Source code for pennylane.templates.subroutines.commuting_evolution

# Copyright 2018-2021 Xanadu Quantum Technologies Inc.

# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at

#     http://www.apache.org/licenses/LICENSE-2.0

# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
r"""
Contains the CommutingEvolution template.
"""
# pylint: disable-msg=too-many-arguments,import-outside-toplevel
import pennylane as qml
from pennylane.operation import Operation, AnyWires


[docs]class CommutingEvolution(Operation): r"""Applies the time-evolution operator for a Hamiltonian expressed as a linear combination of mutually commuting Pauli words. A commuting Hamiltonian is of the form .. math:: H \ = \ \displaystyle\sum_{j} c_j P_j, where :math:`P_j` are mutually commutative Pauli words and :math:`c_j` are real coefficients. The time-evolution under a commuting Hamiltonian is given by a unitary of the form .. math:: U(t) \ = \ e^{-i H t} \ = \exp(-i t \displaystyle\sum_j c_j P_j) = \displaystyle\prod_j \exp(-i t c_j P_j). If the Hamiltonian has a small number of unique eigenvalues, partial derivatives of observable expectation values, i.e. .. math:: \langle 0 | W(t)^\dagger O W(t) | 0 \rangle, where :math:`W(t) = V U(t) Y` for some :math:`V` and :math:`Y`, taken with respect to :math:`t` may be efficiently computed through generalized parameter shift rules. When initialized, this template will automatically compute the parameter-shift rule if given the Hamiltonian's eigenvalue frequencies, i.e., the unique positive differences between eigenvalues. .. warning:: This template uses the :class:`~.ApproxTimeEvolution` operation with ``n=1`` in order to implement the time evolution, as a single-step Trotterization is exact for a commuting Hamiltonian. - If the input Hamiltonian contains Pauli words which do not commute, the compilation of the time evolution operator to a sequence of gates will not equate to the exact propagation under the given Hamiltonian. - Furthermore, if the specified frequencies do not correspond to the true eigenvalue frequency spectrum of the commuting Hamiltonian, computed gradients will be incorrect in general. Args: hamiltonian (.Hamiltonian): The commuting Hamiltonian defining the time-evolution operator. The Hamiltonian must be explicitly written in terms of products of Pauli gates (:class:`~.PauliX`, :class:`~.PauliY`, :class:`~.PauliZ`, and :class:`~.Identity`). time (int or float): The time of evolution, namely the parameter :math:`t` in :math:`e^{- i H t}`. Keyword args: frequencies (tuple[int or float]): The unique positive differences between eigenvalues in the spectrum of the Hamiltonian. If the frequencies are not given, the cost function partial derivative will be computed using the standard two-term shift rule applied to the constituent Pauli words in the Hamiltonian individually. shifts (tuple[int or float]): The parameter shifts to use in obtaining the generalized parameter shift rules. If unspecified, equidistant shifts are used. .. details:: :title: Usage Details The template is used inside a qnode: .. code-block:: python import pennylane as qml n_wires = 2 dev = qml.device('default.qubit', wires=n_wires) coeffs = [1, -1] obs = [qml.X(0) @ qml.Y(1), qml.Y(0) @ qml.X(1)] hamiltonian = qml.Hamiltonian(coeffs, obs) frequencies = (2, 4) @qml.qnode(dev) def circuit(time): qml.X(0) qml.CommutingEvolution(hamiltonian, time, frequencies) return qml.expval(qml.Z(0)) >>> circuit(1) 0.6536436208636115 """ num_wires = AnyWires grad_method = None def _flatten(self): h = self.hyperparameters["hamiltonian"] data = (self.data[0], h) return data, (self.hyperparameters["frequencies"], self.hyperparameters["shifts"]) @classmethod def _unflatten(cls, data, metadata) -> "CommutingEvolution": return cls(data[1], data[0], frequencies=metadata[0], shifts=metadata[1]) def __init__(self, hamiltonian, time, frequencies=None, shifts=None, id=None): # pylint: disable=import-outside-toplevel from pennylane.gradients.general_shift_rules import ( generate_shift_rule, ) if not isinstance(hamiltonian, qml.Hamiltonian): type_name = type(hamiltonian).__name__ raise TypeError(f"hamiltonian must be of type pennylane.Hamiltonian, got {type_name}") trainable_hamiltonian = qml.math.requires_grad(hamiltonian.coeffs) if frequencies is not None and not trainable_hamiltonian: c, s = generate_shift_rule(frequencies, shifts).T recipe = qml.math.stack([c, qml.math.ones_like(c), s]).T self.grad_recipe = (recipe,) + (None,) * len(hamiltonian.data) self.grad_method = "A" self._hyperparameters = { "hamiltonian": hamiltonian, "frequencies": frequencies, "shifts": shifts, } super().__init__(time, *hamiltonian.parameters, wires=hamiltonian.wires, id=id)
[docs] @staticmethod def compute_decomposition( time, *coeffs, wires, hamiltonian, **kwargs ): # pylint: disable=arguments-differ,unused-argument r"""Representation of the operator as a product of other operators. .. math:: O = O_1 O_2 \dots O_n. Args: time_and_coeffs (list[tensor_like or float]): list of coefficients of the Hamiltonian, prepended by the time variable wires (Any or Iterable[Any]): wires that the operator acts on hamiltonian (.Hamiltonian): The commuting Hamiltonian defining the time-evolution operator. frequencies (tuple[int or float]): The unique positive differences between eigenvalues in the spectrum of the Hamiltonian. shifts (tuple[int or float]): The parameter shifts to use in obtaining the generalized parameter shift rules. If unspecified, equidistant shifts are used. .. seealso:: :meth:`~.CommutingEvolution.decomposition`. Returns: list[.Operator]: decomposition of the operator """ # uses standard PauliRot decomposition through ApproxTimeEvolution. hamiltonian = qml.Hamiltonian(coeffs, hamiltonian.ops) return [qml.ApproxTimeEvolution(hamiltonian, time, 1)]
[docs] def adjoint(self): hamiltonian = qml.Hamiltonian(self.parameters[1:], self.hyperparameters["hamiltonian"].ops) time = self.parameters[0] frequencies = self.hyperparameters["frequencies"] shifts = self.hyperparameters["shifts"] return CommutingEvolution(hamiltonian, -time, frequencies, shifts)