# Least reasonable regularity on Riemannian metric tensor to define a metric [closed]

We assume that $M$ is a smooth manifold of dimension $m$, and we assume that $g$ is a smooth Riemannian metric $g$ on $M$.

If $M$ is moreover path-connected, then $g$ induces a metric (in the sense of metric spaces) on $M$. The distance between two points $a, b \in M$ is the infimum of the lengths of smooth curves in $M$ connecting $a$ and $b$. The length of a smooth curve $c : [0,1] \rightarrow M$ is the integral

$$L(c) = \int_0^1 g( \dot c(t), \dot c(t) ) \; dt.$$

The latter expression makes sense if the Riemannian metric $g$ is not smooth but merely continuous. The raises the following question: how low can the regularity of a smooth curve be so that we still have a well-defined metric in terms of path lengths?

For example, in many applications you assume the Riemannian metric to have merely measurable coefficients. This suffices to define the $L^2$ norms on vector fields. The question is whether you still have a distance function in that case.

## closed as unclear what you're asking by Nate Eldredge, Ben McKay, Willie Wong, Misha, Anton PetruninOct 4 '16 at 17:32

Please clarify your specific problem or add additional details to highlight exactly what you need. As it's currently written, it’s hard to tell exactly what you're asking. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.

• Since the expression for $L(c)$ involves the derivative of $c$, how does it make sense if $c$ is only continuous but not differentiable? It could even be nowhere differentiable. "Almost everywhere differentiable" is plausible but not correct (see the Cantor staircase function). I suspect the condition you are really looking for is "absolutely continuous". – Nate Eldredge Oct 3 '16 at 18:26
• Wait - are you talking about the regularity of the curve $c$ or the metric $g$? It seems you are switching back and forth. – Nate Eldredge Oct 3 '16 at 18:27
• Consider the following example: your metric is standard on the plane, but along one line it is twice smaller. Note that $L^2$-norm on the vector fields is the same as in the standard plane, but you can make a shortcut using the spacial line --- depending on the application you have in mind, try to tell what would be right choice of the induced metric in this case? – Anton Petrunin Oct 3 '16 at 19:43
• Although tangentially relevant, you might be interested to know that there is a way to define a distance for metrics with very low regularity which doesn't require being able to compute the length of curves : $d_g(x,y)=\sup\{|f(y)-f(x)|,\,f:M\xrightarrow{C^1}\mathbb{R},\, |\nabla f|^2_g\leq 1\}$. This in particular makes sense for measurable metrics. However, it won't see the pathology Anton built. – Thomas Richard Oct 4 '16 at 8:24
• @NateEldrege: right, that was just an annoying typo. I corrected it. – shuhalo Oct 5 '16 at 2:43