Let $\mathbb H$ be the upper half plane. The hyperbolic metric comes from a Riemannian metric on $\mathbb H$: at each point $z = x+iy \in \mathbb H$, the tangent space $T_z(\mathbb H)$ has a natural identification with $\mathbb R^2$, and we define a real inner product on $T_z(\mathbb H)$ by the formula

$$\langle v,w \rangle_z = \frac{v \cdot w}{y^2}$$ where $v \cdot w$ denotes the standard dot product in $\mathbb R^2$. This gives a norm $||v||_z$ in $T_z(\mathbb H)$. The hyperbolic metric in $\mathbb H$ is defined by

$$d(z_1,z_2) = \inf\limits_{\gamma} \int_0^1 ||\gamma'(t)||_{\gamma(t)}dt= \inf\limits_{\gamma} \int_0^1 \frac{\sqrt{x'(t)^2 + y'(t)^2}}{y(t)}dt$$

as $\gamma(t) = x(t) + i y(t)$ runs over all smooth curves $[0,1] \rightarrow \mathbb H$ satisfying $\gamma(0) = z_1, \gamma(1) = z_2$.

Some textbooks I've seen have said "We define the hyperbolic metric by the formula $ds = \frac{\sqrt{dx^2 + dy^2}}{y}$." What in the world does this mean? I don't know what $ds$ means in this context (it doesn't appear to be a differential form), or what it a priori has to do with a metric. Is this abuse of notation for the definition I gave above, or does it actually mean something?

  • $\begingroup$ In differential geometry, especially Riemannian geometry, a metric is usually given by some equation governing the length of a line element. The inf is implied (locally) due to the properties of geodesics. The reason the metric is a line element is because in order to talk about length, one has to borrow the inner product structure from the tangent space of a manifold and somehow carry it over to connect disparate points (thus disparate tangent spaces). An infinitesimal length fits that intuition. $\endgroup$ Dec 10, 2019 at 21:31
  • 1
    $\begingroup$ Officially, a Riemannian metric is written as a quadratic thing: $ds^2 = A\, dx\otimes dx + B(dx\otimes dy + dy\otimes dx) + C\,dy\otimes dy = A\,dx^2 + 2B\,dxdy + C\,dy^2$ in "old-fashioned" notation. This is a symmetric $(2,0)$-tensor (whereas differential forms are skew-symmetric tensors). $\endgroup$ Dec 10, 2019 at 22:35

1 Answer 1


On any Riemannian manifold, here's a description of the formal mathematical relations between $ds$ and various other objects, each of which is some kind of function that is defined on the tangent space $T_z$ of each point $z$.

The Riemannian metric $\langle \cdot, \cdot \rangle$ itself is a positive definite inner product denoted $v,w \mapsto \langle v,w\rangle_z \in \mathbb R$ for each pair $v,w \in T_z$.

Next, $ds^2$ is the quadratic form associated to the inner product, defined by $ds^2_z(v) = \langle v,v \rangle_z \in \mathbb R$ for each $v \in T_z$.

And, finally, $ds$ is the line element associated to the quadratic form, defined to be its square root $ds_z(v) = \sqrt{\langle v, v\rangle }_z$. The meaning of the line element is what you wrote: it is the thing which you integrate to compute arc lengths. You might compare this to the formula for the length of any single vector $v$ in any inner product space, namely $\sqrt{\langle v, v \rangle}$.

I added those little $z$ subscripts to $ds^2$ and to $ds$ for clarity and for comparizon to the $z$ subscript you wrote in your post for the Riemannian metric itself. But to be honest, I never see subscripts like that used for the length element in Riemannian geometry.


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.