In his paper "On the Electrodynamics of Moving Bodies", Einstein writes the equation

$$\dfrac{\partial \tau}{\partial x'}+\dfrac{v}{c^2-v^2}\dfrac{\partial \tau}{\partial t}=0$$


  • $\tau=\tau(x',y,z,t)$ is a linear function (i.e. $\tau=Ax'+By+Cz+Dt$)
  • $x'=x-vt$
  • $\dfrac{\partial \tau}{\partial y}=0$ (i.e. $B=0$)
  • $\dfrac{\partial \tau}{\partial z}=0$ (i.e. $C=0$)
  • $c$ is a constant
  • $x,x',y,z,t,v$ are variables

and he derives that


where $a=a(v)$

Could someone please walk me through step-by-step how he derived this? I am not very familiar with integrals invovling partial derivatives, so I would be appreciative if any answers are quite explicit.

Additionally, if I modified the question to say that $\tau$ was an affine function (i.e. $\tau=Ax'+By+Cz+Dt+E$), would it make any difference to the result (I suspect it wouldn't)?

  • $\begingroup$ It seems like this question could be of help for you. $\endgroup$ – mrtaurho Dec 30 '18 at 16:18
  • 6
    $\begingroup$ The question is strictly about the derivation of the differential equation's solution. Hadn't the OP mentioned physics at all, it would still be a perfectly fine question for this site. $\endgroup$ – Niki Di Giano Dec 30 '18 at 16:26
  • $\begingroup$ @NikiDiGiano I guess I overreacted. I retracted my close vote but the question I found on physics stackexchange could be useful nevertheless. $\endgroup$ – mrtaurho Dec 30 '18 at 16:28

From the definitions given: $$\partial_{x'}\tau = A, \quad \partial_t \tau = D$$ Also: $$\partial_{y}\tau = B = 0, \quad \partial_z \tau = C = 0$$ From the differential equation we get: $$A + \frac{v}{c^2 - v^2}D = 0 \\ \implies A = - \frac{v}{c^2 - v^2}D$$ Now using the definition given for $\tau$: $$\tau = - \frac{v}{c^2 - v^2}Dx' + Dt = D\bigg(t - \frac{v}{c^2 - v^2}x'\bigg)$$ So if you define $D=a$ you get the final expression for $\tau$. As you have noticed, requiring $\tau$ to be affine doesn't change the results at all - your $\tau$ would be: $$ \tau = a\bigg(t - \frac{v}{c^2 - v^2}x'\bigg) + E$$ The reason $a=a(v)$ is because $D$, who is actually $a$ in disguise, does not depend on $x', y, z, t$ but is assumed to depend on $v$. Otherwise, the relation wouldn't be linear. Without further information, $a$ is a function of potentially anything except those four variables.

  • $\begingroup$ Thank you for your help so far. I am just thinking through your answer at the moment, trying to get my head round it. I am just trying to see how you got from $A=-\frac{v}{c^2-v^2}D$ to $\tau=a\left(t-\frac{v}{c^2-v^2}x'\right)+E$. Could you perhaps extend your answer a little to show every step? $\endgroup$ – Rational Function Dec 30 '18 at 16:39
  • $\begingroup$ I have added the missing intermediate step. It's just a matter of plugging the results in the original expression. $\endgroup$ – Niki Di Giano Dec 30 '18 at 16:45
  • $\begingroup$ Thank you very much, that makes perfect sense - brilliant answer! $\endgroup$ – Rational Function Dec 30 '18 at 16:52

There is really no need to assume that $\tau$ is linear or affine to derive the general form of $\tau$. Write the equation as $$ \frac{\partial\tau}{\partial x'}+k\frac{\partial\tau}{\partial t} = 0 , $$ where $k={v}/({c^2-v^2})$. Change coordinates from $(x',t)$ to $(\xi,u)$ by $$ \begin{cases}\xi= x'\\ u=t-k x' \end{cases} \qquad \text{or equivalently,}\qquad \begin{cases}x'=\xi\\ t=u+k\xi. \end{cases} $$ This gives $$ \frac{\partial\tau}{\partial\xi} = \frac{\partial\tau}{\partial x'} \frac{\partial x'}{\partial\xi}+ \frac{\partial\tau}{\partial t}\frac{\partial t}{\partial\xi} = \frac{\partial\tau}{\partial x'}+k\frac{\partial\tau}{\partial t} = 0 , $$ meaning that in the new coordinates, $\tau$ is a function of only $u$. Hence $$ \tau=a(u)=a(t-k x'), $$ for some function $a$. At this point, you can use the assumption that $\tau$ is linear (or affine) to deduce that $a$ is linear (or affine).

  • 1
    $\begingroup$ Thank you, this is another interesting approach. I especially like that it finds a general form before invoking the linearity of $\tau$, because it would allow us to proceed in a different way if for some reason we decided that the universe wasn't homogeneous (which forces $\tau$ to be linear/affine). If only I could tick multiple answers! $\endgroup$ – Rational Function Dec 30 '18 at 17:18
  • 1
    $\begingroup$ Indeed this answer is much more elegant than mine in my opinion. Props! $\endgroup$ – Niki Di Giano Dec 30 '18 at 17:35

Making the change of variables

$$ x'=x- vt\\ t'=\alpha t $$

we have

$$ (v^2+\alpha(c^2-v^2))\frac{\partial\tau}{\partial x'}+v\frac{\partial \tau}{\partial t'}=0 $$

so choosing

$$ \alpha = \frac{v^2}{v^2-c^2} $$

we get

$$ \frac{\partial \tau}{\partial t'} = 0\Rightarrow \tau(x',t') = f(x') = f(x-v t) $$


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.