Foliations, orders, representations, L-spaces and graph manifolds (Q520362)

It has recently been conjectured (by Boyer, Gordon, Watson and others) that the following rather diverse conditions are equivalent, for a closed prime 3-dimensional manifold \(W\): (1) \(W\) admits a co-oriented taut codimension-one foliation, (2) \(\pi_1(W)\) is left-orderable, (3) \(W\) is NOT a Heegaard-Floer L-space. Much evidence has been given toward this conjecture. All three conditions hold if the first Betti number of \(W\) is positive. So the question is reduced to the case that \(W\) has zero first Betti number, in other words \(W\) is a rational homology sphere. (1) \(\implies\) (3) is known in all cases. The conjecture has also been verified if \(W\) has a geometric structure, other than hyperbolic, which is still an open case. In particular, it is verified if \(W\) is Seifert-fibred. The present paper considers the more general case in which \(W\) is a rational homology sphere which is also a graph manifold. That means that it can be cut open along tori (the JSJ decomposition) into pieces which are all Seifert-fibred. Each piece has boundary a union of tori, and \(W\) can be reconstructed from the pieces by specifying how the tori are glued together, which is described by slopes in the fundamental group of the torus, isomorphic to \(\mathbb{Z}^2\). Some definitions are in order. A codimension-one foliation of \(W\) is a decomposition of \(W\) into sets (leaves) in such a way that locally the leaves intersect Euclidean neighbourhoods in parallel planes. Every 3-manifold admits such a foliation, but not necessarily one which is taut, meaning some simple closed curve in the manifold intersects each leaf transversely, or co-oriented, meaning there's a consistent choice of normal vector to the leaves. A group is left-orderable if its elements can be given a strict total ordering which is invariant under left-multiplication. An L-space, roughly speaking, is a rational homology 3-sphere whose Heegaard-Floer homology is as simple as possible. Returning to the present paper, the principal results can be described as almost proving the conjecture for graph manifolds. Specifically, the authors show that if \(W\) is a rational homology 3-sphere which is also a graph manifold, then (1) \(\iff\) (2) \(\implies\) (3). The methods are a meticulous study of the possible foliations on the pieces and the possible left-orderings of their fundamental groups, together with a study of whether and how these structures glue together to all of \(W\). In particular, the authors consider families of slopes on the torus boundaries determined in four different ways: using representations, left-orderings, foliations and Heegaard-Floer homology. For example, a left-ordering of a Seifert-fibered piece's fundamental group restricts to a boundary torus to an ordering of \(\mathbb{Z}^2\), which in turn determines a unique line with the property that all elements on one side of the line are greater than the identity: this is the slope determined by the left-ordering. The authors proceed, often by very delicate arguments, to show that each method of detection determines the same family of slopes. Various refinements of their arguments are considered, for example restricting to foliations which are horizontal in the sense that leaves are transverse to the Seifert fibres in each piece. In addition, the authors discuss issues of smoothness of the foliations and propose an L-space gluing conjecture which would provide the final implication (3) \(\implies\) (2) for graph manifolds.