3
$\begingroup$

I am reading the paper "Direct Fidelity Estimation from Few Pauli Measurements".

According to the paper, the entanglement fidelity between the a unitary channel $\mathcal U$ and a quantum channel $\mathcal E$, is defined as \begin{align} F_e = {\rm Tr}[\mathcal U^\dagger \mathcal E]/d^2, \tag{1} \end{align} where $d$ is dimension of the underlying Hilbert space and ${\rm Tr}(\cdot)$ is the superoperator trace. My aim is to verify that $F_e \in [-1,1]$. We know that ${\rm Tr}(\mathcal U^\dagger \mathcal U)=d^2$. After Eq. (41), it is directly provided that for any quantum channel, ${\rm Tr}(\mathcal E^\dagger \mathcal E) \in[0, d^2]$ and ${\rm Tr}(\mathcal U^\dagger \mathcal E) \in[-d^2, d^2]$. How to prove this?

According to Eq. (41), we have \begin{align} {\rm Tr}(\mathcal U^\dagger \mathcal E) = \sum_{k=1,k=1}^{d^2, d^2}\chi_{U}(k,k') \chi_{\mathcal E}(k,k'), \tag{2} \end{align} where $\chi_{\mathcal E}(k,k')=\frac{1}{d}{\rm Tr}(W_k^\dagger \mathcal E(W_{k'}))$ and $W_k$ are Pauli strings. It can be proved that $\chi_{U}(k,k')$ and $\chi_{\mathcal E}(k,k')$ take values in [-1,1] due to Pauli strings. However, following this way, I obtain ${\rm Tr}(\mathcal U^\dagger \mathcal E)\in [-d^4,d^4]$, which is inconsistent with paper.

$\endgroup$
8
  • $\begingroup$ Please can you make the question self-contained. What is $\varepsilon$, $U$ and $d$? I also see no quantum channel in the question at the moment. $\endgroup$
    – Rammus
    Commented Jan 10, 2023 at 10:08
  • 1
    $\begingroup$ What kind of trace is this, if those are channels? $\endgroup$ Commented Jan 10, 2023 at 10:58
  • $\begingroup$ Thank you. I have updated the information. $\endgroup$ Commented Jan 10, 2023 at 11:15
  • $\begingroup$ What is the "superoperator trace"? $\endgroup$ Commented Jan 10, 2023 at 20:56
  • $\begingroup$ @NorbertSchuch I made some edits such that the question is more in line with the paper. I think this is self-explanatory, no? Traces are defined for linear maps and superoperators are linear maps. $\endgroup$ Commented Jan 11, 2023 at 9:54

1 Answer 1

2
$\begingroup$

Note that the bound $\mathrm{Tr}(\mathcal E^\dagger \mathcal E) \geq 0$ is trivial since $\mathrm{Tr}(\mathcal E^\dagger \mathcal E) = \| \mathcal E \|_2^2$ is the square of the Schatten 2-norm of $\mathcal E$.

The following is slightly simpler if we use the cyclicity of the trace $\mathrm{Tr}(\mathcal E^\dagger \mathcal E) = \mathrm{Tr}(\mathcal E \mathcal E^\dagger)$ and consider the superoperator $\mathcal E \mathcal E^\dagger$ instead.

To prove the bound $\mathrm{Tr}(\mathcal E \mathcal E^\dagger) \leq d^2$, note that the trace is the sum of diagonal entries, say in the Pauli basis $W_k$. Thus, we can w.l.o.g. assume that $\mathcal E \mathcal E^\dagger$ is diagonal in the Pauli basis. Note that any quantum channel $\mathcal A$ which is diagonal in the Pauli basis is automatically unital (since $\mathcal A^\dagger = \mathcal A$ and $\mathcal A$ is trace-preserving by definition). It is well-known that the spectral norm of unital and trace-preserving quantum channels is $\|\mathcal A\|_\infty = 1$ (see e.g. the book by Watrous, Thm. 4.27).

Now, $\mathcal E \mathcal E^\dagger$ is completely positive but not necessarily trace-preserving (or unital, which is the same as we assumed that $\mathcal E \mathcal E^\dagger$ is diagonal). This is determined by the first diagonal entry $$ \frac1d(\mathbb 1|\mathcal E \mathcal E^\dagger| \mathbb 1) = \frac1d \|\mathcal E^\dagger(\mathbb 1)\|_2^2. $$ Here, $\| \cdot \|_2$ is the Schatten 2-norm (or Hilbert-Schmidt / Frobenius norm if you prefer). Now recall that $\mathcal E$ is trace-preserving and thus $\mathcal E^\dagger$ is unital. Hence, we have $$ \frac1d(\mathbb 1|\mathcal E \mathcal E^\dagger| \mathbb 1) = \frac1d \|\mathcal E^\dagger(\mathbb 1)\|_2^2 = \frac1d \|\mathbb 1\|_2^2 = \frac1d \mathrm{tr}(\mathbb 1) = 1. $$ Hence, we have shown that $\mathcal E \mathcal E^\dagger$ is a diagonal quantum channel and thus $\| \mathcal E \mathcal E^\dagger \|_\infty = 1$. From this, it immediately follows that $$ \| \mathcal E \|_2^2 = \mathrm{Tr}(\mathcal E \mathcal E^\dagger) \leq d^2 \| \mathcal E \mathcal E^\dagger \|_\infty = d^2. $$

Finally, we have $\| \mathcal U \|_2^2 = \mathrm{Tr}(\mathcal U^\dagger \mathcal U) = \mathrm{Tr}(\mathrm{Id}) = d^2$ for any unitary channel $\mathcal U$. Hence, by the Cauchy-Schwarz inequality, we have $$ |\mathrm{Tr}(\mathcal U^\dagger \mathcal E)| \leq \| \mathcal U \|_2 \|\mathcal E\|_2 \leq d^2, $$ for any unitary channel $\mathcal U$ and arbitrary channel $\mathcal E$.

$\endgroup$

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.