I'm getting confused by the Dirac notation. Suppose I have the following two objects.

$$\rho = \sum_k p_k (\rho_A \otimes \rho_B) = \sum_k p_k |k \rangle \langle k | \otimes |k\rangle \langle k | ,$$

$$A = \sum_{ij} |ii \rangle \langle jj | = \sum_{ij} |i \rangle \langle j | \otimes |i \rangle \langle j | .$$

Here $\rho$ is a seperable density matrix, but $A$ isn't quite a density matrix (missing prefactor).

I wanna write down what $A\rho$ is. Do I just write them all "next to" one another?

$$A\rho \rightarrow \sum_{ijk} p_k |i \rangle \langle j | \otimes |i \rangle \langle j | k \rangle \langle k | \otimes |k\rangle \langle k |$$

This makes no sense, now there's two Kronecker products! So maybe like this?

$$A\rho \rightarrow \sum_{ijk} p_k |i \rangle \langle j |k \rangle \langle k | \otimes |i \rangle \langle j |k \rangle \langle k | = \sum_{ij} p_j |i \rangle \langle j | \otimes |i \rangle \langle j | $$

This looks suspicious to me, but let's go on. Ultimately I want to take the trace:

$$tr A\rho = \sum_{ij} p_j tr \left(|i \rangle \langle j | \otimes |i \rangle \langle j | \right) = \sum_{ij} p_j \delta_{ij} = 1$$

I used the fact that trace of Kronecker product is a product of traces. This seems wrong. I don't like that I got a $1$, because $A$ wasn't a density matrix in the first place. Dividing it by a factor would make it one, but then the trace of the product of two density matrix wouldn't be unity, which it should.

What's the mistake I'm making?


1 Answer 1


Notation-wise, you can use either of \begin{align} A\rho & = \sum_{ijk} p_k \Big(|i \rangle \langle j | \otimes |i \rangle \langle j |\Big) \Big(| k \rangle \langle k | \otimes |k\rangle \langle k |\Big) \\ & = \sum_{ijk} p_k |i \rangle \langle j |k \rangle \langle k | \otimes |i \rangle \langle j |k \rangle \langle k | . \end{align} As you well note, here the $⟨j|k⟩$ give $\delta_{jk}$'s, so the sum over $k$ goes away: \begin{align} A\rho & = \sum_{ij} p_j |i \rangle \langle j | \otimes |i \rangle \langle j |. \end{align} The rest of your formal manipulations are also fine:

$$\mathrm{Tr} \left(A\rho\right) = \sum_{ij} p_j \mathrm{Tr} \Big(|i \rangle \langle j | \otimes |i \rangle \langle j | \Big) = \sum_{ij} p_j \delta_{ij} = 1.$$

This does look vaguely suspicious, but it is reasonable because of the special structure of $A$ and $\rho$. It is not a contradiction because there is nothing that requires the product of density matrices to be anything like a density matrix: if $\rho$ and $\sigma$ don't commute, $\rho\sigma$ is not even hermitian, let alone positive semidefinite; if you want an example of density matrices whose product has trace $≠1$, try $\sigma^2$ for $$ \sigma=\begin{pmatrix}1/2 & 0 \\ 0 & 1/2\end{pmatrix}. $$

  • $\begingroup$ That's all I ever wanted, all I ever needed, thanks! $\endgroup$ Jun 10, 2016 at 14:55

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.