US-12626785-B2 - Methods and systems for quantum computing enabled molecular AB initio simulations

Abstract

The present disclosure provides methods and systems for using a hybrid architecture of classical and non-classical (e.g., quantum) computing to compute the quantum mechanical energy and/or electronic structure of a chemical system, as well as to identify stable conformations of a chemical system (e.g., a molecule) and/or to perform an ab initio molecular dynamics calculation or simulation on the chemical system.

Inventors

• Takeshi Yamazaki
• Arman ZARIBAFIYAN
• Shunji Matsuura
• Rudolf PLESCH

Assignees

• GOOD CHEMISTRY INC.

Dates

Publication Date

20260512

Application Date

20211109

Claims (17)

1. 1 . A method for performing a quantum mechanical energy or electronic structure calculation for a chemical system, said method being implemented by a distributed computing system comprising at least three computers, wherein the at least three computers comprises a central classical computer and multiple electronic structure solvers, each electronic structure solver implementing one or more respective electronic structure solver types, each electronic structure solver type being defined by an algorithmic method, a hardware configuration, or both, the electronic structure solvers comprising quantum computing systems, non-classical computing systems, or hybrid computing units, said method comprising: (a) decomposing, at the central classical computer, at least one conformation within an ensemble of conformations of said chemical system into a plurality of molecular fragments; (b) generating, at the central classical computer and for each of one or more of said plurality of molecular fragments, a respective data structure for the molecular fragment, wherein the data structure specifies (i) a specification of the molecular fragment, (ii) an electronic structure solver type and (iii) solver parameters to be passed to an electronic structure solver of the specified type, the solver parameters comprising a hardware backend type and ansatz type to be used by the electronic structure solver of the specified type to determine quantum mechanical energies or electronic structures for the molecular fragment; (c) identifying, at the central classical computer and for each of the one or more of said plurality of molecular fragments, an electronic structure solver included in the distributed computing system dedicated to the type specified in the data structure for the molecular fragment: (d) dispatching the one or more of said plurality of molecular fragments and respective data structures from the central classical computer to the electronic structure solvers included in the distributed computing system of the types specified in the respective data structures; (e) determining, by the electronic structure solvers included in the distributed system of the types specified in the respective data structures, using the hardware backend types specified in the respective data structures and according to the solver parameters in the respective data structures, quantum mechanical energies or electronic structures of at least a subset of said plurality of molecular fragments; (f) combining, at the central classical computer said quantum mechanical energies or electronic structures determined in (e); and (g) electronically outputting a report indicative of said quantum mechanical energies or electronic structures combined in (f).
2. 2 . The method of claim 1 , wherein the non-classical computers comprise at least one Hitachi Ising solver or a coherent Ising machine based on optical parameters and wherein the quantum computer systems comprise one or more members selected from the group consisting of: a quantum hardware device and a classical simulator of a quantum circuit.
3. 3 . The method of claim 1 , wherein a quantum mechanical energy of said quantum mechanical energies comprises nuclear-nuclear repulsion energy.
4. 4 . The method of claim 1 , further comprising providing an input to said central classical computer, said input comprising a set of atomic coordinates for said chemical system.
5. 5 . The method of claim 1 , further comprising performing (a)-(f) for two or more conformations within said ensemble of conformations of said chemical system.
6. 6 . The method of claim 1 , wherein (a) comprises applying one or more members selected from the group consisting of: a fragment molecular orbital (FMO) method, a divide-and-conquer (DC) method, a density matrix embedding theory (DMET) method, a density matrix renormalization group (DMRG) method, a tensor network, and a method of increments.
7. 7 . The method of claim 1 , wherein (e) comprises: determining a fermionic Hamiltonian of a molecular fragment of said at least said subset of said plurality of molecular fragments; transforming said fermionic Hamiltonian into an equivalent qubit Hamiltonian; transforming said qubit Hamiltonian into a quantum circuit; and determining, using said quantum circuit, a quantum mechanical energy or electronic structure of said molecular fragment.
8. 8 . The method of claim 7 , further comprising determining said quantum mechanical energy or electronic structure using one or more members selected from the group consisting of: a molecular Hamiltonian and an electronic Hamiltonian.
9. 9 . The method of claim 7 , wherein transforming said fermionic Hamiltonian into an equivalent qubit Hamiltonian comprises transforming a fermionic operator of a Hamiltonian to a qubit operator.
10. 10 . The method of claim 1 , further comprising, based on the quantum mechanical energies or electronic structures combined in (f), performing an ab initio molecular dynamics (AIMD) simulation of said chemical system.
11. 11 . The method of claim 10 , wherein said AIMD simulation comprises: prior to (a), obtaining an indication of a chemical system, said indication comprising coordinates of each particle of a plurality of particles in said chemical system and velocities of each particle in said chemical system; and subsequent to (f): (i) determining, from said combined energy or electronic structure, a force on each particle in said chemical system; (ii) updating said coordinates of said each particle in said chemical system and said velocities of said each particle in said chemical system; and (iii) electronically outputting a report indicative of said coordinates or said velocities.
12. 12 . The method of claim 11 , wherein (i) comprises applying Jordan's quantum algorithm for numerical gradient estimation to said quantum mechanical energy or electronic structure.
13. 13 . The method of claim 11 , wherein (ii) comprises applying one or more members selected from the group consisting of: a Verlet procedure, a velocity Verlet procedure, symplectic integration, Runge-Kutta integration, and Beeman integration.
14. 14 . The method of claim 1 , further comprising dispatching one or more of said plurality of fragments to one or more remote endpoints and receiving said quantum mechanical energies or electronic structures from said one or more remote endpoints, wherein said one or more remote endpoints comprise portions of a cloud computing system.
15. 15 . The method of claim 14 , wherein at least one of said one or more remote endpoints comprises a non-classical computer.
16. 16 . The method of claim 1 , further comprising, prior to (a), receiving said at least one conformation from a client-side library and dispatching said at least one conformation to a first remote endpoint, dispatching one or more of said plurality of fragments to one or more remote second endpoints and receiving said quantum mechanical energies or electronic structures from said second one or more remote endpoints, and transmitting said report to said client-side library, and wherein at least one of (a) and (f) occur at said first remote endpoint.
17. 17 . A distributed computing system for performing a quantum mechanical energy or electronic structure calculation for a chemical system, the distributed computing system comprising a central classical computer operatively coupled to a memory, said memory comprising instructions for performing a quantum mechanical energy or electronic structure calculation for a chemical system, wherein the distributed computing system further comprises multiple electronic structure solvers, each electronic structure solver implementing one or more respective electronic structure solver types, each electronic structure solver type being defined by an algorithmic method, a hardware configuration, or both, the electronic structure solvers comprising quantum computing systems, non-classical computing systems, or hybrid computing units, wherein said distributed computing system is configured to at least: (a) decompose, at the central classical computer, at least one conformation within an ensemble of conformations of said chemical system into a plurality of molecular fragments; (b) generate, at the central classical computer and for each of one or more of said plurality of molecular fragments, a respective data structure for the molecular fragment, wherein the data structure specifies (i) a specification of the molecular fragment, (ii) an electronic structure solver type and (iii) solver parameters to be passed to an electronic structure solver of the specified type, the solver parameters comprising a hardware backend type and ansatz type to be used by the electronic structure solver of the specified type to determine quantum mechanical energies or electronic structures for the molecular fragment; (c) identify, at the central classical computer and for each of the one or more of said plurality of molecular fragments, an electronic structure solver included in the distributed computing system of the type specified in the data structure for the molecular fragment; (d) dispatch the one or more of said plurality of molecular fragments and respective data structures to electronic structure solvers included in the distributed computing system of the types specified in the respective data structures; (e) determine, by the electronic structure solvers included in the distributed system of the types specified in the respective data structures, using the hardware backend types specified in the respective data structures and according to the solver parameters in the respective data structures, quantum mechanical energies or electronic structures of at least a subset of said plurality of molecular fragments; (f) combine said quantum mechanical energies or electronic structures determined in (e); and (g) electronically output a report indicative of said quantum mechanical energies or electronic structures combined in (f).

Description

CROSS-REFERENCE This application is a continuation of International Application No. PCT/CA2020/050641, filed May 12, 2020, which claims the benefit of U.S. Provisional Application Ser. No. 62/949,263, filed Dec. 17, 2019, and U.S. Provisional Application Ser. No. 62/847,141, filed May 13, 2019, each of which is entirely incorporated herein by reference for all purposes. BACKGROUND In chemistry and biology, the identification and the prediction of the electronic structure and the most energetically stable conformers of a molecule have significant importance as molecular function is inherently embedded in molecular conformation. For example, the reaction rate in a catalyzed reaction can vary significantly based on which of several different conformations of the catalyst are used. As another example, a protein is more functional or functional at all when it forms a certain tertiary structure. In order to accurately identify and predict the electronic structure and the most stable conformers, highly accurate quantum chemistry methods, such as Coupled-Cluster theory (CC) or Full Configuration Interaction (Full CI), may be performed. However, the computational costs of such methods can exponentially increase with the size of a molecule, and they often become intractable in cases where the size of a molecule exceeds about 50 atoms for CC, and about 10 atoms for Full CI, even when performed on some current state-of-the-art classical computers. Therefore, a highly efficient and accurate computational framework is needed to identify the most stable conformers of industry-relevant chemical compounds and biologically-relevant large molecules. Quantum computing (QC) technology may be capable of computing the quantum mechanical energy and/or electronic structure of a molecule with exponentially less computational resources compared to classical computing. Thus, high-accuracy quantum chemistry calculations that are intractable using classical computing may become tractable using the QC approaches. However, QC approaches may face challenges, such as the high expense and rarity of QC resources. In addition, increasing the number of qubits in a quantum computer is a technologically challenge, which has limited the size of quantum computing devices. In addition, qubits are very sensitive to noise and environmental effects, which may cause them to decohere in a very short amount of time, thereby providing a relatively small window for running meaningful calculations. SUMMARY Recognized herein is the need for quantum algorithms and circuits that efficiently leverage current and near-term quantum computing systems to solve complex quantum chemistry problems. One approach is to decompose an industry-sized problem into subproblems, identify the more complex subproblems, and then use quantum computers to process a subset of problems, for example, those subproblems that are challenging for classical computers. Systems and methods provided herein utilize problem decomposition (PD) techniques in quantum chemistry toward identification and prediction of the electronic structure and a set of the most energetically stable conformers of a molecule. Such PD techniques may include the fragment molecular orbital (FMO) method, the divide-and-conquer (DC) method, the density matrix embedding theory (DMET) method, the density matrix renormalization group (DMRG) method, tensor networks, the method of increments, and others, as described herein. In quantum chemistry, PD techniques have been developed to efficiently compute molecular energies and/or electronic structures with reasonable accuracy using classical computing. In PD techniques, the molecule may be decomposed into smaller fragments such that the quantum mechanical energy and/or electronic structure computation becomes tractable for each fragment. The quantum mechanical energy and/or electronic structure computation may then be performed individually for each fragment. The quantum mechanical energy and/or electronic structure computations resulting from each fragment may be recombined into a solution for the original molecule. Systems and methods provided herein to perform PD techniques on a QC platform may enable quantum mechanical energy and/or electronic structure computations to be performed with a high level of accuracy for each fragment. Further, the small size of each fragment may allow highly accurate computations to be performed on QC devices on which the scale of computations is rather restricted, thereby obtaining the energies and/or electronic structures of complex, industry-relevant molecules efficiently and accurately. The identification of the electronic structure and the most energetically stable conformers of a molecule is a fundamental process in chemistry- and biology-related research and development. While such processes may be performed by actually synthesizing the molecule and using a variety of physicochemical measurements to identify its electronic structure and c