I wanted to calculate the radius of the inscribed circle of an ideal triangle.

and when i dat calculate it i came to $\ln( \sqrt {3}) \approx 0.54 $ (being arcos(sec (30^o)) but then at https://en.wikipedia.org/w/index.php?title=Ideal_triangle&oldid=668440011

it says that the equilateral triangle that is made by connecting the intersections of the ideal triangle and its [[inscribed circle]] has a side length of $ 4\ln \left( \frac{1+\sqrt 5}{2}\right) \approx 1.925 $

but this would mean that the radius is less than half the side which is impossible

so didI make a mistake, or is wikipedia wrong?

PS the wikipedia page is corrected now and gives the right values)

  • $\begingroup$ Your calculation of $\ln(\sqrt{3})$ looks right to me. Another way to get it is to use the ideal triangle in the upper half plane having vertices $0,1,\infty$. Its inscribed circle has center $\frac{1}{2} + \frac{\sqrt{3}}{2}i$, and that circle intersects the hyperbolic line with endpoints $0,1$ at the point $\frac{1}{2} + \frac{1}{2} i$. The distance between these two points is $\ln(\sqrt{3})$. $\endgroup$ – Lee Mosher Dec 8 '15 at 17:53
  • $\begingroup$ Thanks I did it in a similar way (only using the points $-1,1,\infty i $, Am I correct wikipedia is wrong and the sides of equilateral triangle that is made by connecting the intersections of the ideal triangle and its inscribed circle is $\ln ( (3+\sqrt{5}) / 2) $ also could you transform your comment as answer then I can accept it $\endgroup$ – Willemien Dec 8 '15 at 20:30

I think the Beltrami-Klein model is particularly useful here since a hyperbolic line is just a straight chord there. For simplicity use a regular triangle as the inscribed ideal triangle. Its inscribed circle has an Euclidean radius which is half that of the fundamental circle. So the chord along the radius gets divided $1:1$ by the midpoint but $3:1$ by the point on the circle, therefore the radius is


just as you computed.

Now to the edge length of the inscribed triangle. A bit of messing around with the Pythagorean theorem will tell you that the ratio between the Euclidean edge length and the length of its supporting chord is $1:\sqrt5$. Which means one endpoint divides the chord in a ratio of $\sqrt5-1:\sqrt5+1$ and the other in the reciprocal ratio $\sqrt5+1:\sqrt5-1$ so the length between them is

$$\frac12\ln\left(\frac{\sqrt5+1}{\sqrt5-1}\cdot\frac{\sqrt5+1}{\sqrt5-1}\right) =\ln\frac{\sqrt5+1}{\sqrt5-1} \approx0.96$$

This looks like a factor of two mistake. Some extra computation shows that this is in fact the case, since

$$\frac{\sqrt5+1}{\sqrt5-1} = \frac{\sqrt5 + 3}2 = \left(\frac{\sqrt5 + 1}2\right)^2 = \varphi^2$$

I will edit Wikipedia.

  • $\begingroup$ thanks addeded wikipedia aswell am wondering is the link to golden ratio really meaningfull $\endgroup$ – Willemien Dec 10 '15 at 8:12
  • $\begingroup$ @Willemien: Well, it's nice in terms of en.wikipedia.org/wiki/Special:WhatLinksHere/Golden_ratio to see where this thing does crop up more or less unexpectedly. $\endgroup$ – MvG Dec 10 '15 at 10:37
  • $\begingroup$ curiously math.stackexchange.com/a/114231/88985 gives another value of the edge length of the inscribed triangle. $\endgroup$ – Willemien Jan 7 '16 at 21:24
  • $\begingroup$ @Willemien: There is no contradiction. Have a look at this construction (created with Cinderella). On the left you have the Beltrami-Klein model, on the right the Poincaré disk (for some reason without the inscribed circle). The thing you described is the line $EG$ between incenter contact points. That other question asks about the distance, i.e. requires the normal $EK$ with a right angle at $K$. That's a different line, leading to a different distance. And the numeric result from Cinderella supports Will Jagy's answer there. $\endgroup$ – MvG Jan 7 '16 at 21:45
  • $\begingroup$ @Willemien posted an answer here with radius of the circle and edge length of the inscribed triangle, written using the upper half plane model. $\endgroup$ – Will Jagy Jan 7 '16 at 23:17

In the upper half plane model, we take the ideal triangle as $x=-1,$ $x=1,$ and $x^2 + y^2 = 1.$ For this problem, you want the inscribed circle, which is $$ x^2 + (y-2)^2 = 1, $$ which meets the ideal triangle at $(-1,2),$ $(1,2),$ $(0,1).$

The diameter along the $y$ axis goes from $(0,1)$ to $(0,3).$ The geodesic center of the circle is at the geometric mean of $1$ and $3,$ therefore $(0, \sqrt 3).$ Recall from my 2012 answer that this geodesic is $(0,e^t).$ Therefore the radius is $\log {\sqrt 3} - \log 1$ = $(1/2) \log 3.$

One edge of the inscribed triangle passes through $(1,2),$ $(0,1).$ This segment is unit speed parametrized by $$ 2 + \sqrt 5 \tanh t, \sqrt 5 \operatorname{sech} t $$ and both ending $t$ values will be negative, because $0 < 2$ and $1 < 2.$ Let us just use the reciprocal of $\cosh t$ and be careful a bout signs.

One endpoint has $$ \frac{\sqrt 5}{ \cosh t} = 1 $$ and the other has $$ \frac{\sqrt 5}{ \cosh t} = 2, $$ so $$ \cosh t = \frac{\sqrt 5}{2} $$ and $$ \cosh t = \sqrt 5 $$

Looked it up, $$ \operatorname{arcosh} x = \log \left(x + \sqrt {x^2 - 1}\right) $$ so the absolute value of the $t$ difference is

$$ \log \left( \sqrt 5 + 2 \right) - \log \left( \frac{\sqrt 5 + 1}{2} \right) = \log (\varphi^2 + \varphi) - \log \varphi = \log (\varphi + 1) = \log (\varphi^2) = 2 \log \varphi \approx 0.96242365 $$


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.