Wall's Poincare Duality Group Conjecture

Wall's Poincare Duality Group Conjecture

OpenAI's claim

An integral PD4 group is claimed to be finitely presented with finite classifying space but to admit no closed aspherical topological four-manifold realization, refuting dimension-four manifold realization. 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 a group $G$ with a finite presentation and an integral Poincaré complex, finite, oriented and aspherical, of dimension four $Q$ and \[ H^i(G;\mathbb ZG)=0\ (i\ne4),\qquad H^4(G;\mathbb ZG)\simeq\mathbb Z \] with trivial right $G$-action in degree four, but with no closed aspherical topological four-manifold having fundamental group $G$. The trivial $\mathbb ZG$-module has a finite finitely generated free resolution. Thus the claimed counterexample is integral duality with a finite classifying space, not merely rational duality.

Outline of the argument

Manuscript 1 (family 305, 2026-09-24): A marked smooth chamber becomes a graph-type Poincare pair after attaching cells along a hyperbolic middle-homology basis, while its boundary forbids a marked graph filling. Reflection closes the pair into an aspherical complex. Any manifold realization would provide a sphere-marked stable product; capping and controlled destabilization recover the forbidden filling, contradicting the chamber obstruction.

Formal verification

Family 305 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, A PD4 group without an aspherical manifold model, 2026-09-24; citation key OAI:A-PD4-group-without-an-aspherical-manifold-model-September-24-2026. PDF (GitHub); PDF (direct); Manuscript page and citation; Latest version; Statement source: thm:main.
• OpenAI catalogue entry.

Wall's Poincare Duality Group Conjecture

Definitions and mathematical statement

An integral Poincaré duality group of dimension $n$ is a group of cohomological dimension $n$, of type FP over ${\mathbb{Z}}$, whose cohomology with group-ring coefficients vanishes outside degree $n$ and is infinite cyclic in degree $n$. The action on that cyclic dualizing module defines its orientation character $w:G\to\{\pm1\}$. For every finitely presented such $G$ and every $n\ge3$, ask for a closed aspherical topological $n$-manifold $M$ with \[ \pi_1(M)\cong G, \] inducing the same orientation character. Aspherical means the universal cover is contractible. A finite Poincaré duality complex is not automatically a topological manifold, and that distinction is central to the target.

Formulation and concept references: [S1].

Short English statement

Is every finitely presented integral Poincaré duality group of dimension at least three the fundamental group of a closed aspherical manifold of that dimension?

Research attempt

Current frontier.

A finitely presented integral Poincaré duality group admits a finitely dominated aspherical Poincaré complex. Passing to a finite complex has a $\widetilde K_0({\mathbb{Z}} G)$ obstruction; passing from a Poincaré complex to a manifold has further surgery obstructions. These steps are distinguished in Lück's Theorem 5.4, Lemma 5.7 and Remark 5.8. [S1]. The June 2026 revision of Davis--Hillman classifies compact aspherical four-manifolds with boundary and elementary amenable fundamental group. Its restricted classification does not realize every abstract finitely presented PD group. [E1], Abstract. The present attempt keeps all dimensions $n\ge3$ and the orientation character, and tests whether duality itself could annihilate the first obstruction.

Attack route.

Three approaches suggest themselves. A finite projective resolution might be converted to a finite CW model; this needs vanishing of its reduced projective class, not just duality of the resolution. Stabilizing by a circle kills that class, but then one must destabilize the resulting space or manifold. Finally, surgery can act on a finite Poincaré complex, but requires a normal reduction and the correct vanishing obstruction. We follow the first two routes explicitly and use the third as a precise conditional completion, not as a theorem that the missing inputs exist.

Attempt.

1. The exact restriction imposed by duality. Let $w:G\to\{\pm1\}$ be the orientation character and let $C_*$ be a bounded finite-projective ${\mathbb{Z}} G$ chain model for a finitely dominated $BG$. The construction of such a Poincaré model is the input from [S1], Theorem 5.4. The involution \[ \overline g=w(g)g^{-1} \] turns the dual of a left projective module back into a left module; write $[P]^*=[\operatorname{Hom}_{{\mathbb{Z}} G}(P,{\mathbb{Z}} G)]$ with this convention. The finiteness obstruction is \[ o(G)=\sum_i(-1)^i[C_i]\in\widetilde K_0({\mathbb{Z}} G). \] Chain Poincaré duality compares $C_*$ with the degree-reversed twisted dual. Taking alternating projective classes gives the rigorous identity \begin{equation}\label{wsref-0-33-30-30-30-36-37-39-37-3a-73-65-6c-66-64-75-61-6c} o(G)=(-1)^n o(G)^*. \end{equation} Here a chain homotopy equivalence preserves the alternating class because its contractible mapping cone has zero alternating class; contractible bounded projective complexes split into elementary adjacent pairs.

Equation (30006797:selfdual) is not a vanishing theorem. As an abstract abelian-group calculation, for even $n$ the group ${\mathbb{Z}}$ with identity involution has every element satisfying it. For odd $n$, ${\mathbb{Z}}/2$ with identity involution has a nonzero solution. These are counterexamples to an algebraic inference, not examples of group-ring obstructions realized by PD groups. More generally, decomposing rationally into the $\pm1$ eigenspaces of $*$ still leaves the permitted eigenspace, and rationalizing loses possible integral torsion. Neither maneuver proves $o(G)=0$ in the actual reduced $K$-group.

2. Circle stabilization hides exactly this information. Let $R={\mathbb{Z}} G$ and $R'=R[t,t^{-1}]$. After extending scalars, the chain complex of $BG\times S^1$ is represented by the mapping cone of \[ t-1:C_*\otimes_RR'\longrightarrow C_*\otimes_RR'. \] In degree $i$ it has projective module $(C_i\otimes_RR')\oplus(C_{i-1}\otimes_RR')$. Therefore \begin{equation}\label{wsref-0-33-30-30-30-36-37-39-37-3a-63-69-72-63-6c-65} o(BG\times S^1)=\sum_i(-1)^i([C_i\otimes R']+[C_{i-1}\otimes R'])=0. \end{equation} This computation does not require $o(G)=0$. By the finiteness-obstruction criterion recalled in [S1], Section 5, the stabilized finitely dominated space has a finite CW model. It is an aspherical PD space of dimension $n+1$ and fundamental group $G\times{\mathbb{Z}}$.

The cancellation is useful but dangerous. It gives no way to recover the original class from the stabilized class: the map on this obstruction is identically zero. Even a manifold model for $G\times{\mathbb{Z}}$ would not by itself produce a manifold model for $G$ by taking an arbitrary map to $S^1$ and declaring its fiber a closed aspherical manifold. A continuous map need not be a bundle, its fibers need not be manifolds, and a regular-value fiber in a smooth representative need not be aspherical. A separate fibering or splitting theorem would be necessary.

3. A conditional high-dimensional completion. Fix $n\ge5$. Suppose the following additional facts can be supplied for the given $G$: the obstruction $o(G)$ vanishes; the resulting finite aspherical PD complex $X$ admits a degree-one topological normal map $f:M\to X$ with the specified orientation character; and the associated Wall obstruction in the appropriate $L_n({\mathbb{Z}} G,w)$ group is zero. The high-dimensional surgery theorem then modifies $f$, within its normal bordism class, to a homotopy equivalence $M'\to X$. Consequently $M'$ is closed, has fundamental group $G$, and its universal cover is homotopy equivalent to the contractible universal cover of $X$. It is therefore aspherical and realizes the target, including $w$. This is an application of the normal-map/surgery framework, not a construction of the normal map. [E2], surgery obstruction and existence theorem.

There are two different missing inputs even after finiteness. A spherical Spivak normal fibration need not already have the required stable topological bundle reduction. A normal map, once obtained, may have a nonzero obstruction. A periodic $L$-theory assembly statement cannot simply be identified with every connective surgery condition: Lück's Remark 5.8 explicitly records the missing degree-zero issue. His Lemma 5.7 produces homology ANR manifolds under additional assumptions and in its stated dimensions; such a space is not automatically the closed topological manifold requested here. [S1].

4. Testing simpler numerical invariants. In an oriented odd-dimensional finite PD complex, rational duality gives $b_i=b_{n-i}$ and hence $\chi(X)=0$. This agrees with the manifold identity but follows without a manifold structure. In even dimension, duality supplies the symmetric or skew form on the middle-dimensional homology, but says nothing by itself about a local Euclidean structure. Thus checking Euler characteristic or a self-dual finite resolution cannot replace either obstruction calculation above.

An attempted minimal-counterexample argument also stops at a specific point: passing to $G\times{\mathbb{Z}}$ raises dimension and erases the finiteness obstruction rather than decreasing a complexity parameter. It provides no induction whose hypothesis applies to $G$, and no monotonicity of manifold realizability under removing a direct ${\mathbb{Z}}$ factor was proved.

Stress test.

The toy groups ${\mathbb{Z}}$ and ${\mathbb{Z}}/2$ in part 1 are not asserted to occur as $\widetilde K_0({\mathbb{Z}} G)$ for a counterexample PD group. They only invalidate the inference from an involution identity to zero. Conversely, (30006797:circle) is an exact projective-class cancellation, not evidence that $G$ itself has a finite classifying complex. Orientation twists are retained in both the duality and surgery expressions.

The surgery completion was explicitly restricted to $n\ge5$. Dimensions three and four in the original target require additional arguments and are not swept into high-dimensional surgery. Likewise, the recent elementary-amenable four-manifold classification presupposes and restricts the geometric objects under study. No claim is made that it proves the unrestricted existence assertion.

Outcome.

Outcome: Exploratory approach with unresolved gap.

The attempt derives the exact duality constraint on the finiteness obstruction and shows algebraically why circle stabilization cannot recover its vanishing. It gives a conditional high-dimensional completion with each geometric input named separately. These are obstruction calculations, not new realizations or claims of priority.

The missing steps are vanishing of the actual finiteness obstruction, construction of the appropriate normal data, removal of the surgery obstruction, and separate treatment of dimensions three and four. No counterexample PD group, and no proof for all such groups, is supplied.

Sources

• S1 Wolfgang Lück, “Survey on Aspherical Manifolds,” arXiv:0902.2480v3 $2009$; published in European Mathematical Society proceedings $2010$. https://arxiv.org/pdf/0902.2480. Catalogue formulation source.
• E1 J. F. Davis and J. A. Hillman, Aspherical 4-manifolds with elementary amenable fundamental group, arXiv:2501.12512v2, 29 June 2026, Abstract and Introduction. https://arxiv.org/abs/2501.12512v2. Consulted 27 September 2026; its boundary and group-class restrictions are not omitted here.
• E2 C. T. C. Wall, Surgery on Compact Manifolds, 2nd ed., edited by A. A. Ranicki, Mathematical Surveys and Monographs 69, AMS, 1999, the normal-map obstruction and high-dimensional surgery existence chapters. https://www.maths.ed.ac.uk/~v1ranick/books/scm.pdf. The application is conditional on the normal data and vanishing obstruction.