This may sound like a strange question, but just to be sure: Suppose I have a general Hermitian operator in Hilbert space whose action on an eigenvector is given by $R|r\rangle = r|r\rangle$. Then, I assume that the following is true:

$\frac{1}{R}|r\rangle = \frac{1}{r}|r\rangle$ and similarly for other powers $R^{-2}|r\rangle = r^{-2}|r\rangle$

Does this follow immediately from the action of the operator, as in the case $R^2|r\rangle = RR|r\rangle = r^2|r\rangle$ or does this have to be defined?

  • $\begingroup$ The inverse is already defined, you just have to prove it (e.g. consider the action of R times its inverse on |r>, on the one hand is |r> because the product is the identity, but it is also r R^(-1)|r>) $\endgroup$ Commented Jan 7, 2016 at 19:47
  • $\begingroup$ I don't understand your question. Both identities you ask about follow directly from $R^{-1}$ being the inverse. $\endgroup$
    – ACuriousMind
    Commented Jan 7, 2016 at 20:04
  • $\begingroup$ Observe that $R^{-1}$ might not exist... $\endgroup$
    – Phoenix87
    Commented Jan 7, 2016 at 21:30

1 Answer 1


If you have $R|r\rangle=r|r\rangle$ then you have, where $I$ is the identity operator,

$|r\rangle=I|r\rangle=R^{-1}R|r\rangle=rR^{-1}|r\rangle$ and you immediately have


  • $\begingroup$ Note that this is no a general operator in a general Hilbert space. If $R$ is not compact and normal, it is not warranted to have a complete set of eigenvectors. Also, if there exists a vector with $0$ eigenvalue, then you wouldn't find an $P^{-1}|v\rangle$. $\endgroup$
    – mlainz
    Commented Jan 7, 2016 at 22:18
  • $\begingroup$ This is another case when it fails. Take $H$ to be the space of polynomials of real variable x. If $H p(x) = x p(x)$, there is not $H^{-1} 1$. Note that $H$ is hermitian. $\endgroup$
    – mlainz
    Commented Jan 7, 2016 at 22:38
  • $\begingroup$ Note that $H$ is hermitian, wrt., for example, $\langle p | q\rangle \int_{-1}^{1}p(x)q(x) dx$. $\endgroup$
    – mlainz
    Commented Jan 7, 2016 at 22:44
  • $\begingroup$ Good points, my approach was rather naive, although I was making the assumption that the inverse operator actually exists. At least for the finite dim case we dont have the issue of 1/0 since if 0 is an eigenvalue then the inverse doesn't exist in the first place. $\endgroup$
    – Okazaki
    Commented Jan 7, 2016 at 23:15

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.