I'm doing something very wrong or It seems to me that I can't generate a finite Lorentz transformation using the exponential of the infinitesimal Lorentz boosts.

Let me define $L_{x;v}$ as the operator that produce a Lorentz boost in the $x$-direction with a speed of $v$. This operator acts on the components of the 4-position as follows

$$L_{x;v}(x) =\gamma_{v}(x-vt),$$

$$L_{x;v}(y) =y,$$

$$L_{x;v}(z) =z,$$

$$L_{x;v}(t) =\gamma_{v}(t-vx),$$

where $\gamma_{v}=\frac{1}{\sqrt{1-v^{2}}}$. Now, the infinitesimal generator of a Lorentz boost in the $x$-direction (in the Hamiltonian sense) is $K_{x}=Hx-tp_{x}$, where $H=\sqrt{p^{2}+m^{2}}$. A finite Lorentz transformation should be given via the exponential operator

$$L_{x;v}=\exp\left[\left\{ \circ,K_{x}\right\} \right]=\sum_{n=0}^{\infty}\frac{1}{n!}\left\{ \circ,K_{x}\right\} ^{n}.$$

But I can't see how this operator gives the correct Lorentz transformations.

For the $x$ coordinate we have that $\left\{ x,K_{x}\right\} =x\frac{p_{x}}{H}-t$, hence, the term $\left\{ \left\{ x,K_{x}\right\} ,K_{x}\right\}$ will be independent of time. So the term $\gamma_{v}vt$ will not appear.

Worse yet, $K_{x}$ generates a change in the $y$ coordinate $\left\{ y,K_{x}\right\} =x\frac{p_{y}}{H}\neq0$, in contradiction of what a Lorentz transformation should do.

My take is that the usual tools of Hamiltonian mechanics are unable to give the right answer because in this case we have a transformation that also change the time.


Now usually the generator of (proper orthochronous) LTs (generator of the Lie group; in the fundamental representation) is something like $$ L_{\mu \nu} = i(x_{\mu} \partial_{\nu} - x_{\nu} \partial_{\mu}) $$ and the corresponding finite transformation would be $\Lambda(\omega) = e^{i\omega^{\mu\nu}L_{\mu \nu}}$, where $\omega_{\mu \nu}$ are the parameters (which specify the group element). Acting e.g. on a position four vector: $$ x'^{\rho} \equiv e^{i\omega_{\mu\nu}L^{\mu \nu}} x^{\rho}. $$ To see what this is consider that $$ L^{\mu \nu} x^{\rho} = i(x^{\mu} g^{\nu \rho} - x^{\nu} g^{\mu \rho}) = \underbrace{i(g^{\mu \sigma} g^{\nu \rho} - g^{\nu \sigma} g^{\mu \rho})}_{\equiv (J^{\mu \nu})^{\rho \sigma}} x_{\sigma} $$ and therefore $$ x'^{\rho} = (e^{i\omega_{\mu \nu} J^{\mu \nu}})^{\rho \sigma} x_{\sigma}. $$ Note that $w_{\mu \nu} J^{\mu \nu}$ is a matrix and the LT is the corresponding matrix-exponential. E.g. $\omega_{10} = -\omega_{01} = \psi$ and all others components zero gives a boost in the $x$-direction ($\psi = \tanh^{-1}v/c$ is then the rapidity of the boost). Or $\omega_{12} = -\omega_{21} = \theta$, and all other components zero, corresponds to a rotation around the z-axis by an angle $\theta$. You can check all this by computing the matrix exponential. (See also Lorentz transformation boosts as matrices).

Also the momentum generator is $p_{\mu} = -i \partial_{\mu}$ (this generates space-time translations $e^{ia^{\mu} p_{\mu}} x^{\rho} = x^{\rho} + a^{\rho}$. Putting this into the LT generator gives $$ L_{\mu \nu} = x_{\nu} p_{\mu} - x_{\mu} p_{\nu} $$ and with the canonical Poisson bracket relation $\{ x^{\mu}, p^{\nu} \} = g^{\mu \nu}$ using your formula for the exponential of the Poisson-bracket is basically the same as what I have done. Note that $L_{\mu \nu}$ looks like angular momentum, which is of course expected, since LTs are nothing but rotations in 1+3 dimensional space-time.

Now in your formulation you made the identification $$ -i \partial_0 = p_0 = H = \sqrt{\vec{p}^2 + m^2}, $$ corresponding to a single, free, massive particle. But this messes up the poisson-bracket relation $\{ x^{\mu}, p^{\nu}\} = g^{\mu \nu}$, e.g. $\{ p_0, x^i \} = \frac{p^i}{\sqrt{\vec{p}^2 + m^2}} \neq 0$; $p^0$ and $\vec{p}$ are not independent anymore! This is precisely why your approach does not work. So one needs to set up a proper Hamiltonian formalism of relativistic mechanics and I am not very familiar with this. Maybe Hamiltonian mechanics and special relativity? can help you.


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.