Increasing representation accuracy of quantum simulations without additional quantum resources
Methods, systems and apparatus for simulating physical systems. In one aspect, a method includes the actions of selecting a first set of basis functions for the simulation, wherein the first set of basis functions comprises an active and a virtual set of orbitals; defining a set of expansion operators for the simulation, wherein expansion operators in the set of expansion operators approximate fermionic excitations in an active space spanned by the active set of orbitals and a virtual space spanned by the virtual set of orbitals; performing multiple quantum computations to determine a matrix representation of a Hamiltonian characterizing the system in a second set of basis functions, computing, using the determined matrix representation of the Hamiltonian, eigenvalues and eigenvectors of the Hamiltonian; and determining properties of the physical system using the computed eigenvalues and eigenvectors.
1. A method performed by a hybrid quantum-classical computing device for a quantum simulation of a physical system characterized by a respective Hamiltonian, the method comprising:
selecting, by one or more classical processors included in the hybrid quantum-classical computing device, a first set of basis functions for the quantum simulation, wherein the first set of basis functions comprises i) an active set of orbitals, and ii) a virtual set of orbitals;
defining, by the one or more classical processors, a set of expansion operators for the quantum simulation, wherein expansion operators in the set of expansion operators approximate fermionic excitations in an active space spanned by the active set of orbitals and a virtual space spanned by the virtual set of orbitals and act on both qubits that define the active space and additional qubits that define the virtual space;
performing multiple quantum computations to determine a matrix representation of the Hamiltonian in a second set of basis functions, wherein each basis function in the second set of basis functions comprises a respective expansion operator applied to a wavefunction prepared using the qubits that define the active space, and wherein determining the matrix representation of the Hamiltonian comprises, for each of one or more matrix elements:
determining, by the one or more classical processors, that the matrix element comprises operators that act on the virtual space,
in response to determining that the matrix element comprises operators that act on the virtual space, performing, by the one or more classical processors, a classical computation to contract the matrix element to a matrix element comprising operators that act on the active space only, and
preparing the qubits that define the active space in a quantum state that represents the wavefunction and measuring, by a quantum computing device included in the hybrid quantum-classical computing device and using the quantum state, the operators that act on the active space only to determine a value for the matrix element that comprises contributions from the virtual space that extend beyond the active space without requiring the additional qubits that define the virtual space;
computing, by the one or more classical processors and using the determined matrix representation of the Hamiltonian in the second set of basis functions, eigenvalues and eigenvectors of the Hamiltonian; and
determining, by the one or more classical processors, properties of the physical system using the computed eigenvalues and eigenvectors.
2. The method of claim 1 , further comprising performing multiple computations to determine an overlap matrix in the second set of basis functions, wherein each element of the overlap matrix represents a respective overlap of two basis functions in the second set of basis functions, wherein determining the overlap matrix comprises, for each matrix element:
determining whether the matrix element comprises operators that act on the virtual space or operators that act on the active space only;
in response to determining that the matrix element comprises operators that act on the virtual space, performing a classical computation to contract the matrix element to a matrix element comprising operators that act on the active space only; and
measuring the operators that act on the active space only to determine a value for the matrix element.
3. The method of claim 2 , wherein computing, using the determined matrix representation of the Hamiltonian in the second set of basis functions, eigenvalues and eigenvectors of the Hamiltonian comprises:
computing, using the determined matrix representation of the Hamiltonian in the second set of basis functions and the determined overlap matrix, eigenvalues and eigenvectors of the Hamiltonian.
4. The method of claim 1 , wherein measuring operators that act on the active space only to determine a value for the matrix element comprises:
preparing the wavefunction in the active space; and
measuring a Pauli operator corresponding to a qubit transformation of the matrix element.
5. The method of claim 2 , further comprising, before computing eigenvalues and eigenvectors of the Hamiltonian:
computing eigenvalues and eigenvectors of the determined overlap matrix;
removing eigenvalues of the determined overlap matrix that are equal to zero or smaller than a predetermined threshold to define an updated overlap matrix; and
using the updated overlap matrix to compute the eigenvalues and eigenvectors of the Hamiltonian.
6. The method of claim 1 , wherein the first set of basis functions further comprises a core set of orbitals.
7. The method of claim 1 , wherein defining a set of expansion operators for the quantum simulation further comprises selecting a maximum excitation level for the set of expansion operators.
8. The method of claim 1 , wherein the active space comprises a subspace of total active space that is excited into the virtual space.
9. The method of claim 1 , wherein measuring the operators that act on the active space only to determine a value for the matrix element comprises:
approximating the operators that act on the active space only using cumulant approximations or ensemble variational methods; and
measuring the approximated operators that act on the active space only.
10. The method of claim 1 , wherein the Hamiltonian characterizes an electronic structure of a semiconductor, and wherein simulating the physical system comprises simulating properties of the semiconductor.
11. The method of claim 10 , wherein properties of the semiconductor comprise conductivity or resistance.
12. A method performed by a hybrid quantum-classical computing device for simulating a quantum system, the method comprising:
performing, by a quantum computing device included in the hybrid quantum-classical computing device, a quantum simulation of a quantum system using qubits that define an active space to obtaining a simulation output from the quantum simulation of the quantum system, wherein the quantum simulation comprises a quantum simulation in active space and the simulation output comprises an expected value of the energy of the quantum system in active space; and
classically post-processing the simulation output, comprising adjusting, by classical computation performed by one or more classical processors included in the hybrid quantum-classical computing device, the simulation output using multiple single particle rotations in full space to obtain an estimated energy of the quantum system in full space, the adjusting comprising:
solving a nonlinear optimization problem, wherein the nonlinear optimization problem comprises:
an objective function comprising the expected value of the energy of the quantum system in active space, and
one or more constraints that specify unitary rotations of quantum system orbitals,
wherein solving the nonlinear optimization problem comprises identifying multiple single particle rotations in full space that rotate an underlying basis of the simulation output and minimizes the objective function subject to the one or more constraints; and
applying the identified multiple single particle rotations in full space to rotate the underlying basis of the simulation output in full space and obtain the estimated energy of the quantum system in full space;
wherein classically post-processing the simulation output reintroduces virtual space contributions that extend beyond the active space to the simulation output without requiring additional qubits or gate complexity.
13. The method of claim 12 , wherein the simulation output comprises an estimated energy of the quantum system.
14. The method of claim 12 , wherein the quantum system is characterized by an electronic structure Hamiltonian, and wherein the simulation output comprises a two electron reduced density matrix (2-RDM) for the ground state of the quantum system.
15. The method of claim 12 , wherein solving the nonlinear optimization problem further comprises parameterizing a unitary operator representing the multiple single particle rotations as an exponentiated anti-Hermitian matrix.
16. The method of claim 12 , wherein solving the nonlinear optimization problem comprises parameterizing a unitary operator representing the multiple single particle rotations as a product of Givens rotations.
17. The method of claim 16 , wherein solving the nonlinear optimization problem comprises implementing a multi-configurational self-consistent-field method.
18. The method of claim 12 , wherein the one or more constraints produce normalized and physical wavefunctions.
19. The method of claim 12 , further comprising iteratively, until a change in energy between iterations is lower than a predetermined threshold:
providing the adjusted simulation output as an input for a subsequent quantum simulation in active space of the quantum system; and
obtaining a subsequent simulation output from the subsequent quantum simulation of the quantum system.
20. An apparatus comprising:
quantum hardware; and
one or more classical processors;
wherein the apparatus is configured to perform operations comprising:
selecting, by one or more classical processors included in a hybrid quantum-classical computing device, a first set of basis functions for a quantum simulation, wherein the first set of basis functions comprises i) an active set of orbitals, and ii) a virtual set of orbitals;
defining, by the one or more classical processors, a set of expansion operators for the quantum simulation, wherein expansion operators in the set of expansion operators approximate fermionic excitations in an active space spanned by the active set of orbitals and a virtual space spanned by the virtual set of orbitals and act on both qubits that define the active space and additional qubits that define the virtual space;
performing multiple quantum computations to determine a matrix representation of the Hamiltonian in a second set of basis functions, wherein each basis function in the second set of basis functions comprises a respective expansion operator applied to a wavefunction prepared using the qubits that define the active space, and wherein determining the matrix representation of the Hamiltonian comprises, for each of one or more matrix elements:
determining, by the one or more classical processors, that the matrix element comprises operators that act on the virtual space,
in response to determining that the matrix element comprises operators that act on the virtual space, performing, by the one or more classical processors, a classical computation to contract the matrix element to a matrix element comprising operators that act on the active space only, and
preparing the qubits that define the active space in a quantum state that represents the wavefunction and measuring, by a quantum computing device included in the hybrid quantum-classical computing device and using the quantum state, the operators that act on the active space only to determine a value for the matrix element that comprises contributions from the virtual space that extend beyond the active space without requiring the additional qubits that define the virtual space;
computing, by the one or more classical processors and using the determined matrix representation of the Hamiltonian in the second set of basis functions, eigenvalues and eigenvectors of the Hamiltonian; and
determining, by the one or more classical processors, properties of the physical system using the computed eigenvalues and eigenvectors.