3
$\begingroup$

I am more familiar with Cartan geometry, and in this setting we have a notion of development of curves. As described in Cap & Slovak 1.5.17, on a Cartan geometry $(\mathcal{P} \to M, \omega)$ modelled on $G/P$, we have a "Cartan space" which is an associated bundle $S :=\mathcal{P} \times^P G/P \to M$ that comes equipped with a canonical section $O: x \mapsto [u, eP]$ as well as a connection induced by the Cartan connection. Then there is a development map which takes germs of curves in the base $M$ to germs of curves in the model $G/P$ (technically the fibre of the Cartan space until we pick a frame in the $\mathcal{P}$). Explicitly, for a (germ of a) curve $c$ in $M$ with $c(0)=x$, we get (a germ of a) curve $\mathrm{dev}_c$ in the fibre $S_x$ of the Cartan space given by $$\mathrm{dev}_c(t) := \mathrm{Pt}_{c_t}((O(c(t)),-t)$$ where $c_t(s) := c(t+s)$. After picking a frame $u \in S_x$ we can identify $S_x$ with the model $G/P$.

Since n-dimensional (oriented) Hyperbolic geometry can be viewed as a Cartan geometry modelled on $\mathrm{SO}(n,1)/\mathrm{SO}(n)$, we also have this notion of development of curves in hyperbolic geometry.

But in the hyperbolic geometry literature (which I am mostly unfamiliar with) there is another notion of developing/development: For $(M,g)$ a connected hyperbolic $n$-manifold, there is a developing/development map on the universal cover $\mathrm{dev}: \widetilde{M} \to \mathbb{H}^n$ which is an isometry if $M$ is complete (if I remember correctly).

Is there some nice way to related these two notions of development? Or is the naming mostly coincidence?

$\endgroup$
3
$\begingroup$

Of course it is the same. Any submanifold on which the curvature of a Cartan geometry vanishes has a developing map from its universal covering space. I don't know any great reference, but I have used this in many of my papers, for example: https://arxiv.org/abs/1005.1472

$\endgroup$
  • $\begingroup$ Thanks! I've thought up a sketch of a proof and posted it as a reply; would you mind having a look to see if I got it right? $\endgroup$ – ಠ_ಠ Jul 24 '18 at 1:00
  • 1
    $\begingroup$ @ಠ_ಠ: Yes, that is correct. The same idea works for any smooth map $f \colon Z \to M$ of any smooth manifold $Z$ with a chosen point $z_0 \in Z$: if the curvature on $M$ pulls back to zero as a 2-form on $Z$ valued in the pullback of the adjoint bundle of the Cartan geometry bundle, then the universal covering space of $Z$ has a developing map to the model $G/H$ of the Cartan geometry. The developing map is equivariant for a holonomy morphism, i.e. a group morphism $\pi_1(Z) \to G$. $\endgroup$ – Ben McKay Jul 24 '18 at 8:50
  • $\begingroup$ Sharpe in his Differential Geometry: Cartan's Generalization of Klein's Erlangen Program calls this the fundamental theorem of calculus and proves it in one of the earlier chapters. $\endgroup$ – Vít Tuček Aug 8 '18 at 22:06
0
$\begingroup$

Here's a sketch of a proof I've thought up: Since $M$ is flat, the connection is complete and parallel transport is homotopy invariant (rel endpoints). Thus, the development map $\mathrm{dev}$ takes homotopy classes of paths (rel. endpoints) $[c]: [0, 1] \to M$ from $x$ to $y$ in $M$ to paths starting at $O(x)$ in $S_x$ (or starting at $eP$ in $G/P$ if we pick a frame). By the usual construction, the universal cover $\widetilde{M}$ is given by homotopy classes of paths (rel. endpoints). Thus, by taking the endpoint of the developed paths on $S_x$, the development map lifts to a map $\widetilde{M} \to S_x$. Choosing a frame gives a map $\widetilde{M} \to G/P$.

Since hyperbolic manifolds can be realized as flat Cartan geometries modeled on $\mathbb{H}^n \cong SO(n,1)/SO(n)$ by model mutation of the usual Riemannian structure, this recovers the usual hyperbolic development map.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.