I am working on a project and I expect to have expressions of a bunch of quantum channels of interest. The quantum channels will be in matrix form. For example for a 2 qubit system, the quantum channel will be given by a 16 by 16 matrix.

Is there any systematic way to use the matrix representations of quantum channels and find the corresponding Kraus operators etc? Find which channels correspond to unitary channels etc?

  • $\begingroup$ This might be helpful: quantumcomputing.stackexchange.com/a/5816/13968 $\endgroup$
    – narip
    Nov 21, 2022 at 2:10
  • $\begingroup$ There are multiple representations of channels using a single, necessarily square matrix, most notably the Choi and $\chi$ matrix. See e.g. this previous answer of mine. For both these we have: 1) the eigenvectors correspond to the Kraus operators and 2) if the matrix is single rank it corresponds to a unitary. If you let us know what representation you will work with I can write up a more detailed answer. $\endgroup$
    – JSdJ
    Nov 21, 2022 at 8:28
  • $\begingroup$ you can also find a bunch of explicit examples of matrix representations (with a focus on Chois and Stinespring isometries) in quantumcomputing.stackexchange.com/q/24511/55. Does any of these linked posts answer your question? If not, could you further clarify what you are asking and how it differs from the other related discussions? $\endgroup$
    – glS
    Nov 21, 2022 at 9:32
  • $\begingroup$ I have yet to perform the computation but I expect to have a bunch of matrices in the usual sense. I'll have to throw away the matrices that don't correspond to CPTP maps. I'll then like to identify the remaining matrices/CPTP maps with their Kraus and/or Choi decomposition. $\endgroup$
    – user22511
    Nov 21, 2022 at 16:12
  • $\begingroup$ @user22511 that still doesn't clarify the problem. There are multiple possible matrix representations of a channel. You need to specify which one you are talking about. $\endgroup$
    – glS
    Nov 21, 2022 at 18:13

1 Answer 1


In your latest comments you talked about matrices (channels) that meet some commutation relations. This is best tackled with the representation matrix (also known as "natural representation") with respect to vectorization---as linked in one of the comments, cf. also this answer---which is the unique matrix $\hat\Phi$ such that ${\rm vec}(\Phi(\rho))=\hat\Phi{\rm vec}(\rho)$ for all $\rho$. Given any $\Phi:\mathbb C^{n\times n}\to\mathbb C^{n\times n}$ the explicit expression for $\hat\Phi$ reads $$ \hat\Phi=\sum_{i,j,k,l=1}^n\;\langle l|\Phi(|i\rangle\langle j|)|k\rangle\;|k\rangle\langle j|\otimes|l\rangle\langle i| $$ as is readily verified. The key property here is that $\widehat{(\cdot)}$ is a (linear) homomorphism, that is, $\widehat{\Phi_1\Phi_2}=\widehat{\Phi_1}\widehat{\Phi_2}$ for all $\Phi_1,\Phi_2$; thus commutators of linear maps are in one-to-one correspondence with commutators of their representation matrices. Then in order to check complete positivity of $\Phi$ you have to check the Choi matrix $\mathsf C(\Phi)$ which is just a re-shuffling of $\hat\Phi$. More precisely, it is easy to see that $$ \mathsf C(\Phi)=\sum_{i,j,k,l}\;\langle l|\Phi(|i\rangle\langle j|)|k\rangle\;\color{red}{|i\rangle}\langle j|\otimes|l\rangle\color{red}{\langle k|} $$ so the linear (and even unitary) map $S:\mathbb C^{n^2\times n^2}\to \mathbb C^{n^2\times n^2}$ uniquely defined via $S(|k\otimes l\rangle\langle j\otimes i|):=|i\otimes l\rangle\langle j\otimes k|$ satisfies $S(\hat\Phi)=\mathsf C(\Phi)$ (and even $S(\mathsf C(\Phi))=\hat\Phi$ because $S^2={\rm id}$). From the Choi matrix complete positivity and Kraus operators are obtained as usual.

Finally, because you explicitly mentioned it let me say that as per this result a unitary representation matrix $\hat\Phi$ corresponds to a channel if and only if $\hat\Phi=\overline{U}\otimes U$ for some $U\in\mathsf U(n)$. An easy way to check this would be to look at the partial traces of $\hat\Phi$ with respect to the first and second tensor factor and to check that they are both unitary and coincide up to complex conjugation $\overline{(\cdot)}$.


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.