In numerous discussion of the special conformal transformation, they cite the finite transformation as

$${x^\mu}'=\frac{x^\mu-b^\mu x^2}{1-2x\cdot b+b^2 x^2}$$

This can be found from integrating the infinitesimal conformal transformation

$$\delta x^\mu =2(b\cdot x)x^\mu-x\cdot x b^\mu$$

I found the derivation given as an answer on this site. I completely understand what they did, but at the end of the day they get the answer to be

$$x(t)=\frac{x_0-x^2_0 (tb)}{1-2x_0(tb)+x_0^2(tb)^2}$$

Their starting point was the differential equation $\dot{x}=2(b\cdot x)x-x^2 b$.

Why, exactly, is there a $t$ in the second equation but no $t$ in the first? Is $tb=b^\mu$? Also, where in the derivation I've cited do they only consider the case of $\mu=0$ (i.e., time)? Any clarification would be greatly appreciated.

  • $\begingroup$ @CosmasZachos That was a typo--thank you for pointing that out. I've edited the post appropriately. I followed the derivation out completely and I believe I understand it, but I don't understand why the final result given in that answer differs from the result in my textbooks. $\endgroup$ Feb 10, 2018 at 16:54
  • 1
    $\begingroup$ $t=|b|$, that is, $b^\mu\equiv \hat{b}^\mu t$. $\endgroup$ Feb 10, 2018 at 17:00
  • $\begingroup$ But is it different? why did you not reinstate the 4-indices μ in the last displayed equation you wrote, so $x(t)^\mu =...$ and the parametric diff eqn (s) that led to it? $\endgroup$ Feb 10, 2018 at 17:05
  • $\begingroup$ Also, Josh is not considering $\mu=0$ only. He is suppressing the indices, so the vector equation $x=y$ is equivalent to $x^\mu=y^\mu$. $\endgroup$ Feb 10, 2018 at 17:09
  • $\begingroup$ @CosmasZachos I was quoting the result from the linked answer which didn't have the superscripts. That's why I was confused--they solved the equation in a different notation, and that's what's confusing me. Also, in their solution they have $\dot{y}=-b$, so $y=y_0-bt$. Would this still mean that $b=|t|$? Are all derivatives technically taken with respect to $|b|$? $\endgroup$ Feb 10, 2018 at 17:16

1 Answer 1


@Trimok solved the problem most elegantly in his comment to the question cited, and since you are troubled by @Josh's simplifying changes of variables, $b^\mu\equiv \hat{b}~ t$, I'll avoid them to merely integrate the variation $$\delta x^\mu =2(b\cdot x)x^\mu-x\cdot x~ b^\mu$$ directly. It immediately implies $$\delta \left (\frac{ x^\mu}{x^2}\right) = \frac{ \delta x^\mu}{x^2} -2 x^\mu \frac{ x\cdot \delta x }{x^4} = -b^\mu ~.$$

That is, the vector $x^\mu/x^2$ evolves by shifting along $-\hat{b}^\mu$, linearly in the magnitude $|b|$ of $b^\mu$, so with constant unit speed in this "pseudotime" $|b|$. Integrating this simplest of advections for finite pseudotime, we immediately get $$ \frac{ x'^\mu}{x'^2}= \frac{ x^\mu}{x^2} -b^\mu . $$ Square both sides, to get the normalization, $$ \frac{ 1}{x'^2}= \frac{1}{x^2} +b^2 -2\frac{ x\cdot b}{x^2}= \frac{1-2x\cdot b+b^2 x^2 }{x^2}, $$ which divides the above vector equation to yield your conventional form for it, ∴ $${x^\mu}'=\frac{x^\mu-b^\mu x^2}{1-2x\cdot b+b^2 x^2}~.$$


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.