Implement quantum operations using Paddle Quantum's gate and channel¶

Copyright (c) 2023 Institute for Quantum Computing, Baidu Inc. All Rights Reserved.

The gate and channel modules allow basic quantum operations to be defined. They can act on a state object directly or be inserted into a quantum circuit. Paddle Quantum provides a list of common operations but customized ones can also be created.

In [12]:
import paddle_quantum
from paddle_quantum.ansatz import Circuit
from paddle_quantum.gate.custom import Oracle
from paddle_quantum.channel import *
from paddle_quantum.state.common import zero_state
from paddle_quantum.qinfo import qft_generation, random_channel
from paddle_quantum.linalg import dagger

Custom gates¶

Previously, we've seen how to create a gate object and insert it into a circuit. Users can also implement customized unitary operation using the Oracle class.

For example, consider the n qubit inverse quantum Fourier transform (QFT), it's a unitary operation whose matrix is given by

1√N⎛⎜ ⎜ ⎜ ⎜ ⎜ ⎜⎝111…11ωN−1ωN−2…ω1ω2(N−1)ω2(N−2)…ω2……………1ω(N−1)(N−1)ω(N−1)(N−2)…ωN−1⎞⎟ ⎟ ⎟ ⎟ ⎟ ⎟⎠1N(111…11ωN−1ωN−2…ω1ω2(N−1)ω2(N−2)…ω2……………1ω(N−1)(N−1)ω(N−1)(N−2)…ωN−1)

where N=2nN=2n is the dimension of the Hilbert space and ω=e2πi/Nω=e2πi/N.

The following code creates an oracle that implements a 2-qubit inverse QFT.

In [13]:
unitary = qft_generation(2) # obtain matrix for 2-qubit QFT using Paddle Quantum's built-in function
unitary = dagger(unitary) # obtain matrix for inverse QFT by calculating conjugate transpose of QFT matrix using Paddle Quantum's built-in function
QFT_inverse = Oracle(unitary)

As a gate object, we can insert it into a circuit and print the circuit.

In [14]:
cir = Circuit(3)
cir.h()
cir.cnot([0,1])
cir.append(QFT_inverse) # insert the oracle
cir.cnot([2,1])
cir.plot()
No description has been provided for this image

Note: to print the oracle, it must not be disconnected, otherwise error message will be displayed.

To change how the oracle is printed, users can modify the content of gate_info. Users can change the name of the oracle printed and the width of the gate in the plot.

In [15]:
gate_info = {
    'texname': r'$QFT^{\dagger}$', # change the name of the oracle when printing the circuit
    'plot_width': 1.5, # change the width of the box representing the oracle
}
# update gate_info of the oracle
QFT_inverse.gate_info.update(gate_info)
# create the same circuit as before
cir = Circuit(3)
cir.h()
cir.cnot([0,1])
cir.append(QFT_inverse) # insert the oracle gate
cir.cnot([2,1])
cir.plot()
No description has been provided for this image

Another way of inserting an oracle into a circuit is by calling Circuit.oracle directly when creating the circuit.

In [16]:
cir=Circuit(3)
cir.h()
cir.cnot([0,1])
# call cir.oracle directly to create an oracle gate, its gate_info can also be specified when creating
cir.oracle(unitary, qubits_idx=[0,1], latex_name=r'$QFT^{\dagger}$', plot_width=1.5) 
cir.cnot([2,1])
cir.plot()
No description has been provided for this image

Channel¶

Paddle Quantum also supports channel as the most general quantum operation.

There are two ways of creating a channel. Users can either create one of the common channels provided by Paddle Quantum, or create a custom channel.

Note: Unlike gate object which supports both in 'state_vector' and 'density_matrix' backend, channel must be used in 'density_matrix' backend.

In [17]:
paddle_quantum.set_backend('density_matrix') # set the backend to density matrix

For example, we can create a bit flip channel and apply it to the zero state. For the full list of common channels, see API of Paddle Quantum

In [18]:
bit_flip_channel = BitFlip(prob=0.5, num_qubits=1) # define a single qubit bit flip channel with probability 0.5
input = zero_state(1)
output = bit_flip_channel(input) # apply bit flip channel to input zero state
print(output)

[[0.49999997+0.j 0.        +0.j]
 [0.        +0.j 0.49999997+0.j]]

A custom channel can be created using either its Kraus operators, Stinespring isometry, or Choi matrix. We show how to create a custom channel using these three different representations. We generate them randomly using the built-in function from Paddle Quantum.

Note: the convention for Choi matrix in PaddleQuantum is that the channel acts on the second system, i.e.

J(Λ)=∑i,j|i⟩⟨j|⊗Λ(|i⟩⟨j|).J(Λ)=∑i,j|i⟩⟨j|⊗Λ(|i⟩⟨j|).

where ΛΛ denotes the channel.

In [19]:
# generate Kraus operators of a random channel
kraus = random_channel(num_qubits=1, target='Kraus') 
# create channel using Kraus operators
custom_kraus = KrausRepr(kraus)

# generate Stinespring isometry of a random channel
stinespring = random_channel(num_qubits=1, target='Stinespring')
# create channel using Stinespring isometry
custom_stinespring = StinespringRepr(stinespring)

# generate Choi matrix of a random channel
choi = random_channel(num_qubits=1, target='Choi')
# create channel using Choi matrix
custom_choi = ChoiRepr(choi)

Channel can also be inserted into a circuit using the same syntax as for gate. However, currently printing circuit containing a channel is not supported.

In [20]:
 # by default, the channel would act on qubit with index 0, we can specify which qubit it acts on by providing qubits_idx
custom_choi = ChoiRepr(choi, qubits_idx=[2])
cir = Circuit(3)
cir.append(bit_flip_channel) # insert a channel object at the end of the circuit
cir.h()
cir.insert(index=2, operator = custom_choi) # insert a channel object at specific position
cir.bit_flip(0.5, qubits_idx=[2]) # create a channel when building the circuit