On the developability of Lie subalgebroids (Q869835)

The paper under review is the second half of a two-parts work dedicated to the question of integrability of Lie algebroids (the first half is reviewed in [Adv. Math. 204, No. 1, 101--115 (2006; Zbl 1131.58015)]). If \(G\) is a Lie groupoid with Lie algebroid \(\mathfrak{g}\), given a Lie subalgebroid \(\mathfrak{h}\subset\mathfrak{g}\), there is a Lie groupoid \(H\), with Lie algebroid \(\mathfrak{h}\), and a groupoid homomorphism \(H\to G\) which is an immersion. Unlike the case of Lie groups, in general, we have: {\parindent=6mm \begin{itemize}\item[(i)] it is not possible to choose this immersion to be injective; \item[(ii)] even when such an immersion exists, the closure of its image is not a Lie subgroupoid. \end{itemize}} In [Moerdijk (loc. cit.)], the authors have given criteria for such properties to hold. These criteria are stated in terms of properties of the invariant foliation \(\mathcal{F}(\mathfrak{h})\) of \(G\) associated to Lie subalgebroid \(\mathfrak{h}\) (defined by integrating the invariant vector fields tangent to \(\mathfrak{h}\)). Namely, they show that (i) one can choose an injective immersion \(H\to G\) iff the foliation \(\mathcal{F}(\mathfrak{h})\) has trivial holonomy; and (ii) if the foliation \(\mathcal{F}(\mathfrak{h})\) is transversely complete, then one can choose an injective immersion \(H\to G\) for which the closure of its image is a Lie subgroupoid \(\overline{H}\subset G\). Moreover, the image is closed iff this injective immersion is actually an embedding. In the present paper, the authors complete their study of Lie subalgebroids by considering the question: when is a Lie subalgebroid integrable by a closed Lie subgroup? More precisely, if \(G\) is a Lie groupoid with Lie algebroid \(\mathfrak{g}\), call a Lie subalgebroid \(\mathfrak{h}\subset\mathfrak{g}\) developable if it can be integrated to a closed subgroupoid of the universal covering groupoid \(\tilde{G}\). The reason for this name is the following: recall that a foliation \(\mathcal{F}\) of a manifold \(M\) is called developable if its lift to the universal covering space \(\tilde{M}\) is given by the fibers of a submersion. Now, the authors show that a Lie subalgebroid \(\mathfrak{h}\subset\mathfrak{g}\) is developable iff the lift of \(\mathcal{F}(\mathfrak{h})\) to the universal covering groupoid \(\tilde{G}\) is a developable foliation. The main theorem in this paper states that for a proper, Hausdorff, Lie groupoid \(G\), a Lie subalgebroid \(\mathfrak{h}\subset\mathfrak{g}\) for which \(\mathcal{F}(\mathfrak{h})\) is transversely complete is developable iff a certain Lie algebroid \(\mathfrak{b}(G,\mathfrak{h})\) is integrable. Note that, contrary to the usual Lie theory, not every Lie algebroid is integrable. The first obstruction to integrability was noted by \textit{R. Almeida} and \textit{P. Molino} [C. R. Acad. Sci., Paris, Sér. I 300, 13--15 (1985; Zbl 0582.57015)], while a complete theory was given more recently in [\textit{M. Crainic} and \textit{R. L. Fernandes}, Ann. Math. (2) 157, No. 2, 575--620 (2003; Zbl 1037.22003)]. The classical result of Almeida and Molino states that a transversely complete foliation \(\mathcal{F}\) on a compact manifold \(M\) is developable iff a certain Lie algebroid \(\mathfrak{b}(M,\mathcal{F})\) is integrable. The Almeida and Molino result can be seen as a very special case of the main theorem in this paper, where \(G\) is the pair groupoid \(M\times M\) and \(\mathfrak{h}\subset\mathfrak{g}\) is the subalgebroid \(T\mathcal{F}\subset TM\). A fundamental fact underlying the proof of the main theorem is that if \(G\) is a proper, Hausdorff, Lie groupoid, then for every Lie subgroupoid \(H\subset G\) for which \(\mathcal{F}(\mathfrak{h})\) is transversely complete, the Lie subalgebroid \(\mathfrak{h}\subset\mathfrak{g}\) is given by the kernel of a Lie algebroid valued Maurer-Cartan form.