How to get_statevector() with defined gates in Qiskit?

Question

I'm learning Qiskit for my QC lectures and I'm not able to run the following code:

import numpy as np
from qiskit import QuantumCircuit, QuantumRegister, assemble, Aer

def cnotnot(gate_label = 'CNOTNOT'):
    gate_circuit = QuantumCircuit(3, name = gate_label)
    gate_circuit.cnot(0, 1)
    gate_circuit.cnot(0, 2)

    gate = gate_circuit.to_gate()

    gate.label = gate_label

    return gate

q = QuantumRegister(3, name = 'q')

circuit = QuantumCircuit(q)

# Define initial state
initial_state = [1. / np.sqrt(2.), 1. / np.sqrt(2.)]
circuit.initialize(initial_state, 0)

circuit.append(cnotnot(), [q[0], q[1], q[2]])

circuit.draw(plot_barriers = False)

# Let's simulate our circuit in order to get the final state vector!
svsim = Aer.get_backend('statevector_simulator')

# Create a Qobj from the circuit for the simulator to run
qobj = assemble(circuit)

# Do the simulation, return the result and get the state vector
result = svsim.run(qobj).result().get_statevector()

# Get the state vector for the first qubit
final_state = [result[0], result[1]]

print('a and b coefficients before simulation:', initial_state)
print('a and b coefficients after simulation:', final_state)

without getting errors. In particular:

QiskitError: 'ERROR:  [Experiment 0] Circuit contains invalid instructions {"gates": {CNOTNOT}} for "statevector" method.

I followed this answer to define a CNOTNOT gate and append it to my circuit, but I can't get the Statevector after the simulation. If I don't use a defined gate (just use two separated CNOT), the code runs without problems.

Thanks for your time!

Answer 1

You are missing an important part of the preparation process of the circuit for it to run on the simulator (any backend, actually) : the transpilation.

In order to create the qobj, you first have to transpile your circuit into gates your backend "understands", kind of like a translator between your custom gates and the basis gates of the backend. First you do the transpile of the circuit to transform it into gates the backends knows, then you assemble it into a obj to be able to run it.

So for your code to run here, you would have to do something like this:

# Create a Qobj from the circuit for the simulator to run
test = transpile(circuit, svsim)
qobj = assemble(test)

# Do the simulation, return the result and get the state vector
result = svsim.run(qobj).result().get_statevector()

You can also use the execute function that does it for you:

from qiskit import execute
# Do the simulation, return the result and get the state vector
result = execute(circuit, svsim).result().get_statevector()

Hope this is clear enough, if you have any questions about this please let me know in the comments :)