Back to the Contents section

This section follows on from the section on Euler-Langrange equations. The task here is to find the geodesic equation that describes straight lines in general.

I do hope you recall from the other pages that, one form of the Euler-Langrange equation is

are the conditions that finds a local minimum, maximums or inflection point of an integral of f.

because that was indeed a waste of brain power, we're going to ignore that just for now, and first derive the geodesic equation directly. This is so we can get a better handle on what's going on from more then one point of view.

Consider a Tellytubby playing on a slide chute, i.e. undergoing acceleration

If there are no net forces acting on Po (this is the deeper meaning bit of G.R.) in order to achieve this acceleration then we have, from Newton's laws

In our newly acquired, very impressive tensor notation, this can be written, noting that derivatives go over to covariant derivatives always, as

because,

and noting the obvious extension to the ";" is required

So, to continue with

guess what index's we swapped now

and so the first term can be written as

and subbing in again to all terms gets us

Which is the geodesic equation that we are after.

So, this gives one a bit of a feel, one hopes, of what is happening dude

Now to do the difficult bit and show how things all tie up with the variational principle

Consider the path that light takes

so that, using our prior result for distance, one can write

To make the sums all work out, an "affine parameter " for the time is introduced. This is simply to get rid of all those dx's, bloody annoyance that they are.

and dividing out by dl in our distance formula above gives, well after taking the square root and all

hence

or finding the total time

So, now the job is to minimize this integral, Laa Laa oops, I mean ala this is the celebrated least action integral for our geodesic.

When I was plagiarizing researching for this project on the web I found one or two derivations of this result. However, they were all rather more complicated because it is obvious that whatever locally minimizes f^1/2, will also locally minimize plain old f as well, so we'll drop the square root complication and just consider:

First term, and note we have dropped c because we are equating to 0

Second term

where I've changed the notation to make it a bit clearer what's going on. So mentally ignore the dots on the x's when doing the sums. I have filled in all the steps because they were not done in the derivation where I copied the outline of this from. What these poor excuse's for Tellytubby professors don’t realize is that, precisely because the reader is going through these elementary deviations, it inherently implies that the punter is not familiar with these sorts of calculations, and so more guidance is needed. e.g. Note how the delta swap's index's.

We now have then

Ahmm, getting close, seems familiar? Swap one more time

which, by referring to our Christoffel page, is

and, obviously, we can let tau = lambda

amazing, ain't it. How different methods give the same answer.

Copyright © of Kevin Aylward, 2000-2006, all rights reserved.

This paper may be reproduced so long as it is reproduced in full along with this copyright notice, so that your truly gets full credit for it and no profit is made from it

http://www.kevinaylward.co.uk