0
$\begingroup$

Let me define (orthogonal) projection operator $P$ by its characterization: $P = P^2 = P^\dagger$

How can I show that $P$ can always be represented as $\sum_{i=1}^n |\psi_i\rangle\langle\psi_i|$ (where $\langle\psi_i|\psi_j\rangle = \delta_{ij}, n \le \mathrm{dim} \, \mathcal H$) in braket notation?

I only consider Hilbert spaces of finite dimention here (i.e. $\mathrm{dim} \, \mathcal H < \infty$).

It's easy to show that $\sum_i |\psi_i\rangle\langle\psi_i|$ satisfies the property, but I'm struggling to prove the converse.

Thanks in advance!

$\endgroup$
1
  • 1
    $\begingroup$ Your sum is over a subset of your complete set of kets and bras, in general. You may write any hermitian operator in terms of the complete set. Can you constrain the coefficients? $\endgroup$ Commented Mar 8, 2023 at 20:09

2 Answers 2

1
$\begingroup$

Operators that can be diagonalized look like \begin{align} P = \sum_i p_i \left | \psi _i \right > \left < \psi_i \right | \end{align} in their own eigenbasis. The spectral theorem tells us more than this; that $P$ is hermitian so it can be diagonalized by a unitary matrix. Therefore $\left < \psi_i | \psi_j \right > = \delta_{ij}$.

So far we have used $P^\dagger = P$ but not $P^2 = P$. As suggested above, you should use this to solve for the possible eigenvalues $p_i$.

$\endgroup$
1
$\begingroup$

Let us a assume a Hilbert space $\mathcal{H}$ with finite dimension $n$. The property $P =P^2$ implies that the eigenvalues of $P$ can only be $0$ or $1$. As $P$ is hermitian ($P^\dagger =P$), there exists an orthonormal basis of $\mathcal{H}$, $$\{|\psi_1\rangle, \ldots |\psi_r\rangle, |\psi_{r+1}\rangle, \ldots |\psi_n\rangle \}, \quad \langle \psi_i | \psi_j \rangle = \delta_{ij}, \quad \sum\limits_{i=1}^n |\psi_i \rangle \langle \psi_i | = \mathbf{1}_n,$$ of eigenvectors of $P$, with $P | \psi\rangle = |\psi_i \rangle $ for $i=1, \ldots r$ and $P |\psi_i \rangle =0$ for $i=r+1, \ldots n$ (assuming that the eigenvalue $1$ is $r$-fold degenerate). Consequently, the spectral representation of $P$ is given by $$P= \sum\limits_{i=1}^r |\psi_i \rangle \langle \psi_i |.$$ Obviously the choice of the vectors $\{ |\psi_i \rangle \}_{i=1}^r$ is not unique, any orthonormal basis of the $r$-dimensional subspace $P \mathcal{H}$ will do the job.

$\endgroup$

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.