Considering a qubit $\scr H =\Bbb C^2$ I have seen a proof of the no-cloning theorem for pure states. I wonder how do you prove it for a classical state?

1)That is, how do I prove that there is no quantum channel that clones all classical states?

2) For a state that is both pure and classical my books says that there exists a channel that clones all of them. How do I find it?

Definition of quantum channel: We say that Is that it is a superoperator $\Phi$ (a map in $L(L(\scr {H_A}),L(\scr {H_B})$) that is trace preserving and completely positive( For all $\scr H_R$ and $M_{AR}\ge 0$, it holds that $(\Phi\otimes I_R)[M_{AR}]\ge 0$ )

Definition of clonning:

A quantum channel $\Phi \in C(\scr H,\scr H\otimes \scr H)$ (C is just the set of all quantum channels between the specified spaces)clones a state $\rho \in D(\scr H)$ if $\Phi[\rho]=\rho\otimes \rho$

Definition of classical state

A quantum state $\rho$ on $ C(\scr H^{\Sigma})$ is classical if it is of the form $\sum_{x \in \Sigma}p_x|x\rangle\langle x|$ where $(p_x)_{x \in \Sigma}\in P(\Sigma)$ is an arbitrary probability distribution.

Edit: My try: For the classical case:

I have for the more simple case of the question which is $\scr H =\Bbb C^2$.

$\rho= \begin{pmatrix} p_0 & 0 \\ 0 & p_1 \end{pmatrix}$

and if by the sake of contradiction I assume that it can be cloned, there exists $\Phi$ such that $\Phi(\rho)=\begin{pmatrix} p_0 & 0 & 0 & 0\\ 0 &p_1 & 0 & 0 \\ 0 & 0 & p_0 & 0 \\ 0 & 0 & 0 &p_1\end{pmatrix}$. Now I am supposed to use the lineary of the channel to arrive at a contradiction. OR I can observe that the trace is 2 and not 1 as it should, so it is not trace-preserving. A contradiction. Is this approach correct?

  • $\begingroup$ What do you mean by a ``classical state''? By the standard definition, it should be clonable. $\endgroup$
    – Rammus
    Commented Mar 8 at 10:19
  • $\begingroup$ @Rammus It means diagonal with respect to the standard basis $\endgroup$
    – darkside
    Commented Mar 8 at 10:19
  • $\begingroup$ @Rammus Who are the pure and classical states? $\endgroup$
    – darkside
    Commented Mar 8 at 10:27
  • $\begingroup$ @darkside maybe start by considering a single classical bit. Say you have a pure state $|b\rangle$ where $b \in \{0, 1\}$, what is a map that takes $|b\rangle|0\rangle$ to $|b\rangle|b\rangle$? Is there a map that works for single qubit mixed states? $\endgroup$
    – xzkxyz
    Commented Mar 8 at 19:31

1 Answer 1


Consider a classical state $$ \rho = \sum_x p(x) |x\rangle \langle x|\,. $$ Now take a quantum channel $\Phi(M) = \sum_x K_x M K_x^\dagger$ defined by the Kraus operators $\{K_x\}_x$ where $K_x = |xx\rangle\langle x|$, then $$ \Phi(\rho) = \sum_x K_x \rho K_x^{\dagger} = \sum_x p(x) |xx\rangle \langle xx|\,. $$ Now this is not your definition of cloning, but I would argue it is the correct definition. The state represents two random variables $XX'$ which both have the same marginal distribution $p(x)$ and $X=X'$, i.e. $X'$ is an exact copy of $X$.

Your definition of cloning You are asking for a state representing two random variables $XX'$ that have the same marginal distribution $p(x)$ but are independent, that is you want $$ \Phi\left(\sum_x p(x) |x\rangle \langle x|\right) = \left(\sum_x p(x) |x\rangle \langle x|\right) \otimes \left(\sum_x p(x) |x\rangle \langle x|\right). $$ But you have the immediate problem that this mapping is not linear, you should have $\Phi(\lambda \rho) = \lambda \Phi(\rho)$ but with the above map you get $\Phi(\lambda \rho) = \lambda^2 \Phi(\rho)$. For the case of pure states, $p(x)=1$ for some $x$ and then this is covered by first map I described.

  • $\begingroup$ what is the range of values of lambda here? $\endgroup$
    – darkside
    Commented Mar 8 at 10:54
  • $\begingroup$ is it real in [0,1] only? $\endgroup$
    – darkside
    Commented Mar 8 at 10:55
  • $\begingroup$ Would my trace argument work in my try in the edit? $\endgroup$
    – darkside
    Commented Mar 8 at 11:07
  • $\begingroup$ If it is linear $\lambda \in \mathbb{C}$. $\endgroup$
    – Rammus
    Commented Mar 8 at 14:03
  • $\begingroup$ No your approach is not correct, you have written $\rho \oplus \rho$ and not $\rho \otimes \rho$. $\endgroup$
    – Rammus
    Commented Mar 8 at 14:05

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.