6
$\begingroup$

I am trying to prove the following:

$$\langle\psi|\hat{H}|\phi\rangle\langle\phi|\hat{H}|\psi\rangle-\langle\psi|\hat{H}|\psi\rangle\langle\phi|\hat{H}|\phi\rangle\leqslant0.$$

I tried some ideas but could reach nowhere. I exploited the fact that $\hat{H}$ is hermitian, and thus the first term in the inequality became $\langle\phi|\hat{H}|\psi\rangle^*\langle\phi|\hat{H}|\psi\rangle=|\langle\phi|\hat{H}|\psi\rangle|^2$ and then by Cauchy Schwartz inequality, $|\langle\phi|\hat{H}|\psi\rangle|^2\leqslant\langle\phi|\phi\rangle\langle\psi|\hat{H}\hat{H}|\psi\rangle$,but I can see this just removes $|\phi\rangle$ from the game.

Another idea was to write $\hat{H}$ in the first term in outer product notation, this gives:$$\langle\psi|\hat{H}|\phi\rangle\langle\phi|\hat{H}|\psi\rangle=\sum_{i,j}E_iE_j\langle\psi|i\rangle\langle i|\phi\rangle\langle\phi|j\rangle\langle j|\psi\rangle$$I tried to work on this to get the inequality, but it got me nowhere.

Any help is appreciated.

$\endgroup$
  • $\begingroup$ You're on the right track. But why not split it up in another way with Cauchy-Schwartz? Write $H = \sqrt{H} \sqrt{H}$ and give one square root to each of the vectors Psi and Phi. $\endgroup$ – Luke Jun 20 '17 at 14:59
  • $\begingroup$ You can repeat this proof of the Cauchy–Schwarz inequality inserting $H$ in the middle of all scalar products. $\endgroup$ – Alexey Sokolik Jun 20 '17 at 15:04
7
$\begingroup$
  1. 2D Counterexample: Assume $$ \langle\psi| \psi \rangle ~=~1~=~\langle\phi| \phi \rangle, \qquad \langle\psi| \phi \rangle~=~0, \qquad \hat{H}~=~| \psi \rangle\langle\psi| -| \phi \rangle\langle\phi|. $$

  2. OP's inequality is true for a semi-positive operator $\hat{H}\geq 0$, since then it has a well-defined square root $\sqrt{\hat{H}}$, and it becomes the standard Cauchy–Schwarz inequality.

$\endgroup$
  • $\begingroup$ Yes, right, the solution I proposed only works for Hamiltonians that are positive. This counter-example shows it is not possible for all Hamiltonians. $\endgroup$ – Luke Jun 20 '17 at 15:14
  • 1
    $\begingroup$ Please forgive my ignorance, but that hamiltonian seems to have a negative eigenvalue for $|\phi\rangle$, isn't it supposed to have positive eigenvalues only? $\endgroup$ – Tofi Jun 20 '17 at 15:49
  • 1
    $\begingroup$ A Hermitian operator has real (but not necessarily positive) eigenvalues. $\endgroup$ – Qmechanic Jun 20 '17 at 16:04
  • 1
    $\begingroup$ @VIP Usually the Hamiltonian is assumed to be bounded from below, for physical reasons. The result however holds only for positive operators, and therefore given an Hamiltonian $H$ bounded from below by $-m$, $m>0$, the result holds for the positive operator $H+m$. Physically, $H+m$ is equivalent to $H$, up to a rescaling of the ground state energy. $\endgroup$ – yuggib Jun 20 '17 at 16:25

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.