0
$\begingroup$

I am following a lecture series for Classical Field Theory in which they the lecturer uses the invariance of the length of a vector during a Lorentz transformation to derive the equation $$O^{T}\eta~O=\eta. $$ He derives this equation by the following these steps: Taking a vector $v$ and taking its length $$|v|^{2}=v^{T}\eta~v$$to be constant in a transformation we would get $$|v'|^{2}=v'^{T}\eta~v'$$ where $v'=O~v$, $O$ being the Lorentz transformation matrix. Further he writes $$|v'|^{2}=(v^{T}~O^{T})\eta~(O~v)$$ and then $$O^{T}\eta~O=\eta$$ for the length of the vector to be constant. Now my question is shouldn't the metric tensor $\eta$ itself not change under coordinate transformation and the equation should be $$O^{T}\eta'~O=\eta?$$ Because he then moves forth to derive some properties of $O$ using the equation $$O^{T}\eta~O=\eta.$$

$\endgroup$
1
$\begingroup$

This is a common confusion and it's just due to some tricky notation.

  • Every rank 2 tensor $A$ transforms as $A' = O^T A O$.
  • Your book has shown that for Lorentz transformations, the metric $\eta$ satisfies $\eta = O^T \eta O$.
  • Therefore, the metric transforms as $\eta' = \eta$ under Lorentz transformations. This is a very special condition that isn't true for a general rank 2 tensor.
$\endgroup$
  • $\begingroup$ where does this condition of $\eta=\eta'$ come from? $\endgroup$ – Naman Agarwal Dec 12 '17 at 12:42
  • $\begingroup$ @NamanAgarwal Just combine the first two equations. $\endgroup$ – knzhou Dec 12 '17 at 12:45
  • $\begingroup$ I have learnt the general way in which tensors transform but have never heard of the in variance of the metric tensor...so i am asking why is it so? $\endgroup$ – Naman Agarwal Dec 12 '17 at 12:49
0
$\begingroup$

Let's start with 3-dimensional euclidian world. Consider 3D rotations represented by matrices R. This should be demonstrative enough to show that the length of a 3-dim. vector is invariant under rotations.

$ |v|^2 =v^T v$ is the length of a 3D arbitrary vector. In the rotated system the length is $ |v'|^2 =v'^T v' = (v^T R^T)(Rv) = v^T v = |v|^2 $ since the rotation matrices fulfill $R^T R = id \equiv 1$ It can be also be written like $R_{ij}\delta_{jk}R_{kl}= \delta_{il}$ with $\delta$ as Kronecker symbol which now has the role of the metric, i.e. in component-free language we would write for it : $R^T\delta R=\delta$. Of course one can imagine a coordinate transformation to polar coordinates, then the corresponding transformation matrices P no longer fulfill $P^T P = id$. But of course the length of an arbitrary vector is the same in polar coordinates than in cartesian coordinates.

Now we'll consider Minkowski space. Actually the Lorentz-transformations $O$ are those types of coordinate transformations which fulfill $\eta = O^T \eta O$. The Minkowski metric is not changed by the Lorentztransformation is another phrasing of this fact. And as in euclidian 3D space we can carry out a coordinate transformation $C$ where $\eta' = C^T \eta C$. But imagine a transformation which would only change the diagonal elements in $\eta$.

Normally:

$\eta = \left( \begin{array}{cccc} 1 & 0 & 0 & 0 \\ 0 & -1 & 0 & 0 \\ 0 & 0 &-1 & 0\\ 0 & 0& 0 & -1 \end{array} \right)$

Imagine the matrix element $\eta'_{00}=1.2$ after the coordinate transformation instead of $\eta_{00}=1$. Then the invariant line element would be $ds^2 = 1.2 c^2 dt'^2 -dx'^2 -dy'^2 -dz'^2$. In these coordinates the speed of light would be no longer $c$. Actually such kind of coordinate transformations exist, but evidently these are no longer Lorentz transformations.

$\endgroup$

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.