5

I am looking for a inner metric on $\mathbb{R}^2$ (that induces the standard topology) which is not Finsler.

By "Finsler" here I mean a metric that is obtained by the following construction:

  1. pick a smooth structure on $\mathbb R^2$ and take a suitable continuous function $\mu$ on $T\mathbb R^2$
  2. define a metric as $d(x,y):=\inf_\gamma \int \mu(\dot\gamma) dt $ over all piecewise smooth paths $\gamma$ connecting two points $x,y$.

If exists, it is necessarily non-invariant (Berestovski theorem).

YCor
  • 60,149
Dmitrii Korshunov
  • 2,147
  • 4
    What about the geodesic metric on the cone $x^2+y^2=z^2$, $z\ge 0$ (which is homeomorphic to the plane?) Or similarly, the quotient of $\mathbf{R}^2$ by $x\mapsto -x$, with distance induced by $\min(d(x,y),d(x,-y))$. – YCor Nov 29 '20 at 19:25
  • 3
    The real question (originally due to Busemann) is "what is the characterization of Finsler distance functions among path-metrics on surfaces?" See "Foundations of singular Finsler geometry" by P.Andreev for some recent progress in this direction. – Moishe Kohan Nov 29 '20 at 22:31
  • @MoisheKohan it's somewhat open-ended. Characterization in which terms? – YCor Nov 29 '20 at 22:59
  • @YCor: Similar to Nikolaev's theorem characterizing Riemannian metrics among path metrics. The first condition would be local extendibility of geodesics, which fails in both examples that you gave. Compare my answers here and here. – Moishe Kohan Nov 29 '20 at 23:33
  • See also 2nd answer here. – Moishe Kohan Nov 30 '20 at 01:32
  • Will you be interested in a 3-dimensional example? – Anton Petrunin Dec 17 '23 at 19:21
  • @AntonPetrunin very much interested! in this case i would like to have something which is not sub-Finsler too, and not of a cone-type singularity type suggested by YCor in the first comment. – Dmitrii Korshunov Dec 17 '23 at 20:32

1 Answers1

4

Your metric has to be geodesic, in particular $$d(x,y)=\min \{\,1,|x-y|\,\}$$ is not Finsler in your sense.

Now let $d$ be a geodesic metric on $\mathbb{R}^n$.

Suppose $n=2$. In "Two counterexamples..." by Burago, Ivanov, and Shoenthal, it was conjectured that a neighborhood of any point in $(\mathbb{R}^2,d)$ admits a Lipschitz embedding into the Euclidean plane.

Suppose $d$ is a Finsler metric in your sense. Since $\mu$ is continuous, the natural map $(\mathbb{R}^n,d)\to \mathbb{R}^n$ is a locally lipschitz homeomorphism. So this conjecture is closely related to your question.

Now suppose $n=3$. The same paper provides an example of a metric $d$ on $\mathbb{R}^3$ that (locally) does not admit a Lipschitz embedding into $\mathbb{R}^3$. In particular, $d$ is not a Finsler metric in your sense. (The construction is interesting --- it is worth reading.)

Anton Petrunin
  • 43,739