Given a classical observable, $a(x,p),$ Weyl quantization gives the correspondent QM observable as: $$\langle x | \hat{A} | \phi \rangle=\hbar^{-3}\int \int a \left(\frac{x+y}{2},p\right)\phi(y) e^{2\pi ip(x-y)/\hbar} dy dp.$$ The correspondent inverse transformation (Wigner transform; which if I understand correctly would be the symbol of the pseudo-differential operator defined above, but maybe this is not very important) is: $$a(x,p)= \int \langle x-\frac{y}{2} \vert \hat{A} \vert x+\frac{y}{2} \rangle e^{-2\pi i yp/\hbar}dy.$$

I wanted to test this in a few simple examples. It is always easy, except for when $\langle x \vert \hat{A} \vert x^\prime \rangle $ is actually a distribution and cannot be understand as a function.

A basic case which is confusing me (even ignoring the mathematical subtleties of the matter) is the case in which $\hat{A}$ is $\hat{T},$ the quantum mechanical kinetic operator. If you start with $a(x,p)=\dfrac{p^2}{2},$ it is very easy to see that the first formula above gives the correct $\hat{T},$ i.e, that $\langle x \vert \hat{T} \vert \phi \rangle$ is just the laplacian. But what if want to check the second formula, the inversion of the quantization? Well, in position representation, we have $\langle x \vert \hat{T} \vert x^\prime \rangle=-\dfrac{1}{2}\Delta_{x} \delta(x-x^\prime),$ where $\Delta$ is the laplacian. Therefore, one would expect: $$\dfrac{p^2}{2}=\int -\dfrac{1}{2}\Delta_{x-\frac{y}{2}} \delta(x-\frac{y}{2}-x-\frac{y}{2}) \exp \left(-2\pi i yp/\hbar \right) dy,$$ but I don't what to do with this; I don't even undertand what is the exact meaning of this. This can be simplified: $$-\dfrac{1}{2}\int\Delta_{x-\frac{y}{2}} \ \delta(y) \exp \left(-2\pi i yp/\hbar \right) dy,$$ but I don't know how to continue from this.

  • 1
    $\begingroup$ by translation invariance, is $\Delta_{x-y/2}\delta(y)=\Delta_{-y/2}\delta(y)$? ($x$ is a dummy variable). By the chain rule, this equals $=4\Delta_y\delta(y)$, from which your identity follows, right? $\endgroup$ – AccidentalFourierTransform Oct 6 '17 at 10:20
  • $\begingroup$ You are totally right, it is easy from translational invariance of the derivative... I still have some issues with the precise value of the premultiplicative constants, but that is of no much relevance $\endgroup$ – Qwertuy Oct 6 '17 at 12:59

I'll tweak your notations to correct the normalizations and conform with the sound Wikipedia conventions, instead:

The Weyl map is $$\langle x | \hat{A} | y \rangle=\frac{1}{2\pi\hbar}\int dp ~ a \left(\frac{x+y}{2},p\right) e^{ ip(x-y)/\hbar},$$ with the inverse, Wigner map, $$a(x,p)=2 \int dy ~\langle x+y \vert \hat{A} \vert x-y \rangle ~e^{-2 i yp/\hbar}.$$

For $a=p^2$, the Weyl map yields, $$ \frac{1}{2\pi\hbar}\int dp ~ p^2 e^{ ip(x-y)/\hbar}\\ =\frac{-\hbar^2}{2\pi\hbar}\int dp ~ \partial_x^2 ~ e^{ ip(x-y)/\hbar}\\ =-\hbar^2 \partial_x^2 \delta(x-y)= -\hbar^2\partial_x^2 \langle x| y\rangle =\langle x| \hat{p}^2|y\rangle ~. $$

Conversely, in the Wigner map, $$2 \int dy ~\langle x+y \vert \hat{p}^2 \vert x-y \rangle ~e^{-2 i yp/\hbar}\\ =2 \int dy dp' ~\langle x+y \vert \hat{p}^2|p'\rangle\langle p' \vert x-y \rangle ~e^{-2 i yp/\hbar}\\ =2 \int dy dp' ~p'^2 ~\langle x+y \vert p'\rangle\langle p' \vert x-y \rangle ~e^{-2 i yp/\hbar}\\ =\frac{2}{2\pi\hbar} \int dy dp' ~p'^2 ~ ~e^{i(-2 yp +2p'y)/\hbar}\\ =\int dp' ~p'^2 ~ \delta (p-p')=p^2.$$

But for the wrong normalization and exponential factors, etc... your first equation follows from this first equation here for the Weyl map, upon operating on the latter with $\int dy \langle y|\phi\rangle$.

  • $\begingroup$ A very nice answer indeed! I must have the premultiplicative constants wrong, because I managed to prove the result up to some silly factor $4\pi^2.$ Thanks for your clarity $\endgroup$ – Qwertuy Oct 7 '17 at 18:05

Hint: Use Integration by parts. You have $\Delta = \nabla^2$ and you can use Substitution $z = - \frac{y}{2}$ and then for any functions $X,Y$

$\int (\nabla_z X) Ydz = \int (\nabla_z(XY) - X \nabla_z Y)dz$.

The first term on the right Hand side gives a boundary term and These will always give you Zero.

You do the Integration by part twice and finally you can evaluate the Delta function easily (at $z = 0$).


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.