I'm reading about the variational method in Shankar's Principles of Quantum Mechanics, page 433. The author states that if we have the trial ket $$|\psi\rangle = |E_0\rangle + |\delta\psi\rangle$$ where $|E_0\rangle$ is the eigenket corresponding to the ground state energy and we decompose $|\delta\psi\rangle$ into parts parallel and perpendicular to $|E_0\rangle$ such that $$|\psi\rangle = |E_0\rangle + \alpha|E_0\rangle + |\delta\psi_{\perp}\rangle$$ then we obtain $$\langle\psi|H|\psi\rangle = \frac{E_0|1+\alpha|^2 + \langle\delta\psi_{\perp}|H|\delta\psi_{\perp}\rangle}{|1+\alpha|^2 + \langle\delta\psi_{\perp}|\delta\psi_{\perp}\rangle}$$ $$\langle\psi|H|\psi\rangle = E_0 + O[(\delta\psi_{\perp})^2]$$ I don't know how the author arrives at the final line from the one above it. In particular, how do we know that $\langle\delta\psi_{\perp}|H|\delta\psi_{\perp}\rangle \sim O[(\delta\psi_{\perp})^2]$ ?


1 Answer 1


Let $\Bbb I$ denote the identity on the Hilbert space so$$\langle\psi|H|\psi\rangle-E_0=\frac{\langle\delta\psi_\perp|H-E_0\Bbb I|\delta\psi_\perp\rangle}{|1+\alpha|^2+\langle\delta\psi_\perp|\delta\psi_\perp\rangle}.$$The numerator (denominator) is $O\left((\delta\psi_\perp)^2\right)$ ($O(1)$). Your question is about the numerator's behaviour. This $O$ characterization is correct in the sense that, if we scale $\delta\psi_\perp$, we scale $\langle\delta_\perp|M|\delta_\perp\rangle$ the same way for all operators $M$ from the Hilbert space to itself, including $\Bbb I$.

  • $\begingroup$ Is this $O[(\delta\psi_{\perp})^2]$ characterisation due to the fact that we can write $\langle\delta\psi_{\perp}|M|\delta\psi_{\perp}\rangle = \int\delta\psi_{\perp}^*(x)M\delta\psi_{\perp}(x)dx$ ? $\endgroup$
    – Nitram
    Aug 24, 2021 at 6:20
  • $\begingroup$ @Nitram RE your comment: this is just a definition that you have written! $\endgroup$
    – Jacob A
    Aug 28, 2021 at 2:15

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.