5
$\begingroup$

Given two (Euclidean or hyperbolic) triangles $ T = ABC $ and $ T' = A'B'C' $, the natural map is the one that sends $ A' \mapsto A $, $ B' \mapsto B $, $ C' \mapsto C $ and maps affinely each side of $T'$ onto the corresponding side of $T$. We say that the triangle $ T' $ dominates the triangle $ T $ if the natural map is a short map (Lipschitz with constant $1$) with respect to the distance in the (Euclidean or hyperbolic) plane.

My question is, given a triangle $T$ with side lengths $(a, b,c)$, is it true that the triangle $T'$ with side lengths $(a+\epsilon, b+\epsilon, c+\epsilon)$ dominates $T$ for all $\epsilon>0$ small enough?

I can prove this statement for Euclidean triangles by some calculations involving the law of cosines, but I couldn't manage to do the same in the hyperbolic plane.

$\endgroup$
6
  • 1
    $\begingroup$ Your map is between the boundaries of the triangles, and not defined in the interior, right? $\endgroup$ Sep 28, 2016 at 11:12
  • 1
    $\begingroup$ Yes, the map is only defined on the boundary. If we know it is a short map we can extend it to the interior of the triangle with the Kirszbraun theorem. $\endgroup$
    – FMB
    Sep 28, 2016 at 11:16
  • $\begingroup$ Can you compute the derivative with respect to $\epsilon$ of the distance between pairs of corresponding points (could be easier than comparing distances)? $\endgroup$ Sep 28, 2016 at 11:19
  • $\begingroup$ This is what I did for the Euclidian case, but the calculation of the derivatives for the hyperbolic gives messy formulas involving hyperbolic sines and cosines that I was unable to manipulate. $\endgroup$
    – FMB
    Sep 28, 2016 at 11:24
  • $\begingroup$ Perhaps try working in barycentric coordinates using metric pulled back from the Klein model. $\endgroup$
    – Neal
    Sep 28, 2016 at 12:54

2 Answers 2

5
$\begingroup$

No, this is not true. Here is an indirect argument (if I made no mistake).

If the statement would be true for $\epsilon > 0$ small enough, then it would be true for all $\epsilon > 0$ (since the maps commute, the set of "good" epsilons for a given $(a,b,c)$ is open; this set is also closed because the Lipschitz condition is closed). We will construct two triangles $(a,b,c)$ and $(a+M, b+M, c+M)$ for which the latter does not dominate the former.

Take $a=b=1$ and $c=2$ (or slightly smaller if you need). Take points $A_1$ and $B_1$ on the sides $BC$ and $AC$ at distance $\frac13$ from $C$. The distance between $A_1$ and $B_1$ is $\frac23$. Now consider a triangle $(x, x, x+1)$ for a very large $x$. I claim that the distance between the corresponding points $A'_1$ and $B'_1$ tends to $0$ as $x$ tends to $\infty$.

By the sine law in a right triangle we have $$\frac{\sinh \frac{A'_1B'_1}{2}}{\sinh\frac{x}{3}} = \sin\frac{\gamma'}{2} = \frac{\sinh\frac{x+1}2}{\sinh x}$$ which implies that $\sinh\frac{A'_1B'_1}2$ goes down as $e^{-\frac{x}6}$.

This means that somewhere inbetween there is an isosceles triangle for which the map is not $1$-Lipschitz even for small $\epsilon$...

$\endgroup$
2
  • $\begingroup$ Thank you, that is a nice argument. Does your argument still work if we allow arbitrary perturbation of the vertices of the triangle ? $\endgroup$
    – FMB
    Sep 28, 2016 at 16:21
  • $\begingroup$ I don't know. Maybe there is a "side elongation" formula different from $(a,b,c) \mapsto (a+\epsilon, b+\epsilon, c+\epsilon)$ that works. I wouldn't be surprised. $\endgroup$ Sep 28, 2016 at 17:06
1
$\begingroup$

We can answer the question positively if we adapt the definition of domination to the hyperbolic setting. Given two hyperbolic triangles $ T = ABC $ and $ T' = A'B'C' $, the natural map is the one that sends $ A' \mapsto A $, $ B' \mapsto B $, $ C' \mapsto C $ and maps each side of $T'$ onto the corresponding side of $T$ with the following parametrization:

A geodesic $[x,y]$ of length $l$ is in bijection with the interval $[0,1]$ via the map that sends $t$ to the point in $[x,y]$ at distance $\sinh^{-1} ( t \sinh (l))$ from $x$; now maps each side of $T'$ to the corresponding side of $T$ with the mapping which sends a point to the point with the same parameter in $[0,1]$.

With this definition of domination it can be check directly using the hyperbolic law of cosines that for small epsilon, the triangle with sides $(a+\epsilon, b+\epsilon, c+\epsilon)$ dominates the triangle with sides $(a,b,c)$.

$\endgroup$
1
  • $\begingroup$ Nice! Do you have an explicit description of an "expanding" extension of this map to the interiors of the triangles? $\endgroup$ Nov 21, 2016 at 15:59

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.