I'm trying to understand a step in a derivation of the Lorentz transformation, that my professor gave in class. We start assuming the homogeneity and the isotropy of the 4 dimensional spacetime and then we consider two inertial frames of reference $S$ and $S'$, with $S'$ moving at speed $v$ along the $x$-axis of $S$. We also assume that $S$ and $S'$ have parallel axes and their origins coincide at time $t=0$ in $S$. So a general transformation between the coordinates of $S$ and $S'$ respectively is, \begin{align} t\qquad &\to\qquad t'=T(t,x,y,z,v)\\ x\qquad &\to\qquad x'=X(t,x,y,z,v)\\ y\qquad &\to\qquad y'=Y(t,x,y,z,v)\\ z\qquad &\to\qquad z'=Z(t,x,y,z,v) \ . \end{align} Then, applying homogeneity we find that the transformation must be linear, so \begin{equation} \begin{pmatrix}t'\\x'\\y'\\z'\end{pmatrix}=A(v)\begin{pmatrix}t\\x\\y\\z\end{pmatrix} \ , \end{equation} where $A(v)$ is a $4\times4$ matrix. Using the principle of relativity we find that directions perpendicular to the motion don't change, so \begin{equation} A(v)=\begin{pmatrix}A_1(v)&A_2(v)\\\mathbf{0}&\mathbf{1}\end{pmatrix} \ , \end{equation} where $A_1(v)$ and $A_2(v)$ are $2\times 2$ matrices, and $\mathbf{0}=\begin{pmatrix}0&0\\0&0\end{pmatrix}$, $\mathbf{1}=\begin{pmatrix}1&0\\0&1\end{pmatrix}$. Now comes the step that I don't understand, it says that:

From the isotropy of space, follows that $A_2(v)=\mathbf{0}$

Can please someone help me to understand how the isotropy of space have this implication?

  • $\begingroup$ $A(v)$ is a boost in the $x$ direction. $\endgroup$ Jan 22, 2020 at 20:53

1 Answer 1


I have a feeling there should be a physical reason for the off-diagonal elements both having to be zero, but I can't think of one off-hand. Here's another way to show it, though:

Consider, for example, a point particle whose rest frame is $S^\prime$. To you, the observer sitting in $S$, this particle would be moving away along your $x$ axis. Now, what about the $y$ and $z$ axes? Well, they shouldn't be important here, as these directions are orthogonal to the motion of the particle. In other words, once you have chosen your $x$ axis to be along the direction of motion of the particle, you have an infinite number of $y$ and $z$ axes that can be chosen -- all related by simple rotations around the $x$-axis -- which must all give the same $A(v)$ matrix. This is one of the assumptions of isotropy.

Suppose, instead of the $(t,x,y,z)$ you used $(t,x,Y,Z)$, where $Y$ and $Z$ are two different mutually perpendicular directions that are also perpendicular to $x$. Since space is isotropic, your definition of $y$ and $z$ should not affect your transformation matrix, and so

$$\begin{pmatrix}t^\prime\\x^\prime\end{pmatrix} = A_1(v) \begin{pmatrix}t\\x\end{pmatrix} + A_2(v) \begin{pmatrix}y\\z\end{pmatrix}$$

$$\begin{pmatrix}t^\prime\\x^\prime\end{pmatrix} = A_1(v)\begin{pmatrix}t\\x\end{pmatrix} + A_2(v) \begin{pmatrix}Y\\Z\end{pmatrix}$$

Or $$A_2(v) \begin{pmatrix}y\\z\end{pmatrix} = A_2(v) \begin{pmatrix}Y\\Z\end{pmatrix}$$

It should be intuitively clear that since $Y$ and $Z$ could be any possible orthogonal set (also orthogonal to $x$), this must mean that $A_2(v)=\mathbf{0}$, but if you wish to be a little more rigorous, these new $Y,Z$ axes can be obtained from $y,z$ by rotation of some angle $\theta$ around the $x$ axis, and so

$$\begin{pmatrix}Y\\Z\end{pmatrix} = R(\theta) \begin{pmatrix}y\\z\end{pmatrix},$$ where $R(\theta)$ is the usual rotation matrix. The above equality then means that for any arbitrary value of $\theta$,

$$A_2(v) \left(\mathbf{1} - R(\theta) \right) = \mathbf{0}.$$

Since $\theta$ and $v$ are both arbitrary, it must be that $A_2(v) = \mathbf{0}$.

  • $\begingroup$ So if the $A(v)$ is a boost in a general direction, space is no longer isotropic? $\endgroup$ Jan 22, 2020 at 21:14
  • $\begingroup$ I don't quite understand, the Lorentz transformations do distinguish between the direction parallel to the boost and those perpendicular to it. If the boost were in a general direction, then one could always rotate one's coordinate system to be along that direction and call it $x$, and the same argument I made would hold. I think the exact argument could also be made for a general transformation, by defining a direction $\hat{n}$ and two orthogonal basis vectors $Y,Z$ that are perpendicular to $\hat{n}$... $\endgroup$
    – Philip
    Jan 22, 2020 at 21:25
  • 1
    $\begingroup$ Thank you! Very clear! $\endgroup$ Jan 23, 2020 at 7:41
  • 1
    $\begingroup$ @Philip: you just need to careful when considering more then one boost since two boost create a rotation. And I don't believe the OP can prove space is isotropic mathematically. Locally, it can inferred from conservation of linear and angular momentum. But in general relativity, it isn't that simple - homogeneity and the isotropy of space are typically assumed and conservation of linear and angular momentum demonstrated. $\endgroup$ Jan 23, 2020 at 22:00
  • 1
    $\begingroup$ @Philip - actually, I just realized I misread the post - the OP assumes the isotropy of space - my bad. In which case, I presume the OP just didn't realize when the boost is in the $x$ direction, the boost plane is $t-x$ - and it doesn't include the other coordinates. You were right. $\endgroup$ Jan 23, 2020 at 22:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.