I am studying quantum mechanics and I encountered the famous Ehrenfest Theorem, which states that given an observable $A$, its expectation value time evolution is governed by $\partial_t\langle A\rangle=\frac{1}{ih}\langle[A,H]\rangle$. This is famously analogous through canonical quantization of Poisson Brackets to the fact that in Hamiltonian mechanics one has $\frac{dF}{dt}=\{F,H\}$.

One has that the Hamilton equations have the following analogous form:$$\partial_t\langle x\rangle=\frac{1}{ih}\langle[x,H]\rangle=\frac{1}{m}\langle p_j\rangle$$ $$\partial_t\langle p\rangle=\frac{1}{ih}\langle[p,H]\rangle=-\langle\partial_x U\rangle$$ Whereas the first is completely analogous to the first Hamilton eq., to have the complete analogy one needs to have $$\langle\partial_x U\rangle=\partial_x\langle U\rangle$$ and this happens for at most quadratic potentials. But this must mean that at the classical level $\langle x^2\rangle=\langle x\rangle\langle x\rangle$, whereas $\langle x^2\rangle\neq\langle x\rangle\langle x\rangle$ at the quantum level. In some sense then at quantum level we have a kind of correlation.

What justifies this asymmetry in the two equations? Is there anything to this idea of correlation?

  • $\begingroup$ How are you defining $\partial_x \langle U\rangle$? $\langle U \rangle = \int dx \;\psi^*(x) U(x) \psi(x)$, so $x$ is an internal integration varaible. There is no external $x$ variable to differentiate with respect to. $\endgroup$ Commented Jun 12, 2019 at 11:32
  • $\begingroup$ @BySymmetry you're correct, I mean $\partial_{<x>} U(<x>)$ $\endgroup$ Commented Jun 12, 2019 at 13:53

1 Answer 1


FWIW, the asymmetry is apparently directly tied to the assumption that the Hamiltonian $\hat{H}$ is quadratic in momenta $\hat{p}_i$ but not necessarily quadratic in positions $\hat{x}^i$. Generically, the rule $\langle f(\hat{A})\rangle= f(\langle \hat{A}\rangle)$ is only expected to hold for affine functions $f$.

  • $\begingroup$ Ok, this is indeed true, but affinity is only sufficient condition for the rule you stated. What I mean is, classically we do have that non affine functions' expectation factorizes. $\endgroup$ Commented Jun 11, 2019 at 20:55

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.