Syndrome Extraction Circuits

An ancilla qubit and a few controlled gates measure a multi-qubit stabilizer as one classical bit without collapsing the logical state.

Reading a Pauli with an ancilla

To measure a stabilizer such as Z Z Z Z, prepare an ancilla in |0>, apply a controlled-Z-parity from each data qubit onto the ancilla (implemented with CNOTs), then measure the ancilla in the computational basis. The ancilla outcome is the parity: 0 for +1 eigenvalue, 1 for -1. The data qubits are projected onto the corresponding stabilizer eigenspace but are otherwise undisturbed.

# Measure Z0 Z1 Z2 Z3 into ancilla a
circuit.reset(a)
for q in (0,1,2,3):
    circuit.cx(q, a)   # accumulate Z-parity onto ancilla
circuit.measure(a, syndrome_bit)

For an X-type stabilizer such as X X X X, sandwich the ancilla coupling in Hadamards, or use an ancilla prepared in |+> with the control and target roles swapped, so the ancilla accumulates X-parity instead.

Order and hazards

The order of the controlled gates matters. A poorly scheduled circuit lets a single ancilla fault spread into a multi-qubit error on the data, defeating the code. This is why practical designs use carefully ordered CNOT schedules and, for fault tolerance, more elaborate ancilla preparations.

• Z-checks: CNOTs from data to ancilla, ancilla measured in Z basis.
• X-checks: Hadamard-wrapped couplings, or ancilla in |+> measured in X basis.
• One noisy ancilla can propagate to many data qubits if gates are naively ordered.
• Fault-tolerant designs limit this spread; see fault-tolerant syndrome extraction.

Fault tolerance of the extraction itself

Because a single ancilla shared across a high-weight stabilizer can turn one gate error into many data errors, fault-tolerant protocols use verified cat states, flag qubits, or Shor- and Steane-style ancillas so that any single fault produces a correctable error. See fault-tolerant syndrome extraction. In the surface code, each stabilizer touches only four data qubits with a fixed CNOT schedule, keeping extraction simple and local.

Because measurements are noisy, extraction is repeated many times and the time-ordered syndromes are decoded jointly.