For CNOT gate with control qubit set to 1, the measured state of the second qubit unexpectedly depends on the measurement

Question

I have the following circuit in qiskit. It is a simple CNOT operation with the control qubit set to 1, by running the control qubit through an X-gate.

However, when I measure the 2nd qubit, I get the state $|00\rangle$, not $|11\rangle$.

If I add a measurement of the control qubit, I do see $|11\rangle$ as the final result.

I fully expected $|11\rangle$ as the outcome regardless of whether I measure a single or both qubits

from qiskit import ClassicalRegister, QuantumRegister, QuantumCircuit
from qiskit import execute
from qiskit import BasicAer
from math import *
q0 = QuantumRegister(2, "q0") # one ancilla qubit
c0 = ClassicalRegister(2, "c0") # n classical bits for output
qc = QuantumCircuit(q0,c0)
qc.x(q0[0])
qc.cx(q0[0], q0[1])
  #qc.measure(q0[0],c0[0]) <= when commented out I measure |00>
qc.measure(q0[1],c0[1])
qc.draw()
 job = execute(qc,backend=BasicAer.get_backend('qasm_simulator'))
 job.result().get_counts()

Answer 1

Thanks for pointing this out. This is a bug that occurs when only a subset of the qubits are measured.

It's being fixed. Until then, workarounds are:

• Use Aer instead of BasicAer (always the best thing to do when possible).

• Use LegacySimulators instead of BasicAer (this will give a deprecation warning).

• Install Qiskit from the master branch, where the issue has been fixed (here be dragons!).