Let G be a finitely presented group. Then there is a 4-dimensional manifold M such that G ≅ π1(M).
Let $G = \gen{x_{1}, \ldots, x_{m} \mid r_{1}, \ldots, r_{n}}$ . Construct a CW complex as follows:
The problem is that the base point is a singularity. We fix this by thickening everything
Implications: Because the word problem
is unsolvable for finitely presented groups, this shows that fundamental
groups can't be used to algorithmically classify 4-manifolds.
Higher groups:
Every finitely presented group is the fundamental group of a compact, symplectic 4-manifold
The result carries into higher dimensions: you can find a manifold of any dimension ≥ 4 with a given fundamental group.
You basically look at M4 × Dk where Dk is a k-dimensional disk. The fundamental group is a monoidal functor over the product, so this preserves the fundamental group.
But this is a manifold with boundary. To remove the boundary, you attach handles.
In dimension ≥ 5 you can show that manifolds with a fixed fundamental group G are classifyable. So for example, you can classify simply-connected manifolds in dimension ≥ 5.
On Insolubility of the Homeomorphy Problem
This is one of my favourite mathematical results of all time.
Classifying compact manifolds up to homeomorphism is undecidable in dimension ≥ 4.
First, Markov gives an algorithm that takes any group presentation and gives you a corresponding manifold. Your group is trivial if and only if the manifold you get is a homotopy 4-sphere (it has the same fundamental group and homology groups as S4 ). If your space is homeomorphic to S4 then you have a homotopy 4-sphere, and thus G is trivial. But the rest of the proof actually relies on the other direction: if you start with the trivial group, is the space you get from the algorithm actually homeomorphic to S4, or just a homotopy 4-sphere?
Markov then assumes the topological Poincare conjecture holds in dimension 4 (which wasn't proved at the time): that every homotopy 4-sphere is homeomorphic to $S4.$
Assuming that's true (which it turned out to be), if there was some invariant that could distinguish every manifold from S4, that would solve the word problem as follows:
If you do everything smoothly, the trivial group gives you a smooth manifold that is homeomorphic to S4. But the Smooth 4D Poincare Conjecture (whether this implies that manifold is diffeomorphic to S4) is still unsolved.
Every higher homotopy group is a $ℤπ1$-module
This is essentially due to the Whitehead products
4-dimensional simply-connected manifolds are determined by Freedman's intersection form
Homology is decidable both on the set of compact manifolds, and on finite CW complexes. Homotopy is undecidable on both of these.
We can compute the Smith normal form in polynomial time
Consequently: There's no algorithm for
determining whether a finite 4d CW complex has π2 ≅ 0. But there is one
to check if $H2 ≅ 0.$
This is rational homotopy theory.
Since π1 ≅ 0, from Hurewicz's theorem we get that $π2 ≅ H2.$ This is computable.
We then construct the higher homotopy groups with the Postnikov Tower:
For any simply connected finite CW complex, there is a primitive recursive algorithm for $πn.$
However the complexity of the algorithm scales exponentially with n
Let Σg be a Riemann surface of genus g. Then $$ \pi_{1}(\Sigma_{g}) = \gen{a_{1}, b_{1}, \ldots, a_{g}, b_{g} \mid [a_{1},b_{1}]\ldots [a_{g}, b_{g}] = 1} $$