Calculation mistake in variation of length functional?

228 Views Asked by At

This should be pretty simple to check if you know the basics of variational calculus. I feel like I am making an obvious mistake somewhere like not using chain rule somewhere. Let $g : \mathbb{R}^n \times \mathbb{R}^n \rightarrow \mathbb{R}$ be a smooth metric on some domain in $\mathbb{R}^n$, and let $\gamma : [0,1] \rightarrow \mathbb{R}^n$ be a smooth path in this domain. I'm trying to show that minimising the length functional $$A(\gamma) = \int_0^1 \sqrt{g(\dot{\gamma},\dot{\gamma})} dt$$ is equivalent to minimising the action functional $$E(\gamma) = \frac{1}{2} \int_0^1 g(\dot{\gamma},\dot{\gamma}) dt.$$ I can do this in coordinates no problem, but this calculation is a bit messy. In general, the Lagrangian $L$ is not linear, but in this case $g$ is bilinear.

Let $v \in C^1$, $v(0) = v(1) = 0$. Question: is this use of the metric $g$ correct, and can we conclude the result via this calculation, or am I forgetting to use the chain rule somewhere? \begin{align*} A(\gamma+tv) &= \frac{d}{dt}\bigg|_{t=0} \int_0^1 \sqrt{g(\dot{\gamma}+t\dot{v},\dot{\gamma}+t\dot{v})} dt \\ &= \frac{d}{dt}\bigg|_{t=0} \int_0^1 \sqrt{g(\dot{\gamma},\dot{\gamma}) + 2t g(\dot{v},\dot{\gamma}) + t^2 g(\dot{v},\dot{v})} dt \\ &= \int_0^1 \frac{1}{\sqrt{g(\dot{\gamma},\dot{\gamma})}}g(\dot{v},\dot{\gamma})dt \end{align*}

By comparison, the first variation of the action functional is $$\frac{d}{dt}\bigg|_{t=0} E(\gamma+tv) = \int_0^1 g(\dot{\gamma},\dot{v})dt.$$ So if $\gamma$ has constant speed, i.e. if $g(\dot{\gamma},\dot{\gamma})$, then it seems that minimizing these functionals is identical. But this seems too easy (no derivatives of $g$ coming out)! Also, I can't get the Euler-Lagrange equations for geodesics to come out of this computation (integration by parts, right? -- so that's where the derivatives of $g$ would come out?).

A simple "incorrect" will be appreciated, and an explanation why even more so. Thanks!

1

There are 1 best solutions below

0
On

Thanks to user98130's comment. Came back later and it seemed obvious. Anyway, because there is no dependence on $\gamma$ (which I'll call $f$ for simplicity) or $v$, the Euler-Lagrange equation is "homogeneous": \begin{align*} \int_0^1 g(\dot{f},\dot{v}) dt &= \int_0^1 g_{ij}\dot{f}^i\dot{v}^j dt \\ &= g_{ij}\dot{f}^iv(t)\bigg|_{t=0}^{t=1} - \int_0^1 \frac{d}{dt}(g_{ij}\dot{f}^i)v dt \\ \frac{d}{dt}(g_{ij}\dot{f}^j) &= \frac{\partial g_{ij}}{\partial f^k}\dot{f}^k\dot{f}^i + g_{ij}\ddot{f}^j = 0, \end{align*} which are the geodesic equations when $g(\dot{f},\dot{f}) = 0$.