add ability to specify open controls.

[ewinston]

This PR adds the ability to control on '0' qubit states as described in issue #3726. Visualizations are not updated in this PR.

Summary

Details and comments

Add ability to specify control conditioned on a qubit being in the ground state. The state of the control qubits is represented by a decimal integer. For example:

  from qiskit import QuantumCircuit
  from qiskit.extensions.standard import XGate

  qc = QuantumCircuit(4)
  cgate = XGate().control(3, ctrl_state=6)
  qc.append(cgate, [0, 1, 2, 3])

Creates a four qubit gate where the fourth qubit gets flipped if the first qubit is in the ground state and the second and third qubits are in the excited state. If ctrl_state is None, the default, control is conditioned on all control qubits being excited.

Alternatively the ctrl_state may be specified as a bit string:

 cgate = XGate().control(3, ctrl_state='110').

[ajavadia]

can you make ctrl_state also accept a bitstring too, as it may be more natrual to specify like that ('110' or 6)

[Cryoris]

@ajavadia What's the status on this PR? This implements a functionality we would like to use in #3714, which is part of the Aqua 0.7 goal

[ajavadia]

There are no tests for a controlled Unitary gate (specified by a matrix). Can you add one please?

There are no modifications to the ControlledGate class as far as I can tell. It would be nice if the ctrl_state was also stored in instances of ControlledGate, so that e.g. it could be visualized with open controls. The conjugation with X gates should be done when decomposing that controlled gate.

[ajavadia]

This is from before, but why is Rz a special case here, and all other gates are further down?

[ewinston]

This is due to the global phase issue. When num_ctrl_qubits=1 CrzGate is properly retrieved. However when num_ctrl_qubits > 1, the algorithm computes the control gate from the definition of the base gate. Since RzGate is defined by U1Gate, this leads to the wrong phase. By reducing the number of controls by 1 and using instead the definition from CrzGate, which is phase correct, the proper controlled version is generated.

[ajavadia]

Ok, so maybe a comment here so someone reading the code would know why this is the way it is

[ewinston]

@ajavadia Controlled unitaries are tested in the method test_controlled_random_unitary.
I'll add ctrl_state as an attribute.

[ajavadia]

The changes look good, thanks.
Were you going to test open control unitaries too?
Also I just saw the test for controlled unitaries and this is how random unitaries are made:
UnitaryGate(scipy.stats.unitary_group.rvs(num_target), but a simpler way is to just use qiskit.quantum_info.random.random_unitary()