What is the method to create arbitrary controlled circuits in Qiskit?

Question

I am working on an implementation of an algorithm which uses ampitude estimation and amplification as a subroutine but we are not using qiskit but QuEST. QuEST doesn't have any direct method to create a controlled circuit. So we're trying to find out some efficient method for that using only the basic functionality present in most quantum computing languages.

I am looking into this paper - https://www.nature.com/articles/ncomms1392.pdf. But I am having a hard time understanding how the particular thing is working, I also looked at the backend code of qiskit but that is far more confusing as it has a lot of inter-dependencies from different classes and all.

Answer 1

First, create a custom gate. A possible way to do that is from a circuit:

from qiskit import QuantumCircuit

subcircuit = QuantumCircuit(3, name='myGate')
subcircuit.h(0)
subcircuit.h(1)
subcircuit.ccx(0,1,2)
my_gate = subcircuit.to_gate()

Then, create a controlled version of that gate. You need to specify how many controls you have. For example, 1:

controlled_my_gate = my_gate.control(1)

You can add this controlled custom gate to a circuit with append:

circuit = QuantumCircuit(4)
circuit.append(controlled_my_gate, [0,1,2,3])
circuit.draw('text')

q_0: ─────■─────
     ┌────┴────┐
q_1: ┤0        ├
     │         │
q_2: ┤1 myGate ├
     │         │
q_3: ┤2        ├
     └─────────┘

The method decompose will tell you how this controlled gate is built:

circuit.decompose().draw('text', fold=-1) 

q_0: ───────■───────────────■───────────────■──────────■───■──────────■───■─────────■───■────────■──────────■────■───■────────■──────────■───■──────────────■────────
     ┌──────┴───────┐       │               │          │   │          │   │         │   │        │          │    │   │P(π/4)  │          │   │              │        
q_1: ┤ U(π/2,0,π,0) ├───────┼───────────────┼──────────┼───┼──────────■───┼─────────┼───┼────────┼──────────■────■───■────────┼──────────■───┼──────────────┼────────
     └──────────────┘┌──────┴───────┐       │          │   │          │   │         │   │P(π/4)  │          │  ┌─┴─┐          │P(-π/4) ┌─┴─┐ │              │        
q_2: ────────────────┤ U(π/2,0,π,0) ├───────┼──────────■───┼──────────┼───┼─────────■───■────────┼──────────┼──┤ X ├──────────■────────┤ X ├─┼──────────────┼────────
                     └──────────────┘┌──────┴───────┐┌─┴─┐ │P(-π/4) ┌─┴─┐ │P(π/4) ┌─┴─┐          │P(-π/4) ┌─┴─┐└───┘                   └───┘ │P(π/4) ┌──────┴───────┐
q_3: ────────────────────────────────┤ U(π/2,0,π,0) ├┤ X ├─■────────┤ X ├─■───────┤ X ├──────────■────────┤ X ├──────────────────────────────■───────┤ U(π/2,0,π,0) ├
                                     └──────────────┘└───┘          └───┘         └───┘                   └───┘                                      └──────────────┘

I think it roughly follows the method in the attached paper.