Figure 3: Generation of two- and three-qubit entanglement by stabilizer measurements.
From: Detecting bit-flip errors in a logical qubit using stabilizer measurements
Starting with the data qubits in the state 
, we selectively perform stabilizer measurements by activating the corresponding ancilla, i.e., preparing it in a maximal superposition state. (a,b), Performing one parity measurement generates entanglement between the paired data qubits. Measured average 
of the four witness operators 
and 
involving the data qubits paired by activating the top (a) or bottom (b) ancilla only and postselection on Pt=o and Pb=o, respectively. Entanglement is witnessed whenever 
(shaded area). The weak oscillations in 
result from false positives, which we have partially reduced here by postselecting more strongly than the threshold maximizing the average parity assignment fidelity. Standard deviations of 
(∼0.007, smaller than the symbol size) are estimated by bootstrapping 31. (c), Measured average of the Mermin operator 
with both ancillas activated and data strongly postselected on PtPb=oo (black circles). Three-qubit entanglement is witnessed whenever 
(shaded area). A stronger violation of the Mermin inequality is observed when targeting the GHZ state 
using unitary gates only (white circles). The average standard deviations of 0.1 (encoding by measurement) and 0.08 (encoding by gates) are smaller than the symbol size. (d), Tomography (absolute value of the density matrix elements) of the 
-maximizing state generated by double-parity measurement. The fidelity F=〈GHZ|ρ|GHZ〉 is 73%. For comparison, targeting this state with gates achieves F=82%.
