It is well-known that the Poisson bracket can be recovered out of the Moyal bracket under the limit when $\hbar$ goes to zero $$\lim_{\hbar\rightarrow 0} \lbrace f,g\rbrace_M=\lbrace f,g \rbrace_P.$$ This is easy to verify in the differential form of the brackets, but the Moyal bracket admit an integral representation as $$\lbrace f,g\rbrace_M\\ =\frac{2}{\hbar^3\pi^2}\int dp' dp'' dx' dx'' f(x+x',p+p')g(x+x'',p+p'')\sin\left (\frac{2}{\hbar}(x'p''-x''p')\right ).$$ How to recover an integral form of the Poisson bracket performing the limit here?
-
$\begingroup$ The integral representation is equivalent to the differential form, cf. e.g. this Phys.SE post, so essentially nothing left to prove. $\endgroup$– Qmechanic ♦Commented Apr 5, 2023 at 17:55
-
$\begingroup$ Obviously, the aim is not to prove that a classical limit exists but more to find an integral expression for the Poisson bracket from the integral form of Moyal, as I stated it in the question. $\endgroup$– Nicolas Medina SanchezCommented Apr 5, 2023 at 18:00
-
1$\begingroup$ Wouldn't that just produce the standard formula $(df/dx) (dg/dp)-(df/dp) (dg/dx)$ for the PB? $\endgroup$– Qmechanic ♦Commented Apr 5, 2023 at 18:36
-
$\begingroup$ I’m not familiar with the standard integral formula for the Poisson bracket, do you have any reference for that? $\endgroup$– Nicolas Medina SanchezCommented Apr 6, 2023 at 5:34
-
$\begingroup$ Crossposted to mathoverflow.net/q/445419/13917 $\endgroup$– Qmechanic ♦Commented Aug 16, 2023 at 18:14
1 Answer
There are no compact integral forms of the PBs, of course, unless you consider conversions of derivatives into powers of the Fourier conjugate variable, as you might be insinuating in the comments. The genius of G Baker's 1958 construction is his appreciation of the $\star$-product as the basis of phase-space translations, which results in the MB integral expression you have (corrected by me).
One may reduce the classical limit of that expression to the standard differential rep of the PBs (only), provided for the curious student of the future, in case you yourself were not interested anymore, as commented.
You must first non-dimensionalize the primed and double primed dummy variables of the integral (only), so, for $$ (x',x'',p',p'')\mapsto \sqrt{\hbar}~(x',x'',p',p''), $$ you have, for the Baker MB, $$ =\frac{2}{\hbar\pi^2}\!\!\int\!\! dp' dp'' dx' dx'' f(x+\sqrt{\hbar}x',p+\sqrt{\hbar}p')g(x+\sqrt{\hbar}x'',p+\sqrt{\hbar}p'')\sin\left ({2}(x'p''-x''p')\right )\\ = \frac{1}{i\hbar\pi^2}\!\!\int\!\!\! dp' dp'' dx' dx'' \left ( e^{{2i}(x'p''-x''p')}-e^{{-2i}(x'p''-x''p')}\right )e^{\sqrt{\hbar}(x'~_f\partial_x + p'~_f\partial_p+ x''~_g\partial_x + p''~_g\partial_p)} f(x,p)g(x,p) . $$ The last line reflects the Baker translation operator, where the left subscript of the partial derivatives signifies the respective functions f or g on which these partial derivatives act! note all four terms of the last exponential commute among themselves, so there are no ambiguities.
Now performing the x' and p' integrals nets two evident δ-functions, which then collapse the x'' and p'' integrals $$ = \frac{1}{i\hbar}\!\!\int\!\! dp'' dx'' \left ( \delta(p''-i\sqrt\hbar ~_f\partial_x /2)\delta(x''+i \sqrt\hbar ~_f\partial_p /2)-\hbox{ c.c.}\right ) e^{\sqrt{\hbar}( x''~_g\partial_x + p''~_g\partial_p)} f(x,p)g(x,p)\\ = \frac{2i}{i\hbar}\sin \left({\hbar \over 2}(_f\partial_x~_g\partial_p -_g\partial_x ~_f\partial_p)\right )~~fg. $$
It is evident that this expression collapses to the the PB $\{f,g\}$, for small $\hbar$, and it should be evident why this differential expression is 37000 times simpler than any indirect integral expression, in sharp contrast to the MB...
-
$\begingroup$ Thanks for your insight! Though I was interested, precisely, on how to avoid the $\hbar\right arrow 0$ approach to the classical limit and try to find a way from the integral expressions only. $\endgroup$ Commented Jun 11 at 9:20