Consider a quantum channel $\Phi$ that maps from density operators $\mathcal{S}(\mathcal{H}_A)$ to itself, that commutes with any unitary channel $\mathcal{U}$ on $\mathcal{S}(\mathcal{H}_A)$, i.e. $\Phi(\mathcal{U}(\rho))=\mathcal{U}(\Phi(\rho)),\,\,\forall\rho\in\mathcal{S}(\mathcal{H}_A)$.

Is there a general characterisation of such channels? (As far as I can think of it, $\Phi$ can either be a unitary channel or a depolarization channel. What are the other possibilities?)

  • $\begingroup$ This seems to me related to reference frame preserving maps iopscience.iop.org/article/10.1088/1367-2630/10/3/033023/pdf $\endgroup$
    – R.W
    Commented Apr 11, 2022 at 13:48
  • $\begingroup$ Thanks for the reference! @R.W Could you help point to the exact part where the relevant discussion is given? $\endgroup$
    – Shadumu
    Commented Apr 11, 2022 at 14:05
  • $\begingroup$ The given answer is great, check it out. My reference is only an application of this idea of using commutations to define interesting classes of channels, therefore only of secondary relevance to your question. $\endgroup$
    – R.W
    Commented Apr 11, 2022 at 14:11

1 Answer 1


Yes, there is, you're asking about the commutant of the representation $U\mapsto U(\cdot)U^\dagger$ of the unitary group $U(d)$. From a representation-theoretic point of view, the superoperators in the commutant are easy to characterize: By Schur's lemma, such an operator has to be proportional to the identity on every irrep of $U\mapsto U(\cdot)U^\dagger$.

These irreps are well-known and are given by the following orthogonal decomposition of the vector space of linear operators $L(\mathbb C^d)$: $$ L(\mathbb C^d) = \langle \mathbb I \rangle \oplus \{\text{traceless operators}\}. $$ The projection on $\langle \mathbb I \rangle$ is the completely depolarizing channel: $$ \mathcal{D}(X) := \mathrm{tr}(X) \frac{\mathbb I}{d}. $$ The projection on the traceless operators is simply $\mathrm{id}-\mathcal{D}$. Hence, every element $X$ in the commutant can be written as $$ X = a \mathcal{D} + b (\mathrm{id}-\mathcal{D}) = a' \mathcal{D} + b' \mathrm{id}. $$ The latter decomposition is not orthogonal, but has to advantage to answer your question: The only quantum channels which commute with unitary channels are the completely depolarizing channel and the identity channel, and of course any convex combination of those.

For more details on the rep theory, see my answer here: Why does the twirl of a quantum channel give a depolarizing channel?


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.