Suppose I have two quantum channels. Assume they they consist of $r\in \mathbb{Z}$ applications of unitaries, $U$ and $V$ respectively. Let the error between the channels acting on some state $\rho$ be: $$\varepsilon(U,V) = U^{\dagger r} \rho U^r - V^{\dagger r} \rho V^r. $$

Ideally, I want an expression for $\epsilon(U,V)$ in terms of the difference between these unitaries and $r$. Ideally I want an expression for $\varepsilon(U,V)$ as defined above, not the norm of the expression.

My first attempt was to write that $U^r - V^r = \sum_{k=1}^{r-1} U^{r-k}(U-V)V^k$ and substitute into the expression for $\varepsilon(U,V)$. However, this introduces a lot of cross terms which make things messy, and one ends up with $O(r^2)$ many terms when it feels like there should only be $O(r)$. Is there a better way to do this?

  • $\begingroup$ What do you mean by "in terms of the difference between these unitaries"? Will you only accept functions of $U-V$? $\endgroup$ Commented Sep 18, 2023 at 18:47
  • $\begingroup$ Or something like $(U^\dagger\rho U - V^\dagger\rho V)$ (i.e. a single application of the channel). Or potentially something similar. $\endgroup$ Commented Sep 18, 2023 at 19:03
  • 1
    $\begingroup$ Hm. I was thinking along the lines of $\epsilon(U,V)=U^{\dagger r}[\rho - W \rho W^\dagger ]U^r$ for $W=U^r V^{\dagger r}$, and finding $W$ in terms of $U V^\dagger$, but even that requires lots and lots of arbitrary commutators $\endgroup$ Commented Sep 18, 2023 at 20:23
  • $\begingroup$ To be honest, I think I'd be happy with anything which is linear in $r$ and some characterisation of the error between a single application of the separate unitaries. $\endgroup$ Commented Sep 18, 2023 at 21:25
  • $\begingroup$ @QuantumMechanic yeah, if $U$ and $V$ are simultaneously diagonalizable, then it would make things a lot easier. $\endgroup$
    – FDGod
    Commented Sep 18, 2023 at 23:01

1 Answer 1


The following should meet OP's expectations for a suitable expansion as specificed in their comment: Defining the difference channel $\Phi_{U,V}(\rho):=U^*\rho U-V^*\rho V$ one can indeed express $\varepsilon(U,V)$ as a sum of $r$ terms involving just $\Phi_{U,V}$ (and, of course, the original unitaries $U,V$): $$ \boxed{\varepsilon(U,V)=\sum_{j=0}^{r-1}(U^*)^{r-j-1} \Phi_{U,V}\big( (V^*)^{j}\rho V^{j} \big) U^{r-j-1}}\tag1 $$ While this holds for arbitrary $U,V$ and all $\rho$, in the special case where $[U,V]=0$ Eq.(1) simplifies to $$ \varepsilon(U,V)=\sum_{j=0}^{r-1}(U^*)^{r-j-1} (V^*)^{j}\Phi_{U,V}( \rho ) V^{j}U^{r-j-1}\tag2 $$ so there the only "error information" one needs is $\Phi_{U,V}( \rho )=U^*\rho U-V^*\rho V$; for the general case, however, I don't see a way around evaluating the error not only on $\rho$ but also on $(V^*)^{j}\rho V^{j}$ for all $j$. In other words I think this is the best one can do given the desired constraints on the expansion.

Proof of Eq.(1): \begin{align*} \sum_{j=0}^{r-1}&(U^*)^{r-j-1} \Phi_{U,V}\big( (V^*)^{j}\rho V^{j} \big) U^{r-j-1}\\ &=\sum_{j=0}^{r-1}(U^*)^{r-j-1} U^* (V^*)^{j}\rho V^{j} U U^{r-j-1}-\sum_{j=0}^{r-1}(U^*)^{r-j-1} V^*(V^*)^{j}\rho V^{j} V U^{r-j-1}\\ &=\sum_{j=0}^{r-1}(U^*)^{r-j} (V^*)^{j}\rho V^{j} U^{r-j}-\sum_{j=0}^{r-1}(U^*)^{r-j-1} (V^*)^{j+1}\rho V^{j+1}U^{r-j-1}\\ &=\sum_{j=0}^{r-1}(U^*)^{r-j} (V^*)^{j}\rho V^{j} U^{r-j}-\sum_{j=1}^{r}(U^*)^{r-j} (V^*)^{j}\rho V^{j}U^{r-j}\\ &=(U^*)^{r-j} (V^*)^{j}\rho V^{j} U^{r-j}\Big|_{j=0}-(U^*)^{r-j} (V^*)^{j}\rho V^{j}U^{r-j}\Big|_{j=r}\\ &=(U^*)^r\rho U^r-(V^*)^r\rho V^r=\varepsilon(U,V)\tag*{$\square$} \end{align*}


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.