If $(X,d)$ is a metric space and $f : X \rightarrow \mathbb{R}$ is a Lipschitz function with Lipschitz constant $k < 1$, then the function $$ D(x,y) := d(x,y) + f(y) - f(x) $$ defines an asymmetric metric on $X$. I'll denote this procedure---for reasons I'll explain later---as adding a coboundary to $d$

Note that if $x,y$, and $z$ are three points in $X$ satisfying $d(x,y) + d(y,z) = d(x,z)$, then they also satisfy $D(x,y) + D(y,z) = D(x,z)$, which entails that the geodesics in $(X,d)$ are also geodesics in $(X,D)$.

Question. Are there interesting examples of finite or infinite graphs $\Gamma$ (perhaps some Cayley graph) such that if $D$ is an asymmetric metric with the same geodesics as the standard metric on $\Gamma$, then $D$ is obtained by adding a coboundary to a multiple of the standard metric?

Motivation for the problem and its title

I am looking for a wider setting and/or interpretation of a theorem of mine that generalizes what Gautier Berck and I did for projective Finsler metrics:

Theorem. If $D$ is a continuous asymmetric metric on the $n$-sphere $S^n$ for which all great circles are geodesics, there exists a (continuous) function $f : S^n \rightarrow \mathbb{R}$ such that $$ D(x,y) - D(y,x) = f(y) - f(x). $$

In other words, $D$ is obtained from the (symmetric) metric $d(x,y) := (D(x,y) + D(y,x))/2$ (for which all great circles are also geodesics) by adding a coboundary. This settles Hilbert's fourth problem for continuous asymmetric metrics in $n$-dimensional projective space by reducing the problem to continuous symmetric metrics treated by Busemann, Pogorelov, and Szabo.

Reading an old paper of Kolmogorov (Skew-symmetric forms and topological invariants in Vol.1. of his collected works), which is basically Kolmogorov's early version of the Alexander-Spanier cohomology, suggested the following setup:

Given an asymmetric metric $D$, consider the anti-symmetric function $g(x,y) := D(x,y) - D(y,x)$ as a 2-cochain. If $D$ has the same geodesics as its symmetrization, $d(x,y) := D(x,y) + D(y,x)$, then the differential of $g$, defined as $$ \delta g \, (x,y,z) = g(y,z) - g(x,z) + g(x,y) \, , $$ vanishes on all triples $(x,y,z)$ in which $y$ is between $x$ and $z$ (i.e., $d(x,y) + d(y,z) = d(x,z)$). I guess one could say somewhat loosely that "the differential of $g$ vanishes on geodesics". In any case, this is exactly the condition for $D$ and its symmetrization to share the same geodesics

The very loosely stated question would then be

Under what general conditions can we guarantee that if the 2-cochain $g(x,y) := D(x,y) - D(y,x)$ vanishes on all geodesics of the symmetric metric $D(x,y) + D(y,x)$, then $g$ is a cocycle or even a coboundary?

One can even forget about metrics, prescribe a system of "geodesics" on $X$ by giving a subset
$$ G \subset X \times X \times X $$ (i.e. $(x,y,z) \in G$ if $y$ is between $x$ and $z$) and ask for the condition that a 2-cochain that vanishes on $G$ be a cocycle or a coboundary.


This is a further comment on the question, which I add for the sake of completeness.

A simple remark that came out of a discussion with Jimmy Dillies:

Proposition. If $D$ is any (asymmetric) length metric on the real line, then $D(x,y)-D(y,x)$ is a coboundary.

Proof. First we show that $g(x,y) = D(x,y) - D(x,y)$ is a cocycle (i.e., $\delta g \equiv 0$). Indeed, if $x \leq y \leq z$, then $g(x,y,z) = 0$ because $D(x,y) + D(y,z) = D(x,z)$ and $D(z,y) + D(y,x) = D(z,x)$. However, the order of $x$, $y$, and $z$ is not important because $\delta g$ is alternating.

Setting $f(x) = g(0,x)$ and using the equation $\delta g(0,x,y) = 0$, we obtain $$ g(x,y) = f(y) - f(x) = \delta f(x,y). $$


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.