Source code for qutip.qip.device.processor

from collections.abc import Iterable
import warnings
from copy import deepcopy

import numpy as np
from scipy.interpolate import CubicSpline

from qutip.qobj import Qobj
from qutip.qobjevo import QobjEvo
from qutip.operators import identity
from qutip.qip.operations.gates import expand_operator, globalphase
from qutip.tensor import tensor
from qutip.mesolve import mesolve
from qutip.mcsolve import mcsolve
from qutip import Options
from qutip.qip.circuit import QubitCircuit
from qutip.qip.noise import (
    Noise, RelaxationNoise, DecoherenceNoise,
    ControlAmpNoise, RandomNoise, process_noise)
from qutip.qip.pulse import Pulse, Drift, _merge_qobjevo, _fill_coeff


__all__ = ['Processor']


[docs]class Processor(object): """ A simulator of a quantum device based on the QuTiP solver :func:`qutip.mesolve`. It is defined by the available driving Hamiltonian and the decoherence time for each component systems. The processor can simulate the evolution under the given control pulses. Noisy evolution is supported by :class:`.Noise` and can be added to the processor. Parameters ---------- N: int The number of component systems. t1: list or float, optional Characterize the decoherence of amplitude damping for each qubit. A list of size `N` or a float for all qubits. t2: list of float, optional Characterize the decoherence of dephasing for each qubit. A list of size `N` or a float for all qubits. dims: list, optional The dimension of each component system. Default value is a qubit system of ``dim=[2,2,2,...,2]`` spline_kind: str, optional Type of the coefficient interpolation. Default is "step_func" Note that they have different requirement for the length of ``coeff``. - "step_func": The coefficient will be treated as a step function. E.g. ``tlist=[0,1,2]`` and ``coeff=[3,2]``, means that the coefficient is 3 in t=[0,1) and 2 in t=[2,3). It requires ``len(coeff)=len(tlist)-1`` or ``len(coeff)=len(tlist)``, but in the second case the last element of `coeff` has no effect. - "cubic": Use cubic interpolation for the coefficient. It requires ``len(coeff)=len(tlist)`` Attributes ---------- N: int The number of component systems. pulses: list of :class:`.Pulse` A list of control pulses of this device t1: float or list Characterize the decoherence of amplitude damping of each qubit. t2: float or list Characterize the decoherence of dephasing for each qubit. noise: :class:`.Noise`, optional A list of noise objects. They will be processed when creating the noisy :class:`qutip.QobjEvo` from the processor or run the simulation. drift: :class:`qutip.qip.pulse.Drift` A `Drift` object representing the drift Hamiltonians. dims: list The dimension of each component system. Default value is a qubit system of ``dim=[2,2,2,...,2]`` spline_kind: str Type of the coefficient interpolation. See parameters of :class:`.Processor` for details. """ def __init__(self, N, t1=None, t2=None, dims=None, spline_kind="step_func"): self.N = N self.pulses = [] self.t1 = t1 self.t2 = t2 self.noise = [] self.drift = Drift() if dims is None: self.dims = [2] * N else: self.dims = dims self.pulse_mode = "discrete" self.spline_kind = spline_kind @property def num_qubits(self): return self.N @num_qubits.setter def num_qubits(self, value): self.N = value
[docs] def add_drift(self, qobj, targets, cyclic_permutation=False): """ Add a drift Hamiltonians. The drift Hamiltonians are intrinsic of the quantum system and cannot be controlled by external field. Parameters ---------- qobj: :class:`qutip.Qobj` The drift Hamiltonian. targets: list The indices of the target qubits (or subquantum system of other dimensions). """ if not isinstance(qobj, Qobj): raise TypeError("The drift Hamiltonian must be a qutip.Qobj.") if not qobj.isherm: raise ValueError("The drift Hamiltonian must be Hermitian.") num_qubits = len(qobj.dims[0]) if targets is None: targets = list(range(num_qubits)) if not isinstance(targets, list): targets = [targets] if cyclic_permutation: for i in range(self.N): temp_targets = [(t + i) % self.N for t in targets] self.drift.add_drift(qobj, temp_targets) else: self.drift.add_drift(qobj, targets)
[docs] def add_control(self, qobj, targets=None, cyclic_permutation=False, label=None): """ Add a control Hamiltonian to the processor. It creates a new :class:`.Pulse` object for the device that is turned off (``tlist = None``, ``coeff = None``). To activate the pulse, one can set its `tlist` and `coeff`. Parameters ---------- qobj: :class:`qutip.Qobj` The Hamiltonian for the control pulse.. targets: list, optional The indices of the target qubits (or subquantum system of other dimensions). cyclic_permutation: bool, optional If true, the Hamiltonian will be expanded for all cyclic permutation of the target qubits. label: str, optional The label (name) of the pulse """ # Check validity of ctrl if not isinstance(qobj, Qobj): raise TypeError("The control Hamiltonian must be a qutip.Qobj.") if not qobj.isherm: raise ValueError("The control Hamiltonian must be Hermitian.") num_qubits = len(qobj.dims[0]) if targets is None: targets = list(range(num_qubits)) if not isinstance(targets, list): targets = [targets] if cyclic_permutation: for i in range(self.N): temp_targets = [(t + i) % self.N for t in targets] if label is not None: temp_label = label + "_" + str(temp_targets) temp_label = label self.pulses.append( Pulse(qobj, temp_targets, spline_kind=self.spline_kind, label=temp_label)) else: self.pulses.append( Pulse(qobj, targets, spline_kind=self.spline_kind, label=label) )
def find_pulse(self, pulse_name): if isinstance(pulse_name, str): try: return self.pulses[self.pulse_dict[pulse_name]] except (KeyError): raise KeyError( "Pulse name {} undefined. " "Please define it in the attribute " "`pulse_dict`.".format(pulse_name)) elif isinstance(pulse_name, int): return self.pulses[pulse_name] else: raise TypeError( "pulse_name is either a string or an integer, not " "{}".format(type(pulse_name)) ) @property def ctrls(self): """ A list of Hamiltonians of all pulses. """ result = [] for pulse in self.pulses: result.append(pulse.get_ideal_qobj(self.dims)) return result @property def coeffs(self): """ A list of the coefficients for all control pulses. """ if not self.pulses: return None coeffs_list = [pulse.coeff for pulse in self.pulses] return coeffs_list @coeffs.setter def coeffs(self, coeffs_list): for i, coeff in enumerate(coeffs_list): self.pulses[i].coeff = coeff @property def pulse_mode(self): if self.spline_kind == "step_func": return "discrete" elif self.spline_kind == "cubic": return "continuous" else: raise ValueError( "Saved spline_kind not understood.") @pulse_mode.setter def pulse_mode(self, mode): if mode == "discrete": spline_kind = "step_func" elif mode == "continuous": spline_kind = "cubic" else: raise ValueError( "Pulse mode must be either discrete or continuous.") self.spline_kind = spline_kind for pulse in self.pulses: pulse.spline_kind = spline_kind
[docs] def get_full_tlist(self, tol=1.0e-10): """ Return the full tlist of the ideal pulses. If different pulses have different time steps, it will collect all the time steps in a sorted array. Returns ------- full_tlist: array-like 1d The full time sequence for the ideal evolution. """ full_tlist = [pulse.tlist for pulse in self.pulses if pulse.tlist is not None] if not full_tlist: return None full_tlist = np.unique(np.sort(np.hstack(full_tlist))) # account for inaccuracy in float-point number full_tlist = np.concatenate( (full_tlist[:1], full_tlist[1:][np.diff(full_tlist) > tol])) return full_tlist
[docs] def get_full_coeffs(self, full_tlist=None): """ Return the full coefficients in a 2d matrix form. Each row corresponds to one pulse. If the `tlist` are different for different pulses, the length of each row will be same as the `full_tlist` (see method `get_full_tlist`). Interpolation is used for adding the missing coefficient according to `spline_kind`. Returns ------- coeffs: array-like 2d The coefficients for all ideal pulses. """ # TODO add tests self._is_pulses_valid() if not self.pulses: return np.array((0, 0), dtype=float) if full_tlist is None: full_tlist = self.get_full_tlist() coeffs_list = [] for pulse in self.pulses: if pulse.tlist is None and pulse.coeff is None: coeffs_list.append(np.zeros(len(full_tlist))) continue if not isinstance(pulse.coeff, (bool, np.ndarray)): raise ValueError( "get_full_coeffs only works for " "NumPy array or bool coeff.") if isinstance(pulse.coeff, bool): if pulse.coeff: coeffs_list.append(np.ones(len(full_tlist))) else: coeffs_list.append(np.zeros(len(full_tlist))) continue if self.spline_kind == "step_func": arg = {"_step_func_coeff": True} coeffs_list.append( _fill_coeff(pulse.coeff, pulse.tlist, full_tlist, arg)) elif self.spline_kind == "cubic": coeffs_list.append( _fill_coeff(pulse.coeff, pulse.tlist, full_tlist, {})) else: raise ValueError("Unknown spline kind.") return np.array(coeffs_list)
[docs] def set_all_tlist(self, tlist): """ Set the same `tlist` for all the pulses. Parameters ---------- tlist: array-like, optional A list of time at which the time-dependent coefficients are applied. See :class:`.Pulse` for detailed information` """ if isinstance(tlist, list) and len(tlist) == len(self.pulses): for i, pulse in enumerate(self.pulses): pulse.tlist = tlist[i] else: for pulse in self.pulses: pulse.tlist = tlist
[docs] def add_pulse(self, pulse): """ Add a new pulse to the device. Parameters ---------- pulse: :class:`.Pulse` `Pulse` object to be added. """ if isinstance(pulse, Pulse): if pulse.spline_kind is None: pulse.spline_kind = self.spline_kind self.pulses.append(pulse) else: raise ValueError("Invalid input, pulse must be a Pulse object")
[docs] def remove_pulse(self, indices=None, label=None): """ Remove the control pulse with given indices. Parameters ---------- indices: int or list of int The indices of the control Hamiltonians to be removed. label: str The label of the pulse """ if indices is not None: if not isinstance(indices, Iterable): indices = [indices] indices.sort(reverse=True) for ind in indices: del self.pulses[ind] else: for ind, pulse in enumerate(self.pulses): if pulse.label == label: del self.pulses[ind]
def _is_pulses_valid(self): """ Check if the pulses are in the correct shape. Returns: bool If they are valid or not """ for i, pulse in enumerate(self.pulses): if pulse.coeff is None or isinstance(pulse.coeff, bool): # constant pulse continue if pulse.tlist is None: raise ValueError( "Pulse id={} is invalid. " "Please define a tlist for the pulse.".format(i)) if pulse.tlist is not None and pulse.coeff is None: raise ValueError( "Pulse id={} is invalid. " "Please define a coeff for the pulse.".format(i)) coeff_len = len(pulse.coeff) tlist_len = len(pulse.tlist) if pulse.spline_kind == "step_func": if coeff_len == tlist_len-1 or coeff_len == tlist_len: pass else: raise ValueError( "The length of tlist and coeff of the pulse " "labelled {} is invalid. " "It's either len(tlist)=len(coeff) or " "len(tlist)-1=len(coeff) for coefficients " "as step function".format(i)) else: if coeff_len == tlist_len: pass else: raise ValueError( "The length of tlist and coeff of the pulse " "labelled {} is invalid. " "It should be either len(tlist)=len(coeff)".format(i)) return True
[docs] def add_noise(self, noise): """ Add a noise object to the processor Parameters ---------- noise: :class:`.Noise` The noise object defined outside the processor """ if isinstance(noise, Noise): self.noise.append(noise) else: raise TypeError("Input is not a Noise object.")
[docs] def save_coeff(self, file_name, inctime=True): """ Save a file with the control amplitudes in each timeslot. Parameters ---------- file_name: string Name of the file. inctime: bool, optional True if the time list should be included in the first column. """ self._is_pulses_valid() coeffs = np.array(self.get_full_coeffs()) if inctime: shp = coeffs.T.shape data = np.empty((shp[0], shp[1] + 1), dtype=np.float64) data[:, 0] = self.get_full_tlist() data[:, 1:] = coeffs.T else: data = coeffs.T np.savetxt(file_name, data, delimiter='\t', fmt='%1.16f')
[docs] def read_coeff(self, file_name, inctime=True): """ Read the control amplitudes matrix and time list saved in the file by `save_amp`. Parameters ---------- file_name: string Name of the file. inctime: bool, optional True if the time list in included in the first column. Returns ------- tlist: array_like The time list read from the file. coeffs: array_like The pulse matrix read from the file. """ data = np.loadtxt(file_name, delimiter='\t') if not inctime: self.coeffs = data.T return self.coeffs else: tlist = data[:, 0] self.set_all_tlist(tlist) self.coeffs = data[:, 1:].T return self.get_full_tlist, self.coeffs
[docs] def get_noisy_pulses(self, device_noise=False, drift=False): """ It takes the pulses defined in the `Processor` and adds noise according to `Processor.noise`. It does not modify the pulses saved in `Processor.pulses` but returns a new list. The length of the new list of noisy pulses might be longer because of drift Hamiltonian and device noise. They will be added to the end of the pulses list. Parameters ---------- device_noise: bool, optional If true, include pulse independent noise such as single qubit Relaxation. Default is False. drift: bool, optional If true, include drift Hamiltonians. Default is False. Returns ------- noisy_pulses: list of :class:`.Pulse` A list of noisy pulses. """ pulses = deepcopy(self.pulses) noisy_pulses = process_noise( pulses, self.noise, self.dims, t1=self.t1, t2=self.t2, device_noise=device_noise) if drift: noisy_pulses += [self.drift] return noisy_pulses
[docs] def get_qobjevo(self, args=None, noisy=False): """ Create a :class:`qutip.QobjEvo` representation of the evolution. It calls the method `get_noisy_pulses` and create the `QobjEvo` from it. Parameters ---------- args: dict, optional Arguments for :class:`qutip.QobjEvo` noisy: bool, optional If noise are included. Default is False. Returns ------- qobjevo: :class:`qutip.QobjEvo` The :class:`qutip.QobjEvo` representation of the unitary evolution. c_ops: list of :class:`qutip.QobjEvo` A list of lindblad operators is also returned. if ``noisy==Flase``, it is always an empty list. """ # TODO test it for non array-like coeff # check validity self._is_pulses_valid() if args is None: args = {} else: args = args # set step function if not noisy: dynamics = self.pulses else: dynamics = self.get_noisy_pulses( device_noise=True, drift=True) qu_list = [] c_ops = [] for pulse in dynamics: if noisy: qu, new_c_ops = pulse.get_noisy_qobjevo(dims=self.dims) c_ops += new_c_ops else: qu = pulse.get_ideal_qobjevo(dims=self.dims) qu_list.append(qu) final_qu = _merge_qobjevo(qu_list) final_qu.args.update(args) # bring all c_ops to the same tlist, won't need it in QuTiP 5 full_tlist = self.get_full_tlist() temp = [] for c_op in c_ops: temp.append(_merge_qobjevo([c_op], full_tlist)) c_ops = temp if noisy: return final_qu, c_ops else: return final_qu, []
[docs] def run_analytically(self, init_state=None, qc=None): """ Simulate the state evolution under the given `qutip.QubitCircuit` with matrice exponentiation. It will calculate the propagator with matrix exponentiation and return a list of :class:`qutip.Qobj`. This method won't include noise or collpase. Parameters ---------- qc: :class:`.QubitCircuit`, optional Takes the quantum circuit to be implemented. If not given, use the quantum circuit saved in the processor by ``load_circuit``. init_state: :class:`qutip.Qobj`, optional The initial state of the qubits in the register. Returns ------- U_list: list A list of propagators obtained for the physical implementation. """ if init_state is not None: U_list = [init_state] else: U_list = [] tlist = self.get_full_tlist() coeffs = self.get_full_coeffs() # Compute drift Hamiltonians H_drift = 0 for drift_ham in self.drift.drift_hamiltonians: H_drift += drift_ham.get_qobj(self.dims) # Compute control Hamiltonians for n in range(len(tlist)-1): H = H_drift + sum( [coeffs[m, n] * self.ctrls[m] for m in range(len(self.ctrls))]) dt = tlist[n + 1] - tlist[n] U = (-1j * H * dt).expm() U = self.eliminate_auxillary_modes(U) U_list.append(U) try: # correct_global_phase are defined for ModelProcessor if self.correct_global_phase and self.global_phase != 0: U_list.append(globalphase( self.global_phase, N=self.num_qubits) ) except AttributeError: pass return U_list
[docs] def run(self, qc=None): """ Calculate the propagator of the evolution by matrix exponentiation. This method won't include noise or collpase. Parameters ---------- qc: :class:`.QubitCircuit`, optional Takes the quantum circuit to be implemented. If not given, use the quantum circuit saved in the processor by `load_circuit`. Returns ------- U_list: list The propagator matrix obtained from the physical implementation. """ if qc: self.load_circuit(qc) return self.run_analytically(qc=qc, init_state=None)
[docs] def run_state(self, init_state=None, analytical=False, states=None, noisy=True, solver="mesolve", **kwargs): """ If `analytical` is False, use :func:`qutip.mesolve` to calculate the time of the state evolution and return the result. Other arguments of mesolve can be given as keyword arguments. If `analytical` is True, calculate the propagator with matrix exponentiation and return a list of matrices. Noise will be neglected in this option. Parameters ---------- init_state: Qobj Initial density matrix or state vector (ket). analytical: bool If True, calculate the evolution with matrices exponentiation. states: :class:`qutip.Qobj`, optional Old API, same as init_state. solver: str "mesolve" or "mcsolve" **kwargs Keyword arguments for the qutip solver. Returns ------- evo_result: :class:`qutip.solver.Result` If ``analytical`` is False, an instance of the class :class:`qutip.solver.Result` will be returned. If ``analytical`` is True, a list of matrices representation is returned. """ if states is not None: warnings.warn( "states will be deprecated and replaced by init_state", DeprecationWarning) if init_state is None and states is None: raise ValueError("Qubit state not defined.") elif init_state is None: # just to keep the old parameters `states`, # it is replaced by init_state init_state = states if analytical: if kwargs or self.noise: raise warnings.warn( "Analytical matrices exponentiation" "does not process noise or" "any keyword arguments.") return self.run_analytically(init_state=init_state) # kwargs can not contain H or tlist if "H" in kwargs or "tlist" in kwargs: raise ValueError( "`H` and `tlist` are already specified by the processor " "and can not be given as a keyword argument") # construct qobjevo for unitary evolution if "args" in kwargs: noisy_qobjevo, sys_c_ops = self.get_qobjevo( args=kwargs["args"], noisy=noisy) else: noisy_qobjevo, sys_c_ops = self.get_qobjevo(noisy=noisy) # add collpase operators into kwargs if "c_ops" in kwargs: if isinstance(kwargs["c_ops"], (Qobj, QobjEvo)): kwargs["c_ops"] += [kwargs["c_ops"]] + sys_c_ops else: kwargs["c_ops"] += sys_c_ops else: kwargs["c_ops"] = sys_c_ops # set the max step size as 1/2 of the smallest gate time. options = kwargs.get("options", Options()) if options.max_step == 0.: options.max_step = np.min(np.diff(self.get_full_tlist())) / 2 kwargs["options"] = options # choose solver: if solver == "mesolve": evo_result = mesolve( H=noisy_qobjevo, rho0=init_state, tlist=noisy_qobjevo.tlist, **kwargs) elif solver == "mcsolve": evo_result = mcsolve( H=noisy_qobjevo, psi0=init_state, tlist=noisy_qobjevo.tlist, **kwargs) return evo_result
[docs] def load_circuit(self, qc): """ Translate an :class:`.QubitCircuit` to its corresponding Hamiltonians. (Defined in subclasses) """ raise NotImplementedError("Use the function in the sub-class")
[docs] def eliminate_auxillary_modes(self, U): """ Eliminate the auxillary modes like the cavity modes in cqed. (Defined in subclasses) """ return U
[docs] def get_operators_labels(self): """ Get the labels for each Hamiltonian. It is used in the method``plot_pulses``. It is a 2-d nested list, in the plot, a different color will be used for each sublist. """ label_list = [] for pulse in self.pulses: label_list.append(pulse.label) return [label_list]
[docs] def plot_pulses( self, title=None, figsize=(12, 6), dpi=None, show_axis=False, rescale_pulse_coeffs=True, num_steps=1000): """ Plot the ideal pulse coefficients. Parameters ---------- title: str, optional Title for the plot. figsize: tuple, optional The size of the figure. dpi: int, optional The dpi of the figure. show_axis: bool, optional If the axis are shown. rescale_pulse_coeffs: bool, optional Rescale the hight of each pulses. num_steps: int, optional Number of time steps in the plot. Returns ------- fig: matplotlib.figure.Figure The `Figure` object for the plot. ax: matplotlib.axes._subplots.AxesSubplot The axes for the plot. Notes ----- ``plot_pulses`` only works for array_like coefficients """ import matplotlib.pyplot as plt import matplotlib.gridspec as gridspec color_list = plt.rcParams['axes.prop_cycle'].by_key()['color'] # create a axis for each pulse fig = plt.figure(figsize=figsize, dpi=dpi) grids = gridspec.GridSpec(len(self.pulses), 1) grids.update(wspace=0., hspace=0.) tlist = np.linspace(0., self.get_full_tlist()[-1], num_steps) dt = tlist[1] - tlist[0] # make sure coeffs start and end with zero, for ax.fill tlist = np.hstack(([-dt*1.e-20], tlist, [tlist[-1] + dt*1.e-20])) coeffs = [] for pulse in self.pulses: coeffs.append(_pulse_interpolate(pulse, tlist)) pulse_ind = 0 axis = [] for i, label_group in enumerate(self.get_operators_labels()): for j, label in enumerate(label_group): grid = grids[pulse_ind] ax = plt.subplot(grid) axis.append(ax) ax.fill(tlist, coeffs[pulse_ind], color_list[i], alpha=0.7) ax.plot(tlist, coeffs[pulse_ind], color_list[i]) if rescale_pulse_coeffs: ymax = np.max(np.abs(coeffs[pulse_ind])) * 1.1 else: ymax = np.max(np.abs(coeffs)) * 1.1 if ymax != 0.: ax.set_ylim((-ymax, ymax)) # disable frame and ticks if not show_axis: ax.set_xticks([]) ax.spines['bottom'].set_visible(False) ax.spines['top'].set_visible(False) ax.spines['right'].set_visible(False) ax.spines['left'].set_visible(False) ax.set_yticks([]) ax.set_ylabel(label, rotation=0) pulse_ind += 1 if i == 0 and j == 0 and title is not None: ax.set_title(title) fig.tight_layout() return fig, axis
def _pulse_interpolate(pulse, tlist): """ A function that calls Scipy interpolation routine. Used for plotting. """ if pulse.tlist is None and pulse.coeff is None: coeff = np.zeros(len(tlist)) return coeff if isinstance(pulse.coeff, bool): if pulse.coeff: coeff = np.ones(len(tlist)) else: coeff = np.zeros(len(tlist)) return coeff coeff = pulse.coeff if len(coeff) == len(pulse.tlist)-1: # for discrete pulse coeff = np.concatenate([coeff, [0]]) from scipy import interpolate if pulse.spline_kind == "step_func": kind = "previous" else: kind = "cubic" inter = interpolate.interp1d( pulse.tlist, coeff, kind=kind, bounds_error=False, fill_value=0.0) return inter(tlist)