0
$\begingroup$

Consider $X = (\mathbb{R^n},c)$, where $c$ is the equivalence class of all torsion free affine connections having straight lines as unparameterized geodesics. What is the group of symmetries of $X$? This ought to be known. Thank you.

$\endgroup$
2
  • 7
    $\begingroup$ Google “The fundamental theorem of projective geometry“. $\endgroup$
    – Misha
    Commented Jul 28, 2019 at 6:34
  • $\begingroup$ @Misha, thank you. I did not know this theorem. Thank you. $\endgroup$
    – Malkoun
    Commented Jul 28, 2019 at 8:21

1 Answer 1

1
$\begingroup$

Long answer taking a complicated but rewarding detour: This is example that fits into the framework of Cartan geometries. Category of manifolds with an equivalence class of torsion free affinne connections that have the same unparametrized geodesics is equivalent to (properly normalized) Cartan geometry modeled on a homogeneous space $G / P$ where $G = SL(n+1, \mathbb{R})$ and $P$ is the (parabolic) subgroup of $G$ that stabilizes a line in $\mathbb{R}^{n+1}$. By the fundamental theorem of Cartan geometries, if the Cartan curvature is zero then the structure is locally isomorphic to open subset of the homogeneous space $G/P$ (and the automorphism group is maximal up to "discrete part"). If you calculate the Cartan curvature in your case you indeed get zero. The proof of the fundamental theorem of Cartan geometries even gives you the isomorphism kind of explicitly via development map. Since your manifold is simply connected, the development map should provide global isomorphism onto an open subset of $G/P \simeq \mathbb{R}P^n.$ It remains to check which subgroup of $G$ fixes this subset.

$\endgroup$
7
  • $\begingroup$ I think I get it, except you are doing the projective version, rather than the affine version. Could you maybe indicate that for future readers? Thank you for your answer. $\endgroup$
    – Malkoun
    Commented Jul 29, 2019 at 12:05
  • $\begingroup$ @Malkoun I don't understand. Projective structure on a manifold is by definition a class of torsion-free affinne connections that have the same unparametrized geodesics. $\endgroup$ Commented Jul 29, 2019 at 12:20
  • $\begingroup$ I mean I asked about real affine $n$-space, so I think $G$ should be the subgroup of $SL(n+1,\mathbb{R})$ which preserves the hyperplane at infinity. $\endgroup$
    – Malkoun
    Commented Jul 29, 2019 at 12:26
  • $\begingroup$ Thus the group of affine transformations, rather than the group of projective transformations. A minor modification. $\endgroup$
    – Malkoun
    Commented Jul 29, 2019 at 12:30
  • $\begingroup$ It is not correct that the automorphism group of any Cartan geometry modelled on $G/P$ (even with $G=SL(n+1,\mathbb{R})$) is a subgroup of $G$. For example, the translation invariant flat projective structure on the torus (induced by the standard flat Riemannian metric) contains the torus of translations as a subgroup, but there is no $n$-dimensional torus subgroup inside $G$. $\endgroup$
    – Ben McKay
    Commented Jul 29, 2019 at 13:01

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .