Considering a sandwich of the form $\langle{\psi}|{\hat{O}}|{\phi}\rangle$, where $\hat{O}$ is an operator and $\psi, \phi$ are state vectors, I know that if $\psi=\phi$ then the sandwich represents the expectation value of that operator $\langle \hat{O} \rangle$ i.e. the average value of all the possible outcomes of a measurement under identical conditions. But if $\psi \ne \phi$ does the sandwich have a physical meaning?

Moreover if the operator $\hat{O}$ is not hermitian, then the expectation value does not represent an obserable, so does $\langle{\psi}|{\hat{O}}|{\psi}\rangle$ still have a physical interpretation?

  • 4
    $\begingroup$ You are asking about the "physical meaning" of an arbitrary matrix element of an arbitrary matrix? $\endgroup$ Dec 7, 2022 at 15:23
  • $\begingroup$ Do people really call such matrix elements, such inner products, not only overlaps but also sandwiches? As for interpreting them, they're fixed by more easily-understood diagonal ($\psi=\phi$) elements if $\hat{O}$ is positive-definite; indeed, $\hat{O}$ is the difference of two such operators. $\endgroup$
    – J.G.
    Dec 7, 2022 at 16:26
  • $\begingroup$ @J.G. I've heard people use the word sandwich for them a lot in the US. Also as a verb e.g. "to sandwich the operator between $\langle \phi |$ and $|\psi \rangle$ $\endgroup$ Dec 7, 2022 at 17:42
  • $\begingroup$ @doublefelix Thanks for identifying a relevant nation. My experience with "overlap" is UK-based. $\endgroup$
    – J.G.
    Dec 7, 2022 at 19:15

1 Answer 1


In general there is no physical meaning I am aware of even if $O$ is hermitian. Even when $O$ is the momentum operator, for example, taking $\langle \psi | \hat{p} | \phi \rangle$ has no specific interpretation for arbitrary $\psi, \phi$.

There is a specific set of rules at the foundation of QM which lead to interpretable connections to reality. Some examples which have a similar flavor to your question:

  • The probability of transitioning from an initial state $|\psi \rangle$ at $t=0$ to an eigenstate $|\lambda \rangle$ of some operator $O$ such that $O |\lambda \rangle = \lambda |\lambda \rangle$, is (if $\lambda$ has only 1 orthonormal eigenvector) $$|\langle \lambda | e^{-itH/\hbar }|\psi \rangle|^2$$
  • Letting $P_\lambda$ be a projector to the eigenspace corresponding to eigenvalue $\lambda$, the probability of getting outcome $\lambda$ when measuring a state $|\psi \rangle$ without evolution in between is $$\langle \psi | P_\lambda | \psi \rangle$$
  • The expectation value of an operator $O$ when measuring a particle in state $|\psi \rangle$ is $$\langle \psi | O | \psi \rangle$$

But aside from the specific connections to reality specified in the postulates of QM (see for example chapter 4 in Shankar's textbook principles of QM) there is no reason for an arbitrary mathematical expression to have a clear physical interpretation.

  • $\begingroup$ "there is no reason for an arbitrary mathematical expression to have a clear physical interpretation". Especially in quantum mechanics, where the inferential gap between the mathematical formalism and the interpretational apparatus is huge. $\endgroup$
    – march
    Dec 7, 2022 at 17:28

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.