1

I'd like some assistance in demonstrating $d_{\mathbb{H}}(p,q)=|\log(pq;rs)|$ where $\mathbb{H}$ is a Poincare Upper Half-Plane, and $(pq;rs)$ is a cross ratio of $p,q,r,s$. ($p,q$ are on a geodesic in $\mathbb{H}$ and $r,s$ are the intersection of $x$-axis and the geodesic.)

I tried as follows.

For some $f \in PSL(2,\mathbb{R})$, $p,q,r,s$ moves to $0,i,\infty, \lambda i$ for some $\lambda >0$.

$f$ is an isometry map, so $d_{\mathbb{H}}(p,q)=d_{\mathbb{H}}(i,\lambda i)$.

And, I moved $i,\lambda i$ back to the Poincare disk model $D$ and moved to $B$, the Beltrami model.

By a conformal map that maps $\mathbb{H}$ to $D$, $0 \mapsto -1, \infty \mapsto 1, i \mapsto 0=O, \lambda i \mapsto \frac{\lambda -1}{\lambda +1}=q'$.

And these points get mapped to $-1,1,O,q' \in B$

$$d_{\mathbb{H}}(i,\lambda i)=d_D(O,q')=d_B(O,q')=\frac{1}{2}\left| \log(Oq';(-1) (1))\right|=\frac{1}{2}\log\frac{1\cdot \left\| 1-q'\right\|}{\left\|q'+1 \right\|\cdot 1} $$

And, $\left\|q'+1 \right\|=\frac{\lambda -1}{\lambda +1}+1, \left\| 1-q'\right\|= 1- \frac{\lambda -1}{\lambda +1}$, so $\frac{\left\| 1-q'\right\|}{\left\|q'+1 \right\|}=\frac{1}{\lambda}$.

Therefore, $\frac{1}{2}\log\frac{1\cdot \left\| 1-q'\right\|}{\left\|q'+1 \right\|\cdot 1}=\frac{1}{2}\log\frac{1}{\lambda}=d_{\mathbb{H}}(p,q)$.

$(pq;rs)=((i) (\lambda i);0 \infty)=\frac{i(\lambda i- \infty)}{\lambda i(i-\infty)}=\frac{1}{\lambda}$ because cross-ratio is preserved by $f$.

However, this is different from my $d_\mathbb{H}(p,q)$ because there is $\frac{1}{2}$.

I have no clue where I made this mistake. Maybe it's because of the equation $d_D(O,q')=d_B(O,q')$, but why is this wrong?

john
  • 1,268
  • 1
    Following conventions I'm more accustomed to $d_{\Bbb{H}}(ai, bi) = \log b/a$ and $d_{\Bbb{H}}(p,q) = \log [r,q,p,s]$ likewise $d_{\Bbb{D}}(p,q) = \log [r,q,p,s]$ with $[a,b,c,d]=\frac{a-b}{a-c}\cdot\frac{d-c}{d-b}$, and the first formula for distance would give $d_{\Bbb{H}}(i,\lambda i) = \log \lambda$ while the latter using your $-1,q',0,1$

    $$ \log \frac{(-1-q')(1-0)}{(-1-0)(1-q')}=\log \frac{-(1+q')}{-(1-q')}= \log \frac{1+q'}{1-q'}=\log \lambda $$ Where does that $\frac{1}{2}$ come from in your distance formula for the Poincare disk?

    – sharding4 May 02 '22 at 01:14
  • @sharding4 It is from the distance function of the Beltrami-Klein model. I wanted to bring this distance function as it is defined with cross-ratio. However, I am not sure of the relationship between $D$ and $B$. I thought $d_D(p,q) = d_B(p',q')$ where $p',q'$ are the representative points of $p,q$ in the Beltrami model. – john May 02 '22 at 01:26
  • 1
    I see that now. That's the problem. The distances aren't the same in the Poincare model and the Beltrami-Klein model. I think you'll be good after you resolve $d_D(O,q')=d_B(O,q')$ – sharding4 May 02 '22 at 01:28
  • @sharding4 Yes, then it seems like $d_D(p,q)=2d_B(p',q')$ for any $p,q \in D$. Is it true? – john May 02 '22 at 01:31
  • 1
    Yes. The stereographic projection which creates the Poincare disk model is onto a disk of twice the radius of the hemisphere attached to the Klein model. – sharding4 May 02 '22 at 01:36
  • @sharding4 Thanks a lot for your help! – john May 02 '22 at 01:37
  • 1
    I feel reminded of my answer https://math.stackexchange.com/a/633924/35416 which also explores distance formulas in different models, and how the coefficient in those formulas relates to curvature. – MvG May 02 '22 at 07:34

0 Answers0