My book is Connections, Curvature, and Characteristic Classes by Loring W. Tu (I'll call this Volume 3), a sequel to both Differential Forms in Algebraic Topology by Loring W. Tu and Raoul Bott (Volume 2) and An Introduction to Manifolds by Loring W. Tu (Volume 1).

I refer to Section 2.1, Section 2.2, Volume 1 Section 8.6 (Part 1) and Volume 1 Section 8.6 (Part 2).

Use $t$ to denote the standard coordinate on $[a,b]$, and use $t_0$ to denote a point in $[a,b]$. Let $x$ be the standard coordinate on $[0,l]$. The speed of a curve $c: [a,b] \to M$ into a Riemannian manifold $M$ at a point $t_0 \in [a,b]$ is defined $\|c'(t_0)\| := \sqrt{\langle c'(t_0), c'(t_0) \rangle_{t_0}}$. Then we can define speed as a map by $\|c'\|: [a,b] \to [0, \infty), (\|c'\|)(t_0) := \|c'(t_0)\|$. Here, it seems to be claimed that this map $\|c'\|$ is the derivative of arc length function $s$ of $c$.

Question: In the first place, is $c$ supposed to be assumed regular/an immersion for the definition of speed $\|c'\|$, arc length $l$ or arc length function $s$, and why/why not?

My thoughts:

  1. If $c$ is regular/an immersion, then $\|c'\|$ is smooth by this, but I think it's possible to define $\|c'\|$, $l$ and $s$ for continuous $\|c'\|$. I can't think of a condition on $c$ to make $\|c'\|$ continuous but not necessarily smooth (see thought (2) below).

    • 1.1. Edit: I actually didn't mention earlier: Observe that in the paragraph before Proposition 2.3, Tu uses the fundamental theorem of calculus. Based on the version of FTC on Wikipedia, I think the rule behind FTC is something like
    • "continuous $\mathbb R$-valued functions defined on a closed interval $[a,b]$ of $\mathbb R$ are Riemann integrable on $[a,t]$ for any $a<t\le b$"
    • Without such rule, I don't think we can define the "$F$" in the version of FTC on Wikipedia. With such rule, if $\|c'\|$ (the "f") were continuous, then we could define $s$ (the "F") and thus define $l$. If $c$ is regular/an immersion, then $\|c'\|$ is smooth and thus continuous. If $c$ were irregular/not an immersion, then $\|c'\|$ is not necessarily smooth, I think (see thought (2) below). But we can still define $s$ (and thus define $l$) by the rule if $\|c'\|$ is somehow at least continuous.
  2. It could be possible $\|c'\|$ is actually continuous or even smooth for an irregular/a non-immersion, but still smooth, $c$ because in this question, Paulo Mourão can prove the smoothness part without immersion.

  3. Update: I think we can still define $\|c'\|$, $l$ and $s$ for an irregular/a non-immersion $c$ because there's this exercise: Exercise 2.6, which asks for the arc length of a parametrized curve that was shown in Example 2.2 (see here) to be irregular/not an immersion. At the very least $l$ and $\|c'\|$ are defined. Not sure if $s$ is.


  • 1
    $\begingroup$ One generalization you might find interesting is for absolutely continuous curves. I believe you should be able to define $c^\prime$ in the sense of the metric derivative (en.wikipedia.org/wiki/Metric_derivative) using local coordinates, and then you can use the Riemannian inner product to compute $\Vert c^\prime (t) \Vert$. $\endgroup$ – pseudocydonia Aug 5 at 3:17
  • $\begingroup$ @pseudocydonia Thanks. I assume you mean define $\|c'\|$ (with manifold implies Riemannian manifold implies metric space) and not $c'$ (asked about here). Anyway, so indeed, $\|c'\|$, arc length $l$ or arc length function $s$ can be defined for irregular/non-immersion $c$? $\endgroup$ – Selene Auckland Aug 5 at 3:26
  • 1
    $\begingroup$ Yes you're right, I was mistaken, the metric derivative should be able to give you the speed directly - it does not of course actually give you a vector $c^\prime$ in the tangent space. $\endgroup$ – pseudocydonia Aug 5 at 3:28

Here is a good reference that goes into some detail, of how to work with absolutely continuous curves on Riemannian manifolds: http://nyjm.albany.edu/j/2015/21-12v.pdf In other words, there is a reasonable extension of notions like the speed of a curve on a Riemannian manifold so that the answer to your question is negative.

Notably, a similar strategy sometimes allows you to work even with curves which are defined in an abstract metric space, with no manifold structure at all. For this, a good reference is the first half of the book by Ambrosio, Gigli, and Savaré.

  • 1
    $\begingroup$ Sorry, I misspoke again. Absolute continuity should be sufficient to define arc length and the arc length function - no smoothness is required. $\endgroup$ – pseudocydonia Aug 5 at 3:49
  • 1
    $\begingroup$ This is because the definition of absolute continuity directly gives you an arc length function if you look carefully. $\endgroup$ – pseudocydonia Aug 5 at 3:50
  • 1
    $\begingroup$ 7.1 correct, 7.2 see comment below, 7.3 probably not correct. One issue, which I'm sure you've seen, is that non-immersive curves can cross themselves. At these points, the derivative is no longer well-defined. Therefore the notion of speed no longer makes any sense. $\endgroup$ – pseudocydonia Aug 5 at 4:27
  • 1
    $\begingroup$ Sorry, for 7.2: the point is that one can give various, more general definitons of length/speed which do not require the curve to be an immersion, but which nonetheless coincide with the definition you have given, for every curve which actually is an immersion. $\endgroup$ – pseudocydonia Aug 5 at 4:32
  • 1
    $\begingroup$ And I misread, you are right for (7.3). It does not work for non-immersive curves, simply because the derivative tangent vector does not generally exist at every point. $\endgroup$ – pseudocydonia Aug 5 at 4:33

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.