In the article "Identification of the velocity operator for an irreducible unitary representation of the Poincaré group", there is formula to implement a Lorentz transformation of the velocity as a unitary transformation.


It goes as follows: consider a Lorentz transformation upon the components of the velocity operator $\mathbf{V}\rightarrow\mathbf{V}'$. The velocity operator transform just as the velocity components in classical relativistic physics. The unitary transformation that gives the Lorentz transformation of $\mathbf{V}$ by a boost in the $z$ direction with a factor $\beta$ is given by


What I don't understand is why the velocity appears in a factor $\tanh^{-1}\beta$. It seem quite different from the other space-time transformation of the Lorentz group (or the Galilei group).

  • $\begingroup$ Have you tried computing those matrix exponentials? (Use the Taylor series.) They should end up looking like familiar Lorentz boosts with nonzero matrix elements like $\gamma$ and $\gamma\beta$. This is a very instructive exercise to do. $\endgroup$ – G. Smith Oct 27 '20 at 21:20
  • $\begingroup$ So far I've been able to show the equality of the first terms of the series expansion, but I yet have to find have to find how to prove the equality of the whole series. $\endgroup$ – AndresB Oct 28 '20 at 0:19
  • $\begingroup$ The paper is behind a paywall. What does it have for the boost generator $K_z$ as an explicit matrix? I’m thinking of writing an answer to show you how the exponential is just an ordinary Lorentz transformation, and I want to use the same sign conventions as in the paper. $\endgroup$ – G. Smith Oct 28 '20 at 3:01
  • $\begingroup$ Also see: physics.stackexchange.com/questions/99051/… $\endgroup$ – SRS Oct 28 '20 at 3:46
  • 1
    $\begingroup$ It turns out that they're not very similar. I've posted two answers working out both of them. $\endgroup$ – G. Smith Oct 30 '20 at 20:55

The velocity transformation you presented,


is equivalent to the usual Lorentz transformation of velocity, which for a boost by $\beta\hat z$ is

$${V_x}'=\frac{\sqrt{1-\beta^2}\,V_x}{1-\beta V_z}\tag{2a}$$

$${V_y}'=\frac{\sqrt{1-\beta^2}\,V_y}{1-\beta V_z}\tag{2b}$$

$${V_z}'=\frac{V_z-\beta}{1-\beta V_z}\tag{2c},$$

but it's a messy calculation. I believe that the equivalence can only be shown term-by-term in an expansion in powers of $\beta$.

In comments, you explained that (1) is to be interpreted in terms of various operators which are possibly non-commuting, namely the relativistic velocity operator


the Lorentz boost generator


and the relativistic energy operator


Here $\mathbf X$ is the position operator and $\mathbf P$ the relativistic momentum operator with the canonical commutation relations


To show the equivalence of (1) and (2), use the identity


which follows from the Taylor series for the function $f(t)=e^{tA}Be^{-tA}$. Equation (1) becomes


So the next step is to calculate the commutator $[K_z,V_i]$ and further nestings. We have


To get this it was necessary to calculate


and this required calculating

$$[z,H^{-1}]=[z,(\mathbf{P}^2+m^2)^{-1/2}]=[z,P_z]\frac{\partial}{\partial P_z}(\mathbf{P}^2+m^2)^{-1/2}=-i\frac{P_z}{H^3}.\tag{7c}$$

That last commutator required the identity

$$[A,f(B)]=[A,B]\frac{\partial f}{\partial B}\tag8$$

which holds when $[A,B]$ commutes with both $A$ and $B$.

From (7a) we have




We're boosting along $z$, so the transverse $y$ direction works the same as the transverse $x$ direction. We'll just do $x$.

From (9a) and (9b) it is possible to grind out the nested commutators. I used Mathematica to do it up to ten nestings. I'll show four of them.

For $V_x$ the nestings are

$$\begin{align} [K_z,V_x]&=-iV_xV_z\tag{10a}\\ [K_z,[K_z,V_x]]&=V_x \left(1-2 V_z^2\right)\tag{10b}\\ [K_z,[K_z,[K_z,V_x]]]&=i V_x V_z \left(-5+6 V_z^2\right)\tag{10c}\\ [K_z,[K_z,[K_z,[K_z,V_x]]]]&=V_x \left(5-28 V_z^2+ 24 V_z^4\right)\tag{10d} \end{align}$$

and for $V_z$ the nestings are

$$\begin{align} [K_z,V_z]&=i(1-V_z^2)\tag{11a}\\ [K_z,[K_z,V_x]]&=2 V_z \left(1-V_z^2\right)\tag{11b}\\ [K_z,[K_z,[K_z,V_z]]]&=2 i \left(1-4 V_z^2+3 V_z^4\right)\tag{11c}\\ [K_z,[K_z,[K_z,[K_z,V_z]]]]&=8 V_z \left(2-5 V_z^2+3 V_z^4\right).\tag{11d} \end{align}$$

As you can see, there is no obvious pattern. It’s basically a mess.

You can then insert these nested commutators into (6) and expand $\tanh^{-1}\beta$ in powers of $\beta$. The result through order $\beta^4$ is

$$\begin{align} {V_x}'&=\beta V_x V_z +\beta ^2 \left(V_x V_z^2-\frac{V_x}{2}\right) +\beta ^3 \left(V_x V_z^3-\frac{V_x V_z}{2}\right)\\ &+\beta ^4 \left(V_x V_z^4-\frac{1}{2} V_x V_z^2-\frac{V_x}{8}\right) +O\left(\beta ^5\right)\tag{12a} \end{align}$$

$$\begin{align} {V_z}'&=V_z +\beta \left(V_z^2-1\right) +\beta ^2 \left(V_z^3-V_z\right) +\beta ^3 \left(V_z^4-V_z^2\right)\\ &+\beta ^4 \left(V_z^5-V_z^3\right) +O\left(\beta ^5\right).\tag{12b} \end{align}$$

(12a) is the Taylor expansion of (2a), and (12b) is the Taylor expansion of (2c). (I verified it through order $\beta^{10}$.) This shows the equivalence of the two ways, (1) and (2), of doing a relativistic velocity transformation.

I would be impressed if anyone can prove the equivalence without doing it order-by-order in powers of $\beta$.

This does not motivate why $\tanh^{-1}\beta$ appears in (1). But if it didn't appear in that way, (1) would not be equivalent to (2). I will write a separate answer explaining why you might expect $\tanh^{-1}\beta$ to appear as the factor multiplying the boost generator in the exponential.

  • $\begingroup$ Amazing answer. I already had checked the expansion up to order $\beta^{2}$, but after that the terms become quite annoying. I guess the equality between terms of the same power could be worked out using Mathematical induction, but It does not seem so a simple task. In any case, I consider the question completely solved. $\endgroup$ – AndresB Oct 31 '20 at 0:33

If we write $\beta\equiv v/c= \tanh s$, then $s$ is the rapidity. If we combine two parallel boosts with rapidity $s_1$ and $s_2$ then the combined boost has rapidity $s_1+s_2$ so $s=\tanh^{-1} \beta$ is the natural variable to describe boosts.


To motivate why $\tanh^{-1}\beta$ appears in your formula as the factor multiplying the boost generator in the exponential, consider a much simpler transformation, namely the Lorentz transformation of coordinates. For a boost $\beta\hat z$, this is

$$ \begin{pmatrix}t'\\x'\\y'\\z' \end{pmatrix}= \begin{pmatrix}\gamma & 0 & 0 & -\gamma\beta\\ 0 & 1 & 0 & 0\\ 0 & 0 & 1 & 0\\ -\gamma\beta & 0 & 0 & \gamma \end{pmatrix} \begin{pmatrix}t\\x\\y\\z \end{pmatrix}\tag1 $$

where $\gamma\equiv(1-\beta^2)^{-1/2}$.

It is easy to show that this boost matrix, which I'll call $B(\beta,\hat z)$, is just

$$B(\beta,\hat z)=e^{-K_z \tanh^{-1}\beta}\tag2$$


$$K_z=\begin{pmatrix} 0 & 0 & 0 & 1 \\ 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 \\ 1 & 0 & 0 & 0 \end{pmatrix}\tag3$$

is a the generator of boosts in the $z$-direction.

(How do you get this generator? By observing that an infinitesimal boost by velocity $\epsilon \hat z$ is $B(\epsilon,\hat z)\approx I-\epsilon K_z$.)

The matrix exponential is defined by its power series,

$$e^{-\zeta K_z}\equiv I-\zeta K_z+\frac{\zeta^2}{2!}K_z^2-\frac{\zeta^3}{3!}K_z^3+\frac{\zeta^4}{4!}K_z^4+\dots,\tag4$$

where I'll write $\zeta\equiv\tanh^{-1}\beta$ for convenience.

Now $K_z$ has the nice property that


and thus all higher powers of $K_z$ reduce to either $K_z$ or $K_z^2$. Thus (4) becomes

$$\begin{align} e^{-\zeta K_z}&=I-\left(\zeta+\frac{1}{3!}\zeta^3+\frac{1}{5!}\zeta^5+\dots\right)K_z\\ &+\left(\frac{1}{2!}\zeta^2+\frac{1}{4!}\zeta^4+\frac{1}{6!}\zeta^6+\dots\right)K_z^2\tag{6a}\\ \\ &=I-(\sinh\zeta)K_z+(\cosh\zeta-1)K_z^2\tag{6b}\\ \\ &=\begin{pmatrix} \cosh\zeta & 0 & 0 & -\sinh\zeta\\ 0 & 1 & 0 & 0\\ 0 & 0 & 1 & 0\\ -\sinh\zeta & 0 & 0 & \cosh\zeta \end{pmatrix}.\tag{6c} \end{align}$$

Finally, $\tanh\zeta=\beta,$ implies that $\cosh\zeta=\gamma$ and $\sinh\zeta=\gamma\beta$, so the boost in (2) is equivalent to that in (1).

  • $\begingroup$ These are certainly Lorentz's transformations into hyperbolic form. :-) $\endgroup$ – Sebastiano Oct 30 '20 at 22:28
  • 2
    $\begingroup$ Yes, and the cosh and sinh arise from the exponential series separating into terms of odd and even power because the boost generator satisfies $K_z^3=K_z$. $\endgroup$ – G. Smith Oct 30 '20 at 22:44

To expand on mike stone's answer, a Lorentz boost is simply a rotation, and the traditional $dx/dt$ velocity is a slope.

In 2D Euclidean space, you could parametrize origin-preserving rotations by the slope $m=dy/dx$ of the line to which they rotate the line $y=0$. If you did that, then the angle of the rotation would be related to the slope by $θ=\tan^{-1} m$, and the "sum" of two slopes $m$ and $m'$, meaning the slope corresponding to the composition of the rotations given by $m$ and $m'$, would be $\frac{m+m'}{1-mm'}$. In spacetime the situation is exactly the same, except that because of the change of sign in the metric, you end up needing $\tanh$ instead of $\tan$, and the sign in the denominator of the slope "addition" formula is flipped.

Writing a Lorentz boost as $\exp i K_z α$ is closely related to writing a rotation in the complex plane as $\exp i θ$. The generalization of this formula (the Euler formula) to other continuous symmetries is the theory of Lie algebras and Lie groups.


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.