I think this is the dual question to my previous question.

From Gudmundsoon notes, page 60:

We can model the hyperbolic space $\mathbb H^m$ as the super half plane space $\mathbb R^+ \times \mathbb R^{m-1}$ with the metric

$$g_p(X, Y) = \frac{1}{x_1^2} \langle X, Y \rangle_{\mathbb R^m}, $$

where $x = (x_1, \dots, x_m) \in H^m$.

My question is: what is the relationship between the metric and the inner product? What is the "right" way of multiplying those vectors in a space that has this curvature? Clearly the "flat" one inherited from $\mathbb R ^m$ won't make much sense here.

Follow up question: more generally, what's the right way of constructing the inner product on $(\mathbb R^ m, g)$ for an arbitrary metric?

  • $\begingroup$ I'm no expert in Riemannian geometry, but I would assume that the inner product here is not of points from the manifold, but of tangent vectors in the tangent space at point $x$. Since the idea of Riemannian geometry is that the surface is locally Euclidean, using the regular inner product on those makes a lot of sense to me. $\endgroup$ – MvG Feb 13 at 20:24

You seem to be missing some key concepts regarding tangent spaces and Riemannian metrics.

In general, a Riemannian metric on a manifold $M$ is a family of inner products, one for each point $p \in M$.

Given $p,q \in M$, the tangent spaces at $p$ and $q$ are disjoint vector spaces $T_p M$ and $T_q M$. The disjoint union of all these vector spaces is called the tangent bundle and is denoted $$TM = \coprod_{p \in M} T_p M $$

What a Riemannian metric does is to assign an inner product on each $T_p M$. There is not just one inner product.

In the example of $\mathbb H^n$, $T \mathbb H^n = \mathbb H^n \times \mathbb R^n$ and $T_p \mathbb H^n = \{p\} \times \mathbb R^n$. The vector space structure on $T_p \mathbb H^n$ is $(p,v) + (p,w) = (p,v+w)$, and $r (p,v) = (p,rv)$. Given $p = (x_1,...,x_n) \in \mathbb H^n$, and given $\vec v,\vec w \in \mathbb R^n$, what $g_p$ does is to assign the inner product $$\langle (p,\vec v) , (p,\vec w) \rangle = \frac{\langle \vec v, \vec w \rangle_{\mathbb R^n}}{x_1^2} $$

  • $\begingroup$ So you are right I was skipping some details. I finally found what I was looking for: the Gyrovector space, a way of carrying over to the hyperbolic space some sort of euclidean space vector space structure. $\endgroup$ – meto Feb 14 at 5:29
  • $\begingroup$ Then my follow up question would be: is there a standard way of carrying over this construct to an arbitrary Riemann manifold? $\endgroup$ – meto Feb 14 at 5:29
  • $\begingroup$ The best thing to do with a followup question is to close out the current question and post a new question. No-one but me, or someone who reads these comments, will be aware of a followup question buried in the comments of my answer. $\endgroup$ – Lee Mosher Feb 14 at 13:35

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.