Suppose we are given a certain Riemannian metric $$\frac{(dx)^2 + (dy)^2}{y^2}$$

How do I calculate geodesics?

I know that we have the geodesic equation $$\ddot{x}^k(t) + \dot{x}^i(t)\dot{x}^j\Gamma_{ij}^k(x(t)) = 0$$

and that we have only two components, hence a geodesic $\gamma$ should be of the form $\gamma(t) = (x^1(t),x^2(t))$. Writing out the case $k = 1$ we get $$\ddot{x}^1(t) + (\dot{x}^1(t))^2\Gamma_{11}^1(x(t)) + 2\dot{x}^1(t)\dot{x}^2(t)\Gamma_{12}^1(x(t)) + (\dot{x}^2(t))^2\Gamma_{22}^1(x(t)) = 0$$

I also calculated $\Gamma^1_{11} = \Gamma^1_{22} = 0$ and $\Gamma^1_{12} = -1/y$. Now my problem is that I do not really know how to plug in the Christoffel symbols since they are evaluated in $x(t)$.

I know that there is a similar problem here, but I do know how to solve the resulting ODE. My problem is, that I would like to get to the initial equation.

Edit. The geodesic equation is taken from Riemannian Manifolds by John M. Lee, p. 58.

  • $\begingroup$ Just plug $\gamma(t)=(x^1(t),x^2(t))$ in Christoffel symbols. For example, $\Gamma_{12}^1=-1/y$, so on $\gamma$ we have $\Gamma_{12}^1=-1/x^2(t)$. $\endgroup$ – Cave Johnson Dec 18 '16 at 12:31
  • $\begingroup$ Sorry but I don't quite understand what you mean :( $\gamma$ is of course not a vector: it's a curve, or a map, but $\gamma(t)$, i.e., the value of $\gamma$ at some $t$, can be considered as a vector(more precisely, a point), if this is what you are wondering. And Lee actually meant $\ddot{x}^k(t) + \dot{x}^i(t)\dot{x}^j\Gamma_{ij}^k(\gamma(t)) = 0$ in the equation, since $\gamma(t)=(x^1(t),\cdots,x^n(t))$, while $x(t)$ is usually short for the point $(x^1(t),\cdots,x^n(t))$. $\endgroup$ – Cave Johnson Dec 18 '16 at 12:44
  • $\begingroup$ @CaveJohnson Thanks, sorry for the confusion (deleted my comment because I wrote too fast). You are right. This was also my initial thought, but you will end up with $\ddot{x}^1(t) - 2\frac{\dot{x}^1(t)\dot{x}^2(t)}{x^2(t)} = 0 $ wheras for example here people.maths.ox.ac.uk/hitchin/hitchinnotes/Geometry_of_surfaces/… we just find the ODE $d/dt (x'/y^2) = 0$ which is a bit simpler to solve. Thus I thought I was wrong in the beginning. $\endgroup$ – TheGeekGreek Dec 18 '16 at 12:48
  • $\begingroup$ But in our situation we also come to solve the same ODE. Since $\Gamma_{11}^1=\Gamma_{22}^1=0$ and $\Gamma_{12}^1=-1/y$, the equation comes to $\ddot x^1(t)-2\dot x^1(t)\dot x^2(t)/x^2(t)=0$, or equivalently $\dfrac d {dt}\left(\frac{\dot x^1(t)}{(x^2(t))^2}\right)=0$, which is the same as what you mentioned, isn't it? $\endgroup$ – Cave Johnson Dec 18 '16 at 12:56
  • $\begingroup$ @CaveJohnson How to resolve $x'=cy^2$ ? $\endgroup$ – lanse7pty Dec 18 '16 at 13:03

$x(t)$ means the same thing as $\gamma(t)$ in our situation, where $\gamma(t)=(x^1(t),\ldots,x^n(t))$, since $x(t)$ is usually short for $(x^1(t),\ldots,x^n(t))$. So we only need to plug $(x^1(t),\ldots,x^n(t))$ in the Christoffel symbol $\Gamma_{ij}^k(x^1,\ldots,x^n)$. For example, in this particular problem we have $$\Gamma_{11}^1=\Gamma_{22}^1=\Gamma_{12}^2=\Gamma_{21}^2=0,~\Gamma_{11}^2=1/y,~\Gamma_{22}^2=\Gamma_{12}^1=\Gamma_{21}^1=-1/y$$ Then plug in $x=x(t),y=y(t)$ to see that along the geodesic $\gamma$, we have $$\Gamma_{11}^1=\Gamma_{22}^1=\Gamma_{12}^2=\Gamma_{21}^2=0,~\Gamma_{11}^2=1/y(t),~\Gamma_{22}^2=\Gamma_{12}^1=\Gamma_{21}^1=-1/y(t)$$ The geodesic equation becomes \begin{align} \ddot x(t)-2\dot x(t)\dot y(t)/y(t)=0\\ \ddot y(t)+(\dot x(t))^2/y(t)-(\dot y(t))^2/y(t)=0 \end{align} The first equation can be integrated easily to $$\dot x(t)=c(y(t))^2$$ While the second equation is more complicated to deal with. However, it can be checked that $$\dfrac d{dt}\frac{\dot x^2+\dot y^2}{y^2}=\frac{2}{y^3}((\dot x\ddot x+\dot y\ddot y)y-(\dot x^2+\dot y^2)\dot y)=0$$ which leads to another integration. Now it's not hard to solve those equations. The complete solution can be found here, which is a semicircle centered on the real axis.

  • $\begingroup$ Sorry to ask again, but I do not yet really see, why I have to calculate $d/dt ...$ in the last equation. I mean, I do not see this from the system of geodesic equations. $\endgroup$ – TheGeekGreek Dec 18 '16 at 14:28
  • $\begingroup$ @TheGeekGreek It can be infered from a geometric point of view (geodesic curvature is $0$, or tangent vector is orthogonal to its covariant derivative) that $\langle\dot\gamma(t),\dot\gamma(t)\rangle$ remains constant along the geodesic $\gamma$. This is where the calculation orignates. $\endgroup$ – Cave Johnson Dec 18 '16 at 14:44
  • $\begingroup$ Sorry, if I was not clear. This was not exactly what I meant: I mean, how does $\dfrac d{dt}\frac{\dot x^2+\dot y^2}{y^2} = 0$ helps to solve $\ddot y(t)+(\dot x(t))^2/y(t)-(\dot y(t))^2/y(t)=0$? $\endgroup$ – TheGeekGreek Dec 18 '16 at 14:54
  • $\begingroup$ @TheGeekGreek We do not need the second equation any more, since it is already contained in $\dot x=cy^2$ and $\dot x^2+\dot y^2=c'y^2$. We only need to consider the latter two equations, which is quite easy. $\endgroup$ – Cave Johnson Dec 18 '16 at 14:59

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.