The question was motivated by this question of Anton Petrunin.

By a metric continuum we understand a connected compact metric space.

Let $p$ be a positive real number. A metric continuum $X$ is called $\ell_p$-almost path-connected if for any points $x,y\in X$ and any $\varepsilon>0$ here exists a family $\big((a_n,b_n)\big)_{n\in\omega}$ of pairwise disjoint open intervals in the unit segment $[0,1]$ and a continuous map $\gamma:[0,1]\setminus\bigcup_{n\in\omega}(a_n,b_n)\to X$ such that $\gamma(0)=x$, $\gamma(1)=y$ and $\sum_{n=0}^\infty d_X(\gamma(a_n),\gamma(b_n))^p<\varepsilon$.

It is easy to see that each almost $\ell_p$-connected metric continuum is $\ell_q$-almost connected for any $q\ge p$.

By my answer to the question of Anton Petrunin, each plane continuum is almost $\ell_1$-connected. By analogy it can be shown that each continuum in $\mathbb R^3$ is $\ell_2$-connected.

Problem. Is there a metric continuum which is not almost $\ell_1$-path connected? not almost $\ell_p$-connected for every $p<\infty$?

  • 1
    $\begingroup$ An easy observation is that if we have an example of a non-almost $\ell_p$-path connected continuum for some $p$, then we can turn it into one which is not almost $\ell_q$-path connected for every $q<\infty$. Specifically if we take a convex function that is 'sharp enough' at 0 and compose it with the metric we'll get the required counterexample. So for instance letting the new metric be $\rho = \sum_{n>0} 2^{-n} d^{1/n}$ (after scaling the diameter of $X$ to be $1$) will work. $\endgroup$ Aug 14, 2018 at 16:31
  • $\begingroup$ @JamesHanson Very good! Thanks. So, it remains to consider some specific metrics like the metrics inherited from the Euclidean spaces. $\endgroup$ Aug 14, 2018 at 17:34

1 Answer 1


Yes there are such examples.

Assume the sequence $\varepsilon_n$ is very fast converging to $0$. Consider a sequence of short $\varepsilon_n$-crooked maps between intervals $\mathbb{J}_n\to \mathbb{J}_{n-1}$. Its inverse limit is a pseudoarc $\mathbb{J}_\infty$; denote by $\phi_n\colon\mathbb{J}_\infty\to \mathbb{J}_n$ the projections.

Equip $\mathbb{J}_\infty$ with the maximal metric such that $$|x-y|_{\mathbb{J}_\infty}\le \tfrac1{2^{n/p}}+|\phi_n(x)-\phi_n(y)|_{\mathbb{J}_n}.$$

Assume that $\mathbb{J}_\infty$ is $\ell_p$-almost path-connected. Let $\gamma$ be the path that connects the ends of $\mathbb{J}_\infty$. It can not have more then one jump of size $1$; the jump brakes $\alpha$ in two arcs one of which has diameter at least $1-\varepsilon_1$; this arc can not have more than 2 jumps of length $\tfrac12$, so one of the subarcs has diameter at least $1-\varepsilon_1-2\cdot \varepsilon_2$ and so on. At the end of the day you see that $\mathbb{J}_\infty$ contains a subarc of $\alpha$ of positive diameter. But $\mathbb{J}_\infty$ has no arcs, a contradiction.

The given construction is nearly identical to Example 4.2 in my paper on intrinsic isometries.

  • $\begingroup$ Great! This example opens a door to considering $\ell_p$-connected components, $\ell_p$-metrics on the space of such components etc. $\endgroup$ Aug 16, 2018 at 20:13
  • $\begingroup$ Can such a pseudoarc be realized in some Euclidean space $\mathbb R^n$? I means a pseudoarc which is not $\ell_1$-connected (since any continuum in $\mathbb R^n$ is $\ell_p$-connected for $p>n$). $\endgroup$ Aug 16, 2018 at 20:17
  • $\begingroup$ @TarasBanakh I also would like to know if it can be realized. By the way, you might be interested in the following geometric reformulation: Given a compact set $P\subset\mathbb{R}^k$, denote by $\mathbb{R}^k/P$ the quotient space with quotient metric. If there is a pseudoarc $P$ that is not $\ell_1$-almost path connected, then the length-metric induced by the map $ \mathbb{R}^k\to \mathbb{R}^k/P$ coinsides with the original metric on $\mathbb{R}^k$. $\endgroup$ Aug 16, 2018 at 20:46
  • $\begingroup$ I am afraid that the situation with reformulation is not so optimistic. The $\ell_1$-almost path-connectedness of $P$ is equivalent to the triviality on $P$ of the length-metric induced by the quotient map.p $\endgroup$ Aug 16, 2018 at 21:34
  • $\begingroup$ @TarasBanakh Right, but what do you mean by "optimistic" (?) I simply want to understand the induced length-metric. According to your argument, in dimesion 2 the length metric for the quotient map is isometric to the quotient (assuming that the inverse images are connected +...). In dimension 3 it is unknown, and I want to know. $\endgroup$ Aug 16, 2018 at 23:20

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.