The fault-tolerant procedures for state preparation and logical operation consider the procedure successful if only a one-qubit error occurs in each encoded block of the output.

So, I thought that the procedure for fault-tolerantly measuring the observable M, as shown in the figure below, should also assume pre-existing errors on the encoded data qubits.

Figure 10.28.

I understand that this circuit is not fault-tolerant, so we have to repeat it three times and take a majority vote to make the error probability $O(p^2)$.

But with a probability $O(p)$, there is a single-qubit pre-existing error, and it may propagate through the controlled-$M'$ gate. And the measurement on the ancilla is likely to be incorrect. The single-qubit pre-existing error will still exist in the second and third measurement trials. So I think the measurement error probability after the majority vote will also be $O(p)$.

What am I misunderstanding? Or do we go through the error recovery process for each of the three measurement procedures for the majority vote?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.