It is often claimed in the literature of random quantum hybrid circuits that a measurement induced phase transition is invisible in trajectory averages of linear observables and can only be seen in trajectory averages of non-linear observables (see here, here and here). Romain Vasseur in his BSS lectures (see here), makes the following argument to justify this claim:

Let's consider a brick-wall circuit in which we have the following pattern of evolution: first we apply a Haar random unitary upon even links, then a few measurements are randomly made followed by Haar random unitaries applied to the odd links. The trajectory average i.e. average over different measurement outcomes can be written as a quantum channel: $$\bar{\rho}_t = \sum_{\{m\}} |\psi_{m} \rangle \langle \psi_{m} | = N_t(\rho)=\sum_{\{m\}} K_m \rho K_m^{\dagger}$$ where the Kraus operators are of the form: $$K_m = \prod_{\tau=1}^t \hat{\Pi}^{m_T^{'}} U_\tau^{\text {odd }} \hat{\Pi}^{m_T} U_\tau^{\text {even }}$$ here,

  • $\hat{\Pi}^{m_T^{'}}$: the projector onto measurement outcome $m_T^{'}$
  • $U_\tau^{\text {odd }}$: Haar-random unitary acting on odd links
  • $U_\tau^{\text {even }}$: Haar-random unitary acting on even links

The central claim is that linear quantities such as $\langle O \rangle$, $$\langle O \rangle = \sum_{\{m\}} p_m \frac{\left\langle\psi_m|O| \psi_m\right\rangle}{\left\langle\psi_n \mid \psi_n\right\rangle}=\text{Tr}{\left(\bar{\rho}*t O\right)}$$ cannot "see" measurement induced phase transitions as at long times, the trajectory averaged density matrix becomes an infinite temperature state: $$\rho_t \rightarrow \rho_{\infty}=\frac{1}{d^L} \mathbb{I}$$ This can be shown by taking the following limit: $$\lim_{t\rightarrow \infty} \bar{\rho}_t = \lim_{t\rightarrow \infty} \sum_{\{m\}} K_m \rho K_m^{\dagger}$$

My questions are the following:

  1. How does one compute this limit?
  2. How does one generalize this argument beyond brick-wall circuits? Is there rather a more intuitive way to understand this?

1 Answer 1


By definition of the Kraus operators $K_m$, trajectory average is given by the channel defined by the equation, $$ \bar{\rho}_t = \sum_{\{m\}} \prod_{\tau = 1}^{t} \left(\hat{\Pi}^{m_T^{'}} U_\tau^{\text{odd }} \hat{\Pi}^{m_T} U_\tau^{\text{even }}\right) \rho \left( U_\tau^{\dagger \text{ even }} \hat{\Pi}^{m_T} U_\tau^{\dagger \text{ odd }} \hat{\Pi}^{m_T^{'}} \right).$$

Now to evaluate the limit for $t \to \infty$, we must consider the Haar averages as the specifics of the random unitaries essentially wash out at this limit, leaving only the averaged effect. Given a Haar-random unitary $U$, the expectation value of the linear operator $U \mathcal{O} U^\dagger$ with respect to the Haar-measure is given by (see here), $$ \mathbb{E}_U (U \mathcal{O} U^\dagger) \equiv \int_{\mathbf U(d)} d\mu(U) U \mathcal{O} U^\dagger = \mathrm{Tr}(\mathcal{O})\frac{\mathbb{I}}{d}. $$

Using the above expression we compute Haar averages over even and odd links to get, $$ \lim_{t \to \infty} \bar{\rho}_t = \lim_{t \to \infty} \sum_{\{m\}} \prod_{\tau = 1}^{t} \frac{\mathbb{I}}{d} \hat{\Pi}^{m_T^{'}} = \lim_{t \to \infty} \frac{\mathbb{I}}{d^t} = \frac{\mathbb{I}}{d^L} $$

As you can see from the above expression, with increasing $t$, the state gets increasingly mixed. At $t \to \infty$, we can physically intuit that this must converge to the maximally mixed state.

I believe this answer should give some intuition to the result, as well as possible extensions therof, as it has been derived mostly using physical principles.


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.