# Affine Parameters and Euler-Lagrange Equations

Earlier today I was struggling to see why I couldn’t derive the general geodesic equation (*) $\frac{\textrm{d}}{\textrm{d}\lambda}(g_{ab}\dot{x}^b)-\frac{1}{2}g_{bc,a}\dot{x}^b\dot{x}^c=f(\lambda)g_{ab}\dot{x}^b$

where $f$ a general nonzero function. I had been trying to do this by varying the action $S=\int L \textrm{d}\lambda$ with Lagrangian $L=\frac{\textrm{d}s}{\textrm{d}\lambda}=\sqrt{g_{ab}\dot{x}^a\dot{x}^b}$.

The standard technique is use as a Lagrangian $L'=L^2$ instead, and claim that this produces the same results as we would have got with $L$. I’d always accepted this as gospel, but a simple calculation shows that we need an additional assumption for this to be true.

Indeed the Euler-Lagrange equations for $L'$ give $\frac{\textrm{d}}{\textrm{d}\lambda}(2L\frac{\partial L}{\partial\dot{x}^a})=2L\frac{\partial L}{\partial x^a}$

To regain the Euler-Lagrange equations for $L$ we want to cancel our $2L$ factors. Clearly a sufficient condition for this is that $L$ is constant. But by the definition of $L$ this means that $\frac{\textrm{d}s}{\textrm{d}\lambda}$ is constant, which precisely says that $\lambda$ is an affine parameter.

Hence, by the use of this method we lose the generality needed to obtain equation (*).

Nevertheless it is easy to derive (*) from the affine geodesic equation (**) $\frac{\textrm{d}}{\textrm{d}\lambda}(g_{ab}\dot{x}^b)-\frac{1}{2}g_{bc,a}\dot{x}^b\dot{x}^c=0$

Indeed let $\mu$ be a general parameter and write $\lambda = \lambda(\mu)$. Then rewriting the geodesic equation (**) in terms of $\mu$ using the chain rule yields (*) with $f(\mu)=\frac{\lambda ''}{\lambda '}$ (try it).

Alternatively you can reach a manifestly reparameterisation invariant version of (*) by directly varying our original action $S$. It’s not pretty though!