I have been studying the geometry of special relativity through the Minkowski diagrams. I wanted to derive the figure shown in an article by using simple change of basis methods using the Lorentz transformation $\mathbf{x}'=\Lambda \mathbf{x}$ for observer $S'$ moving at speed $v$ relative to $S$ (in one dimension),

enter image description here

$$ \begin{pmatrix} ct' \\ x' \end{pmatrix} = \begin{pmatrix} \gamma_v & -\beta \gamma_v \\ -\beta \gamma_v & \gamma_v\end{pmatrix} \begin{pmatrix} ct \\ x \end{pmatrix} $$

I wanted to define the basis vectors in $S$ to be the Euclidean basis vectors $$ \mathbf{s}_t = \mathbf{s}_0 = (0,1) \; \text{and} \; \mathbf{s}_x = \mathbf{s}_1 = (1,0) $$

But under the transformation, $$\mathbf{s}'_0=\Lambda \mathbf{s}_0 = (-\beta\gamma_v,\gamma_v) $$ $$\mathbf{s}'_1=\Lambda \mathbf{s}_1 = (\gamma_v,-\beta\gamma_v)$$

This basis has the same angle $\alpha$ between the two axes as given in the figure, namely $\tan \alpha = \beta$. However, this transformation would make the $S'$ basis have axes that are at an obtuse angle relative to the $S$ basis, whereas most Minkowski diagrams I've seen have $S'$ axes within the $S$ axes at an acute angle. Am I doing something wrong here? Thanks.


Compare the matrix of pure rotation:

\begin{pmatrix} \cos \theta & -\sin \theta \\ \sin \theta & \cos \theta \\ \end{pmatrix}

which rotates both axes anti-clockwise by $\theta$.


$$\Lambda=\sqrt{\sec 2\alpha} \begin{pmatrix} \cos \alpha & -\sin \alpha \\ -\sin \alpha & \cos \alpha \\ \end{pmatrix} $$

which rotates the $x$-axis anti-clockwise by $\alpha$ whereas the $y$-axis clockwise by $\alpha$.

Further points to be noticed:

  • $0\le \alpha < 45^{\circ}$

  • $(x',ct')=\sqrt{\cos 2\alpha}(1,0) \iff (x,ct)=(\cos \alpha, \sin \alpha)$

  • $(x',ct')=\sqrt{\cos 2\alpha}(0,1) \iff (x,ct)=(\sin \alpha, \cos \alpha)$

  • $\det \Lambda=1$

  • $\Lambda^{-1}=\sqrt{\sec 2\alpha} \begin{pmatrix} \cos \alpha & \sin \alpha \\ \sin \alpha & \cos \alpha \end{pmatrix}$

| cite | improve this answer | |
  • $\begingroup$ Great answer, thanks. However, the first matrix you have directly takes the Cartesian unit vectors and rotates them counterclockwise, so that using the matrix from a vector x in S to x' in S' we can construct the new basis vectors for the S' coordinate. Here, however, you're starting with the basis vectors in S'. What is the reasoning behind this? $\endgroup$ – ChrisL Aug 30 '18 at 3:52
  • $\begingroup$ Basing on your choice of illustration, $S'$ frame is not orthogonal, so the coordinate grids are parallelograms instead of rectangles. It'll be difficult to convert $(x,y)=(1,0)$ to $(x',y')$, etc. Alternatively, we may use the eigenvectors of the Lorentz matrix, which are $x'+ct'$ and $x'−ct'$, as the bases. See another question here. $\endgroup$ – Ng Chung Tak Aug 30 '18 at 5:36

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.