Let $G$ be a connected Lie group. Then by a theorem of Cartan there is a diffeomorphism $$ G \cong K \times \mathbb{R}^n $$ where $K$ is a maximal compact subgroup of $G$. Now, let $M$ be a homogeneous manifold. In other words, there exists a Lie group $G$ acting transitively on $M$. Is it true that $M$ deformation retracts onto a compact submanifold? Slightly stronger, is it true that there is a diffeomorphism $$ M \cong K \times \mathbb{R}^n $$ where $K$ is a compact submanifold of $M$?
-
1$\begingroup$ You can find information about the topology of homogeneous spaces of Lie groups in A. Borel, Les bouts des espaces homogènes des groupes de Lie, Ann. Math. 58(3), 1953, 443-457, behind JSTOR's rapacious paywall here. $\endgroup$– YCorCommented Nov 13, 2019 at 20:15
-
$\begingroup$ Crossposted from MathSE: math.stackexchange.com/questions/3433385 $\endgroup$– YCorCommented Nov 13, 2019 at 23:01
1 Answer
Mostow-Karpelevich theorem says that if $G/G^\prime$ is a homogeneous space where $G$, $G^\prime$ are Lie groups with finitely many connected components, and maximal compact subgroups $K\supset K^\prime$, respectively, then $G/G^\prime$ is a vector bundle over $K/K^\prime$. In fact, it is a homogeneous $K$-vector bundle. A more precise statement is in [Mostow, G. D., Covariant fiberings of Klein spaces II, Amer. J. Math. 84 (1962), 466–474]. There are examples where the bundle is nontrivial.
Another result in this direction addresses the situation when $G$ is solvable, and $G^\prime$ is any closed subgroup. Then $G/G^\prime$ is a vector bundle over a compact of the form $H/K^\prime$ where $H$ is solvable Lie group and $H^\prime$ is a closed subgroup. See [Auslander, L. and Tolimieri, R., Splitting theorems and the structure of solvmanifolds, Ann. of Math. (2) 92 (1970), 164–173].
In order for the quotient to be a vector bundle some assumptions are clearly necessary. For example consider a nonabelian free discrete subgroup of $SL(2,\mathbb R)$, and lift it to a nonabelian free discrete subgroup $F$ of the universal cover $\widetilde{SL}(2,\mathbb R)$. The latter is diffeomorphic to $\mathbb R^3$, so the quotient $\widetilde{SL}(2,\mathbb R)/F$ is homotopy equivalent to a wedge of circles, and in particular it is not a vector bundle over a manifold.
-
$\begingroup$ Last sentence of your first paragraph. Would you happen to have an example where the bundle is nontrivial? The paper says such an example is due to Samelson but doesn't give a reference $\endgroup$ Commented Dec 8, 2021 at 21:09
-
$\begingroup$ Moebius band is a solvmanifold. This is a very special case of the corollary on p.263 of "Vector Bundles Over Tori and Noncompact Solvmanifolds" by L. Auslander and R. H. Szczarba, and of course can also be seen directly. $\endgroup$ Commented Dec 9, 2021 at 0:57
-
$\begingroup$ I specifically meant an example where $ G $ and $ G' $ both have finitely many connected components (as in the first sentence of the first paragraph) but $ G/G' $ is still a nontrivial vector bundle. To the best of my knowledge there is no way to realize the Moebius band as $ G/G' $ where $ G' $ has finitely many connected components. $\endgroup$ Commented Dec 9, 2021 at 4:47
-
$\begingroup$ Maybe, Mostow meant the hyperquadric example mentioned on p.28 of Samelson's survey "Topology of Lie groups", ams.org/journals/bull/1952-58-01/S0002-9904-1952-09544-6/…. I have not checked that $G$, $G^\prime$ have finitely many components. $\endgroup$ Commented Dec 9, 2021 at 13:00