A finite complex whose third dimension cannot be removed

← All digests

Family 321 · Topology
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-22

1. The question

A cell complex is built by attaching points, edges, disks, and higher-dimensional cells. Sometimes the higher cells are unnecessary: a complex may be homotopy equivalent to one of lower dimension. Algebraic invariants often detect when such a reduction should be possible.

Wall’s finite $D(2)$ problem asks whether particular homology and cohomology vanishing conditions force a finite complex to have a finite two-dimensional homotopy model. The conditions remove the obvious algebraic need for dimension three. The question is whether geometry can still prevent the reduction.

Family 321 claims a negative answer. Its example is a finite three-dimensional complex satisfying an even stronger coefficient-vanishing condition than required, yet having no finite model of dimension at most two. Its fundamental group is infinite; “finite” refers to the cell structure.

2. Background and essential definitions

A CW complex is formed by attaching cells along their boundaries. A presentation complex for a group has one vertex, one edge for each generator, and one disk for each relator. Its fundamental group is the presented group.

For a connected complex $X$, let $G=\pi_1(X)$ and let $\widetilde X$ be its universal cover. The group ring $\mathbb ZG$ consists of finite integer combinations of group elements. Modules over it allow coefficients that change when transported around loops.

Such local coefficients contain more information than constant coefficients. A particularly accessible example is a character

$$\rho:G\longrightarrow\mathbb C^{\times}.$$

It gives a one-dimensional complex local system $L_\rho$: traversing a loop representing $g$ multiplies a coefficient by $\rho(g)$. Twisted homology is calculated by cellular boundary matrices whose entries include these transport factors.

For a finite three-dimensional complex, the $D(2)$ condition is

$$H_i(\widetilde X;\mathbb Z)=0\quad(i>2), \qquad H^3(X;M)=0$$

for every finitely generated $\mathbb ZG$-module $M$. The first condition concerns the universal cover; the second concerns cohomology with all the specified local systems. Neither condition says that $X$ is aspherical.

A group is perfect if it equals its commutator subgroup, or equivalently if its abelianization is zero. Quillen’s plus construction kills a chosen perfect normal subgroup of the fundamental group by adding two- and three-dimensional cells while preserving homology with coefficients from the quotient group.

A homotopy model preserves the entire homotopy type, not merely ordinary homology. The question here also requires that the lower-dimensional model have finitely many cells.

3. Development history and previous partial results

Wall introduced these dimension conditions in his work on finiteness of CW complexes in the 1960s and returned to the question in his 1979 problem list. His reductions bring the $D(2)$ case to a finite three-dimensional model. The unresolved step was removing the remaining three-cells up to homotopy.

A stable version has a positive answer. Work going back to Cohen shows that adding sufficiently many two-spheres by a wedge makes a finite $D(2)$ complex homotopy equivalent to a finite two-dimensional complex. Hambleton gives a bound in terms of the number of three-cells. But adding spheres changes the space, so stable reduction does not settle the original problem.

There are positive results for particular groups. Hambleton established the $D(2)$ property for finite subgroups of $SO(3)$. M1 also cites Hofmann–Nicholson’s results for generalized quaternion groups, including the resolution of a longstanding candidate involving the group of order 32.

Other candidate counterexamples arose from relation gaps: the relator subgroup may appear to need more normal generators than its abelianization needs as a module. Bridson–Tweedale constructed relevant finite $D(2)$ complexes, with failure of a two-dimensional model conditional on a relation-gap assertion.

Mannan connected the problem directly with the plus construction. Applied to a finite two-dimensional complex along a finitely normally generated perfect subgroup, it produces a finite $D(2)$ complex. He also obtained a converse description for finite cohomologically two-dimensional three-complexes.

M1 uses this established production method. Its claimed new obstruction is a restriction on rank-one twisted homology in a finite two-dimensional complex, obtained by adapting Howie’s cyclic-tower method. The construction and obstruction then meet in a small explicit presentation.

4. What the new manuscripts claim

M1, dated 6 October 2026, states that there is a finite connected three-dimensional CW complex $X$ with

$$H_i(\widetilde X;\mathbb Z)=0\quad(i>2), \qquad H^3(X;M)=0\quad\text{for every }\mathbb ZG\text{-module }M,$$

but no homotopy equivalence from $X$ to any finite CW complex of dimension at most two. The cohomology assertion covers all modules, not merely finitely generated ones.

The example starts with the presentation

$$\begin{aligned} P=\langle x_1,a_1,x_2,a_2,s\mid\;&x_1a_1x_1^{-1}=a_1^4,\quad x_2a_2x_2^{-1}=a_2^3,\\ &x_1=a_2^{13},\quad sx_2s^{-1}=a_1^5\rangle. \end{aligned}$$

Its presentation complex $B$ has one vertex, five edges, and four two-cells. Let $K$ be the normal closure of $x_1^2$ and $x_2^3$ in $P$. The manuscript shows that $K$ is perfect and forms the finite plus construction with group $G=P/K$.

This adds two two-cells and two three-cells. Sending $s$ to $1\in\mathbb Z$ and the other generators to zero gives a surjection $G\to\mathbb Z$, so the group is infinite.

The general obstruction states that for a finite two-dimensional complex $Y$,

$$H_2(Y;L_\rho)=0 \quad\Longrightarrow\quad \rho(g)=1\text{ for every finite-order }g\in\pi_1(Y).$$

The constructed $X$ has a character violating the conclusion while retaining the vanishing hypothesis. Homotopy invariance would transfer both properties to any proposed finite two-dimensional model, which is therefore excluded.

5. Why the advance matters

Importance: 5/5, provisionally. If correct, the example gives a negative answer to a classical dimension-reduction problem and separates algebraic cancellation from geometric realization in a particularly low dimension.

The obstruction is finer than ordinary homology. A one-dimensional coefficient system remembers how loops act on coefficients, and that small extra piece of information detects the impossibility of a finite two-dimensional model.

The explicit size is also informative. The construction begins with five generators and four relators, then adds a finite plus-construction package. The difficulty is not an uncontrolled infinite complex; it is a mismatch between finite algebraic conditions and finite geometric realizability.

6. The main idea

A tower that cannot keep growing

Suppose a finite two-dimensional complex $Y$ has $H_2(Y;L)=0$. The manuscript proves a stronger statement than the torsion obstruction: a map from a finite complex $F$ with $b_1(F;\mathbb Q)=0$ cannot pull $L$ back to a system with nontrivial monodromy.

Assume such a map exists and replace its image by a finite simplicial subcomplex $A$. In dimension two, second homology is the kernel of the second boundary map, with no third boundary to quotient out. Its vanishing therefore passes to subcomplexes.

Nontrivial rank-one monodromy gives $H_0(A;L)=0$. Together with $H_2(A;L)=0$, the Euler-characteristic formula implies $\chi(A)\leq0$, hence $b_1(A;\mathbb Q)\geq1$. There is consequently an infinite cyclic cover of $A$.

The map from $F$ lifts to that cover because every homomorphism from $\pi_1(F)$ to $\mathbb Z$ is zero. The manuscript shows that the twisted second-homology vanishing also passes to the cover. Its boundary matrix has Laurent-polynomial entries; a nonzero full-rank minor after setting the deck variable to one remains a nonzero polynomial upstairs.

Replace the cover by the image of the lifted map and repeat. Each new image has strictly more simplices, while every image is still the image of the same fixed triangulated $F$. The number of simplices can neither grow indefinitely nor exceed that finite bound. This yields the claimed contradiction.

For a finite-order loop, a suitable finite Moore complex provides $F$. In the order-two case used below, it is simply the real projective plane $\mathbb RP^2$, whose first rational Betti number is zero.

Arrange the forbidden character

The exponents in the presentation are chosen so that

$$4^2-1=15,\qquad3^3-1=26,\qquad\gcd(15,26)=1.$$

The induced relations in the abelianization of $K$ impose coprime annihilators on its normal generators. M1 uses this to show that the abelianization vanishes, making the finite plus construction available.

Choose a primitive cube root of unity $\zeta$. The character is

$$\rho(x_1)=-1,\quad\rho(a_1)=\zeta,\quad \rho(x_2)=\zeta^2,\quad\rho(a_2)=-1,\quad\rho(s)=1.$$

It respects the relations and kills the two generators of $K$, so it descends to $G$. The element represented by $x_1$ has order exactly two: its square is killed, and its character value $-1$ shows that it is not the identity.

The paper computes the twisted cellular boundary of $B$. A four-by-four minor has determinant $-4(1-\zeta^2)\neq0$, giving $H_2(B;L_\rho)=0$. This is the manuscript’s reported boundary calculation, used here to explain how the example is arranged.

The plus construction preserves this twisted homology while supplying the $D(2)$ conditions. A finite two-dimensional model would therefore have vanishing twisted second homology and a character nontrivial on an order-two element. The tower obstruction forbids that combination.

7. Scope, limitations, and open questions

The theorem concerns Wall’s finite $D(2)$ problem. Both the initial complex and the excluded two-dimensional model have finitely many cells. The statement should not be expanded into an exclusion of every possible infinite two-dimensional model.

The example is not assumed aspherical, and its group contains torsion. It is therefore a different construction from the cohomological-versus-geometric dimension problem discussed in family 249.

The positive stabilized theorem remains compatible with this counterexample: adding two-spheres changes the homotopy type and can remove the obstruction. Likewise, positive $D(2)$ results for specified finite groups are unaffected by this infinite-group example.

The claimed general rank-one obstruction may be useful in other realization problems, but the manuscript does not classify every group or every finite complex for which dimension reduction fails. Its conclusion is one explicit obstruction and one resulting counterexample.

8. Sources and further reading

M1 supplies the example, tower obstruction, and twisted boundary calculation. The established construction framework is Mannan’s Quillen’s plus construction and the D(2) problem. M1’s bibliography gives Wall’s 1965 finiteness paper and 1979 problem list, Cohen’s stable result, Hambleton’s positive cases, Howie’s cyclic-tower method, and the relation-module work of Bridson–Tweedale. These sources distinguish the existing method for producing finite $D(2)$ complexes from the new obstruction claimed for this one.

Source manuscripts

• M1. OpenAI, A Counterexample to Wall’s D(2) Problem (October 6, 2026).

Sources consulted on 22 October 2026 at repository revision 1581d4c. Download the manuscript citations.