1
$\begingroup$

I know how the operator norms $\| X\|_{1}$,$\| X\|_{2}$, and $\| X\|_{\infty}$ are defined for any operator $X\in B(\mathcal{H})$. My question is about how to view$\| T(X)\|_{1}$,$\| T(X)\|_{2}$, and $\| T(X)\|_{\infty}$, when T is a quantum channel.

Espescially when we use the open-system representation $T(X)=Tr_{AE}(U(X\otimes |\psi_{AE}\rangle\langle\psi_{AE}|))U^{\dagger})$. For the 1-norm my intuition is:

I assume that since T is CP, it is self-adjoint. So for example the one norm becomes: $$\| T(X)\|_{1} =\| Tr_{AE}(U(X\otimes |\psi_{AE}\rangle\langle\psi_{AE}|))U^{\dagger})\|_{1}=Tr(Tr_{AE}(U(X\otimes |\psi_{AE}\rangle\langle\psi_{AE}|))U^{\dagger}))=Tr(U(X\otimes |\psi_{AE}\rangle\langle\psi_{AE}|))U^{\dagger})=Tr(X\otimes |\psi_{AE}\rangle\langle\psi_{AE}|)=Tr(X)Tr(|\psi_{AE}\rangle\langle\psi_{AE}|)=Tr(X). $$

But for the 2-norm and sup-norm I have no idea.

PS:

The question stems from trying to show $\|T(X)\|_{1}\leq \|X\|_{1}$ using open system representation, and then if this also holds for the 2 and the sup-norm.

$\endgroup$

1 Answer 1

3
$\begingroup$

First, let us prove that every quantum channel is trace-norm contractive using the unitary Stinespring representation. For this we need the following lemma.

Lemma. For all $m,n\in\mathbb N$ and all $Z\in\mathbb C^{n\times n}\otimes\mathbb C^{m\times m}$ it holds that $\|{\rm tr}_2(Z)\|_1\leq\|Z\|_1$.

Proof. The easiest way to show this is to go via the adjoint channel/dual channel, that is, the fact that ${\rm tr}({\rm tr}_2(Z)Y)={\rm tr}(Z(Y\otimes{\bf1}))$ for all $Z\in\mathbb C^{n\times n}\otimes\mathbb C^{m\times m}$, $Y\in\mathbb C^{n\times n}$. This simplifies the norm computation by the duality of the trace norm and the $\infty$-norm: \begin{align*} \|{\rm tr}_2(Z)\|_1&={\rm sup}_{Y\in\mathbb C^{m\times m},\|Y\|_\infty=1}|{\rm tr}({\rm tr}_2(Z)Y) )|\\ &={\rm sup}_{Y\in\mathbb C^{m\times m},\|Y\|_\infty=1}|{\rm tr}(Z(Y\otimes{\bf1}))|\\ &\leq{\rm sup}_{Y\in\mathbb C^{m\times m},\|Y\|_\infty=1}\|Z\|_1\|Y\otimes{\bf1}\|_\infty\\ &={\rm sup}_{Y\in\mathbb C^{m\times m},\|Y\|_\infty=1}\|Z\|_1\|Y\|_\infty\|{\bf1}\|_\infty=\|Z\|_1\,. \end{align*} In the second-to-last line we used this inequality and in the last line we used that the $\infty$-norm factorizes under tensor products. $\square$

With this the statement we want to prove follows at once; for all $X\in\mathbb C^{n\times n}$, all states $\omega\in\mathbb C^{m\times m}$ and all unitaries $U\in\mathbb C^{n\times n}\otimes\mathbb C^{m\times m}$ we compute \begin{align*} \|{\rm tr}_2(U( X\otimes\omega )U^*)\|_1&\leq \|U( X\otimes\omega )U^*\|_1\\ &=\|X\otimes\omega\|_1\\ &=\|X\|_1\|\omega\|_1\\ &=\|X\|_1{\rm tr}(\omega)=\|X\|_1\,. \end{align*} In the second step we used that the trace norm is invariant under unitary channels (because the singular values do not change) and in the last line we used that because $\omega\geq 0$ its trace norm is equal to its trace (obvious from the definition of $\|\cdot\|_1$).

One can re-write this result in terms of superoperator-norms as $\|T\|_{1\to 1}\leq 1$ for all channels $T$ where $\|T\|_{1\to 1}:=\sup_{\|X\|_1=1}\|T(X)\|_1$; in fact because every channel has a fixed point one even gets $\|T\|_{1\to 1}=1$ for all channels.


The follow-up question now is whether this result continues to hold if the underlying trace norm is replaced by another norm. For this one defines the general Schatten $p$-norm as $$\|X\|_p:=\big({\rm tr}((X^\dagger X)^{p/2})\big)^{1/p}$$---or, equivalently, $\|X\|_p:=(\sum_j\sigma_j(X)^p)^{1/p}$ where $\sigma_j(X)$ are the singular values of $X$---as well as the corresponding operator norm $\|\Phi\|_{p\to p}:=\sup_{\|X\|_p=1}\|\Phi(X)\|_p$. Note that this Schatten-$p$ norm is the natural generalization of $\|\cdot\|_2$ and $\|\cdot\|_\infty$ to other indices.

In this language, asking whether $\|T(X)\|_p\leq\|X\|_p$ for all $X$ is equivalent to $\|T\|_{p\to p}\leq 1$ ("$T$ is $p$-norm contractive"). Indeed, one can characterize when this is the case: as shown in this paper (arXiv) the following statements are equivalent:

  • $\Phi$ is unital, i.e. $\Phi({\bf1})={\bf1}$
  • $\|\Phi\|_{p\to p}=1$ for all $p\in[1,\infty]$
  • $\|\Phi\|_{p\to p}=1$ for some $p\in(1,\infty]$

In other words:

For all channels that are not unital and all $p>1$ there exists $X$ such that $\|T(X)\|_p>\|X\|_p$.

$\endgroup$
6
  • $\begingroup$ I didn't find it that enlightening, could you please elaborate? $\endgroup$ yesterday
  • $\begingroup$ @PinkElephants Elaborate on what exactly, is there something you don't understand or rather: in your opinion what aspect is missing form this answer? In the current version I show that, in general, channels do not satisfy $\|T(X)\|_2\leq\|X\|_2$ (actually: $\|T(X)\|_p\leq\|X\|_p$, $p>1$) for all $X$ meaning any proof attempt is futile. If there's something else you want to know about please let me know and I*ll gladly try to help out! $\endgroup$ yesterday
  • $\begingroup$ I think my confusion stems from the fact that the proof uses the Hahn-decompositions and not the open open-system repr. aswell as we have end up with a equality and not an inequality. And also i still dont see how we define $\|𝑇(𝑋)\|_{p}$ ? $\endgroup$ yesterday
  • $\begingroup$ @PinkElephants Got it, I adjusted my answer accordingly. Hope this helps! $\endgroup$ yesterday
  • $\begingroup$ Thank you very much. What do you mean by $Tr_{2}$? Is it the partial trace to the second space? $\endgroup$ yesterday

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.