Can you numerically solve the geodesic equation in the most general setting?

227 Views Asked by At

I'm currently trying to solve the geodesic equation in the most general setting possible. A curve $\gamma(s) = (\gamma^0(s), \gamma^1(s), \gamma^2(s), \gamma^3(s))$ satisfies the geodesic equation if

$$\frac{d^2\gamma^i}{ds^2} + \sum_{a,b}\Gamma_{ab}^i(\gamma(s))\frac{d\gamma^a}{ds}\frac{d\gamma^b}{ds} = 0$$ for $i\in\{0,1,2,3\}$. Making the substitution $\tau^i = d(\gamma^i)/ds$ you can reduce the system to one twice as big of the form $$\frac{d\tau^i}{ds} + \sum_{a,b}\Gamma_{ab}^i(\gamma(s))\frac{d\gamma^a}{ds}\frac{d\gamma^b}{ds} = 0$$ I tried to solve it as one usually does (that is, writing it in the form $y' = f(s, y)$) but, although factoring $\tau'$ is easy, one cannot do such a thing for $d\gamma^i/ds$: assuming one has the Levi-Civita connection (that is, $\Gamma^i_{ab} = \Gamma^i_{ba}$) one ends up with something of the form $$a(s) + b(s)\frac{d\gamma^i}{ds} + \Gamma_{ii}^i(\gamma(s))\left(\frac{d\gamma^i}{ds}\right)^2 = 0$$ Is there a method for solving one of these ODEs?; in summary, can one numerically solve an ODE of the form $(y')^2 + p(t)y' + q(t) = 0$?