Let $(M^n,g)$ be a smooth complete Riemannian manifold. Let $p\in M$ be a point. Recall that the cut locus of $p$ is the set of vectors $v$ in the tangent space $T_pM$ such that $\exp(t v)$ is a minimizing geodesic for any $t\in [0,1]$, but not for $t\in [0,1+\varepsilon )$ for any $\varepsilon >0$.

Question. Does the cut locus have Lebesgue measure 0 in $T_pM$? If yes, does it have Hausdorff dimension at most $n-1$?

If the above questions have negative answers, one may ask the same questions about the exponential image of the cut locus.

  • 9
    $\begingroup$ See Sakai's "Riemannian geometry", lemma III.4.4. Namely if $\tilde C_p$ is the cut locus in $T_pM$ , and $C_p$ its emage under $exp$, then both $\tilde C_p$, $C_p$ are null sets, $dim (\tilde C_p)= n-1$ or $\tilde C_p$ is empty, and the equality becomes an inequality after applying $exp$. Here I think $dim$ is the covering dimension (but I have not thought enough about this). $\endgroup$ Nov 28, 2016 at 13:03
  • $\begingroup$ @IgorBelegradek: Thanks! Why not to write it as a final answer? $\endgroup$
    – asv
    Nov 28, 2016 at 14:39
  • 4
    $\begingroup$ Because you asked about Hausdorff dimension and I do not know much about this. I know Itoh, Tanaka and et el. worked on this, see projecteuclid.org/euclid.tmj/1178224899 but I have no time now to trace the references and write an intellegent answer. Would you do that yourself? $\endgroup$ Nov 28, 2016 at 14:46

1 Answer 1


The key fact is that the cut time $t_c : UM \to \mathbb{R}$, defined on the unit tangent bundle $UM$ of a complete, $n$-dimensional Riemannian manifold, is locally Lipschitz continuous around all $v \in UM$ such that $t_c(v) < +\infty$. Hence the tangential cut locus at $p \in M$, that is $$ \tilde{C}_p = \{t_c(v)v\mid v \in UM,\quad t_c(v) < +\infty\} \subset T_q M, $$ either is empty, or it has Hausdorff dimension exactly $n-1$ (being the graph of a locally Lipschitz function). The exponential map being smooth, it cannot increase the Hausdorff dimension, hence $\dim(C_p) = \dim(\exp_p(\tilde{C}_p)) \leq n-1$.

All of this is proved here by Itoh and Tanaka.

  • 1
    $\begingroup$ Is the cut-locus ever a sub-manifold? $\endgroup$
    – ABIM
    May 19, 2020 at 22:57
  • $\begingroup$ @ABIM Just found this question. Even though I've forgot a lot of the differential geometry I studied, I'd day yes, as here: math.stackexchange.com/questions/364745/cut-locus-of-mathbbcpn, you'll see a proof of the cutlocus of $CP^n$ is $CP^{n-1}.$ Similar case for $RP^n.$ So yes to your question(?) $\endgroup$ Aug 5, 2023 at 16:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.