Quantum-Pulse-OPT

Quantum-Pulse-OPT is the code for the quantum pulse optimization problem.

Install dependencies

You need to have Python with version 3.6 or later. You need also to install the package QuTip by the following command.

pip install qutip

Examples

There are two examples, the Ising model with one single controller and the CNOT model with two controllers.

Ising model

This model includes only one Hamiltonian controller. The users can run this example by the following command line.

cd example
python Ising.py

To do the rounding, please set

if_rounding = True

The users can set rounding parameters by the following code.

rounding_type = "BnB" # Another choice: "SUR"
min_up_time = 10

To use the warm start, please set

if_warm_start = True

The users can choose the warm start control you want from a .csv file by setting the variable initial_control as the file name. Here is an example.

initial_control = "../control/" + "{}_evotime{}_n_ts{}_ptype{}_offset{}_obj{}".format(
        example_name, 1, 16, "ZERO", 0.5, obj_type) + ".csv"

The users can change the evolution time and number of time steps by setting the following variables.

# change the value 
evo_time = 1
n_ts = 16

CNOT model

This model include two Hamiltonian controllers to estimate the CNOT gate. The users can run this example by the following command line.

cd example
python Ising.py

To do the rounding, please set

if_rounding = True

The users can set rounding parameters by the following code.

rounding_type = "BnB" # Another choice: "SUR"
min_up_time = 10

To use the warm start, please set

if_warm_start = True

The users can choose the warm start control you want from a .csv file by setting the variable initial_control as the file name. Here is an example.

initial_control = "../control/" + "{}_evotime{}_n_ts{}_ptype{}_offset{}_obj{}".format(
        example_name, 10, 200, "ZERO", 0.5, obj_type) + ".csv"

The users can add a constraint indicating that the summation of all the control variables should be 1 at each time step by setting

sum_cons_1 = True

The users can change the evolution time and number of time steps by setting the following variables.

# change the value 
evo_time = 20
n_ts = 400

Build New Example

Here are the following steps of solving the quantum optimal control by our code based on the QuTip.

Before starting, please import the needed functions.

import os
from qutip import Qobj, identity, sigmax, sigmaz, sigmay, tensor
from qutip.qip.operations.gates import cnot

from optcontrol import optcontrol
from rounding import rounding
from evolution import time_evolution, compute_obj

Step 1: set Hamiltonian matrices

# The control Hamiltonians
H_c = [tensor(sigmax(), identity(2)), tensor(sigmay(), identity(2))]
# Drift Hamiltonian
H_d = tensor(sigmax(), sigmax()) + tensor(sigmay(), sigmay()) + tensor(sigmaz(), sigmaz())
# start point for the gate evolution
X_0 = identity(4)
# Target for the gate evolution
X_targ = cnot()

Step 2: set number of time slots and time interval

# Number of time slots
n_ts = 40
# Time allowed for the evolution
evo_time = 2

Step 3: set convergence parameters

# Fidelity error target
fid_err_targ = 1e-10
# Maximum iterations for the optisation algorithm
max_iter = 500
# Maximum (elapsed) time allowed in seconds
max_wall_time = 120
# Minimum gradient (sum of gradients squared)
# as this tends to 0 -> local minima has been found
min_grad = 1e-50

Step 4: set initial values of control variables

p_type = "ZERO"
offset = 0.5
obj_type = "UNIT"
initial_control = None  # no warm start

# if you want to use warm start
if_warm_start = 1
if if_warm_start:
    p_type = "WARM"
    offset = 0
    initial_control = "../control/" + "{}_evotime{}_n_ts{}_ptype{}_offset{}_obj{}".format(
        example_name, 10, 200, "ZERO", 0.5, obj_type) + ".csv"

Step 5: set the output files

# output file of the numerical results
output_num = "../output/" + "{}_evotime{}_n_ts{}_ptype{}_offset{}_obj{}".format(
    example_name, evo_time, n_ts, p_type, offset, obj_type) + ".log"
# output file of the figures
output_fig = "../output/" + "{}_evotime{}_n_ts{}_ptype{}_offset{}_obj{}".format(
    example_name, evo_time, n_ts, p_type, offset, obj_type) + ".png"
# output file of the controls
output_control = "../control/" + "{}_evotime{}_n_ts{}_ptype{}_offset{}_obj{}".format(
    example_name, evo_time, n_ts, p_type, offset, obj_type) + ".csv"

Step 6: run optimization and output results

# choose to add the summation as 1 constraint or not
sum_cons_1 = True
# run the optimization
optcontrol(example_name, H_d, H_c, X_0, X_targ, n_ts, evo_time, p_type, initial_control, output_num, output_fig, output_control,
           sum_cons_1, fid_err_targ, max_iter, max_wall_time, min_grad, offset)

Step 7: do the rounding and compute the error (optional)

if_rounding = True
if if_rounding:
    # set the rounding type and minimized up time
    rounding_type = "BnB"
    min_up_time = 1

    # do the rounding
    bin_amps = rounding(output_control, rounding_type, min_up_time)
	
    # compute and output the objective value after rounding
    rounding_result = time_evolution(
        H_d.full(), [h_c.full() for h_c in H_c], n_ts, evo_time, bin_amps.T, X_0.full(), sum_cons_1)
    f = open(output_control.split(".csv")[0] + "_binary_" + str(min_up_time) + ".log", "w+")
    print("Rounding result: ", file=f)
    print(rounding_result, file=f)
    print("Final objective value: ", file=f)
    print(compute_obj(X_targ, rounding_result), file=f)

xinyufei/Quantum-Pulse-OPT