These descriptions report the scope of the linked declarations, rather than a new verification of the proofs.

Status caveat

OpenAI attributes these results to an internal model. Its release notice cautions that results lacking full formalization can contain mistakes. This site has not independently assessed the proofs; a solved label records the scope of the reported claim.

Sources

• OpenAI, Nonhomeomorphic closed aspherical four-manifolds with the same homotopy type, 2026-10-04; citation key OAI:Nonhomeomorphic-closed-aspherical-four-manifolds-with-the-same-homotopy-type-October-4-2026. PDF (GitHub); PDF (direct); Manuscript page and citation; Latest version; Statement source: thm:pair; Statement source: thm:self-map.
• OpenAI catalogue entry.

The Borel Conjecture on Topological Rigidity of Aspherical Manifolds

Definitions and mathematical statement

A connected manifold is aspherical if its universal cover is contractible, equivalently its higher homotopy groups vanish. Given closed aspherical manifolds $M,N$ and a homotopy equivalence $f:M\to N$, the conjecture asks for a homeomorphism $h:M\to N$ with \[ f\simeq h. \] The specific homotopy class must contain a homeomorphism; merely knowing that the two manifolds happen to be homeomorphic is weaker. The category is topological, so smooth rigidity or a diffeomorphism is not requested.

Formulation and concept references: [S1].

Short English statement

Does every homotopy equivalence between closed aspherical manifolds come from a homeomorphism up to homotopy?

Research attempt

Current frontier and exact scope.

The source's Section 2 distinguishes rigidity of a given homotopy class from mere classification of manifolds. It also distinguishes topological rigidity from smooth rigidity: exotic smooth structures are not counterexamples here. Bartels--Lück prove the requested conclusion in dimension at least five for, among others, hyperbolic and finite-dimensional $\operatorname{CAT}(0)$ groups. Their Proposition 0.3 states the precise $K$- and $L$-theoretic hypotheses, including orientation twists; its four-dimensional extension requires a good fundamental group in Freedman's sense. The original target imposes none of these restrictions. [S1][E1]

A useful recent caution comes from Gu's August 2026 revision on $\mathcal Z$-compactifications: removing an interval factor requires additional compatibility, and applying compactifications to universal covers requires further control. This is adjacent work, not a proof of Borel rigidity. Its corrected hypotheses warn against a proposed shortcut from contractibility to a controlled homeomorphism. [E2]

Attack route.

Three routes are tested. First, retain exactly the two integral assembly properties that would kill the geometric obstruction. Second, try to construct a homeomorphism on universal covers and descend it equivariantly. Third, exploit the explicit linear realization of automorphisms for tori. The first produces a conditional reduction; the second fails without small-scale control; the third supplies a restricted verification, not a classification of all manifolds homotopy equivalent to a torus.

Attempt: the two-degree obstruction calculation.

Let $N$ be a closed aspherical topological $n$-manifold, $n\geq5$, $\pi=\pi_1(N)$, and $w:\pi\to\{\pm1\}$ its orientation character. Assume $\operatorname{Wh}(\pi)=0$. A homotopy equivalence to $N$ is then simple, so it defines an element of the topological simple structure set $\mathcal S^{\mathrm{TOP}}(N)$. Its basepoint is the identity of $N$; being the basepoint means precisely that the homotopy class contains a homeomorphism, not just that its source happens to be homeomorphic to $N$. Use the orientation-twisted quadratic $L$-spectrum throughout. The relevant part of the surgery sequence is \begin{equation}\label{wsref-0-65-71-3a-33-30-30-30-36-37-33-34-3a-73-75-72-67-65-72-79} \begin{split} H_{n+1}(N;\mathbf L\langle1\rangle)&\xrightarrow{A^c_{n+1}} L^s_{n+1}({\mathbb{Z}}\pi,w)\longrightarrow\mathcal S^{\mathrm{TOP}}(N)\\ &\xrightarrow{\eta}H_n(N;\mathbf L\langle1\rangle) \xrightarrow{A^c_n}L^s_n({\mathbb{Z}}\pi,w). \end{split} \end{equation} Here $\mathbf L\langle1\rangle$ retains the positive homotopy groups of the periodic spectrum and kills degrees at most zero. The middle sequence is interpreted with the usual structure-set action, not as an exact sequence of abelian groups. This is the geometric/algebraic surgery input used in Bartels--Lück's Proposition 0.3. [E1]

Lemma 1 (a sufficient two-degree criterion). If $A^c_n$ is injective and $A^c_{n+1}$ is surjective, then $\mathcal S^{\mathrm{TOP}}(N)$ has one element.

Proof. For every structure $s$, exactness gives $A^c_n\eta(s)=0$. Injectivity gives $\eta(s)=0$. Exactness at the structure set says that every such $s$ lies in the orbit of the basepoint under $L^s_{n+1}({\mathbb{Z}}\pi,w)$. Elements coming from $A^c_{n+1}$ fix that basepoint. Surjectivity therefore makes the entire orbit a singleton. Every $s$ is the basepoint. $\square$

This identifies why an injectivity theorem alone is not enough for this argument: it eliminates the normal invariant but does not by itself eliminate the residual surgery action. It also identifies why rational injectivity is weaker still: it need not kill integral torsion in the normal invariant.

Attempt: check the connective cutoff rather than dropping it.

Let $\mathbf C$ be the cofiber of $\mathbf L\langle1\rangle\to\mathbf L$. Then $\pi_j\mathbf C=0$ for $j>0$. The homological dimension of an $n$-manifold, with the local coefficients occurring here, is at most $n$. The generalized-homology spectral sequence \[ H_a(N;\pi_b\mathbf C)\Longrightarrow H_{a+b}(N;\mathbf C) \] therefore has no nonzero terms of total degree greater than $n$: $a\leq n$ and $b\leq0$. The base dimension is bounded, so no unbounded filtration or convergence interchange is being used. Consequently \[ H_{n+2}(N;\mathbf C)=H_{n+1}(N;\mathbf C)=0. \] The cofiber long exact sequence now gives \begin{equation}\label{wsref-0-65-71-3a-33-30-30-30-36-37-33-34-3a-63-6f-6d-70-61-72-69-73-6f-6e} \begin{aligned} H_{n+1}(N;\mathbf L\langle1\rangle)&\xrightarrow{\ \simeq\ } H_{n+1}(N;\mathbf L),\\ H_n(N;\mathbf L\langle1\rangle)&\hookrightarrow H_n(N;\mathbf L). \end{aligned} \end{equation} Thus isomorphisms of the periodic simple assembly maps in degrees $n,n+1$ imply the hypotheses of Lemma 1. The degree-$n$ comparison need not be surjective; claiming an isomorphism there would be unnecessary and unjustified.

Asphericity supplies $N\simeq B\pi$. To obtain these periodic simple assembly isomorphisms from the usual ultimate $L^{\langle-\infty\rangle}$ version of Farrell--Jones, one must also justify the change of decoration. The vanishing conditions \[ \operatorname{Wh}(\pi)=0,\qquad \widetilde K_0({\mathbb{Z}}\pi)=0,\qquad K_i({\mathbb{Z}}\pi)=0\quad(i<0) \] provide that justification in the established reduction. They are not consequences of the contractibility of the universal cover by elementary homotopy theory. Nor may the orientation character be suppressed for nonorientable manifolds. The needed implications, for torsion-free groups, are exactly those checked in Proposition 0.3 of the cited primary paper. [E1]

The calculation above re-derives a known sufficient route and makes its two-degree use explicit. It does not establish the required assembly properties for a new class of groups. In particular, replacing them by an unresolved global Farrell--Jones assertion makes the argument conditional, not a proof of the catalogue's target.

Attempt: a constructive test on tori.

For $f:T^n\to T^n$ a homotopy equivalence, the induced homomorphism is some $A\in\operatorname{GL}_n({\mathbb{Z}})$. The linear map \[ h_A:{\mathbb{R}}^n/{\mathbb{Z}}^n\longrightarrow{\mathbb{R}}^n/{\mathbb{Z}}^n, \qquad [x]\longmapsto[Ax] \] is a homeomorphism. Since $T^n$ is a $K({\mathbb{Z}}^n,1)$, maps to it are classified up to homotopy by their induced fundamental-group homomorphisms, up to conjugacy. The group is abelian, so $f\simeq h_A$. This checks every homotopy class of self-equivalences of an actual torus. It does not show that an arbitrary closed manifold with fundamental group ${\mathbb{Z}}^n$ is a torus; that requires the rigidity step we are studying.

Could bounded displacement of an equivariant lift yield the missing homeomorphism directly? Already for the torus, take \[ F(x_1,\ldots,x_n)= (x_1+\sin(2\pi x_1),x_2,\ldots,x_n). \] It is proper, is ${\mathbb{Z}}^n$-equivariant, and stays within distance one of the identity. It descends to a map homotopic to the identity. Nevertheless its first-coordinate derivative is $1+2\pi\cos(2\pi x_1)$, which is positive at $0$ and negative at $1/2$. The real first-coordinate function is not monotone and hence is not injective. Thus $F$ is not a homeomorphism. A coarse or bounded-distance lift construction does not by itself give a one-to-one map; a separate controlled replacement theorem is essential.

Stress test.

The source permits every dimension. Lemma 1 explicitly used the high-dimensional topological surgery theorem, so it says nothing new about arbitrary four-manifold groups. Even an unrestricted assembly theorem would not allow that dimension to be silently included in this proof. Conversely, a failure of our sufficient assembly criterion would not by itself exhibit a nontrivial structure: actions can have stabilizers. The calculation is deliberately one-way. [S1][E1]

Contractible universal covers need not even be Euclidean; the primary rigidity paper discusses Davis examples with that feature. Requiring an unproved Euclidean or nonpositively curved model would exclude legitimate inputs. Similarly, a homeomorphism after passing to finite covers is not a descent theorem unless it conjugates the relevant deck actions. The controlled-compactification route needs that equivariant control, not just an abstract compactification. [E1][E2]

Outcome.

Outcome: Conditional reduction.

The notebook isolates an integral, two-degree surgery criterion, verifies its connective-to-periodic comparison, and checks the homotopy-class requirement in the torus test. It also gives an explicit folded equivariant lift defeating a proposed coarse-geometric shortcut. These are clarifications and re-derivations, not a claimed new rigidity theorem. A complete solution still requires the requisite obstruction vanishing for every allowed fundamental group, with a valid argument in dimension four as well. No counterexample to the stated topological conjecture has been constructed.

\Needspace{14\baselineskip}

Sources

• S1 U. Hamenstadt and F. Jaeckel, Rigidity of geometric structures, Geometriae Dedicata 218, article 16 $2024$, Section 2. https://link.springer.com/article/10.1007/s10711-023-00861-4. Catalogue formulation source.
• E1 Arthur Bartels and Wolfgang Lück, The Borel Conjecture for hyperbolic and CAT$0$-groups, Annals of Mathematics 175 $2012$, 631--689. Theorems A--B and Proposition 0.3, including its proof and the orientation-twisted version. https://arxiv.org/html/0901.0442v2.
• E2 Shijie Gu, On Z-compactifiability of manifolds, arXiv:2312.02527v3, 25 August 2026. Revised abstract and author's revision note concerning product compatibility and additional control on universal covers. Reported as a manuscript caution, not used as a rigidity theorem. https://arxiv.org/abs/2312.02527v3.